Endocrine (2018) 59:39–49DOI 10.1007/s12020-017-1440-0

META-ANALYSIS

Effects of growth hormone therapy on bone density and fracturerisk in age-related osteoporosis in the absence of growth hormonedeficiency: a systematic review and meta-analysis

Maya Barake 1,2● Asma Arabi1,3 ● Nancy Nakhoul1,4 ● Ghada El-Hajj Fuleihan1,3 ●

Sarah El Ghandour5 ● Anne Klibanski6 ● Nicholas A. Tritos6

Received: 6 August 2017 / Accepted: 25 September 2017 / Published online: 13 October 2017© Springer Science+Business Media, LLC 2017

AbstractPurpose In adults, growth hormone deficiency (GHD) hasbeen associated with low bone mineral density (BMD), aneffect counteracted by growth hormone (GH) replacement.Whether GH is beneficial in adults with age-related boneloss and without hypopituitarism is unclear.Methods We conducted a systematic literature search usingMedline, Embase and the Cochrane Register of ControlledTrials. We extracted and analyzed data according to thebone outcome included [bone mineral content (BMC),BMD, and bone biomarker, fracture risk]. We performed ameta-analysis when possible.Results We included eight studies. Seven randomized 272post-menopausal women, 61–69 years, to GH or control, for6–24 months, and the eighth was an extension trial. Exceptfor one study, all women received concurrent osteoporosistherapies. There was no significant effect of GH, as com-pared to control, on BMD at the lumbar spine (Weightedmean difference WMD=−0.01 [−0.04, 0.02]), total hip

(WMD= 0 [−0.05, 0.06]) or femoral neck (WMD= 0[−0.03, 0.04]). Similarly, no effect was seen on BMC. GHsignificantly increased the bone formation marker pro-collagen type-I carboxy-terminal propeptide (PICP) (WMD= 14.03 [2.68, 25.38]). GH resulted in a trend for increasein osteocalcin and in bone resorption markers. Patients whoreceived GH had a significant decrease in fracture risk ascompared to control (RR= 0.63 [0.46, 0.87]). Reportedadverse events were not major, mostly related to fluidretention.Conclusion GH may not improve bone density in womenwith age-related bone loss but may decrease fracture risk.Larger studies of longer duration are needed to furtherexplore these findings in both genders, and to investigatethe effect of GH on bone quality.

Keywords Bone mineral density ● Fracture ● Growthhormone ● Meta-analysis ● Osteoporosis

Introduction

Growth hormone (GH) is a peptide hormone that plays acentral role in bone metabolism, both directly and indir-ectly, through the local and systemic production of insulin-like growth factor-1 (IGF-1) [1, 2]. The importance of GHin bone physiology is illustrated in patients with growthhormone deficiency (GHD), who have low bone mineraldensity (BMD) and increased fracture risk [3–7]. We haveshown in a previous meta-analysis that GH replacementin adults with GHD for more than 12 months improvedBMD [8].

* Maya Barakemaya.barake@cmc.com.lb

1 Scholars in HeAlth Research Program, American University ofBeirut, Beirut, Lebanon

2 Clemenceau Medical Center, Beirut, Lebanon3 Calcium Metabolism and Osteoporosis Program, WHO

Collaborating Center for Metabolic Bone Diseases, AmericanUniversity of Beirut, Beirut, Lebanon

4 Department of Internal Medicine, American University of Beirut,Beirut, Lebanon

5 Bareen International Hospital, Abu Dhabi, United Arab Emirates6 Massacchusetts General Hospital and Harvard Medical School,

Boston, MA, USA

Whether GH therapy can increase BMD and decreasefracture risk in adults with age-related bone loss who do nothave organic GHD remains an unanswered question. Therationale for such therapy is that GH secretion by thepituitary, measured as IGF-1, normally declines with aging[9–11]. Serum IGF-1 levels were significantly lower inpostmenopausal women referred for osteoporosis screeningand in men with idiopathic age-related osteoporosis, than inhealthy controls. In addition, serum IGF-1 levels positivelycorrelated with BMD [12, 13], and decreased IGF-1 hasbeen associated with an increased risk of osteoporoticfractures [14, 15].

The role of GH treatment in this patient population hasbeen evaluated, albeit with conflicting results. In a rando-mized placebo-controlled trial in women with post-menopausal osteoporosis with the longest follow-up of 3years, GH therapy resulted in increased bone mineral con-tent (BMC) by 14% as compared to controls [16], a resultthat was not reproduced in another trial of similar design[17]. Available systematic reviews did not specificallyevaluate our population of interest. Yang et al. addressedthe effect of GH therapy on bone only in patients with hipfractures [18]. Another systematic review studied the effectof GH on multiple outcomes, including BMD, in healthyelderly, who were not specifically selected to have osteo-porosis at study entry [19]. A more recent meta-analysisevaluated the anabolic role of GH in elderly with bone loss,however included healthy individuals with normal BMDand individuals on hemodialysis who have secondary boneloss. Moreover, it included certain trials with GH therapyfor a short period (less than six months) [20].

