A Cost-Utility Analysis of Sequential Treatment
with Teriparatide/Alendronate Compared to
Alendronate Alone in Danish Women Suffering from
Osteoporosis
Medicine with Industrial Specialisation, Medical Market Access
Master Thesis
Aalborg University
29th of May 2020
Conducted by
Louise Borup and Maria Hahn Holst
Group 10022
Supervisor
Anne Sig Sørensen
Abstract
Background: Osteoporosis and fragility fractures cause patients to utilise more
healthcare services and thereby impose an increasingly great economic burden on
society. An obvious opportunity of lowering the burden is for osteoporosis patients to
avoid sustaining fractures which increases the demand for cost-effective treatments.
Currently, the most frequently sold drug for treating osteoporosis in Denmark is the
bisphosphonate, alendronate. However, studies have shown promising results of the
effects of the bone anabolic drug, teriparatide. The patent of teriparatide expired in
August 2019, resulting in lowered prices of the compound, which has increased the
relevancy of more recent cost-effectiveness analyses.
Methods: A cost-utility analysis was carried out to investigate the health-economic
consequences of treating Danish women above the age of 50 with sequential
teriparatide/alendronate treatment compared to alendronate alone in terms of costs
and QALYs over a lifetime horizon using Markov modelling.
Results: For the base case analysis the cost of teriparatide/alendronate treatment was
DKK 761,057 per QALY when compared to alendronate alone. A WTP threshold of
DKK 250,000 per QALY was employed, meaning teriparatide/alendronate was not
considered cost-effective when compared to alendronate alone in the base case. This
result was robust to all sensitivity analyses performed except when no discounting
was applied as, in this sensitivity analysis, the ICER was DKK 241,191 per QALY.
Conclusion: This study indicated treating Danish women above the age of
50 suffering from osteoporosis with teriparatide was slightly more effective yet
more expensive when compared to alendronate, and teriparatide/alendronate was
therefore not considered cost-effective when compared to alendronate alone.
ii
Contents
1 Problem Analysis 1
2 Introduction 2
2.1 Osteoporosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Treatment of Osteoporosis . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2.1 Bisphosphonates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2.2 Teriparatide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3 Methods 10
3.1 Information Search . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.1 Fracture Incidence Rates . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1.2 Effects of the Compounds . . . . . . . . . . . . . . . . . . . . . . . 11
3.1.3 Mortality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.1.4 Obtainment of Cost Estimates . . . . . . . . . . . . . . . . . . . . . 12
3.1.5 Utility Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 Markov Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3 The Markov Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3.1 Model Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3.2 Model Inputs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.4 Sensitivity Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4 Results 27
4.1 Base Case Cost-Utility Analysis . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Sensitivity Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2.1 Deterministic Sensitivity Analyses . . . . . . . . . . . . . . . . . . 27
4.2.2 Probabilistic Sensitivity Analysis . . . . . . . . . . . . . . . . . . . 29
4.2.3 Scenario Analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.3 State Probability Charts for Teriparatide and Alendronate . . . . . . . . 31
5 Discussion 34
5.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
5.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.2.1 The Included Estimates of the Effects of Teriparatide and
Alendronate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
5.2.2 The Inclusion of Forearm and "Other" Fractures in the Markov
Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
iii
Aalborg University
5.2.3 Half-Cycle Correction . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.2.4 Ethical Considerations . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.2.5 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5.3 What’s Next? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.3.1 Restricted Prescription of Teriparatide in Denmark . . . . . . . . 41
6 Conclusion 42
References 43
A Appendix 50
A.1 Cost Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
A.1.1 Costs of Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
A.1.2 Costs of a Hip Fracture . . . . . . . . . . . . . . . . . . . . . . . . . 50
A.1.3 Costs of a Vertebral Fracture . . . . . . . . . . . . . . . . . . . . . 51
A.1.4 Costs of a Forearm Fracture . . . . . . . . . . . . . . . . . . . . . . 51
A.1.5 Cost of an "Other" Fracture . . . . . . . . . . . . . . . . . . . . . . 53
A.2 The Markov Model - The Teriparatide Branch . . . . . . . . . . . . . . . 54
A.3 The Markov Model - The Alendronate Branch . . . . . . . . . . . . . . . 55
A.4 Full Tornado Diagram - CUA . . . . . . . . . . . . . . . . . . . . . . . . . 56
A.5 Summary Reports - Base Case . . . . . . . . . . . . . . . . . . . . . . . . . 57
A.6 Tables Used in TreeAge When Altering the Durations of the Effects of
the Two Drugs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
A.7 Tables Used in TreeAge Containing the Baseline Probabilities of Dying
and Baseline Utility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
A.8 Tables Used in TreeAge Containing Fracture Incidence Probabilities for
Hip, Vertebral, Forearm, And Other Fractures . . . . . . . . . . . . . . . 61
A.9 Tables Used in TreeAge for Excess Mortality Following Hip and
Vertebral in First and Subsequent Years . . . . . . . . . . . . . . . . . . . 63
iv
1. Problem Analysis
Osteoporosis is a systemic skeletal disorder causing an increased risk of sustaining
fractures [1, 2, 3] leading to patients utilising more healthcare services which
ultimately causes them to impose increased healthcare costs [4]. The condition is
most prevalent among women [5] above the age of 50 due to menopause [6]. In fact,
about 30% of all European postmenopausal women suffer from the condition [7]. The
economic burden of osteoporosis is substantial and is expected to keep increasing in
the following years due to the ageing population [1]. Burge et al. [5] has estimated
that the fracture-associated costs will increase by 50% from 2005 to 2025 causing the
burden on healthcare systems to increase further, necessitating the need for effective
treatment methods with the ability to lower the occurrence of fractures provided at
low costs. The costs of osteoporotic fractures in Denmark were estimated at DKK
11.638 billion in 2011 [4] and, in a US study, it has been estimated that women account
for 75% of cost related to osteoporosis related fractures [5].
Currently, the most frequently sold pharmacological treatment in Denmark is
the antiresorptive bisphosphonate, alendronate [8], however, studies have shown
promising results of the effects of anabolic compounds such as teriparatide [9, 10, 11].
Unfortunately, treating patients with teriparatide during the patented period was
quite expensive, and therapy with teriparatide was previously not proven to be cost-
effective when compared to the most frequently used bisphosphonate, alendronate
[12], resulting in the use of teriparatide being restricted [13]. In August 2019, Eli
Lilly’s patent on teriparatide expired resulting in biosimilars entering the market
carrying lower prices for the highly effective treatment [14]. Taking into account the
promising effects of teriparatide and the lowered price of the treatment, a need for
more recent investigations of the cost-effectiveness of treating osteoporotic patients
with teriparatide has arised.
1
2. Introduction
2.1. Osteoporosis
Osteoporosis is characterised by an imbalance in bone remodelling leading to low
bone mineral density (BMD) and skeletal fragility which increases the risk of
sustaining fractures [1, 2, 3]. The employed diagnostic criteria for being diagnosed
with osteoporosis in Denmark mean an individual with a fragility fracture to the
vertebra or hip or a T-score of less than -2.5 in the back or hip can be diagnosed
with osteoporosis [15]. The T-score used in the definition correlates to the number of
standard deviations by which the BMD differs from the mean value in healthy young
individuals [3, 16] Approximately 3% of the Danish population has been diagnosed
with osteoporosis, corresponding to roughly 172,000 individuals. However, the
prevalence is estimated to be two to three times higher than currently identified since
many people suffering from osteoporosis remain undiagnosed.[17] In Denmark, it has
been estimated that almost 500,000 undiagnosed individuals suffer from osteoporosis
[2]. As can be seen in Figure 2.1, osteoporosis is traditionally divided into primary
osteoporosis, which is idiopathic, and secondary osteoporosis in which another
disease or pharmacological treatment has been contributing to the development of
osteoporosis. Primary osteoporosis includes postmenopausal osteoporosis, senile
osteoporosis, and a few more rare conditions such as juvenile osteoporosis.[18]
Figure 2.1: The flowchart depicts the criteria of being diagnosed with osteoporosis along withthe different types of osteoporosis.
Osteoporosis is a condition that is most often developed after the age of 50 [6] and out
of all postmenopausal women in Europe, about 30% suffer from osteoporosis [7]. The
occurrence of osteoporosis can be caused by a failure in achieving peak bone mass
2
2 Introduction Aalborg University
or severe bone resorption or/and decreased bone formation in remodelling [3]. It
entails having a greater risk of sustaining fractures compared to people not suffering
from osteoporosis. The characteristic of fractures related to osteoporosis is that they
occur spontaneously after low-energy trauma or from activities during everyday life.
Such fractures are termed fragility fractures and comprise fractures that a strong
skeleton would tolerate without fracturing.[19] Fracture types typically occurring
when suffering from osteoporosis are hip (ICD-10: S720, S721 and S722), vertebral
(ICD-10: M484, M485, M495, S220, S221, S320, and S327), of which compression
fractures are most common, forearm (ICD-10: S522, S525, and S526), and other
fractures such as fractures of the upper arm [20, 21, 19]. A fragility fracture of the hip
or vertebra is diagnostic for osteoporosis, whereas fragility fractures elsewhere are
signs of osteoporosis and should be further examined [19]. Fractures of the vertebra
are divided into clinical fractures, which are fractures leading to clinical attention, and
morphometric fractures, which are fractures without clinical manifestations which are
therefore rarely discovered and treated [22, 23].
Osteoporosis-related fractures impose a great burden on both the individual and
society [3]. In 2011, Hansen et al. [4] estimated the costs of osteoporotic fractures
in Denmark to be EUR 1.563 billion which equals about DKK 11.65 billion in 2020
value. Their study took into account costs regarding general practices, hospitals,
the municipalities, regions, and patients [4]. The prevalence of osteoporosis and
the number of osteoporosis-related fractures are numbers expected to increase and
become substantial public health problems. This is expected as the population is
ageing [1], and osteoporosis is more prevalent among the elderly [24]. Moreover, it
has been projected that the number of annual fractures and the associated costs were
to increase by nearly 50% from 2005 to 2025. In the same study, the division of costs
accounted for by men and women was estimated to be 25% and 75%, respectively.[5]
Patients suffering from symptomatic osteoporotic fractures experience emotionally
strong reactions to their conditions, among which the most frequently mentioned
psychological problems are depression and anxiety. Spine and hip fractures are two
of the most serious types of fractures and are associated with severe pain, disability,
decreased quality of life with an up to 50% reduction in patients who are left with a
permanent impairment, and even death.[3, 1] Osteoporotic asymptomatic individuals
are likely to have no knowledge of their condition until a fracture occurs, which often
brings both limited mobility, pain, and a feeling of helplessness [25].
As an extensive burden is related to fragility fractures, an obvious opportunity of
lowering the burden is to avoid sustaining fractures. Both non-pharmacological
3
2 Introduction Aalborg University
and pharmacological initiatives can reduce the risk of fractures occurring. Non-
pharmacological initiatives include, among other things, maintaining a healthy
weight, not smoking, taking D-vitamin and calcium supplements, and being
physically active.[15, 24]
2.2. Treatment of Osteoporosis
For the time being, no national clinical guidelines for the treatment of osteoporosis
have been developed for use in Denmark. However, guidance for the treatment of
postmenopausal osteoporosis in women is available through the Danish Endocrine
Society [15]. During the conduction of this study, it is assumed most of the women
suffering from osteoporosis after the age of 50 suffer from the condition due to
menopause [26], and therefore, the guidance of treatment from the Danish Endocrine
Society will be applied as a background for the treatment recommendations presented
in this study.
Before initiating the treatment of a patient, it is advised to determine whether the
patient possesses any indications of secondary causes of osteoporosis. If secondary
osteoporosis is diagnosed, the task is often left for a specialist to take care of, since
the treatment will depend largely on the underlying cause. The goal of treating
osteoporosis is to prevent fractures for people with an increased risk of fractures,
to prevent individuals who already suffered a prior fracture from suffering another
fracture, and ease of symptoms for individuals with established osteoporosis.[15, 1]
As a basis it is recommended to choose a drug with documented effect on the
specific type of fracture one has suffered or one is at risk of suffering. Using
more than one drug in the treatment regimen has not been proven effective, and
therefore, monotherapy should be planned.[24, 15] According to the guidance, it is
recommended that patients who are suffering from osteoporosis or who are at risk of
developing osteoporosis take a daily supplement of 20 µg of vitamin D and 800-1200
mg of calcium [15]. The frequently used pharmacological treatments for osteoporosis
are bisphosphonates (BPs), denosumab, selective estrogen receptor modulators, and
parathyroid hormone/analogues [24]. However, only the BP, alendronate, and the
bone anabolic drug, teriparatide, will be included in the following description due to
the aim of this study which was to investigate the health-economic consequences of
treating patients with teriparatide compared to alendronate.
4
2 Introduction Aalborg University
2.2.1. Bisphosphonates
It is well-documented that BPs are effective in treating most cases of osteoporosis in
both men and women [24]. The optimal length of treatment with antiresorptive drugs
is not known, however, it has, in all probability, the best effect during the first five
years of use, and the effects and safety of BPs have been documented for seven to ten
years [27]. After five years of use, the effects and side effects of the treatment for the
individual should be evaluated, however, in general, a treatment duration of five to
seven years is recommended for BPs [24].
BPs are part of the group antiresorptive agents [28]. They can be divided into
compounds that do not contain nitrogen, such as etidronate, and the ones that do,
such as alendronate, which is the most frequently used compound. Their affinity for
calcium is high, and within the body, they concentrate at active bone remodelling
sites. Both groups of BPs exert their effects by reducing the activity or the number of
osteoclasts resulting in osteoclastic bone resorption decrease. Since BPs are embedded
into the bone during the anabolic phase, these compounds reside in the body long
after cessation of treatment. The half-life of elimination of BPs from the skeleton is
up to ten years.[29] The BPs have in common that they decrease the activity of the
osteoclasts and bone turn over which increases BMD [30]. BPs comprise the first
group of approved drugs for the treatment of osteoporosis. The first BP, etidronate,
was approved in 1980 followed by the approval of alendronate in 1995.[3]
2.2.1.1. Alendronate
In Denmark, the recommended first choice for the treatment of postmenopausal
osteoporosis is alendronate [15], and it is the most frequently sold drug for treating
osteoporosis in Denmark [8]. Thus, in 2017, alendronate comprised 84% of the sold
amount of drugs for treating osteoporosis within the primary sector in Denmark.[19]
The patent of the active compound, alendronic acid, has expired, which has resulted
in generic versions of the drug being available in Europe since 2006 [3].
Alendronate has been approved for the treatment of postmenopausal osteoporosis
and the prevention of postmenopausal osteoporosis, and both prevention and
treatment of glucocorticoid-induced osteoporosis [13]. In Denmark, alendronate has
reached general reimbursement. Alendronate has been documented to progressively
increase bone mass in the hip, spine, and total body and decrease the progression
of deformities of the vertebra, and decrease the incidence of vertebral fractures and
height loss in women suffering from postmenopausal osteoporosis [31, 32]. It is
administered as weekly doses and requires overnight fasting and should always
be consumed 30 minutes prior to the first meal of the day or other drinks than
5
2 Introduction Aalborg University
water. The patient is furthermore instructed to swallow the tablet along with a
glass of water while being in an upright position and remaining in an upright
position for 30 minutes following administration.[13] The most common side effects
of alendronate are bone pain, arthralgia, and muscle pain affecting more than ten out
of 100 individuals [33]. Both persistence and compliance rates for oral BPs have, in
general, been documented to be suboptimal and this has been proven to be associated
with the inconveniences of consuming BPs [34].
2.2.2. Teriparatide
In Denmark, parathyroid hormone and analogues, such as teriparatide, are indicated
for the treatment of patients who have suffered a spinal fracture within the last
three years which resulted in more than 25% height reduction of a vertebra with a
concurrent T-score of less than -3 in the spine or hip, or patients that have suffered
more than two spinal fractures in the past three years with a more than 25% height
reduction of each vertebra. The treatment can be prescribed for no more than
18-24 months, and after finishing treatment, antiresorptive drugs such as the BP,
alendronate, are recommended for future use to maintain the effects of teriparatide,
however, as mentioned earlier, they should not be prescribed at the same time.[24]
Teriparatide, which comprises the 34-amino acid N-terminal of the native parathyroid
hormone (PTH) [14], is the first approved anabolic agent that improves bone mass
and quality by stimulating osteoblastic bone formation [35, 9]. In Denmark, treatment
with teriparatide is only prescribed for patients suffering from severe osteoporosis
[15] and it has been granted single reimbursement. However, the drug is clinically
indicated for treating postmenopausal women at increased risk of fracture [36] and
not only for treating women suffering from severe osteoporosis. Firstly, the drug was
approved for a treatment duration of up to 18 months in postmenopausal women
suffering from osteoporosis. Later, the drug received its approval for use for up to 24
months.[37, 36] However, more than two years of use during a patient’s lifetime is not
recommended, since the efficacy and safety of the drug have not yet been evaluated
beyond two years [38].
Teriparatide requires refrigeration and is administered using a pre-filled pen
containing 28 daily doses. The drug is injected subcutaneously into the abdominal
wall or the thigh daily [38] and most patients can be taught to self-administer their
daily dose [39]. PTH has a direct effect on osteoblast lineage cells as well as it
indirectly affects the regulation of certain skeletal growth factors and growth factor
antagonists. The mitogenic properties for osteoblastic cells decrease the osteoblastic
apoptosis [40], of which the consequence is an increased number of bone-forming
6
2 Introduction Aalborg University
cells [41, 42]. However, the precise mechanisms behind the anabolic effects of PTH
have yet to be fully elucidated.
The effects of teriparatide have been compared to placebo in a study by Neer et al. [10].
They found teriparatide has a good effect in terms of reducing the risk of sustaining
vertebral fractures. Furthermore, the risk of sustaining one or more new fractures
of the vertebra was reduced by 65% and the risk of sustaining at least two or more
fractures was lowered by 77% at a 20 µg dose.[10] Side effects of using teriparatide
include hypercalcemia which 1-3% of the treated patients report [10, 43], however,
the most frequently reported side effects are pain, nausea, and arthralgia which
approximately 10% of patients treated with teriparatide experience [38].
The firstly approved teriparatide, Forsteo by Eli Lilly, was granted marketing
authorisation by the European Medicines Agency in June 2003. In August 2019, the
patent expired, making way for biosimilars to enter the market. The two currently
available biosimilars of Forsteo, Terrosa and Movymia, are being used alongside
Forsteo in Denmark [44]. They were both granted marketing authorisation in January
2019 [45, 46] awaiting the expiration of the patent of Eli Lilly’s Forsteo. The price of
one package of Forsteo (item number 014094) was DKK 3,454.85 in July 2019 before
the patent expired in August 2019 [47]. The presence of biosimilars has enabled
competition between Forsteo, Terrosa, and Movymia which has resulted in varying
but lower prices of the drugs compared to when the patent had not yet expired. The
prices of Forsteo (item number (014094), Terrosa (126559), and Movymia (428393) were
DKK 2,539 per package as of the 20th of April 2020.[47] Since the price of teriparatide
was DKK 3,455 before the expiration of the patent, the price in April 2020 is 26.5%
lower compared to when the patent was still valid. For years, teriparatide has only
been prescribed for those at very high risk of sustaining especially vertebral fractures,
since the high price restricted the use of the drug [13]. In Denmark, this has meant that
in order to be prescribed teriparatide, the demands for the severity of the individuals’
osteoporosis are high, while no demands for the severity of osteoporosis are required
to be treated with alendronate. A flow-chart of the recommended treatment options
can be seen in Figure 2.2. However, the treatment guidance is not always followed
slavishly, as treatment choice is rather adapted to the individual patient and the
disease is often undiscovered until a fracture is sustained [25], and at that time the
condition might be advanced.
7
2 Introduction Aalborg University
Figure 2.2: The flowchart depicts the course of treatment recommended by The DanishEndocrine Society [15], however, the figure mainly illustrates the options of relevance duringthis study i.e. several pharmacological treatment options have been omitted due to the theirlack of relevance for this study.
8
2 Introduction Aalborg University
2.3. Aim
Osteoporosis among women over the age of 50 is a substantial financial problem
worldwide and the burden is expected to increase rapidly in the near future.
Moreover, the ultimate consequence of osteoporosis therapies is on the quality and
length of life [48] which enhances the importance of the availability of cost-effective
treatments since patients suffering from osteoporosis are heavily affected by their
conditions both physically and mentally.
For the time being, teriparatide has not been proven cost-effective compared to
alendronate, due to the high price of teriparatide [12]. However, after the patent
of teriparatide has expired, the price has been lowered, leaving the opportunity of
change in the cost-effectiveness of teriparatide in comparison with other treatment
options.
The study aimed to investigate the health-economic consequences of treating female
patients over the age of 50 suffering from osteoporosis with teriparatide compared to
alendronate in terms of costs and accumulation of quality-adjusted life years (QALYs).
This was done by constructing a decision-analytic model build on the occurrence of
four types of fractures (hip, vertebral, forearm, and "others") and their associated
utility decrements resulting in the conduction of a cost-utility analysis (CUA).
During the analyses, a Danish setting was applied to take a healthcare perspective
extended to include patient-paid costs of medication. A Markov model was
constructed to model the course of each individual until the age of 100 years or death.
Alendronate was chosen as comparator since it is the most frequently used drug for
treating osteoporosis in Denmark and it can be prescribed for all women suffering
from osteoporosis.
In Denmark, the use of QALYs and an explicitly stated WTP threshold for one unit
has not been established, however, UK has implemented a WTP threshold of £20.000
to £30.000 per QALY, which means if an intervention exceeds this level, likely, the
intervention will not be adopted for widespread use by the National Health Service
[49]. During this study, NICE’s upper WTP threshold of £30.000 which is about DKK
250,000 per QALY in 2020 value was adopted to evaluate the cost-effectiveness of
teriparatide compared to alendronate in terms of accumulation of costs and QALYs.
9
3. Methods
3.1. Information Search
To gather the needed background information and obtain knowledge about osteo-
porosis, pharmacological treatment options, and financial perspectives concerning
these, initial free text searches were conducted online to create an overview of the
subjects involved. Moreover, focused literature searches were conducted to collect the
needed information regarding the effects of teriparatide and alendronate, fracture in-
cidence rates in women over the age of 50, utility values, and mortality rates. These
were primarily obtained from literature found via the database, PubMed. Both free-
text and MeSH terms were included in the searches on PubMed to ensure thorough
approaches. In addition to the focused literature searches, snowballing was carried
out to expand our field of knowledge. Inclusion criteria used for assessing the ap-
plicability of the found literature were that the studies had to include women above
the age of 50 and they had to include a study population somewhat comparable to
the Danish population. In addition, the literature of a later date was preferred over
older literature. Overall, the literature was assessed based on the transparency of the
studies and descriptions of the applied methodological considerations.
3.1.1. Fracture Incidence Rates
A literature search was carried out to obtain the incidence rates for hip, vertebral,
forearm, and "other" fractures. The search was conducted on the 8th of March 2020
on PubMed and gave a total result of 4,544 articles. The terms used for the search can
be seen in Table 3.1. Only articles published within the past 15 years were considered
to ensure the data and literature were not obsolete. The main part of the articles was
screened to point out the relevant articles. After the screening, the remaining articles
were read in their full-text versions and out of these, two articles were used to obtain
the needed values for the model. The first of these was a study by Hiligsmann et al.
[50] which provided the fracture incidence rates for women, however, to represent a
population of women suffering from osteoporosis, each incidence rate was adjusted
using relative risks of sustaining fractures when having a T-score of -2.5 or less found
in another study by Hiligsmann and Reginster [51]. The majority of the excluded
studies were deselected due to missing information such as insufficient information
regarding their rates and because not all the fracture types included in our study
were included in the literature found. Moreover, a considerable portion of the located
studies was not carried out in women, specifically.
10
3 Methods Aalborg University
AND
Hip Prevalence Women Osteoporosis Fracture
Vertebral Incidence Female OsteoporoticOR
Forearm Rate Postmenopausal
Table 3.1: The table lists the search terms used for the literature search conducted with theaim of obtaining incidence rates for the Markov model.
3.1.2. Effects of the Compounds
The literature searches for the effects of both alendronate and teriparatide were
conducted collectively in one search on PubMed on the 9th of March 2020 and
provided a total result of 289 articles. The included search terms can be seen in
Table 3.2. Following the screening by titles and abstracts, 19 articles were left, and
these were read as full-text. Out of the 19, two were used for building the model
[52, 53]. Most studies were excluded because of missing information regarding
methods and references and some were excluded due to them investigating the effects
in incomparable populations regarding age and race.
AND
Hip Teriparatide Fracture Probability Osteoporosis
Vertebral Alendronate RateOR
Forearm
Table 3.2: The table shows an overview of the search terms used for the literature searchregarding the effects of the two investigated compounds, alendronate and teriparatide.
3.1.3. Mortality
To procure information on mortality related to osteoporosis, a literature search was
conducted on PubMed on the 9th of March 2020 using the terms seen in Table 3.3. The
literature search resulted in 951 articles. The articles were assessed by title and further
assessed by abstract if the title was of relevance. 90 articles were of interest, however,
only seven articles met the criteria of inclusion and were read in the full-text form.
Out of the seven articles, only one article by Johnell et al. [54] was used for including
mortality related to fragility fractures in the model. Moreover, an age-dependent
mortality baseline for Denmark was calculated from data obtained through Statistics
Denmark [55].
11
3 Methods Aalborg University
AND
Osteoporosis Mortality ProbabilityOR
Osteoporotic fracture Excess mortality Rate
Table 3.3: The table shows an overview of the search terms used for the literature searchregarding mortality data for the model.
3.1.4. Obtainment of Cost Estimates
Regarding the costs included in this study, several sources were utilised to obtain
these. When diagnosis-related group (DRG) charges were deemed appropriate and
the best and most precise available estimates of costs, these were utilised. They were
obtained through both the interactive DRG web-page and the spreadsheet containing
all DRG charges both by The Danish Health Data Authority [56]. The prices of the
drugs were obtained through The Danish Medicines Agency [47] and The Danish
Health Data Authority [8]. The specific assumptions needed regarding the estimations
of costs will be further described in section 3.3.2.3.
3.1.5. Utility Values
For the utility values needed in the model, a literature search was conducted on
PubMed on the 6th of March 2020 using the terms seen in Table 3.4. This literature
search gave a total of 60 articles. The 60 articles were initially screened by titles and
later by abstracts which resulted in 16 out of the 60 articles being read in the full-
text form. Out of the 16, one systematic review was used for obtaining utility values
for the model [57]. Two systematic reviews on utility multipliers were of interest,
however, Peasgood et al. [58] suffered from a lack of estimates in utility multipliers
for "other" and clinical vertebral fractures in the subsequent years. Therefore, the
utility multipliers from the study by Hiligsmann et al. [57] were used in the model,
as there was no inadequacy in data, though the review was of an older date than the
study by Peasgood et al. [58]. The majority of the articles that were sorted out were
deselected due to a lack of the needed estimates of utility values for both hip, clinical
vertebral, forearm, and "other" fractures, due to improper time spans for the utility
values, or due to the estimates being of an older date than other identified estimates.
Moreover, meta-analyses and systematic reviews were preferred due to their inclusion
of data from several studies producing aggregated estimates.
12
3 Methods Aalborg University
AND
Hip Utility Quality of life Osteoporosis Fracture
Vertebral Utility values QALY OsteoporoticOR
Forearm Disutility
Table 3.4: The tables gives en overview of the terms used for the literature search conductedin order to obtain utility values for the model.
3.2. Markov Modelling
Modelling has become an important tool in economic evaluations particularly when
a decision involving resource allocation must be made, as it helps decision-makers
identify the optimal intervention under conditions of uncertainty [59]. Moreover,
it enables extrapolation of results from trials, a combination of data from multiple
sources, and generalisation of results to other contexts. However, the quality of a
model will always be limited by the quality of the utilised data and assumptions.[60,
61, 62] Since osteoporosis is characterised as a chronic disease in which a recurrence
of fractures is present, a state transition model was appropriate for use to achieve the
aim of this study [63, 64]. Moreover, Markov models are very useful when significant
events may happen more than once and when the timing of each event is important,
as is the case for osteoporosis and it is furthermore the recommended approach for
modelling osteoporosis [48]. A state transition model also called a Markov model, is a
model in which a cohort of patients can be characterised based on "health states" that
are mutually exclusive [65]. Markov models do not naturally incorporate memory,
however, this can be done by creating unique health states able to capture the history
of patients [66].
3.3. The Markov Model
A model-based cost-effectiveness analysis was conducted in order to compare
teriparatide and alendronate for the treatment of osteoporosis. The modelled cohort
consisted of women with a start age of 50 years suffering from osteoporosis, as
women above the age of 50 comprise the subgroup of the general population in
which osteoporosis is most prevalent. A Danish setting, with a healthcare perspective
extended to include patient-paid costs of medication, was applied. All costs included
in the model were given in 2020 values, and a gross costing approach was applied.
A time horizon modelling the patients up to and including their 100th life year or
death was employed as has frequently been done by other authors when modelling
osteoporosis processes using Markov modelling [67, 68, 69, 70]. This approach was
deemed appropriate due to the chronic nature of osteoporosis causing it to often
project onto the lifetime of individuals. Since it was assumed people rarely experience
13
3 Methods Aalborg University
several fractures a year in line with the studies by Hiligsmann et al. [63] and Mori et al.
[53], the length of each cycle in the model was one year. Since the model ran for 51
cycles of one year each, a discount rate of 4% during the first 35 cycles in the model
and a discount rate of 3% in the remaining 16 cycles were applied to health outcomes
and costs in the model as recommended by The Danish Ministry of Finance [71]. This
economic evaluation used QALYs as the measure of effect as the mean to compare
the two methods of treatment and utilised a synthesis of evidence to perform the
model-based economic evaluation.
3.3.1. Model Structure
A decision-analytic model (DAM) was constructed in TreeAge Pro 2020 (Healthcare
Version) comprising a decision node followed by two identically constructed Markov
nodes, one for teriparatide and one for alendronate. The construction of the model
is illustrated in appendix A.2 and A.3. The model comprised six health states which
can be seen in the state transition diagram in Figure 3.1. The incorporated health
states were "well", "dead", "post-fracture hip", "post-fracture vertebral", "post-fracture
forearm", and "post-fracture "other"". The "dead" health state was the only absorbing
one.
Figure 3.1: The figure depicts the state transition diagram for the Markov model. Allindividuals start in the "well" health state. Each arrow represents the opportunity oftransitioning from one state to another or staying in the same health state (when a roundedarrow is applied).
Moreover, six events were incorporated into the model. Each patient had a probability
of either staying well, sustaining a vertebral, hip, or forearm fracture, sustaining an
"other fracture", or dying. The event comprising "other" fractures cover all other
fractures than hip, vertebral, and forearm fractures as has previously been done in
several other studies [72, 51, 68]. Moreover, only fractures of the vertebra leading
to clinical attention were included in the model. All patients began in the health
14
3 Methods Aalborg University
state "well" meaning they were not suffering from a previous fracture, however, only
patients diagnosed with osteoporosis entered into the model.
As a treatment duration of 5-7 years is recommended for BPs [24, 27], the treatment
duration of alendronate was modelled as six years. Likewise, the treatment duration
was modelled as two years for teriparatide, as the effects and safety of the drug have
only been studied for up to two years [38]. However, since maintaining the effects
of teriparatide can be done by preventing a decline in bone density and thereby
making sure to sustain the fracture efficacy, treatment with alendronate is usually
recommended following the two years of treatment with teriparatide which was
the approach modelled. It was therefore assumed that the patients in the model
maintained their effects of both the treatment with teriparatide/alendronate and
alendronate until they were 100 years old or dead. In the teriparatide/alendronate
branch, the effects of teriparatide were given for all years. The assumptions
regarding courses of treatment were made to reflect the real-life course of treatment
recommended in the treatment guideline [15].
It was assumed that all patients consume vitamin D and calcium supplements since
in Denmark the recommended approach is to use these supplements as a basis of
treatment both in conjunction with other treatments and on their own [24]. However,
the supplements are not reimbursed, the patients acquire them over the counter,
and no effects of the supplements were incorporated into the model. Moreover,
the associated costs were considered irrelevant for the aim of this study since no
difference between the two Markov nodes was assumed to be present. Therefore, the
costs of acquiring them were omitted during this study.
3.3.1.1. "Post-Fracture" Health States
Four "post-fracture" health states were incorporated into the model for hip, vertebral,
forearm, and "other" fractures. The post-fracture health states for hip and vertebral
fractures were incorporated into the model to be able to assign patients experiencing
a fracture in a previous cycle higher mortality and lower utility values since it
was found in the literature that during the year after a hip or vertebral fracture
the mortality rates were higher and the utility values were lower [54, 57]. In the
post-fracture health states for forearm and "other" fractures, individuals would be
given a decrement in utility during the first year after the fracture. In addition, no
utility decrements were given in these two post-fracture health states unless a new
fracture was sustained since a forearm or an "other" fracture does not carry any
utility decrement except in the year the fracture was sustained [57]. Moreover, in
all of the post-fracture health states in the model, patients were at higher risk of
15
3 Methods Aalborg University
suffering another fracture of any type. This method using post-fracture health states
was utilised to overcome the memoryless feature of Markov models known as the
"Markovian assumption" [73] which refers to the fact that whenever a patient has
moved on from a cycle, the model has no memory of where the patient was previously
in the model [64, 66].
A hierarchy of the included fracture types was modelled, with hip fractures being on
top followed by vertebral fractures. Forearm and "other" fractures were given equal
status and placed below vertebral fractures in the hierarchy. This construction was
modelled due to the included fracture types having diverse impacts on utility and
mortality according to the applied evidence. Further explanation will be provided in
the section regarding mortality, section 3.3.2.2, and utility, 3.3.2.4.
3.3.2. Model Inputs
In the following, the model inputs and assumptions regarding fracture risk, mortality,
costs, and utility will be presented.
3.3.2.1. Fracture Risk
Articles from Mori et al. [53] and Freemantle et al. [52] provided data on the effects
of teriparatide and alendronate on the occurrence of osteoporotic fractures and the
included estimates can be seen in Table 3.5. The data was provided in relative risks
compared to placebo, and therefore the risk of sustaining each type of fracture in a
placebo group was needed. Age-dependent rates of sustaining each type of fracture
in Belgian women above the age of 50 from Hiligsmann et al. [50] were converted
into probabilities and used for generating the data for the placebo-group. However, it
was necessary to adjust these to make them reflect the fracture risk in an osteoporotic
population accurately as low BMD is associated with increased fracture risk [51].
Therefore, estimates on the relative risks of sustaining a fracture for women with
a T-score of -2.5 or less were obtained from Hiligsmann and Reginster [51] and
multiplied onto the age-dependent probabilities of sustaining each type of fracture.
This data was used to reflect the fracture risk in a placebo group. The probabilities of
sustaining fractures when not being treated can be seen in Tables A.8, A.9, A.10, and
A.11 in appendix A.8 and, in the model, these were multiplied by the relative risks
of sustaining fractures while treated which were obtained from the studies by Mori
et al. [53] and Freemantle et al. [52]. The data obtained from Hiligsmann et al. [50]
and Hiligsmann and Reginster [51] originated from Belgium and not specifically from
Denmark, however, it has been found that the prevalence of osteoporosis in women
over the age of 50 in Belgium was the same as in Sweden and only slightly higher than
16
3 Methods Aalborg University
in Denmark [3]. Therefore, it was assumed the data was representative of a Danish
population.
It was incorporated into the model that the risk of sustaining another fracture was
increased after experiencing one of the included fracture types. This was in line with
other studies [74, 63]. The relative risks of sustaining a new fracture of each of the
four fracture types after having sustained either a hip, vertebral, forearm, or "other"
fracture, were obtained from a study by Klotzbuecher et al. [74] and can be seen
in Table 3.5. The relative risks were included in the post-fracture health states and
multiplied onto the risk of experiencing a fragility fracture for the first time while
treated with alendronate or teriparatide. Regarding the probability of experiencing
an "other" fracture in the post hip, post vertebral, or post "other" health states, pooled
estimates of the relative risks of sustaining all types of fractures from Klotzbuecher
et al. [74] were incorporated.
Relative risks1 Value 95% CI Ref
Post forearm - new hip fracture 1.9 1.6-2.2 [74]
Post forearm - new vertebral fracture 1.7 1.4-2.1 [74]
Post forearm - new forearm fracture 3.3 2.0-5.3 [74]
Post forearm - new other fracture 2 1.7-2.4 [74]
Post hip - new hip fracture 2.3 1.5-3.7 [74]
Post hip - new vertebral fracture 2.5 1.8-3.5 [74]
Post hip - new forearm fracture 2.4 1.9-3.2 [74]
Post hip - new other fracture 2.4 1.9-3.2 [74]
Post vertebral - new hip fracture 2.3 2.0-2.8 [74]
Post vertebral - new vertebral fracture 4.4 3.6-5.4 [74]
Post vertebral - new forearm fracture 1.4 1.2-1.7 [74]
Post vertebral - new other fracture 1.9 1.7-2.3 [74]
Post other - new hip fracture 2 1.7-2.3 [74]
Post other - new vertebral fracture 1.9 1.3-2.8 [74]
Post other - new forearm fracture 1.8 1.3-2.4 [74]
Post other - new other fracture 1.9 1.7-2.2 [74]
Efficacies of drugs as RRs compared to placebo
Alendronate
Effect on incidence of hip fracture 0.45 0.27-0.68 [53]
Effect on incidence of vertebral fracture 0.5 0.33-0.79 [53]
Effect on incidence of forearm fracture 0.82 0.25-1* [52]
Effect on incidence of other fracture 0.78 0.66-0.92 [53]
Teriparatide
Effect on incidence of hip fracture 0.42 0.1-1* [53]
Effect on incidence of vertebral fracture 0.3 0.16-0.55 [53]
Effect on incidence of forearm fracture 0.24 0.02-1* [52]
Effect incidence of other fracture 0.5 0.32-0.78 [53]
Table 3.5: The table lists the parameters used in the model. 1; relative risks of sustaininganother fracture when having sustained a prior fracture compared to the risk of experiencinga fracture when not having sustained a prior fracture. RRs; relative risks. The 95% CIs(confidence intervals) shown in the table were found through the literature and used whenperforming deterministic sensitivity analyses. For three parameters marked with * in the table,the CIs were cut off at an upper limit of 1, meaning the CIs were considered uncertainty rangesrather than CIs, hence an uncertainty range of ±20% of the mean was used for these threeparameters when performing deterministic sensitivity analyses.
17
3 Methods Aalborg University
3.3.2.2. Mortality
Age-specific baseline mortality probabilities for Danish women were calculated from
data obtained through Statistics Denmark [75]. The risk of dying at each age was
calculated from the total number of women dying at ages 50 through 100 in 2019
divided by the total number of women at each age in 2019. In the model, it
was assumed that women suffering from osteoporosis died at a faster rate due to
sustaining fractures and not from suffering from osteoporosis in general, which means
if they did not sustain any fractures they experienced no excess mortality in addition
to the age-dependent baseline mortality probabilities.
Only a total of the number of women dying at the age of 99 years old or older was
reported for 2019, leaving no specific number of people dying at the ages 99 and 100
years. Therefore, the likelihood of a woman at the age of 99 and 100 years dying was
modelled as the same likelihood as the probability of a 98-year-old woman dying.[76]
It is well-documented that a patient suffering a vertebral or hip fracture experiences
excess mortality compared to if that patient did not experience one of the two fracture
types [54]. Therefore, to incorporate this into the model, the probabilities of dying in
the first year after a hip or vertebral fracture were calculated from the annual rate of
mortality during the first year found in a study by Johnell et al. [54]. However, to
prepare the rates for insertion into the model they were converted into probabilities.
Since the study provided estimates of rates for individuals aged 60 and 80 years,
the rate for individuals at 60 years was used as the estimate of mortality for people
aged 50-70 years. Likewise, the 80-year estimate was used for people aged 70-100
years in the model. An assumption that excess mortality after a hip fracture lasts
lifelong was applied, as was done in another study by Mori et al. [53]. The increase in
mortality following a vertebral fracture was the same or even higher than the increase
in mortality following a hip fracture [63], and therefore, the assumption of lifelong
excess mortality was applied for vertebral fractures as well. To reflect the fact that
either a hip or vertebral fracture increases mortality for life, the annual rate of dying
from a hip or vertebral fracture in the fifth year after a fracture was added as the
mortality estimate in the post-fracture health states for hip and vertebral fractures
which will be applied for all subsequent years following the fractures [54].
Several studies have found that experiencing a forearm fracture or an "other" fracture
does not alter the risk of dying, thus the same probability of dying was used for
women who had not suffered a fracture as for women who had suffered a forearm or
an "other" fracture during each given cycle [77, 54, 63, 68]. A ranking of the fracture
types was modelled, as suffering a hip fracture was associated with the highest
18
3 Methods Aalborg University
probability of dying in the model, while the mortality associated with suffering a
vertebral fracture was the second highest. This means e.g. a patient in the post-
fracture hip health state suffering a fracture of the forearm will remain in the post-
fracture hip health state.
For patients experiencing two fractures in two consecutive cycles, only the excess
mortality related to the latest occurring fracture was incorporated in the model,
assuming more fractures would not lead to a higher rate of mortality. This was in
accordance with other studies.[63, 51]
3.3.2.3. Costs
All the estimated costs used in the model are seen from a healthcare sector perspective
with the inclusion of patient-paid medication costs and can be seen in Table 3.6.
Treatment Costs
Costs related to the acquisition of medication were included due to osteoporosis
medication being reimbursed when a patient reaches a certain limit of yearly
medication costs resulting in the costs being relevant seen from a healthcare sector
perspective.
As previously mentioned, six years of treatment with alendronate and the associated
costs were modelled. Likewise, treatment with teriparatide and the related costs
were modelled to occur in the first two cycles, followed by six subsequent cycles
of treatment with alendronate, cf. the treatment guideline.
The annual costs of the treatments with teriparatide and alendronate included in
the Markov model were obtained through The Danish Medicines Agency [47]. This
source was chosen as the information is updated continuously every two weeks. The
utilised prices of the drugs were of the 17th of March 2020. On this specific date, the
price of one package of teriparatide containing 28 doses was DKK 2,539 regardless
of brand chosen. One dosage is injected every day, which means the cost of being
treated with teriparatide was DKK 33,097 annually. More options were available for
alendronate as several companies sell drugs comprising the same active ingredient.
Alendronate "Sandoz" 70 mg and the package containing 14 units was chosen as
a reference as it is the product most frequently sold in Denmark.[8] The price of
one package containing 14 units was DKK 128.75, which corresponds to DKK 479.53
annually as each patient consumes one unit once a week [47]. All calculations related
to the cost estimates of treatments costs can be seen in appendix A.1.1.
Estimations of Mean Fracture Costs
Inpatient costs related to hip, vertebral, and forearm fractures were gathered through
19
3 Methods Aalborg University
the Danish DRG browser [78]. A DRG-charge includes all services performed and
covers charges that are associated with an inpatient stay. The DRG-charges are
based on the given care and used resources related to a typical patient within a
DRG group.[79] Patients suffering from osteoporosis are assumed to have a DXA-
scan regularly and attend one consultation at the general practitioner (GP) annually
[4]. However, the costs related to this were omitted for this study since no difference
between the two comparators was assumed to be present.
Estimation of the Mean Cost of a Hip Fracture
According to Kold et al. [80], all hip fractures require a surgical treatment approach.
It has been found that 75.4% of hip fractures are treated by internal fixation surgery
and the remaining by alloplasty [4]. Therefore, 75.4% of hip fractures were assigned
the 08MP28 2020 DRG-charge of DKK 67,991 which includes fracture surgery with
internal fixation near the hip, and the remaining 24.6% were assigned the DRG-charge
08MP20 including primary alloplasty of the hip of DKK 51,979. This resulted in an
average weighted cost of DKK 64,052 for treating a hip fracture surgically. Some
costs related to follow-up are accumulated after discharge when having suffered a
hip fracture. As the treatment course after a hip fracture varies depending on the
individual and the procedure performed, the treatment guideline for a hip alloplasty
was applied as the basis for the post-surgery activities included in the cost calculation
[81]. It was assumed that patients attend a follow-up visit 10 days after surgery, which
includes the removal of surgical staples at a GP [81]. The cost of consultation at a GP
in 2020 was found through the Danish Medical Association [82] to be DKK 146, and
an add-on service of DKK 199 was added for the first treatment of bigger wounds and
was assumed to include removal of surgical staples. Moreover, a follow-up visit at
the orthopaedic ward carried out by a doctor three months after surgery was assumed
to take place [81] and the 2020 DRG-charge 23MA04 of DKK 1,512 [78] was applied
for this follow-up. These assumptions resulted in an estimated cost for each hip
fracture of DKK 65,909 when adding up the weighted costs of surgical treatment,
follow up at a GP, removal of surgical staples, and the DRG-charge of a follow-up at
the ambulatory within the first year following the fracture. In the model, the cost of
DKK 65,908.8 was applied to the same cycle as a hip fracture was sustained in. The
full calculations of hip fracture costs can be seen in section A.1.2 in the appendix.
Estimation of the Mean Cost of a Vertebral Fracture
Only vertebral fractures leading to clinical attention were included in the model, as
the estimates of the drugs’ effects were evaluated on fractures that come to clinical
attention. This means the patients being seen at the hospital are typically the patients
with clinical manifestations requiring surgical treatment [83, 19]. Hence, the DRG-
20
3 Methods Aalborg University
charge of fracture surgery, back/neck 08MP22 of DKK 92,307 in 2020 [78] was applied.
A report by North Jutland Orthopaedic Surgery [84] was used to estimate the costs
related to post-surgery activities not included in the DRG-charge of fracture surgery
of the back or neck [84].
The surgical staples were assumed to be removed at a GP or by a nurse 10-13 days
after the surgery, for which the cost of a consultation and the add-on fee for removal
of staples of DKK 146 and DKK 199, respectively, were applied [82]. Moreover, follow-
up visits including X-rays at an orthopaedic surgical ward 3, 12, and 24 months after
the surgery were included [84], to each of which the DRG-charges of a consultation
at the ambulatory for orthopaedic surgery of DKK 1,512 and the DRG charge of a
complicated X-ray of DKK 747 were applied [78]. These costs were all added onto one
cycle due to the Markovian assumption. This resulted in an estimated average cost of
each vertebral fracture leading to clinical attention of DKK 99,429. The calculations
can be seen in appendix A.1.3.
Estimation of the Mean Cost of a Forearm Fracture
Since not all fractures of the forearm are treated surgically, the proportions of the
numbers of forearm fractures treated surgically and with a splint or plaster was
needed to estimate the mean cost of a forearm fracture. Data on the number of
fragility fractures of the forearm distributed by ICD-10 codes occurring in the general
population and people with osteoporosis in 2016 was obtained through The Danish
Health Data Authority [19] and can be seen in appendix A.1.
It was assumed that out of all forearm fractures occurring in the general population,
50%, occur in women while the rest occur in male individuals. Moreover, it was
assumed that the distribution of forearm fractures occurring due to osteoporosis
between men and women was the same as for hip fractures resulting in twice as many
fractures for women as for men [85]. When applying these assumptions it emerged
that 13.7% of forearm fractures occur in women suffering from osteoporosis. However,
as not every forearm fracture requires surgery, the proportion of patients appearing
in the emergency ward, assuming these fractures are treated non-surgically, and the
proportion of patients being hospitalised and discharged, assuming these patients
have been operated, were calculated. Data on men and women discharged from the
hospital after a forearm fracture and patients visiting the emergency ward in 2018
was obtained through the National Patient Register [86]. The 13.7% was applied to
calculate the proportion of women with osteoporosis experiencing a forearm fracture
out of the patients treated surgically and in the ward. It emerged, according to the
latest data from 2018, that women with a forearm fracture were treated 3.1 times
21
3 Methods Aalborg University
more often in the emergency ward than with surgery (the calculations can be seen
in appendix A.1.4). The cost of treatment with splint or plaster was obtained via
interactive DRG-charges 2020 and was found to be DKK 1,952 [56]. The cost of
surgery was the DRG-charge 08MP24 for surgery of the forearm which was DKK
34,854 in 2020 [78]. This resulted in an intermediate estimate of the average weighted
cost of treatment of a forearm fracture of DKK 9,957.
Moreover, for all forearm fractures, one follow-up visit at the ambulatory for
orthopaedic surgery including removal of plaster for each fracture of DKK 1,512,
was applied [78]. However, some fractures of the forearm necessitate an X-ray as well
during the follow-up to ensure the correct position of the bones [87]. Hence, it was
assumed that the patients who were treated surgically also needed an X-ray as part
of their follow-up visit at the ambulatory. The DRG-charge 30PR18 from 2020 of DKK
507 was applied as the cost of an uncomplicated X-ray, however, this was applied as
a weighted average since only the surgically treated fractures demand an X-ray. The
final weighted average cost of a forearm fracture was estimated to be DKK 11,593.
The calculations can be seen in appendix A.1.4.
Costs Value (DKK) Ref
Yearly costs of drugs
Treatment w. alendronate 479.53 [8]
Treatment w. teriparatide 33,097.68 [8]
Hip fracture 65,908.8
Internal fixation, DRG 67,991 [78]
Hip alloplasty, DRG 51,979 [78]
Average weighted cost of hip surgery 64,052
Consultation (GP) 145.46 [82]
Add-on service (treatment of wound) 199.34 [82]
Follow-up at the ambulatory, DRG 1,512 [78]
Vertebral fracture 99,428.8
Surgery, back/neck, DRG 92,307 [78]
Consultation (GP) 145.46 [82]
Add-on service (treatment of wound) 199.34 [82]
Follow-ups at the ambulatory, DRG (3 units, DKK 1,512 each) 4,536 [78]
X-ray, complicated, DRG (3 units, DKK 747 each) 2,241 [78]
Forearm fracture 11,592.91
Surgery, elbow/forearm, DRG 34,854 [78]
Treatment with splint/plaster, Interactive-DRG 1,952 [56]
Average weighted cost of forearm surgery 9,957.55
Follow-up, splint/plaster 1,512 [78]
X-ray, DRG, weighted* 123.36 [78]
Other fracture 25% of costs for a hip fracture 16,477.2
Table 3.6: The table lists the costs included in the mean cost estimates along with the meancosts used in the model (in bold). Ref; reference. GP; general practitioner. *X-rays only applyto the proportion of patients suffering a forearm fracture being treated surgically.
Estimation of the Mean Cost of an "Other" Fracture
For "other" fractures, the assumption that 25% of the costs related to a hip fracture
can be used to estimate the cost of "other" fractures, was applied as has previously
been done in a study by Hiligsmann et al. [63] based on an other study by Gabriel
22
3 Methods Aalborg University
et al. [88]. This assumption was further verified by Peter Vestergaard, professor and
chair of endocrinology, Department of Clinical Medicine and consultant MD PhD
DrMedSc at Aalborg University Hospital [89]. Therefore, 25% of the average cost of a
hip fracture of DKK 65,909 which is equal to DKK 16,477, was used as the estimated
mean cost of an "other" fracture. No extra costs were added as the mean hip fracture
cost includes both rehabilitation, wound treatment, follow-up, and surgical treatment
cost estimates.
3.3.2.4. Utility
The recommended outcome measure for osteoporosis models is QALYs [48]. To
investigate the accumulation of QALYs, utility values reflecting the health-related
quality of life were needed for incorporation into the model. This approach was
desired since fracture events are associated with decrements in health state utility
values that differ between fracture types [90].
Specific utility values can be assessed using several tools such as EQ-5D. When
assessing utility values for each health state, it results in a single value ranging
from 0 to reflect death and up to 1 which corresponds to perfect health.[91] However,
negative values from 0 to -1 can occur when health states worse than dead exist [92].
As utility is known to decrease by age and differs between men and women [93],
Danish age- and gender-specific baseline utility scores were desired, but no utility
data was available for Denmark. It was assumed, however, that applying age- and
gender-specific baseline from another country was more appropriate than assuming
a constant utility score of 1, as it would not reflect the reality and would overestimate
the effects of the treatments. It was assumed that baseline utility scores from the UK
were adequately transferable to the Danish setting, and therefore, age- and gender-
specific baseline utility values from Briggs et al. [94] were applied. Moreover, to
take into account all patients in the model are suffering from osteoporosis, which is a
chronic condition, all baseline utility values were reduced by -0.0418 which was found
in the study by Sullivan et al. [95] to reflect the disutility associated with suffering
from osteoporosis. As an example, this resulted in the baseline utility value for a 50
years-old woman being 0.7572.
A ranking of the fracture types was applied due to the different types of fractures
having various impacts on utility values. A hip fracture was ranked as the fracture
having the most extensive impact on utility values. Vertebral fractures had the second
most extensive impact followed by forearm and "other" fractures, which were given
equal status.
23
3 Methods Aalborg University
In the model, it was assumed that the incurrence of several fractures does not decrease
the utility values further than just one fracture would cause it to. The estimated
disutilities occurring in the years after a hip or vertebral fracture persisted for all years
beyond the year the fracture occurred in [57]. Moreover, as a hip fracture is known
to have the biggest impact on utility, a patient in the post-fracture hip health state
suffering for an example a new forearm fracture will remain in the post-fracture hip
health state. The patient will thereby continue to be given a utility multiplier equal to
the utility multiplier given to any patient suffering a hip fracture in subsequent years
(0.899, see Table 3.7), instead of jumping to the post-fracture forearm health state
which is associated with less disutility. Otherwise, the disutility associated with hip
fracture would be underestimated. This assumption was also applied for vertebral
fractures. This approach was chosen since the utility values associated with a fracture
of the hip or vertebra were not equal to the baseline values (which can be seen in
Table A.7 in appendix A.7) following the first year after the fracture [57]. For forearm
and "other" fractures the decrements in utility scores were only assumed to occur in
the year following the fracture [57]. All utility multipliers included in the model were
obtained from a study by Hiligsmann et al. [57] and can be seen in Table 3.7.
Utility Value 95% CI Reference
Well 1 [57]
Hip
First year 0.797 0.77-0.825 [57]
Subsequent years 0.899 0.885-0.91 [57]
Vertebral
First year 0.72 0.66-0.775 [57]
Subsequent years 0.931 0.916-0.946 [57]
Forearm
Forearm - first year 0.94 0.91-0.96 [57]
Subsequent years 1 [57]
Other
First year 0.91 0.88-0.94 [57]
Subsequent years 1 [57]
Dead 0 [57]
Table 3.7: The table lists the utility multipliers used in the model. CI; confidence interval.
3.4. Sensitivity Analyses
To test the uncertainty and sensitivity of the resulting ICER of this study, deterministic
sensitivity analyses were conducted to test the structural and methodological
uncertainties in the model and to investigate the impact each parameter has on
the base case ICER when varied one at a time. For the deterministic sensitivity
analyses, one-way analyses were performed and presented as a tornado diagram
and three additional one-way analyses, one on the annual cost of teriparatide and
two investigating the consequences of the inclusions of different discount rates (0%
and 7%), were conducted. The means, highs, and lows of the uncertainty intervals
24
3 Methods Aalborg University
provided in the literature were used in the deterministic analyses. However, for
three parameters of efficacy for which the uncertainty intervals were cut off at 1,
ranges of uncertainty characterised as their means ±20% were applied, since they
were considered uncertainty ranges rather than real CIs. Moreover, for each of the
cost estimates an uncertainty range of ±10% of the mean was applied for both the
deterministic and the probabilistic sensitivity analyses.
A probabilistic sensitivity analysis (PSA) was performed to evaluate all parameter
uncertainty included in the model simultaneously and indicate the degree of decision
uncertainty. By assigning each parameter an appropriate distribution, a PSA output
can be computed using second-order Monte Carlo simulation resulting in an expected
cost and effect for each simulation. PSA results can be presented in an incremental
cost-effectiveness (ICE) scatter plot on the cost-effectiveness plane or as a cost-
effectiveness acceptability curve (CEAC). These representations of the PSA output
can aid in assessing the uncertainty related to the decision in question.[96]
The PSA output was simulated by running 10,000 iterations and was presented as a
CEAC and ICE scatter plot. A PSA using the beta, gamma, and normal distributions
requires both means and standard errors (SEs). For the relative risks of suffering
a new fracture of any of the four types after having suffered a previous fracture
and the relative risk efficacy estimates, the SE calculations were performed by first
converting their means, lows, and highs onto a logarithmic scale to approach normal
distributions. These SEs were then calculated using absolute difference measures. In
the PSA, these relative risk estimates were provided with normal distributions using
the natural logarithm to the means and SEs which were exponentiated when used
in the model. However, as mentioned, for three of the efficacy estimates provided
along with CIs, the CIs from the literature were cut off at 1 resulting in a non-
normally distributed uncertainty interval meaning an SE could not be calculated.
Therefore, for the PSA, these estimates were included as beta distributions assuming
an uncertainty interval of±20% of their respective mean. Each of the three parameters
is marked with a in Table 3.8. A gamma distribution was applied to each cost
parameter to reflect a likely distribution of costs as they are often right-skewed and
are characterised as being fixed between zero and infinity [59, 97]. For the utility
parameters which were all provided with 95% CIs, the SEs were calculated from
the CIs using absolute difference measures. For each of the utility parameters, a
beta distribution was applied since they all range from zero to one. For the three
utility multipliers of 1 (well, forearm in subsequent years, and "other" in subsequent
years) and the utility multiplier of 0 for being dead, no distributions were included
in the PSA, since they were not expected to differ and therefore they were modelled
25
3 Methods Aalborg University
accordingly for the PSA. All parameters and their associated distributions used in the
PSA including their means and SEs can be seen in Table 3.8.
Three scenario-analyses were conducted. These three comprised one altering the
durations of effects, one investigating a shorter time-horizon of eight years, and a
scenario-analysis on the inclusion of only hip and vertebral fractures in the model.
A state probability chart along with a state probability summary report for each of the
two branches in the model will be presented to perform qualitative assessments of the
modelled progressions of the disease when treated with alendronate and teriparatide.
Parameter Mean value SE Distribution
Yearly costs (DKK)
Treatment w. alendronate 479.53 47.94∗ Gamma
Treatment w. teriparatide 33097.68 3309.77∗ Gamma
Hip fracture 65908.8 6590.88∗ Gamma
Vertebral fracture 99428.8 9942.88∗ Gamma
Forearm fracture 11592.91 1159.29∗ Gamma
Other fracture 16477.2 1647.72∗ Gamma
Utility multipliers
Hip - first year 0.797 0.01 Beta
Hip - subsequent years 0.899 0.01 Beta
Vertebral - first year 0.72 0.03 Beta
Vertebral - subsequent years 0.931 0.01 Beta
Forearm - first year 0.94 0.01 Beta
Other - first year 0.91 0.02 Beta
Parameter Ln(mean) Ln(SE) Distribution
Relative risks4
Post forearm - new hip fracture 0.64 0.08 Normal
Post forearm - new vertebral fracture 0.53 0.10 Normal
Post forearm - new forearm fracture 1.19 0.25 Normal
Post forearm - new other fracture 0.69 0.09 Normal
Post hip - new hip fracture 0.83 0.23 Normal
Post hip - new vertebral fracture 0.92 0.17 Normal
Post hip - new forearm fracture 0.88 0.13 Normal
Post hip - new other fracture 0.88 0.13 Normal
Post vertebral - new hip fracture 0.83 0.09 Normal
Post vertebral - new vertebral fracture 1.48 0.10 Normal
Post vertebral - new forearm fracture 0.34 0.09 Normal
Post vertebral - new other fracture 0.64 0.08 Normal
Post other - new hip fracture 0.69 0.08 Normal
Post other - new vertebral fracture 0.64 0.20 Normal
Post other - new forearm fracture 0.59 0.16 Normal
Post other - new other fracture 0.64 0.07 Normal
Efficacies of drugs expressed as relative risks compared to placebo
Alendronate
Effect on incidence of hip fracture -0.80 0.24 Normal
Effect on incidence of vertebral fracture -0.69 0.22 Normal
Effect on incidence of forearm fracture 0.82 0.164 Beta
Effect on incidence of other fracture -0.25 0.09 Normal
Teriparatide
Effect on incidence of hip fracture 0.42 0.084 Beta
Effect on incidence of vertebral fracture -1.20 0.32 Normal
Effect on incidence of forearm fracture 0.24 0.048 Beta
Effect incidence of other fracture -0.69 0.23 Normal
Table 3.8: ∗; these SEs correspond to 10% of the mean. 4; Relative risks of sustaining anew fracture when having sustained a fracture in the previous cycle compared to not havingsustained a prior fracture. RR; relative risk. ; for these three parameters the CIs (confidenceintervals) from the literature were cut off at 1, and therefore they were provided with betadistributions and an SE of 20% of the mean (neither means nor SEs were on a logarithmicscale). All Ln(mean) and Ln(SE) values in the table are only presented with two decimals.
26
4. Results
4.1. Base Case Cost-Utility Analysis
The expected values of the costs in the model following all 51 cycles were DKK
75,505 for teriparatide and DKK 15,001 for alendronate. For the utilities, the expected
values were 12.9115 QALYs for teriparatide and 12.8320 QALYs for alendronate. The
difference in costs and utilities between the two drugs was DKK 60,504 and 0.0795
QALY, respectively. Using the expected values for each Markov branch in the tree, an
ICER was calculated comparing teriparatide to alendronate.
ICER =Cteri − CalenEteri − Ealen
=75, 505− 15, 001
12.9115− 12.8320=
60, 5040.0795
= 761, 057 DKK/QALY
The base case resulted in an ICER of DKK 761,057 per QALY.
4.2. Sensitivity Analyses
4.2.1. Deterministic Sensitivity Analyses
The ICER tornado diagram in Figure 4.1 shows the parameters included in the base
case able to affect the ICER the most. However, the utility values for well, forearm, and
"other" in subsequent years, and dead were not included in the one-way analyses for
the tornado diagram, since they were assumed not to be able to vary. The parameters
are listed from the parameter that can affect the ICER the most when varied within the
range of uncertainty, to the parameter least able to affect the ICER. The parameters
able to affect the ICER the most when varied one by one within their uncertainty
ranges were the efficacy of teriparatide on the incidence of vertebral fractures and the
efficacies of alendronate on the incidences of hip and vertebral fractures as can be
seen in Figure 4.1.
Figure 4.1: The figure shows an ICER tornado diagram using all parameters except fourutility values (well, forearm, "other" (subsequent years), and dead). The red bars represent therange of the ICER when the parameter in question is higher than the base case value, and thecontrary is the case for the blue parts. Only the nine parameters able to affect the ICER themost were plotted. The full ICER tornado diagram including all parameters included in themodel can be seen in Figure A.3 in appendix A.4.
27
4 Results Aalborg University
A threshold one-way analysis on the annual medication cost of being treated with
teriparatide was conducted to identify the price of teriparatide required to make the
costs accumulated during the 51 cycles in the alendronate and teriparatide branches
equal. The one-way sensitivity analysis illustrated in Figure 4.2 provided a threshold
value of DKK 2,222, meaning the annual medication cost of teriparatide had to be
reduced to DKK 2,222 or less before the accumulated costs in the teriparatide branch
were lower than the accumulated costs in the alendronate branch when keeping
constant everything else.
Figure 4.2: The figure shows a one-way sensitivity analysis on the cost of teriparatide.
The required decrease in the annual cost of teriparatide to make teriparatide the
cheapest option in the model corresponded to an at least 93.3% price reduction
compared to the incorporated annual medication cost of being treated with
teriparatide.
Another one-way analysis was performed using no discounting on costs and effects
in the model resulting in an ICER of DKK 241,191 DKK per QALY. Likewise, the CUA
was performed using a higher discount rate of 7% in all cycles resulting in an ICER of
DKK 1,530,802 per QALY. When applying no discounting in the model, the resulting
ICER was below the employed WTP threshold of DKK 250,000 per QALY, meaning
teriparatide would be considered cost-effective when compared to alendronate if no
discounting was incorporated into the model.
28
4 Results Aalborg University
4.2.2. Probabilistic Sensitivity Analysis
A PSA was conducted and resulted in a CEAC and an ICE scatter plot which are
presented in Figures 4.3 and 4.4. When utilising a WTP threshold of DKK 250,000
per QALY and performing a PSA of 10,000 iterations, there was an approximately
3% probability of teriparatide being cost-effective compared to alendronate as can be
seen in Figure 4.3. The intersect of the curves indicated the point at which teriparatide
was just as likely to be cost-effective as alendronate was. In this case, the lowest WTP
threshold required for teriparatide to have the highest probability of being considered
cost-effective was approximately DKK 730,000 per QALY, meaning if a WTP threshold
at or above DKK 730,000 per QALY was utilised, teriparatide could be considered cost-
effective when compared to alendronate. However, when adopting a WTP threshold
of less than DKK 730,000 per QALY, as e.g. the employed WTP threshold of DKK
250,000 per QALY, there is a higher probability of alendronate being considered cost-
effective.
Figure 4.3: The figure shows an acceptability curve depicting the probability of teriparatidebeing cost-effective compared to alendronate at a given WTP threshold ranging from DKK100,000 to 1,300,000.
An ICE scatter plot depicting the placement of each of the 10,000 iterations on the
ICER-plane can be seen below in Figure 4.4. From the scatter plot it can be seen that
when plotting the 10,000 ICER iterations on the ICER plane, the iterations are placed
in two of the quadrants of the plane. One is the more expensive and effective quadrant
(the northeast quadrant to the right of the dotted vertical line), and the other is the
more expensive yet less effective quadrant (the northwest quadrant to the left of the
dotted vertical line). It can be seen that the majority of the ICER iterations comparing
teriparatide to alendronate were not cost-effective when adopting the WTP threshold
of DKK 250,000 since the majority of the iterations were placed to the left and above
29
4 Results Aalborg University
the stippled WTP threshold line, which is in accordance with the approximately 97%
of the iterations the CEAC indicated were not cost-effective at the WTP threshold of
DKK 250,000 per QALY when comparing teriparatide to alendronate. A substantial
amount of the 10,000 iterations was placed in the northwest quadrant which means
these ICER scatters represent simulations in which teriparatide is both more expensive
and less effective compared to alendronate, though the main part of the iterations is
situated in the northeast quadrant meaning teriparatide was more effective and more
expensive in the main part of iterations.
Figure 4.4: The incremental cost-effectiveness scatter plot depicts all 10,000 iterations fromthe PSA. The ellipse in the figure shows the 95% confidence interval.
4.2.3. Scenario Analyses
4.2.3.1. Changes in the Durations of the Effects of the Two Drugs
A scenario analysis was run on the base case to test the influence on the ICER other
durations of the effects of the two drugs than the life-long durations used in the
base case would have. In this scenario analysis, everything in the model was kept
the same as in the base case except for the durations of the effects of alendronate and
teriparatide. In this scenario, the full effects of alendronate were applied to the first six
stages followed by a linear decline of the effect over six years until reaching no effect.
For teriparatide/alendronate, the full effect of teriparatide was modelled for the first
eight years until the cessation of treatment followed by eight years of linear decline
in effect until no effect. This approach was in line with other studies [63, 67, 69] and
resulted in an ICER of DKK 1,044,617 per QALY.
30
4 Results Aalborg University
4.2.3.2. Time Horizon of Eight Years
To test the influence of the employed time horizon on the ICER, the model was run
using a time horizon of eight years instead of 51 years to investigate the influence of a
shorter time horizon, which would result in the women only being modelled from 50
and up to 58 years-old and e.g. ensure the effects of the drugs were not extrapolated
into the unknown.
This scenario analysis resulted in an ICER of DKK 15,201,429 per QALY which was
well above the employed WTP threshold of DKK 250,000 per QALY.
4.2.3.3. Changes in the Included Fracture Types
The structural uncertainty of the base case was tested by removing fracture types
included in the model. The base case was constructed by including four fracture
types. In this scenario analysis, forearm and "other" fractures were removed from
the model to investigate the influence of the inclusion of these two fracture types
on the ICER. When running 51 cycles of the model with hip and vertebral fractures
remaining in the model and forearm and "other" fractures removed, the resulting
ICER was DKK 890,385 per QALY.
4.3. State Probability Charts for Teriparatide and
Alendronate
The two state probability charts in Figures 4.5 and 4.6 show the probabilities of
individuals being in each health state at each stage modelled (0-51) for teriparatide
and alendronate, respectively. The state probability chart and the associated Markov
cohort summary report for teriparatide, which can be seen in Table A.4 in appendix
A.5, show the probability of being in the absorbing state at the end of the modelled
time horizon was 99.50% for teriparatide.
31
4 Results Aalborg University
Figure 4.5: The figure shows the state probability chart for the teriparatide branch depictingthe flow of individuals through each health state in the branch through the 51 cycles.
The state probability chart in Figure 4.6 and the associated Markov cohort summary
report, which can be seen in Table A.3 in appendix A.5, show that in the alendronate
branch, the probability of being in the health state "dead" was 99.55% following the
51 cycles.
Figure 4.6: The figure shows the state probability chart for the alendronate branch depictingthe flow of individuals through each health state in the branch through the 51 cycles.
When the two charts above and their associated summary reports, which can be seen
in Tables A.3 and A.4 in appendix A.5, were compared, it was observed that the
probability of experiencing a hip fracture while treated with alendronate was higher
32
4 Results Aalborg University
compared to the probability of incurring a hip fracture while treated with teriparatide.
The same tendency was observed when investigating the probabilities of sustaining
vertebral fractures since the probabilities of experiencing vertebral fractures were
higher when treated with alendronate compared to when treated with teriparatide.
33
5. Discussion
5.1. Results
The base case resulted in an ICER of DKK 761,057 per QALY. From the sensitivity
analyses, it emerged the result of this study was robust to all sensitivity analyses
performed, except for the sensitivity analysis in which no discounting was applied,
as this factor was the only factor investigated able to lower the ICER enough
for teriparatide/alendronate to be considered the cost-effective treatment when
compared to alendronate alone and applying a WTP threshold of DKK 250,000 per
QALY. However, the concept of discounting costs and health benefits is, seen from a
socio-economic perspective, an important aspect to include as it is considered fallible
not to consider the uncertainties associated with the expected future costs and health
benefits, and therefore, the result of the sensitivity analysis investigating the impact
no discounting would have on the ICER is informative though it is unlikely to provide
a basis for a given decision within this field.
When comparing the base case ICER to the result from another study by Liu et al.
[12] investigating the cost-effectiveness of sequential teriparatide/alendronate therapy
compared with alendronate alone using microsimulation, the ICER of our study was
lower than the ICER of DKK 1.07 million (2020 value) per QALY found in their study.
However, their analysis employed a wider perspective including nursing home costs,
and their study was conducted before the expiration of the patent of teriparatide,
meaning their employed annual cost of being treated with teriparatide was 34.7%
higher than the medication cost estimate included in this study. In addition to their
broader perspective and higher medication cost of teriparatide, their study used
higher costs for forearm and hip fractures, and they furthermore did not include
"other" fractures in their analysis. Hence the study by Liu et al. [12] was not fully
comparable to our study, however, a study fully comparable to our study was not
identified.
It was found from the state probability charts and their associated summary reports
that the patients treated with teriparatide are less likely to experience both hip and
vertebral fractures than patients treated with alendronate which is in accordance with
the effects of teriparatide and alendronate obtained from a study by Mori et al. [53],
since teriparatide was superior in reducing the risk of suffering both hip and vertebral
fractures. The fracture efficacy estimates used in the present study were comparable
to the estimates used in other studies incorporating teriparatide as the drug being
superior in reducing fracture risks [12, 98]. However, some discrepancy exists between
studies regarding the effects the two drugs provide in terms of reducing fracture risk
34
5 Discussion Aalborg University
in relative risks compared to placebo, as e.g. Hiligsmann et al. [69] have employed
effects of alendronate on the occurrence of hip and wrist fractures that were more
effective than for teriparatide. However, the inclusion of the effects from Hiligsmann
et al. [69] would not be expected to lead to a different conclusion regarding which
one of the two treatments was considered cost-effective at a WTP threshold of DKK
250,000 per QALY. On the contrary, it would be expected to strengthen the conclusion
since the effects of teriparatide in their study were less efficacious than the effects
included in our study resulting in an even smaller difference in QALY gain between
teriparatide and alendronate or it may even result in teriparatide being dominated. As
can be seen in the ICER tornado diagram in Figure 4.1, the effects of the drugs were
found to be of importance since it emerged that seven out of eight of the drug effect
estimates on the incidence of each fracture type were among the nine parameters
able to affect the ICER the most when varied one by one. This indicates the choice
of literature providing the effect estimates is of extreme importance and should be
chosen carefully since, in the present study, they were found to be able to influence the
final ICER substantially. However, remembering the drawbacks of one-way sensitivity
analyses is important in this context since the likelihood of each parameter reaching
its high or low value is unknown.
Our model predicted the probability of being in the absorbing "dead" state at the
end of the modelled time horizon was approximately 99.5% for both teriparatide and
alendronate. Another study by Mori et al. [53] found when employing a 105-year time
horizon that the probability of being dead at the end of the model was 99.9%, which
must be considered very similar to the results of our study when taking into account
the difference in time horizons.
5.2. Methods
5.2.1. The Included Estimates of the Effects of Teriparatide and
Alendronate
An advantage of using DAMs was emphasised during the conduction of this study
since the effects of the two drugs obtained in the literature were only identified as the
effect of each drug when compared to placebo expressed as relative risks, meaning
no head-to-head comparison between teriparatide and alendronate was identified.
However, the use of DAMs enabled an indirect comparison of the two drugs by
combining data. This necessitated the inclusion of a baseline fracture risk in an
osteoporotic population which had to be transferable to a Danish setting. When
applying this approach instead of a direct comparison of the effects of the two drugs,
the risk of introducing uncertainties in the model is higher, since a baseline reflecting
35
5 Discussion Aalborg University
the actual fracture risk in an osteoporotic population might be difficult to obtain. It
can be difficult to obtain because it would be ethically wrong not to treat patients
diagnosed with osteoporosis since the disease influences both the quality and length
of an individual’s life [48]. And therefore, the fracture risk baseline in an osteoporotic
population had to be estimated. The employed baseline probabilities of sustaining a
fracture of each of the four fracture types when suffering from osteoporosis can be
seen in Tables A.8 through A.11 in appendix A.8. The probabilities of sustaining a
fracture of any of the four types are surprisingly low when considering the modelled
cohort is a population diagnosed with a condition causing a higher risk of bone
fractures. As an example, the probability of sustaining a hip fracture for a 70-years-
old woman was just 0.01, meaning only 1% incurred a hip fracture in cycle 20 and
this was the probability before the application of the effects of the drugs. However,
the baseline probabilities of sustaining fractures were found similar to the estimated
baselines used in other studies [51, 69] and were therefore deemed reasonable for
inclusion into our model.
The estimates of effects of the drugs obtained from the literature for use in our model
were, in some instances, inexact. In particular, the CIs of the three parameters used
in the model for the effect of alendronate on the occurrence of forearm fractures and
the effects of teriparatide on the occurrence of forearm and hip fractures were, in
the original literature, all assumed to be efficacious in reducing the occurrence of
fractures. However, Mori et al. [53] were aware that these three relative risk estimates
had a CI ranging above 1 meaning they were not guaranteed to be efficacious. This
approach applied by Mori et al. [53] resulted in the CIs in question being considered
ranges of uncertainties rather than real CIs and they therefore no longer represented
the real uncertainty related to these parameters. For the results of the PSA of our
model, this meant these three effect estimates were overestimated since the "CIs" of
these effects should, in fact, incorporate the possibility of the relative risks going
above 1, meaning they could lead to more fractures. In conclusion, our model, when
running the PSA, was more effective in reducing the risk of fractures than would have
been the case if these parameters were not cut off at 1.
For the effectiveness of the two drugs on fracture risks, relative risks of sustaining
a fracture of each of the four fracture types while treated compared to fracture
risks in a placebo group were obtained as efficacy estimates from the literature as
opposed to estimates of effects. Studies estimating efficacy are conducted under
ideal circumstances whereas effectiveness trials are conducted in "real-world" clinical
settings [99]. This inevitably leads to the effects included in the model being better
than what may be observed in a normal population of osteoporotic patients. This
36
5 Discussion Aalborg University
means the modelled approach using efficacy estimates likely led to a smaller amount
of fractures than would be observed in reality resulting in an overestimation of
the effects the drugs have on the reduction of fracture risk which further leads
to an overestimation of the accumulations of QALYs. This approach was chosen
due to a lack of estimates of effectiveness, however, since efficacy estimates were
utilised for both alendronate and teriparatide, no expected difference between the
two comparators was expected to be present.
Moreover, during this study, it was assumed that the effects of the drugs in terms
of reducing fracture risks when treated with alendronate or teriparatide persisted for
all years modelled, which must be considered a quite optimistic approach since no
studies confirming the effects are that long-lasting have been identified. Hence, to test
the influence of the effects of the treatments after cessation of treatment on the final
result of the model, a sensitivity analysis was performed assuming a linear decrease
in the effects of the drugs through eight and six years after treatment cessation of
sequential teriparatide/alendronate and alendronate alone, respectively. The ICER of
this scenario analysis was above DKK 250,000 per QALY resulting in no change of
conclusion regarding the cost-effectiveness of teriparatide compared to alendronate.
5.2.2. The Inclusion of Forearm and "Other" Fractures in the
Markov Model
Regarding the importance of including both the post-fracture health states for forearm
and "other" fractures, the relevance of their inclusion can be discussed. In a review
by Zethraeus et al. [48], it is argued that if only minor changes in costs, utility
values, and mortality are related to specific disease states within the model, it can
be suggested to exclude the health states in question. As an example, the exclusion
of a forearm fracture health state has been shown to be of only minor importance for
the ultimate cost-effectiveness results [48]. However, for the purpose of this study,
the common osteoporotic fractures were included to create a detailed osteoporosis
model, since assuming osteoporotic patients do not experience e.g. any forearm
fractures at all would be a gross oversimplification. The same goes for the "other"
fractures since osteoporotic patients are at higher risks of sustaining more of all types
of fractures than a population free of osteoporosis would be [51]. However, it should
be kept in mind as well that all parameters included in the model related to "other"
fractures inevitably carry an extra amount of uncertainty and roughness because of
the heterogeneity of the fractures included.
To test the influence on the ICER the inclusion of forearm and other fractures
had in the base case analysis, the two fracture types were removed completely
37
5 Discussion Aalborg University
from the model in a scenario analysis. The scenario analysis resulted in an ICER
of DKK 890,385 per QALY which is just about DKK 129,000 per QALY higher
than the base case ICER which can be considered a relatively small increase when
removing the two fracture types. This scenario analysis further supports the finding
that teriparatide/alendronate was not cost-effective compared to alendronate since
regardless of the inclusion of these two fractures, the resulting ICERs were well above
the employed WTP threshold of DKK 250,000 per QALY meaning the base case ICER
was not sensitive to the exclusion of forearm and "other" fractures, which backs the
theory by Zethraeus et al. [48] that the inclusion of e.g. forearm fractures is of minor
importance for the ultimate decision.
5.2.3. Half-Cycle Correction
Including half-cycle correction in the model was considered, however, the final
decision was to omit half-cycle correction completely. Half-cycle correction is usually
performed because computer-based models simulate an occurrence of transitions at
either the end or the beginning of each cycle. However, this often fails to reflect
reality since transitions are often assumed to occur half-way through each cycle on
average.[100] In TreeAge Pro, the full cycle’s state reward is accumulated at the
beginning of each cycle while all transitions occur at the end of each cycle leading
to an overestimation of expected values [101].
In our model, the first years of treatment were quite important in terms of costs, as
they constitute a substantial part of the total costs per patient accumulated through
the 51 cycles, since the medication costs are primarily accumulated during the
first cycles, and therefore it would have been unreasonable to implement half-cycle
correction for these costs. It was considered to omit the costs of the drugs from the
half-cycle correction but still correct all other costs and effects included in the model,
however, this would result in an unwanted in-congruence in the model output and
was therefore avoided. Besides, the accumulation of costs and effects during the
remaining cycles was limited, and a half-cycle correction was expected to be of minor
importance for the result of this study.
5.2.4. Ethical Considerations
An ethical consideration regarding the employment of a WTP-threshold in economic
evaluations concerns the fact that some drugs might not be provided for widespread
use because the cost per QALY is above a given WTP threshold, meaning e.g. an
effective treatment may not be offered for all relevant patients due to the price
level, which would deprive some patients the opportunity of benefiting from the
effective treatment. This might be the case for teriparatide, as it is an effective yet
38
5 Discussion Aalborg University
very expensive treatment which has up until now been restricted for use in patients
suffering from severe osteoporosis in Denmark. However, though a treatment method
may be found to be more effective than others, the implementation of a non-cost-
effective intervention would result in a net QALY loss if the intervention was to be
implemented for widespread use, which would be due to high opportunity costs,
which would not be a socially desirable outcome.
5.2.5. Limitations
5.2.5.1. Perspective
Ideally, a strict societal perspective should be included to ensure consistency with
economic theory, and it is furthermore the recommended approach in the field of
osteoporosis since the ultimate consequences of the condition go beyond any specific
health care system [102, 48]. However, during this study limitations regarding
relevant estimates provided in the literature and access to patient-specific data was
restricted, thereby limiting the inclusion of all the parameters needed to conduct
an analysis with a societal perspective. Therefore, a decision was made to only
include somewhat certain parameters into the model to avoid the incorporation of
unacceptably uncertain parameters meaning costs such as those of side effects due
to the two investigated drugs and the number of people being taken care of in care
homes and the related costs were omitted. Since the applied health care perspective
extended to include patient-paid costs is narrower than a societal perspective, some
aspects of benefits and the associated opportunity costs were excluded, and therefore,
this study is unlikely to provide a basis for social choice. In addition, differences
in the chosen perspectives of various studies within the field of osteoporosis along
with the significance of the chosen setting can add to the difficulty of finding studies
comparable to the present study.
5.2.5.2. Considerations regarding Adherence and Persistence
In the model, a limited number of parameters were included, however, there are
inevitably more relevant aspects of the matter to consider. One of these is adherence
to the treatments in question as well as compliance. Adherence rates to treatments
for osteoporosis have been shown to be poor in general [103] since for instance Imaz
et al. [104] found that the mean amount of days of persistence was just 184 days with
a follow-up of 365 days and several studies have found medical possession rates lower
than the lowest optimal level of 80% [103, 104, 34].
For oral BPs, specifically, both the overall compliance and persistence rates are
suboptimal. However, those using weekly BP medication demonstrate better
39
5 Discussion Aalborg University
compliance and persistence rates than patients prescribed a dosing regimen of daily
BP [34]. Moreover, Cramer et al. [34] state that in addition to osteoporosis being a
silent disorder, BPs carry the inconvenience of needing to be consumed following a
night of fasting and the patient moreover needs to stay in an upright position for
30 minutes as mentioned in section 2.2.1. Keeping in mind the poor adherence to
oral BPs and the aforementioned associated inconveniences related to administration,
injectable agents for treating osteoporosis have furthermore been found to carry
both higher persistence and adherence rates than for oral therapies [105] leaving
the question of whether teriparatide, an injectable agent, should be preferred for a
proportion of patients showing very low adherence and persistence rates despite the
higher costs of teriparatide. However, one factor that might contribute to a higher rate
of persistence for patients taking teriparatide could be patients prescribed teriparatide
may be more motivated to be adherent and persistent to their treatments as they often
suffer from severe osteoporosis, since, for instance in Denmark, the use of teriparatide
is only recommended for patients suffering from severe osteoporosis.
For future investigation, an approach incorporating adherence and persistence into
Markov models should be preferred since Cobden et al. [106] has found the
incorporation of adherence when modelling a chronic disease (diabetes type 2 in
their case) can influence the final ICER found through Markov models markedly.
In our model, the expected influence of incorporating adherence and persistence
would be that the incorporation of these factors would lead to teriparatide looking
more favourable when compared to alendronate than it did in the present model in
which persistence and adherence were omitted. This was expected since the effects
of alendronate would be poorer if patients were not adherent and persistent with
therapy due to the inconveniences associated with administration [34].
5.2.5.3. Utility Decrements Related to Modes of Administration of Teriparatide
and Alendronate
In our model, no utility decrements related to the modes of administration of
the drugs were incorporated, however, this may be a factor able to influence
the result of the base case ICER. Deducing in what way the incorporation of
these types of utility decrements would influence the result of our study was
not straightforward since subcutaneous injections can be associated with some
inconveniences, however, that might be the case for the oral administration of
alendronate as well since the requirements pre- and post-administration are quite
strict. It has been found in a study by Hadi et al. [107] that when the burden associated
with treatment mode of administration increases, it is associated with a subsequent
40
5 Discussion Aalborg University
decline in utility and should, therefore, be included when doable. Unfortunately,
no study characterising the utility decrements for weekly oral alendronate and daily
subcutaneous teriparatide was identified, hence the lack of incorporation of utility
decrements related to modes of administration.
5.3. What’s Next?5.3.1. Restricted Prescription of Teriparatide in Denmark
An underlying assumption during the conduction of this model-based study was that
all individuals passing through the model can be prescribed teriparatide, when, in
reality, a certain degree of the severeness of the osteoporotic condition is demanded
for teriparatide to be prescribed in Denmark. Hence, the two modelled populations
were modelled as being equal in terms of the severity of their osteoporosis when,
in real life, they probably are not, at least not in Denmark. However, it should
be kept in mind that in Denmark, the use of teriparatide is restricted for patients
severely affected by their condition only because of the high price of the compound
and not because teriparatide has not been approved for use in all postmenopausal
women, meaning if the costs of teriparatide were lower, severity might be redundant.
These were the considerations creating the basis for the one-way analysis investigating
what level the medication costs of teriparatide would need to be lowered to approach
redundancy of severity of conditions of osteoporosis. As mentioned earlier in the
result section 4.2.1, for the costs accumulated in the teriparatide branch to be equal
to the costs accumulated in the alendronate branch through the 51 cycles, the annual
cost of teriparatide would need to be just DKK 2,222 as can be seen in Figure 4.2. The
required reduction of the medication costs of teriparatide to DKK 2,222 is in stark
contrast to the currently incorporated medication cost of teriparatide of DKK 33,098.
A reduction of the medication cost of teriparatide of this magnitude would call for
a reduction of the current annual cost of 93.3%. As previously mentioned in section
2.2, biosimilars are now available and therefore a reduction of the price from when
teriparatide was patented and till the patent went off has already taken place and
resulted in a 26.5% reduction. This causes it to be highly unlikely that the price will be
further reduced by 93.3% thereby making it unlikely that teriparatide will be adopted
for wide use in the near future despite the effectiveness of the compound. However, a
recent study demonstrated cyclic teriparatide therapy in patients taking alendronate
reduced the dose of teriparatide with 50% over two years, while still maintaining
the full effects of the drug [12]. This leaves the opportunity of using teriparatide in
a cheaper yet very effective way, however, this alternative way of using teriparatide
should be investigated further before a decision to implement can be made.
41
6. Conclusion
Because the price of teriparatide has been reduced by 26.5% following the
expiration of its patent, the relevance of more recent investigations of the cost-
effectiveness of teriparatide compared to the most frequently used treatment with
the bisphosphonate, alendronate, has increased. Therefore, a cost-utility analysis
comparing the use of sequential teriparatide/alendronate to alendronate alone for
treating women above the age of 50 was performed by constructing a Markov
model utilising a lifetime horizon of 51 years with an applied healthcare perspective
extended to include patient-paid costs.
The base case analysis resulted in an ICER of DKK 761,057 per QALY which means
treating Danish women above the age of 50 with sequential teriparatide/alendronate
was not considered cost-effective compared to treating these women with alendronate
when employing the WTP threshold of DKK 250,000 per QALY.
The result was robust to all sensitivity analyses performed except for the use of no
discounting on costs and effects in the model which led to an ICER of DKK 241,191
per QALY which was just below the employed WTP threshold of DKK 250,000 per
QALY. In general, the parameters most likely to influence the result of this study
when varied one at a time in one-way analyses were the effects of the two drugs
in reducing fracture risk. In addition, it was found from the PSA that there was
only a 3% probability teriparatide/alendronate was cost-effective when compared to
alendronate when employing the WTP threshold of DKK 250,000.
This study concluded that teriparatide/alendronate was more effective since the
amount of QALYs accumulated through 51 cycles was higher than for alendronate,
however, benefiting from the better effect of the sequential teriparatide/alendronate is
quite expensive for the gain in QALYs achieved. The base case ICER of DKK 761,057
per QALY indicated adopting the treatment with teriparatide for wide use would
lead to a displacement of QALYs since the opportunity costs are simply too high if
assuming society is willing to pay no more than DKK 250,000 per QALY, which can
also be interpreted as the acceptable amount of opportunity costs and benefits society
is willing to spend on a QALY. Therefore, this study indicated an expansion of the use
of teriparatide would lead to a net loss of QALYs, thereby suggesting the requirements
for receiving treatment with teriparatide should not be loosened shortly.
42
References
[1] S. K. Sandhu and G. Hampson. The pathogenesis, diagnosis, investigation and management of
osteoporosis. Journal of Clinical Pathology, 64(12):1042–1050, 2011.
[2] Sundhedsstyrelsen. Osteoporose - en afdækning af den samlede ind-
sats mod osteoporose, 2018. URL https://www.sst.dk/da/udgivelser/2018/
osteoporose-en-afdaekning-af-den-samlede-indsats-mod-osteoporose. [Access date:
2020-05-28].
[3] E. Hernlund, A. Svedbom, M. Ivergård, J. Compston, C. Cooper, J. Stenmark, E. V. McCloskey,
B. Jönsson, and J. A. Kanis. Osteoporosis in the European Union: Medical management,
epidemiology and economic burden: A report prepared in collaboration with the International
Osteoporosis Foundation (IOF) and the European Federation of Pharmaceutical Industry
Associations (EFPIA). Archives of Osteoporosis, 8(1-2), 2013.
[4] L. Hansen, A. S. Mathiesen, P. Vestergaard, L. H. Ehlers, and K. D. Petersen. A health economic
analysis of osteoporotic fractures: Who carries the burden? Archives of Osteoporosis, 8(1-2), 2013.
[5] R. Burge, B. Dawson-Hughes, D. H. Solomon, J. B. Wong, A. King, and A. Tosteson. Incidence
and economic burden of osteoporosis-related fractures in the United States, 2005-2025. Journal of
Bone and Mineral Research, 22(3):465–475, 2007.
[6] D. Scarlet. Videnscenter for Knoglesundhed. Journal of Chemical Information and Modeling, 53(9):
1689–1699, 2019.
[7] International Osteoporosis Foundation. Epidemiology, 2020. URL https://www.iofbonehealth.
org/epidemiology. [Access date: 2020-04-22].
[8] The Danish Health Data Authority. Medstat.dk, 2019. URL https://medstat.dk/. [Access date:
2020-03-17].
[9] A. B. Hodsman, D. C. Bauer, D. W. Dempster, L. Dian, D. A. Hanley, S. T. Harris, D. L. Kendler,
M. R. McClung, P. D. Miller, W. P. Olszynski, E. Orwoll, and K. Y. Chui. Parathyroid hormone
and teriparatide for the treatment of osteoporosis: A review of the evidence and suggested
guidelines for its use. Endocrine Reviews, 26(5):688–703, 2005.
[10] R. M. Neer, C. D. Arnaud, J. R. Zanchetta, R. Prince, G. A. Gaich, J. Y. Reginster, A. B. Hodsman,
E. F. Eriksen, S. Ish-Shalom, H. K. Genant, O. Wang, B. H. Mitlak, D. Mellström, E. S. Oefjord,
E. Marcinowska-Suchowierska, J. Salmi, H. Mulder, J. Halse, and A. Z. Sawicki. Effect of
parathyroid hormone (1-34) on fractures and bone mineral density in postmenopausal women
with osteoporosis. New England Journal of Medicine, 344(19):1434–1441, 2001.
[11] D. L. Kendler, F. Marin, C. A. F. Zerbini, L. A. Russo, S. L. Greenspan, V. Zikan, A. Bagur,
J. Malouf-Sierra, P. Lakatos, A. Fahrleitner-Pammer, E. Lespessailles, S. Minisola, J. J. Body,
P. Geusens, R. Möricke, and P. López-Romero. Effects of teriparatide and risedronate on new
fractures in post-menopausal women with severe osteoporosis (VERO): a multicentre, double-
blind, double-dummy, randomised controlled trial. The Lancet, 391(10117):230–240, 2018.
[12] H. Liu, K. Michaud, S. Nayak, D. B. Karpf, D. K. Owens, and A. M. Garber. The cost-effectiveness
of therapy with teriparatide and alendronate in women with severe osteoporosis. Archives of
Internal Medicine, 166(11):1209–1217, 2006.
[13] J. Compston, A. Cooper, C. Cooper, N. Gittoes, C. Gregson, N. Harvey, S. Hope, J. A. Kanis,
E. V. McCloskey, K. E.S. Poole, D. M. Reid, P. Selby, F. Thompson, A. Thurston, and N. Vine. UK
clinical guideline for the prevention and treatment of osteoporosis. Archives of Osteoporosis, 12(1),
2017.
[14] I. Takács, E. Jókai, D. E. Kováts, and I. Aradi. The first biosimilar approved for the treatment of
43
References Aalborg University
osteoporosis: results of a comparative pharmacokinetic/pharmacodynamic study. Osteoporosis
International, 30(3):675–683, 2019.
[15] The Danish Endocrine Society. NBV: Postmenopausal osteoporose, 2016. URL http://www.
endocrinology.dk/index.php/3-calcium-og-knoglemetaboliske-sygdomme/3-osteoporose.
[Access date: 2020-04-21].
[16] Sundhed.dk. Knogletæthedsmålinger, 2019. URL https://www.sundhed.dk/sundhedsfaglig/
laegehaandbogen/undersoegelser-og-proever/undersoegelser/oevrige-undersoegelser/
knogletaethedsmaalinger/. [Access date: 2020-05-28].
[17] P. Vestergaard, L. Rejnmark, and L. Mosekilde. Osteoporosis is markedly underdiagnosed: A
nationwide study from Denmark. Osteoporosis International, 16(2):134–141, 2005.
[18] Dansk Knoglemedicinsk Selskab. Vejledning til udregning og behandling af osteoporose.
Technical report, Dansk Knoglemedicinsk Selskab, 2012.
[19] The Danish Health Data Authority. Prævalens, incidens og behandling i sund-
hedsvæsenet for borgere med osteoporose, 2018. URL https://www.sst.dk/~/media/
73D2379F2BA24D98866A921E9D418900.ashx. [Access date: 2019-02-21].
[20] K. L. Stone, D. G. Seeley, L. Y. Lui, J. A. Cauley, K. Ensrud, W. S. Browner, M. C. Nevitt, and S. R.
Cummings. BMD at Multiple Sites and Risk of Fracture of Multiple Types: Long-Term Results
From the Study of Osteoporotic Fractures. Journal of Bone and Mineral Research, 18(11):1947–1954,
2003.
[21] O. Ström, F. Borgström, N. Zethraeus, O. Johnell, L. Lidgren, S. Ponzer, O. Svensson, P. Abdon,
E. Ornstein, L. Ceder, K. G. Thorngren, I. Sernbo, and B. Jönsson. Long-term cost and effect on
quality of life of osteoporosis-related fractures in Sweden. Acta Orthopaedica, 79(2):269–280, 2008.
[22] S. Boonen and A. J. Singer. Osteoporosis management: Impact of fracture type on cost and
quality of life in patients at risk for fracture I. Current Medical Research and Opinion, 24(6):1781–
1788, 2008.
[23] D. Diacinti and G. Guglielmi. How to define an osteoporotic vertebral fracture? Quantitative
Imaging in Medicine and Surgery, 9(9):1485–1494, 2019.
[24] B. Nygaard, J. K. Kristensen, and H. C. Kjeldsen. Lægehåndbogen - Osteoporose,
2020. URL https://www.sundhed.dk/sundhedsfaglig/laegehaandbogen/endokrinologi/
tilstande-og-sygdomme/knoglevaev-og-vitamin-d/osteoporose/. [Access date: 2019-02-19].
[25] D. T. Gold. The nonskeletal consequences of osteoporotic fractures: Psychologic and social
outcomes. Rheumatic Disease Clinics of North America, 27(1):255–262, 2001.
[26] M. Ji and Q. Yu. Primary osteoporosis in postmenopausal women. Chronic Diseases and
Translational Medicine, 1(1):9–13, 2015.
[27] D. M. Black, A. V. Schwartz, K. E. Ensrud, J. A. Cauley, S. Levis, S. A. Quandt, S. Satterfield,
R. B. Wallace, D. C. Bauer, L. Palermo, L. E. Wehren, A. Lombardi, A. C. Santora, and S. R.
Cummings. Effects of continuing or stopping alendronate after 5 years of treatment: The Fracture
Intervention Trial long-term extension (FLEX): A randomized trial. Journal of the American Medical
Association, 296(24):2927–2938, 2006.
[28] E. Canalis, A. Giustina, and J. P. Bilezikian. Mechanisms of Anabolic Therapies for Osteoporosis.
The New England Journal of Medicine, 357:905–916, 2007.
[29] L. S. Kates and C. L. Ackert-Bicknell. How Do Bisphosphonates Affect Fracture Healing?
Physiology & behavior, 176(10):139–148, 2017.
[30] Pro.medicin. Bisfosfonater, 2019. URL http://pro.medicin.dk/laegemiddelgrupper/grupper/
160520. [Access date: 2019-02-18].
44
References Aalborg University
[31] A. Cranney, G. Wells, A. Willan, L. Griffith, N. Zytaruk, V. Robinson, D. Black, J. Adachi, B. Shea,
P. Tugwell, and G. Guyatt. II. Meta-analysis of alendronate for the treatment of postmenopausal
women. Endocrine Reviews, 23(4):508–516, 2002.
[32] U. A. Liberman, S. R. Weiss, J. Bröll, H. W. Minne, H. Quan, N. H. Bell, J. Rodriguez-Portales,
R. W. Downs, J. Dequeker, M. Favus, E. Seeman, R. R. Recker, T. Capizzi, A. C. Santora,
A. Lombardi, R. V. Shah, L. J. Hirsch, and D. B. Karpf. Effect of oral alendronate on bone
mineral density and the incidence of fractures in postmenopausal osteoporosis. New England
Journal of Medicine, 333(22):1437–1444, 1995.
[33] Min.medicin.dk. Alendronat - min.medicin.dk, 2020. URL https://min.medicin.dk/Medicin/
Praeparater/4579. [Access date: 2019-02-19].
[34] J. A. Cramer, D. T. Gold, S. L. Silverman, and E. M. Lewiecki. A systematic review of persistence
and compliance with bisphosphonates for osteoporosis. Osteoporosis International, 18(8):1023–
1031, 2007.
[35] R. Lindsay, J. H. Krege, F. Marin, L. Jin, and J. J. Stepan. Teriparatide for osteoporosis: Importance
of the full course. Osteoporosis International, 27(8):2395–2410, 2016.
[36] European Medicines Agency. EMA - FORSTEO. Technical report, European Medicines Agency,
2012.
[37] N. Napoli, B. L. Langdahl, Ö. Ljunggren, E. Lespessailles, G. Kapetanos, T. Kocjan, T. Nikolic,
P. Eiken, H. Petto, T. Moll, E. Lindh, and F. Marin. Effects of Teriparatide in Patients
with Osteoporosis in Clinical Practice: 42-Month Results During and After Discontinuation of
Treatment from the European Extended Forsteo® Observational Study (ExFOS). Calcified Tissue
International, 103(4):359–371, 2018.
[38] Lilly USA. Full prescribing information - Forteo. Technical report, Lilly USA, 2019.
[39] ProMedicin.dk. Teriparatid (osteoporose). 2019.
[40] R. L. Jilka, R. S. Weinstein, T. Bellido, P. Roberson, A. M. Parfitt, and S. C. Manolagas. Increased
bone formation by prevention of osteoblast apoptosis with parathyroid hormone. Journal of
Clinical Investigation, 104(4):439–446, 1999.
[41] N. Miyakoshi, Y. Kasukawa, T. A. Linkhart, D. J. Baylink, and S. Mohan. Evidence that anabolic
effects of PTH on bone require IGF-I in growing mice. Endocrinology, 142(10):4349–4356, 2001.
[42] E. Canalis, M. Centrella, W. Burch, and T. L. McCarthy. Insulin-like growth factor I mediates
selective anabolic effects of parathyroid hormone in bone cultures. Journal of Clinical Investigation,
83(1):60–65, 1989.
[43] D. T. Gold, B. S. Pantos, D. N. Masica, D. A. Misurski, and R. Marcus. Initial experience with
teriparatide in the United States. Current Medical Research and Opinion, 22(4):703–708, 2006.
[44] Pro.medicin, L. Rejnmark, and B. L. Hansen. Teriparatid, 2019. URL https://pro.medicin.dk/
Laegemiddelgrupper/grupper/315021. [Access date: 2019-02-17].
[45] European Medicines Agency. Terrosa - EMA, 2019. URL https://www.ema.europa.eu/en/
medicines/human/EPAR/terrosa. [Access date: 2020-05-28].
[46] European Medicines Agency. Movymia - EMA, 2019. URL https://www.ema.europa.eu/en/
medicines/human/EPAR/movymia. [Access date: 2020-05-28].
[47] The Danish Medicines Agency. Medicinpriser.dk, 2020. URL https://medicinpriser.dk/
default.aspx. [Access date: 2020-03-17].
[48] N. Zethraeus, W. B. Sedrine, F. Caulin, S. Corcaud, H. J. Gathon, M. Haim, O. Johnell, and B. Jo.
International Review Article Models for Assessing the Cost-Effectiveness of the Treatment and
Prevention of Osteoporosis. Osteoporosis international : a journal established as result of cooperation
45
References Aalborg University
between the European Foundation for Osteoporosis and the National Osteoporosis Foundation of the USA,
13:841–857, 2002.
[49] D. Andrew. Carrying NICE over the threshold, 2015. URL https:
//www.nice.org.uk/news/blog/carrying-nice-over-the-threshold?fbclid=
IwAR0G6lhBWNB7xflDcehZdaSvmuPC5duLmCITjlOkA2jK-L0XXz79THlbmP8. [Access date: 2020-05-
13].
[50] M. Hiligsmann, O. Bruyère, O. Ethgen, H. J. Gathon, and J. Y. Reginster. Lifetime absolute risk
of hip and other osteoporotic fracture in Belgian women. Bone, 43(6):991–994, 2008.
[51] M. Hiligsmann and J. Y. Reginster. Cost effectiveness of denosumab compared with
oral bisphosphonates in the treatment of post-menopausal osteoporotic women in Belgium.
PharmacoEconomics, 29(10):895–911, 2011.
[52] N. Freemantle, C. Cooper, A. Diez-Perez, M. Gitlin, H. Radcliffe, S. Shepherd, and C. Roux.
Response to comments on: Results of indirect and mixed treatment comparison of fracture
efficacy for osteoporosis treatments: A meta-analysis. Osteoporosis International, 24(6):1931–1932,
2013.
[53] T. Mori, C. J. Crandall, and D. A. Ganz. Cost-Effectiveness of Sequential Teriparatide/Alen-
dronate Versus Alendronate-Alone Strategies in High-Risk Osteoporotic Women in the US: An-
alyzing the Impact of Generic/Biosimilar Teriparatide. JBMR Plus, 3(11), 2019.
[54] O. Johnell, J. A. Kanis, A. Odén, I. Sernbo, I. Redlund-Johnell, C. Petterson, C. De Laet, and
B. Jönsson. Mortality after osteoporotic fractures. Osteoporosis International, 15(1):38–42, 2004.
[55] Statistics Denmark. Consumer Price Index, 2019. URL https://www.dst.dk/da/Statistik/
emner/priser-og-forbrug/forbrugerpriser/forbrugerprisindeks. [Access date: 2020-04-
16].
[56] The Danish Health Data Authority. Interactive DRG, 2020. URL http://interaktivdrg.
sundhedsdata.dk. [Access date: 2020-03-26].
[57] M. Hiligsmann, O. Ethgen, F. Richy, and J. Y. Reginster. Utility values associated with
osteoporotic fracture: A systematic review of the literature. Calcified Tissue International, 82(4):
288–292, 2008.
[58] T. Peasgood, K. Herrmann, J. A. Kanis, and J. E. Brazier. An updated systematic review of health
state utility values for osteoporosis related conditions. Osteoporosis International, 20(6):853–868,
2009.
[59] M. Drummond. Methods for Economic Evaluation of Health Care Programmes. Oxford, fourth edition,
2015.
[60] A. Brennan and R. Akehurst. Modelling in health economic evaluation: What is its place? What
is its value? PharmacoEconomics, 17(5):445–459, 2000.
[61] T. A. Sheldon. Problems of using modelling in the economic evaluation of health care. Health
Economics, 5(1):1–11, 1996.
[62] J. Soto. Health economic evaluations using decision analytic modeling. Principles and practices
- Utilization of a checklist to their development and appraisal. International journal of technology
assessment in health care, 18(1):94–111, 2002.
[63] M. Hiligsmann, O. Ethgen, O. Bruyère, F. Richy, H. J. Gathon, and J. Y. Reginster. Development
and validation of a markov microsimulation model for the economic evaluation of treatments in
osteoporosis. Value in Health, 12(5):687–696, 2009.
[64] F. A. Sonnenberg and J. R. Beck. Markov Models in Medical Decision Making: A Practical Guide.
Medical Decision Making, 13(4):322–338, 1993.
46
References Aalborg University
[65] E. Olariu, K. K. Cadwell, E. Hancock, D. Trueman, and H. Chevrou-Severac. Current
recommendations on the estimation of transition probabilities in Markov cohort models for use
in health care decision-making: A targeted literature review. ClinicoEconomics and Outcomes
Research, 9:537–546, 2017.
[66] A. N.A. Tosteson, B. Jönsson, D. T. Grima, B. J. O’Brien, D. M. Black, and J. D. Adachi.
Challenges for model-based economic evaluations of postmenopausal osteoporosis interventions.
Osteoporosis International, 12(10):849–857, 2001.
[67] T. Yoshizawa, T. Nishino, I. Okubo, and M. Yamazaki. Cost-effectiveness analysis of drugs for
osteoporosis treatment in elderly Japanese women at high risk of fragility fractures: comparison
of denosumab and weekly alendronate. Archives of Osteoporosis, 13(1), 2018.
[68] F. Borgström, O. Ström, F. Marin, A. Kutahov, and Ö. Ljunggren. Cost effectiveness of teriparatide
and PTH(1-84) in the treatment of postmenopausal osteoporosis. Journal of Medical Economics, 13
(3):381–392, 2010.
[69] M. Hiligsmann, S. A. Williams, L. A. Fitzpatrick, S. S. Silverman, R. Weiss, and J. Y. Reginster.
Cost-effectiveness of sequential treatment with abaloparatide vs. teriparatide for United States
women at increased risk of fracture. Seminars in Arthritis and Rheumatism, 49(2):184–196, 2019.
[70] S. Taheri, F. Mirzayeh Fashami, F. Peiravian, and P. Yousefi. Teriparatide in the Treatment of
Severe Postmenopausal Osteoporosis: A Cost-Utility Analysis. Iranian journal of pharmaceutical
research : IJPR, 18(2):1073–1085, 2019.
[71] The Danish Ministry of Finance. Den samfundsøkonomiske diskonteringsrente. Technical
Report november, The Danish Ministry of Finance, 2018. URL file:///C:/Users/Laurynas/
Downloads/Dokumentationsnotat_Densamfundsoekonomiskediskonteringsrente.pdf.
[72] L. J. Melton, M. Thamer, N. F. Ray, J. K. Chan, C. H. Chesnut, T. A. Einhorn, C. C. Johnston, L. G.
Raisz, S. L. Silverman, and E. S. Siris. Fractures attributable to osteoporosis: Report from the
national osteoporosis foundation. Journal of Bone and Mineral Research, 12(1):16–23, 1997.
[73] J. R. Beck and S. G. Pauker. The Markov Process in Medical Prognosis. Medical Decision Making,
3(4), 1983.
[74] C. M. Klotzbuecher, P. D. Ross, P. B. Landsman, T. A. Abbott, and M. Berger. Patients with
Prior Fractures Have an Increased Risk of Future Fractures: A Summary of the Literature and
Statistical Synthesis. Journal of Bone and Mineral Research, 15(4):721–739, 2010.
[75] Statistics Denmark. Statistikbanken, 2019. URL https://statistikbanken.dk/statbank5a/
default.asp?w=1280. [Access date: 2020-05-28].
[76] Statistikbanken. DOD - Døde efter køn og alder, 2019. [Access date: 2020-03-25].
[77] J. A. Cauley, D. E. Thompson, K. C. Ensrud, J. C. Scott, and D. Black. Risk of mortality following
clinical fractures. Osteoporosis International, 11(7):556–561, 2000.
[78] The Danish Health Data Authority. DRG-charges, 2020. URL https://sundhedsdatastyrelsen.
dk/da/afregning-og-finansiering/takster-drg/takster-2020. [Access date: 2020-04-07].
[79] HMSA. HMSA - Provider Resource Center, 2018. URL https://hmsa.com/portal/provider/
zav_pel.fh.DIA.650.htm. [Access date: 2020-04-22].
[80] S. Kold, B. Christensen, J. B. Lauritzen, and H. C. Kjeldsen. Femur, hoftenært brud - Sund-
hed.dk, 2019. URL https://www.sundhed.dk/sundhedsfaglig/laegehaandbogen/ortopaedi/
tilstande-og-sygdomme/knoglebrud/femur-hoftenaert-brud/. [Access date: 2020-05-28].
[81] North Denmark Region. Total primary hip alloplasty, 2018. URL https://pri.rn.dk/Sider/
8929.aspx. [Access date: 2020-04-16].
[82] the Danish Medical Association. Overenskomst om almen praksis. Vasa, 3(april):1–9, 2020.
47
References Aalborg University
[83] Sundhed.dk. Lægehåndbogen, 2019. URL https://www.sundhed.dk/
sundhedsfaglig/laegehaandbogen/ortopaedi/tilstande-og-sygdomme/knoglebrud/
vertebralt-kompressionsbrud/. [Access date: 2020-03-30].
[84] North Jutland Orthopaedic Surgery. Stabilising back surgery. Technical report, North Denmark
Region, 2020.
[85] Osteoporoseforeningen. Knogleskørhed - en almindelig sygdom hos mænd, 2020. URL https:
//www.osteoporose-f.dk/stoette-og-hjaelp/osteoporose-og-maend/. [Access date: 2020-
03-17].
[86] The Danish Health Data Authority. LPR - Landspatientregisteret (National Patient Registry),
2018. URL www.esundhed.dk. [Access date: 2020-03-26].
[87] Lægehåndbogen. Sundhed.dk - håndledsbrud, 2018. URL https://www.sundhed.
dk/sundhedsfaglig/laegehaandbogen/ortopaedi/tilstande-og-sygdomme/knoglebrud/
haandledsbrud/. [Access date: 2020-04-28].
[88] S. E. Gabriel, S. E. Gabriel, A. N.A. Tosteson, C. L. Leibson, C. S. Crowson, G. R. Pond,
C. S. Hammond, and L. J. Melton. Direct medical costs attributable to osteoporotic fractures.
Osteoporosis International, 13(4):323–330, 2002.
[89] P. Vestergaard. Mail correspondance,, 2020.
[90] L. Si, T. M. Winzenberg, B. De Graaff, and A. J. Palmer. A systematic review and meta-analysis
of utility-based quality of life for osteoporosis-related conditions. Osteoporosis International, 25
(8):1987–1997, 2014.
[91] P. Lips and N. M. van Schoor. Quality of life in patients with osteoporosis. Osteoporosis
international : a journal established as result of cooperation between the European Foundation for
Osteoporosis and the National Osteoporosis Foundation of the USA, 16(5):447–455, 2005.
[92] M. C. Weinstein, G. Torrance, and A. McGuire. QALYs: The basics. Value in Health, 12(SUPPL.
1):S5–S9, 2009.
[93] L. Lundberg, M. Johannesson, D. G. L. Isacson, and L. Borgquist. Health-state utilities in a
general population in relation to age, gender and socioeconomic factors. European Journal of
Public Health, 9(3):211–217, 1999.
[94] A. D. M. Briggs, L. J. Cobiac, J. Wolstenholme, and P. Scarborough. PRIMEtime CE: A multistate
life table model for estimating the cost-effectiveness of interventions affecting diet and physical
activity. BMC Health Services Research, 19(1):1–19, 2019.
[95] P. W. Sullivan, J. F. Slejko, M. J. Sculpher, and V. Ghushchyan. Catalogue of EQ-5D scores for the
United Kingdom. Medical Decision Making, 31(6):800–804, 2011.
[96] A. Briggs. Probabilistic analysis of cost-effectiveness models: Statistical representation of
parameter uncertainty. Value in Health, 8(1):1–2, 2005.
[97] L. Zaninetti. A right and left truncated gamma distribution with application to the stars.
Advanced Studies in Theoretical Physics, 7(23):1139–1147, 2013.
[98] M. H. Murad, Matthew T. Drake, Rebecca J. Mullan, Karen F. Mauck, Louise M. Stuart,
Melanie A. Lane, Nisrin O. Abu Elnour, Patricia J. Erwin, Ahmad Hazem, Milo A. Puhan, and
Victor M. Montori. Comparative effectiveness of drug treatments to prevent fragility fractures:
A systematic review and network meta-analysis. Journal of Clinical Endocrinology and Metabolism,
97(6):1871–1880, 2012.
[99] M. Godwin, L. Ruhland, I. Casson, S. MacDonald, D. Delva, R. Birtwhistle, M. Lam, and
R. Seguin. Pragmatic controlled clinical trials in primary care: The struggle between external
and internal validity. BMC Medical Research Methodology, 3:1–7, 2003.
48
References Aalborg University
[100] D. M. J. Naimark, M. Bott, and M. Krahn. The half-cycle correction explained: Two alternative
pedagogical approaches. Medical Decision Making, 28(5):706–712, 2008.
[101] TreeAge Software Inc. TreeAge Pro 2020 User’s Manual. Technical report, TreeAge Pro, 2020.
[102] World Health Organization. Recommendations for Health Economics Evaluations of Interven-
tions in Osteoporosis. Technical report, WHO, 1999.
[103] P. Kothawala, E. Badamgarav, S. Ryu, R. M. Miller, and R. J. Halbert. Systematic Review
and Meta-analysis of Real-World Adherence to Drug Therapy for Osteoporosis. Mayo Clinic
Proceedings, 82(12):1493–1501, 2007.
[104] I. Imaz, P. Zegarra, J. González-Enríquez, B. Rubio, R. Alcazar, and J. M. Amate. Poor
bisphosphonate adherence for treatment of osteoporosis increases fracture risk: Systematic
review and meta-analysis. Osteoporosis International, 21(11):1943–1951, 2010.
[105] E. Durden, L. Pinto, L. Lopez-Gonzalez, P. Juneau, and R. Barron. Two-year persistence
and compliance with osteoporosis therapies among postmenopausal women in a commercially
insured population in the United States. Archives of Osteoporosis, 12(1):1–9, 2017.
[106] D. S. Cobden, L. W. Niessen, F. F. H. Rutten, and W. K. Redekop. Modeling the economic
impact of medication adherence in type 2 diabetes: A theoretical approach. Patient Preference and
Adherence, 4:283–290, 2010.
[107] M. Hadi, P. Swinburn, L. Nalysnyk, A. Hamed, and A. Mehta. A health state utility valuation
study to assess the impact of treatment mode of administration in Gaucher disease. Orphanet
journal of rare diseases, 13(1):159, 2018.
49
A. Appendix
A.1. Cost Calculations
This section provides the full calculations that are the basis of the final cost estimates
used in the model. The full descriptions with the associated sources can be seen in
section 3.3.2.3
A.1.1. Costs of Treatment
The costs of pharmacological treatment with teriparatide and alendronate were
calculated from the prices of one package of each drug from The Danish Medicines
Agency [47]. The price of one package of teriparatide comprising 28 doses was DKK
2,539. One dosage is injected every day. The annual cost of teriparatide treatment was
calculated as:
2,539/28 = DKK 90.67857 a day
90.6786*365 days = DKK 33,097.678 annually
For alendronate the price for one package containing 14 units was DKK 128.75.
A patient consumes one unit once a week. The annual cost of treatment with
alendronate was calculated as:
128.75/14 = DKK 9.19642 per unit
9.19642*(365 days/7) = DKK 479.527 annually
A.1.2. Costs of a Hip Fracture
It was found that all hip fractures require surgery, of which 75.4% of hip fractures
were treated by internal fixation and the remaining 25.6% were treated by primary
alloplasty. The 2020 DRG-charge of internal fixation and primary hip alloplasty of
67,991 and 51,979, respectively, were applied. The average cost of hip surgery was
calculated as:
(67,991*75.4)+(51,979*24.6)/100 = DKK 64,052.048 annually
Following a hip fracture, some costs are accumulated after discharge and are therefore
not included in the DRG-charge. These costs include treatment of the wound at a
50
A Appendix Aalborg University
GP and one follow-up visit at the ambulatory. The costs of treatment of the wound,
removal of staples at the GP, and follow-up at the ambulatory for orthopaedic surgery
of DKK 199.34, DKK 145.46, and DKK 1,512, respectively, were added to the average
cost of hip surgery, and the total mean cost of a hip fracture was calculated:
64,052.048+199.34+145.46+1,512= DKK 65,908.8
DKK 65,908.8 was the final mean cost estimate of a hip fracture.
A.1.3. Costs of a Vertebral Fracture
As every clinical vertebral fracture was assumed to require surgery, the 2020 DRG-
charge of back/neck surgery of DKK 92,307 was applied for each vertebral fracture.
The costs of one consultation (DKK 145.46), removal of staples (DKK 199.34), and
three follow-up visits including X-rays at the ambulatory of DKK 1,512 and DKK 747,
respectively, were added to the cost of back/neck surgery:
92,307+145.46+199.34+(1512*3)+(747*3) = DKK 99,428.8
The final mean cost of a vertebral fracture was estimated at DKK 99,428.8.
A.1.4. Costs of a Forearm Fracture
The numbers of forearm fractures in 2016 in both the general population and in people
suffering from osteoporosis can be seen in Table A.1.
Forearm fractures, 2016 Fractures in osteoporosis patients Fractures in the general population Total
Diagnosis code S522 112 1,266 1,378
Diagnosis code S525 4,442 36,865 41,307
Diagnosis code S526 459 5,512 5,971
Total: 5,013 43,643 48,656
Table A.1: This table gives an overview of the numbers of fractures of the forearm sustainedby people suffering from osteporosis and in the general population. All numbers are from year2016 and the information was drawn from a report by The Danish Health Data Authority[19].
To identify 50% of all the fractures, 48,656 was divided by two, since it was assumed
half of all the forearm fractures occur in women.
48,656/2 = 24,328
Meaning 24,328 out of the 48,656 forearm fractures occurred in women.
51
A Appendix Aalborg University
The number of forearm fractures occurring in women suffering from osteoporosis was
needed as well. It was assumed that 2/3 fractures occured in women and 1/3 was
sustained by men. The total number of forearm fractures sustained by people with
osteoporosis was 5,013 as can be seen from Table A.1. In order to identify 2/3 of the
5,013 fracture the following calculation was performed:
5013/100*(100/3*2) = 3,342
Meaning 3,342 out of the 5012 forearm fractures that occurred in the population
suffering from osteoporosis were sustained by women.
The percentage of the fractures occurring in women suffering from osteoporosis could
then be calculated from the number of fractures occurring in women in the general
population:
3,342/24328*100 = 13.7%
Data on men and women discharged from hospital after a forearm fracture and
patients visiting the emergency ward was obtained through the National Patient
Register [86] and can be seen in Table A.2.
Acute ambulant patients (DS52) 11,550
Discharged patients (DS52) 3,714
Table A.2: Total numbers of both discharged and acute ambulant patients within the diagnosiscode S52.
The 13.7% were applied to the total number of both acute ambulant patients within
the diagnosis code DS52 (fracture of the elbow and forearm) and the total number of
patients that were discharged within the same diagnosis code (DS52) which can be
seen in Table A.2.
Acute ambulant: 13.7% of 11,550 = 1582
Discharged patients: 13.7% of 3,714 = 509
Distribution: 1582/509 = 3.1099
Which means women were being treated 3.1099 times more often as acute ambulant
patients than they were admitted.
The cost of being treating with plaster or splint was estimated using interactive DRG-
charges from 2020 and was found to be DKK 1,952 [56]. And furthermore, the cost
52
A Appendix Aalborg University
of being treated surgically was obtained through The Danish Health Data Authority
[78]. The DRG-charge 08MP24 of forearm fracture surgery was found to be DKK
34,854 in 2020.
It was assumed that all patients that were discharged carried a cost of DKK 34,854
from being treated surgically, and it was furthermore assumed all acute ambulant
patients carried the cost of splint or plaster treatment of DKK 1,952 found through
interactive DRG. The interactive DRG charge was found to be DKK 1,952 since no
matter what fracture of the forearm, gender, and age was chosen within the system,
the cost was still DKK 1,952. When taking into account that women are being treated
3.1099 times more often in the ambulatory, the average estimated inpatient cost was:
34,854+(1,952*3.1099)/4.1099= DKK 9957.55
A follow-up visit at the ambulatory for orthopaedic surgery of DKK 1,512 was
included for each forearm fracture, and moreover, the DRG-charge of DKK 507 for an
uncomplicated x-ray was added for the patients who were treated surgically. When
taking into account that women were being treated 3.1099 times more often in the
ambulatory, the average cost of a follow-up visit was DKK 1,635.36
1,512+(507/4.1099) = 1,635.36
The final weighted mean cost of a forearm fracture was estimated to be:
9,957.55+1,635.36= DKK 11,592.91
A.1.5. Cost of an "Other" Fracture
It was assumed the cost of an "other" fracture could be estimated from taking 25% of
the costs related to a hip fracture:
DKK 65,908.8*0.25 = DKK 16,477.2
The final mean cost of an "other" fracture was therefore estimated at DKK 16,477.2.
53
A Appendix Aalborg University
A.2. The Markov Model - The Teriparatide Branch
Figure A.1: The figure depicts the Markov node and the associated branches for teriparatidein the base CUA. To the left of this image was a decision node called "choice of treatment forosteoporosis".
54
A Appendix Aalborg University
A.3. The Markov Model - The Alendronate Branch
Figure A.2: The figure depicts the Markov node and the associated branches for alendronatein the base case CUA. To the left of this image was a decision node called "choice of treatmentfor osteoporosis".
55
A Appendix Aalborg University
A.4. Full Tornado Diagram - CUA
Figure A.3: The figure shows an ICER tornado diagram using all parameters, except the fourutility values (well, forearm, "other" (subsequent years), and dead). The red parts of the barsrepresent the range of the ICER when the parameter in question is higher than the base casevalue, and the contrary is the case for the blue parts.
56
A Appendix Aalborg University
A.5. Summary Reports - Base Case
Stage % - Well % - Post-fracture (hip) % - Post-fracture (vertebral) % - Post-fracture (forearm) % - Post-fracture (others) % - Dead Cum Costs Cum Utility Cum Effects
0 1 0 0 0 0 0 749.3037673 0.754869815 0.01045602
1 0.987564084 0.000607509 0.00093832 0.003546566 0.00526654 0.002076981 1471.172207 1.479146775 0.020603276
2 0.984002086 0.001192422 0.001845257 0.003577589 0.005291101 0.004091545 2164.511429 2.173347801 0.030358062
3 0.979551164 0.001749472 0.00271351 0.003565476 0.005272914 0.007147463 2829.707854 2.8376377 0.039724247
4 0.974873116 0.002279372 0.003543801 0.003549433 0.005249164 0.010505114 3467.68053 3.473981119 0.048713824
5 0.970012332 0.002782982 0.004337086 0.003532503 0.00522412 0.014110977 4133.249921 4.083628338 0.059123447
6 0.963431729 0.003406806 0.005426599 0.005118945 0.005227242 0.017388678 4399.960398 4.667397149 0.069143165
7 0.957887796 0.00399946 0.006465807 0.005112026 0.005209226 0.021325685 4656.11804 5.225258211 0.078757362
8 0.951791605 0.004559924 0.00745305 0.005083105 0.00517961 0.025932706 4901.932537 5.758938038 0.087974947
9 0.94556223 0.005090469 0.008391692 0.00505085 0.00514672 0.030758039 5137.736117 6.268889008 0.096809404
10 0.938317759 0.00558741 0.009275407 0.005017819 0.005113056 0.036688549 5519.584301 6.741555304 0.109781812
11 0.924660527 0.007541036 0.010739931 0.008941532 0.005342784 0.04277419 5888.186898 7.192243631 0.122311518
12 0.913784087 0.009383199 0.012108669 0.008938792 0.005312819 0.050472434 6241.687744 7.621895971 0.134313286
13 0.903097656 0.011115316 0.013391259 0.008837332 0.005251817 0.05830662 6580.503643 8.030617147 0.14580272
14 0.891942573 0.012735581 0.014583687 0.008734095 0.005190445 0.066813619 6904.914835 8.419513723 0.156791188
15 0.880014892 0.014241288 0.015682378 0.008626451 0.005126436 0.076308555 7256.224794 8.789093537 0.168287281
16 0.865878514 0.016056757 0.016830029 0.007331897 0.007305887 0.086596916 7592.309718 9.139505147 0.179267383
17 0.852147136 0.017734837 0.017876925 0.007216168 0.007204519 0.097820415 7913.109762 9.471786239 0.189737569
18 0.837697405 0.019271926 0.01882031 0.007101955 0.007090498 0.110017907 8218.869096 9.786695352 0.199707353
19 0.823129284 0.02068209 0.01967411 0.00698186 0.006970521 0.122562134 8510.056713 10.08506464 0.209193396
20 0.808705266 0.021976388 0.02044799 0.006860579 0.006849405 0.135160372 8958.748893 10.3473769 0.221693717
21 0.786935559 0.022295218 0.021459838 0.008775442 0.008182735 0.152351208 9384.947559 10.59445881 0.233562834
22 0.769043529 0.022517421 0.022299091 0.008639696 0.008031835 0.169468429 9788.056567 10.82639942 0.244775941
23 0.7491848 0.022574147 0.022906271 0.008446108 0.007851124 0.189037551 10167.85962 11.04323816 0.255329545
24 0.728080488 0.022495398 0.023312178 0.00822952 0.007649462 0.210232954 10524.63639 11.24572782 0.265233766
25 0.705853655 0.022294207 0.023532343 0.007998794 0.007434759 0.232886241 10902.45187 11.433862 0.274969766
26 0.680198356 0.0236129 0.023224545 0.005430133 0.008958402 0.258575663 11254.96152 11.60833203 0.284042133
27 0.65410457 0.024528284 0.022825438 0.005201538 0.008618828 0.284721342 11583.25994 11.7695063 0.292488635
28 0.627676102 0.025085151 0.022333745 0.005001562 0.008287793 0.311615646 11888.14299 11.9181175 0.300330479
29 0.600460308 0.025309894 0.021740895 0.004799949 0.007953579 0.339735376 12170.21538 12.05449843 0.307583767
30 0.571844474 0.02521325 0.021031564 0.004592521 0.007609667 0.369708524 12457.16761 12.1671508 0.314885794
31 0.539742094 0.025596478 0.020139751 0.003926026 0.008531839 0.402063812 12720.27188 12.26931238 0.321578837
32 0.50759633 0.025546967 0.019164681 0.003712686 0.008077898 0.435901438 12959.7874 12.36093367 0.327671174
33 0.472953838 0.025027497 0.018049253 0.003492646 0.00759886 0.472877906 13175.68457 12.44243917 0.333162259
34 0.436524295 0.02410452 0.016821816 0.003255849 0.007083175 0.512210345 13368.33793 12.51418246 0.338061765
35 0.398745667 0.022835018 0.01550256 0.003006579 0.006540408 0.553369767 13621.66784 12.5766638 0.344477795
36 0.354610323 0.023311868 0.014466506 0.003113017 0.009643113 0.594855173 13846.49277 12.63001421 0.350170373
37 0.312957676 0.022844456 0.013217124 0.002830183 0.008764123 0.639386439 14040.60762 12.67524705 0.35508392
38 0.274198511 0.021776515 0.011929103 0.002503153 0.007749657 0.681843061 14206.48228 12.71269829 0.359281457
39 0.23463133 0.020005117 0.010486222 0.002194662 0.006794079 0.725888589 14344.83837 12.74337043 0.36278181
40 0.199142012 0.018005817 0.009109107 0.001881426 0.005823268 0.766038369 14502.94291 12.76784477 0.366780593
41 0.160829301 0.016694499 0.007983605 0.002216406 0.006843129 0.80543306 14632.38222 12.78688913 0.370053583
42 0.128492041 0.014801967 0.006772976 0.001853172 0.005701216 0.842378628 14734.5122 12.801303 0.372634412
43 0.100013405 0.01255798 0.005548264 0.001487681 0.004574387 0.875818283 14812.73832 12.81197327 0.374610092
44 0.076376929 0.010301848 0.004424858 0.001162017 0.003571737 0.904162611 14871.33895 12.81950143 0.376089399
45 0.055440025 0.007971788 0.003343659 0.000889722 0.002734042 0.929620764 14919.81162 12.82452956 0.377312493
46 0.037498594 0.005982084 0.002430434 0.000755507 0.00231425 0.951019131 14952.91599 12.82771612 0.378147581
47 0.024304926 0.004214319 0.001666808 0.000527056 0.001609473 0.967677419 14974.21053 12.8297908 0.378684494
48 0.01635966 0.003010723 0.001165861 0.000345438 0.001053673 0.978064645 14988.2119 12.83096586 0.379037317
49 0.009432306 0.00186213 0.000706799 0.00023151 0.000706447 0.987060808 14996.27155 12.83163071 0.379240389
50 0.005498627 0.00114094 0.000425996 0.000136879 0.000416677 0.99238088 15000.87007 12.83200578 0.379356212
51 0.003208569 0.000692716 0.000255126 7.99742E-05 0.000243389 0.995520227 15000.87007 12.83200578 0.379356212
Table A.3: The summary report related to the alendronate branch in the base case CUA.
57
A Appendix Aalborg University
Stage % - Well % - Post-fracture (hip) % - Post-fracture (vertebral) % - Post-fracture (forearm) % - Post-fracture (others) % - Dead Cum Costs Cum Utility Cum Effects
0 1 0 0 0 0 0 33264.54222 0.75520419 0.00561328
1 0.992406824 0.000567008 0.000562992 0.001038019 0.003375987 0.00204917 65185.08866 1.479858478 0.011030253
2 0.989343891 0.001110376 0.001105576 0.001040001 0.003379016 0.00402114 65781.50182 2.174480739 0.016235769
3 0.985320891 0.001627644 0.001625355 0.001036834 0.003368708 0.007020568 66353.50189 2.839238956 0.021232554
4 0.981061503 0.002119535 0.002122768 0.001032628 0.003355039 0.010308527 66901.91352 3.476101278 0.026027268
5 0.976613873 0.002586873 0.002598372 0.001028167 0.003340544 0.013832172 67457.67183 4.086389647 0.031324242
6 0.97174205 0.003165672 0.003251998 0.001490589 0.003344054 0.017005637 67990.75117 4.670866584 0.03641222
7 0.966771493 0.003714594 0.003875474 0.00148554 0.003330677 0.020822222 68501.65215 5.229502054 0.041292955
8 0.961212761 0.004233469 0.004468393 0.001477965 0.003313682 0.02529373 68649.42406 5.764024119 0.045971373
9 0.95550942 0.004724456 0.005032749 0.001469477 0.003294649 0.029969248 68791.09712 6.274884524 0.050454552
10 0.948777993 0.005184166 0.005564714 0.001460761 0.003275107 0.035737258 69032.97693 6.748637738 0.056950763
11 0.93890916 0.006991803 0.006447896 0.002604584 0.00342434 0.041622217 69265.44622 7.200503165 0.063192575
12 0.928954011 0.008690179 0.007272645 0.00258877 0.003398252 0.049096142 69488.15972 7.631419259 0.069168724
13 0.919059344 0.01028642 0.008047109 0.002561477 0.003362375 0.056683275 69701.44322 8.041487789 0.074888427
14 0.908667964 0.011779191 0.008768845 0.002534198 0.003326565 0.064923236 69905.50479 8.431814186 0.080357719
15 0.897469754 0.013166101 0.00943557 0.002505571 0.00328898 0.074134024 70133.61447 8.802923581 0.08652206
16 0.884087449 0.014838365 0.010133987 0.002131771 0.004692152 0.084116276 70351.72237 9.154941717 0.092412236
17 0.871088686 0.016383783 0.010772992 0.0021035 0.004632959 0.09501808 70559.77055 9.488903007 0.098027943
18 0.857330757 0.01779928 0.011350961 0.002072637 0.004564968 0.106881397 70757.9449 9.805560649 0.103374594
19 0.843413259 0.019097936 0.011876114 0.002039942 0.004492945 0.119079804 70946.57618 10.10574154 0.108461477
20 0.82960724 0.020290057 0.012354133 0.002006843 0.004420038 0.131321688 71238.59684 10.37007146 0.115298583
21 0.811195047 0.020592258 0.01299415 0.00256998 0.005286648 0.147361917 71515.45189 10.61944553 0.121777469
22 0.794644283 0.020800281 0.013528659 0.002524333 0.005188456 0.163313988 71777.37783 10.85392092 0.127903419
23 0.775930017 0.020857911 0.013924822 0.002472951 0.005082796 0.181731503 72024.28332 11.07350137 0.133675016
24 0.755827182 0.020792852 0.014200547 0.002414883 0.004963393 0.201801143 72256.37514 11.27891088 0.139097726
25 0.734462611 0.02061691 0.0143645 0.002352446 0.004835023 0.22336851 72520.99357 11.47009568 0.145043655
26 0.708468883 0.021846526 0.01421127 0.001600597 0.005839137 0.248033587 72768.46119 11.64772898 0.150602311
27 0.682891281 0.02271474 0.014002407 0.001543755 0.00563947 0.273208347 72999.23013 11.81214418 0.155785104
28 0.65688351 0.023255971 0.013735359 0.001488071 0.00543605 0.299201039 73213.83193 11.964046 0.160604151
29 0.629932719 0.023492953 0.0134044 0.001431464 0.005229237 0.326509228 73412.66772 12.10373176 0.165068613
30 0.601390449 0.023434194 0.012999649 0.001372827 0.005015 0.355787882 73620.07175 12.21936403 0.169731363
31 0.56929877 0.02382668 0.01248283 0.001176341 0.005636068 0.387579312 73810.52186 12.32445742 0.174012523
32 0.536814417 0.023817871 0.011910714 0.00111585 0.005348557 0.42099259 73984.19598 12.41892037 0.177916385
33 0.501537298 0.023372304 0.011248012 0.001052339 0.005044084 0.457745964 74141.02628 12.50314821 0.181441472
34 0.464178886 0.022549558 0.010511694 0.000983408 0.004713609 0.497062845 74281.23691 12.57746375 0.184592869
35 0.425189991 0.021400786 0.00971385 0.000910371 0.004363466 0.538421535 74470.28673 12.64238431 0.188875627
36 0.381399175 0.021904704 0.009100687 0.000944951 0.006449905 0.580200577 74637.83469 12.69797942 0.192670505
37 0.338139244 0.021503837 0.008343886 0.000858439 0.005860815 0.625293779 74782.70965 12.74525863 0.195951403
38 0.297510853 0.020537912 0.007557692 0.000761775 0.00520068 0.668431088 74906.75459 12.78453098 0.198760232
39 0.255742404 0.018907522 0.006668342 0.000670464 0.004577236 0.713434032 75010.43994 12.81680246 0.201107788
40 0.218009593 0.017056562 0.005814627 0.000576863 0.003938091 0.754604263 75129.19689 12.84266872 0.203795994
41 0.178535912 0.015865182 0.005123054 0.000682274 0.004646507 0.795147072 75226.19975 12.8628847 0.205991047
42 0.143748946 0.01409602 0.004366159 0.000568511 0.003869166 0.833351199 75302.87687 12.87825774 0.207725621
43 0.112731003 0.011987829 0.003594471 0.000458621 0.003121048 0.868107028 75361.76463 12.88969544 0.209057414
44 0.086731915 0.009861511 0.002882003 0.000360257 0.002451496 0.897712817 75406.02407 12.89781005 0.210058146
45 0.063507259 0.007655937 0.002190554 0.000277516 0.00188837 0.924480363 75442.75609 12.90326661 0.210888389
46 0.043665168 0.005768717 0.001603645 0.000236931 0.001607947 0.947117593 75467.86818 12.90674965 0.211455819
47 0.028664258 0.004079023 0.001107422 0.000165383 0.001121736 0.964862177 75484.08451 12.90903354 0.21182214
48 0.019460104 0.002925491 0.000780028 0.000109096 0.000739827 0.975985453 75494.81175 12.91033871 0.212064409
49 0.011414658 0.001819152 0.000477058 7.38921E-05 0.000501135 0.985714105 75501.01495 12.91108386 0.212204478
50 0.006735068 0.001120515 0.000289972 4.38792E-05 0.000297454 0.991513113 75504.57986 12.91150816 0.21228496
51 0.003974989 0.000684408 0.00017523 2.59047E-05 0.000175602 0.994963866 75504.57986 12.91150816 0.21228496
Table A.4: The summary report related to the teriparatide branch in the base case CUA.
58
A Appendix Aalborg University
A.6. Tables Used in TreeAge When Altering the
Durations of the Effects of the Two Drugs
E teriparatid forearm E teriparatid hip E teriparatid other E teriparatid vertebral
Index Value Index Value Index Value Index Value
0 0.24 0 0.42 0 0.5 0 0.3
1 0.24 1 0.42 1 0.5 1 0.3
2 0.24 2 0.42 2 0.5 2 0.3
3 0.24 3 0.42 3 0.5 3 0.3
4 0.24 4 0.42 4 0.5 4 0.3
5 0.24 5 0.42 5 0.5 5 0.3
6 0.24 6 0.42 6 0.5 6 0.3
7 0.24 7 0.42 7 0.5 7 0.3
8 0.335 8 0.4925 8 0.5625 8 0.3875
9 0.43 9 0.565 9 0.625 9 0.475
10 0.525 10 0.6375 10 0.6875 10 0.5625
11 0.62 11 0.71 11 0.75 11 0.65
12 0.715 12 0.7825 12 0.8125 12 0.7375
13 0.81 13 0.855 13 0.875 13 0.825
14 0.905 14 0.9275 14 0.9375 14 0.9125
15 1 15 1 15 1 15 1
E alendronate forearm E alendronate hip E alendronate other E alendronate vertebral
Index Value Index Value Index Value Index Value
0 0.82 0 0.45 0 0.78 0 0.5
1 0.82 1 0.45 1 0.78 1 0.5
2 0.82 2 0.45 2 0.78 2 0.5
3 0.82 3 0.45 3 0.78 3 0.5
4 0.82 4 0.45 4 0.78 4 0.5
5 0.82 5 0.45 5 0.78 5 0.5
6 0.85 6 0.541666667 6 0.816666667 6 0.583333333
7 0.88 7 0.633333333 7 0.853333333 7 0.666666667
8 0.91 8 0.725 8 0.89 8 0.75
9 0.94 9 0.816666667 9 0.926666667 9 0.833333333
10 0.97 10 0.908333333 10 0.963333333 10 0.916666667
11 1 11 1 11 1 11 1
Table A.5: The table shows the values used for creating a linear decline in the effects of thedrugs. E; efficacy. For teriparatide, the values for all indices following index 15 was 1, andlikewise, all values after index 11 were 1 for alendronate, however, they are not shown in thetable in order to keep the table small. In the model, the teriparatide branch applies 8 stages ofthe effects of treatment with teriparatide (2 stages of teriparatide and 6 stages of alendronateto maintain the effect of teriparatide), and the effect declines linearly following stage 8 for 8stages. Likewise, in the alendronate branch, individuals are treated for 6 years, and the effectdeclines towards 1 during the 6 following cycles.
59
A Appendix Aalborg University
A.7. Tables Used in TreeAge Containing the Baseline
Probabilities of Dying and Baseline Utility
Table A.6: The table lists the baseline proba-bilities of dying at each stage in the model forpeople aged 50 through 100 years.
Probabilities - mortality
Index Value
0 0.00197989644
1 0.00186354407
2 0.00285765775
3 0.00311750600
4 0.00332838202
5 0.00292780860
6 0.00354452813
7 0.00418245189
8 0.00436841301
9 0.00547520334
10 0.00548325087
11 0.00706478695
12 0.00713589639
13 0.00778746315
14 0.00880230413
15 0.00960611707
16 0.01061375173
17 0.01171046876
18 0.01215509116
19 0.01229319096
20 0.01329753632
21 0.01335524350
22 0.01649210798
23 0.01885678782
24 0.02121513944
25 0.02548490594
26 0.02685498309
27 0.02883288881
28 0.03176601617
29 0.03603319676
30 0.04147312541
31 0.04634591602
32 0.05496662740
33 0.06373448858
34 0.07323091618
35 0.08035075161
36 0.09644619941
37 0.10344057194
38 0.12342163902
39 0.13080771980
40 0.15078733895
41 0.17091114884
42 0.19195179686
43 0.20696068013
44 0.24339314845
45 0.28008474576
46 0.31474597274
47 0.29474485910
48 0.38247863248
49 0.38247863248
50 0.38247863248
Table A.7: The table gives the baseline utilityvalues used in the model. The utility decrementassociated with suffering from osteoporosis (-0.0418) has been subtracted.
Baseline - utility
Index Value
0 0.7572
1 0.7572
2 0.7572
3 0.7562
4 0.7562
5 0.7562
6 0.7562
7 0.7552
8 0.7552
9 0.7552
10 0.7332
11 0.7332
12 0.7332
13 0.7322
14 0.7322
15 0.7322
16 0.7312
17 0.7312
18 0.7312
19 0.7312
20 0.6822
21 0.6822
22 0.6822
23 0.6812
24 0.6812
25 0.6812
26 0.6812
27 0.6802
28 0.6802
29 0.6802
30 0.6162
31 0.6162
32 0.6152
33 0.6152
34 0.6152
35 0.6152
36 0.6142
37 0.6142
38 0.6142
39 0.6132
40 0.6132
41 0.6132
42 0.6132
43 0.6122
44 0.6122
45 0.6122
46 0.6122
47 0.6112
48 0.6112
49 0.6112
50 0.6102
60
A Appendix Aalborg University
A.8. Tables Used in TreeAge Containing Fracture
Incidence Probabilities for Hip, Vertebral, Forearm,
And Other Fractures
Table A.8: The baseline probabilities of suffer-ing a hip fracture from 50 years-old (cycle 0)to 100 years-old (cycle 50) when suffering fromosteoporosis.
Incidence - hip fracture
Index Value
0 0.001423501
1 0.001423501
2 0.001423501
3 0.001423501
4 0.001423501
5 0.001762342
6 0.001762342
7 0.001762342
8 0.001762342
9 0.001762342
10 0.005318124
11 0.005318124
12 0.005318124
13 0.005318124
14 0.005318124
15 0.006367213
16 0.006367213
17 0.006367213
18 0.006367213
19 0.006367213
20 0.010079854
21 0.010079854
22 0.010079854
23 0.010079854
24 0.010079854
25 0.01567266
26 0.01567266
27 0.01567266
28 0.01567266
29 0.01567266
30 0.018929063
31 0.018929063
32 0.018929063
33 0.018929063
34 0.018929063
35 0.030549369
36 0.030549369
37 0.030549369
38 0.030549369
39 0.030549369
40 0.042265958
41 0.042265958
42 0.042265958
43 0.042265958
44 0.042265958
45 0.049018851
46 0.049018851
47 0.049018851
48 0.049018851
49 0.049018851
50 0.049018851
Table A.9: The baseline probabilities of suffer-ing a vertebral fracture from 50 years-old (cycle0) to 100 years-old (cycle 50) when sufferingfrom osteoporosis.
Incidence - vertebral fracture
Index Value
0 0.002004678
1 0.002004678
2 0.002004678
3 0.002004678
4 0.002004678
5 0.002723298
6 0.002723298
7 0.002723298
8 0.002723298
9 0.002723298
10 0.004203342
11 0.004203342
12 0.004203342
13 0.004203342
14 0.004203342
15 0.004573196
16 0.004573196
17 0.004573196
18 0.004573196
19 0.004573196
20 0.009109677
21 0.009109677
22 0.009109677
23 0.009109677
24 0.009109677
25 0.008528487
26 0.008528487
27 0.008528487
28 0.008528487
29 0.008528487
30 0.008582517
31 0.008582517
32 0.008582517
33 0.008582517
34 0.008582517
35 0.011432501
36 0.011432501
37 0.011432501
38 0.011432501
39 0.011432501
40 0.015846761
41 0.015846761
42 0.015846761
43 0.015846761
44 0.015846761
45 0.01839958
46 0.01839958
47 0.01839958
48 0.01839958
49 0.01839958
50 0.01839958
61
A Appendix Aalborg University
Table A.10: The baseline probabilities ofsuffering a forearm fracture from 50 years-old (cycle 0) to 100 years-old (cycle 50) whensuffering from osteoporosis.
Incidence - forearm fracture
Index Value
0 0.00432508
1 0.00432508
2 0.00432508
3 0.00432508
4 0.00432508
5 0.006298846
6 0.006298846
7 0.006298846
8 0.006298846
9 0.006298846
10 0.011310583
11 0.011310583
12 0.011310583
13 0.011310583
14 0.011310583
15 0.0097431
16 0.0097431
17 0.0097431
18 0.0097431
19 0.0097431
20 0.012684706
21 0.012684706
22 0.012684706
23 0.012684706
24 0.012684706
25 0.008881206
26 0.008881206
27 0.008881206
28 0.008881206
29 0.008881206
30 0.007970466
31 0.007970466
32 0.007970466
33 0.007970466
34 0.007970466
35 0.009029496
36 0.009029496
37 0.009029496
38 0.009029496
39 0.009029496
40 0.012523289
41 0.012523289
42 0.012523289
43 0.012523289
44 0.012523289
45 0.014555897
46 0.014555897
47 0.014555897
48 0.014555897
49 0.014555897
50 0.014555897
Table A.11: The baseline probabilities ofsuffering an "other" fracture from 50 years-old (cycle 0) to 100 years-old (cycle 50) whensuffering from osteoporosis.
Incidence - other fracture
Index Value
0 0.006751974
1 0.006751974
2 0.006751974
3 0.006751974
4 0.006751974
5 0.006789839
6 0.006789839
7 0.006789839
8 0.006789839
9 0.006789839
10 0.007149515
11 0.007149515
12 0.007149515
13 0.007149515
14 0.007149515
15 0.010326834
16 0.010326834
17 0.010326834
18 0.010326834
19 0.010326834
20 0.012557075
21 0.012557075
22 0.012557075
23 0.012557075
24 0.012557075
25 0.015605279
26 0.015605279
27 0.015605279
28 0.015605279
29 0.015605279
30 0.018368561
31 0.018368561
32 0.018368561
33 0.018368561
34 0.018368561
35 0.029634209
36 0.029634209
37 0.029634209
38 0.029634209
39 0.029634209
40 0.041002382
41 0.041002382
42 0.041002382
43 0.041002382
44 0.041002382
45 0.047537088
46 0.047537088
47 0.047537088
48 0.047537088
49 0.047537088
50 0.047537088
62
A Appendix Aalborg University
A.9. Tables Used in TreeAge for Excess Mortality
Following Hip and Vertebral in First and Subsequent
YearsTable A.12: The probabilities of dying duringthe first year after having suffered a hip fracturefrom 50 years-old (cycle 0) to 100 years-old(cycle 50).
Mortality - hip, first year
Index Value
0 0.051619988
1 0.051619988
2 0.051619988
3 0.051619988
4 0.051619988
5 0.051619988
6 0.051619988
7 0.051619988
8 0.051619988
9 0.051619988
10 0.051619988
11 0.051619988
12 0.051619988
13 0.051619988
14 0.051619988
15 0.051619988
16 0.051619988
17 0.051619988
18 0.051619988
19 0.051619988
20 0.150408812
21 0.150408812
22 0.150408812
23 0.150408812
24 0.150408812
25 0.150408812
26 0.150408812
27 0.150408812
28 0.150408812
29 0.150408812
30 0.150408812
31 0.150408812
32 0.150408812
33 0.150408812
34 0.150408812
35 0.150408812
36 0.150408812
37 0.150408812
38 0.150408812
39 0.150408812
40 0.150408812
41 0.150408812
42 0.150408812
43 0.150408812
44 0.150408812
45 0.150408812
46 0.150408812
47 0.150408812
48 0.150408812
49 0.150408812
50 0.150408812
Table A.13: The probabilities of dying duringthe first year after having suffered a vertebralfracture from 50 years-old (cycle 0) to 100years-old (cycle 50).
Mortality - vertebral, first year
Index Value
0 0.063869136
1 0.063869136
2 0.063869136
3 0.063869136
4 0.063869136
5 0.063869136
6 0.063869136
7 0.063869136
8 0.063869136
9 0.063869136
10 0.063869136
11 0.063869136
12 0.063869136
13 0.063869136
14 0.063869136
15 0.063869136
16 0.063869136
17 0.063869136
18 0.063869136
19 0.063869136
20 0.150408812
21 0.150408812
22 0.150408812
23 0.150408812
24 0.150408812
25 0.150408812
26 0.150408812
27 0.150408812
28 0.150408812
29 0.150408812
30 0.150408812
31 0.150408812
32 0.150408812
33 0.150408812
34 0.150408812
35 0.150408812
36 0.150408812
37 0.150408812
38 0.150408812
39 0.150408812
40 0.150408812
41 0.150408812
42 0.150408812
43 0.150408812
44 0.150408812
45 0.150408812
46 0.150408812
47 0.150408812
48 0.150408812
49 0.150408812
50 0.150408812
63
A Appendix Aalborg University
Table A.14: The probabilities of dying after thefirst year after having suffered a hip fracturefrom 50 years-old (cycle 0) to 100 years-old(cycle 50).
Mortality - hip, subsequent years
Index Value
0 0.04208861
1 0.04208861
2 0.04208861
3 0.04208861
4 0.04208861
5 0.04208861
6 0.04208861
7 0.04208861
8 0.04208861
9 0.04208861
10 0.04208861
11 0.04208861
12 0.04208861
13 0.04208861
14 0.04208861
15 0.04208861
16 0.04208861
17 0.04208861
18 0.04208861
19 0.04208861
20 0.123659005
21 0.123659005
22 0.123659005
23 0.123659005
24 0.123659005
25 0.123659005
26 0.123659005
27 0.123659005
28 0.123659005
29 0.123659005
30 0.123659005
31 0.123659005
32 0.123659005
33 0.123659005
34 0.123659005
35 0.123659005
36 0.123659005
37 0.123659005
38 0.123659005
39 0.123659005
40 0.123659005
41 0.123659005
42 0.123659005
43 0.123659005
44 0.123659005
45 0.123659005
46 0.123659005
47 0.123659005
48 0.123659005
49 0.123659005
50 0.123659005
Table A.15: The probabilities of dying afterthe first year after having suffered a vertebralfracture from 50 years-old (cycle 0) to 100years-old (cycle 50).
Mortality - vertebral, subsequent years
Index Value
0 0.033428495
1 0.033428495
2 0.033428495
3 0.033428495
4 0.033428495
5 0.033428495
6 0.033428495
7 0.033428495
8 0.033428495
9 0.033428495
10 0.033428495
11 0.033428495
12 0.033428495
13 0.033428495
14 0.033428495
15 0.033428495
16 0.033428495
17 0.033428495
18 0.033428495
19 0.033428495
20 0.081487716
21 0.081487716
22 0.081487716
23 0.081487716
24 0.081487716
25 0.081487716
26 0.081487716
27 0.081487716
28 0.081487716
29 0.081487716
30 0.081487716
31 0.081487716
32 0.081487716
33 0.081487716
34 0.081487716
35 0.081487716
36 0.081487716
37 0.081487716
38 0.081487716
39 0.081487716
40 0.081487716
41 0.081487716
42 0.081487716
43 0.081487716
44 0.081487716
45 0.081487716
46 0.081487716
47 0.081487716
48 0.081487716
49 0.081487716
50 0.081487716
64