In the present study, we investigated the possible ana-bolic role of GH as a treatment for osteoporosis. We con-ducted a systematic review and meta-analysis of availableprospective controlled studies examining the efficacy andsafety of GH therapy, as compared to placebo or conven-tional osteoporosis therapies, on bone densitometric end-points, bone turnover markers, and fracture risk in adultswith age-related osteopenia or osteoporosis.

Patients and methods

Literature search

We conducted a systematic search of the literature forcontrolled prospective and randomized studies up toNovember 2015. The online search included the databasesMEDLINE (1946 to present), EMBASE (1947 to present)and the Cochrane Register of Controlled Trials. No lan-guage or year of publication limitation was used. The searchwas performed using the keywords and MESH terms rele-vant to the growth hormone intervention (growth hormone,

human growth hormone, somatotropin, somatropin, soma-totrophin and somatrophin, GH, hGH, h-GH, rhGH, r-hGH,rh-GH, Genotropin, Humatrope, Norditropin, Nutropin,Omnitrope, Saizen, Serostim, Tev-tropin, and Zorbtive) andto the study population (osteoporosis, postmenopausalosteoporosis, osteopenia, bone density, bone loss, bonedemineralization, bone fractures, broken bone and low bonemineral density or content) and the Boolean functions AND,OR. We also performed a manual search of the references ofboth original and review articles collected.

In the present meta-analysis, we included all controlledprospective and randomized human studies examining theeffects of GH in postmenopausal women and men aboveage 50 years, who had osteoporosis or osteopenia. Weexcluded studies involving patients with organic GHD andthose with secondary causes of osteoporosis (such as end-stage renal disease or glucocorticoids), since the effect ofGH on bone in those patient populations has been eitherdefined or is likely confounded by other factors. We alsoexcluded studies where the length of treatment was less than6 months, as there would be no expected benefit within sucha short treatment interval, based on individual trials andmeta-analysis in patients with GHD [8].

The selection of outcome measures was done based onthe Core Outcome Measures in Effectiveness Trials(COMET) initiative, and included clinical health statusoutcomes, namely bone fractures; and intermediate out-comes, namely bone densitometry (BMD and BMC) andbone turnover biomarkers [21, 22].

Data abstraction and analysis

Two authors screened, independently and in duplicate,retrieved abstracts and articles for eligibility, based on ourpre-defined inclusion and exclusion criteria. Similarly, theyextracted data from retained articles using a pre-specifieddata collection form. They abstracted data with regards tostudy design, year of publication and number of subjectsincluded and number withdrawn. They also extracted dataon the study subjects’ demographics, their underlying bonedisease and on the GH intervention, as well as on the studyfunding source.

The primary outcome was predefined as the mean dif-ference in BMD measured using dual-energy X-rayabsorptiometry (DXA) after treatment with GH vs. com-parator. Secondary outcomes included mean difference inBMD measured using double photon absorptiometry(DPA), mean difference in BMC using DXA and singlephoton absorptiometry (SPA). It also included the risk ratioin fracture incidence, the mean difference between groups inindividual bone turnover biomarkers, the risk difference inincidence of adverse events and in all-cause mortality.When the outcome of interest was not available in the text,

40 Endocrine (2018) 59:39–49

the corresponding author was contacted by email. When themissing data could not be obtained (no reply or unavail-ability of records), the absolute values were retrieved fromgraphs [17, 23]. We also contacted by email the corre-sponding authors of three studies with inconsistent data [16,17, 24]. Two of the three contacted authors answered andprovided updated findings [16, 24]. We did not receive ananswer to our query regarding one trial’s outcome [17]. Thatstudy was therefore excluded with regards to that specificoutcome [17].

Data synthesis

For the purpose of analysis, studies were combinedaccording to the outcome measured and, in the case of bonedensitometric endpoints (BMD, BMC), according to thebone site studied (lumbar spine, femoral neck, total hip, andforearm) and to the densitometric technique used. When atleast two publications were available for an outcome, ameta-analysis was conducted for data synthesis. Twoincluded studies reported data on the same patient popula-tion at different time points, one being the non-randomizedextension of the other [16, 25]. Each was then included withregards to its most recent published outcome. Both studieswere not included together for the same outcome. Thepresence of duplicate publications is thus highly unlikely.

Two trials randomized participants to more than twointervention groups; in one trial [24], participants wererandomly assigned to four treatment groups. Each pair-wisecomparison was included separately in our meta-analysis(GH vs. placebo as one pair, and GH plus calcitonin vs.placebo plus calcitonin as another pair). In another trial[16], participants were randomized to one of two GHintervention groups or to a placebo control group. The studyoutcomes for the two GH groups were pooled together.

Statistical analysis

Outcomes were combined in meta-analysis using a random-effects model. This model yields a conservative estimate ofthe overall effect size. Choice of this model over a fixed-effect model was made based on the included studies thatpresumably show a certain level of heterogeneity and withinstudy variability [26].

Using a random-effects model, a weighted mean differ-ence (WMD) with a 95% confidence interval (CI) wasobtained for each of the primary and secondary continuousoutcomes, representing the summary effect. For thedichotomous outcome of fracture incidence, a risk ratio wasderived as the summary effect.

Evaluation for heterogeneity among included studies wasdone using the chi-squared test (Chi2) and I2 [26]. Values of30, 50, and 75% were considered as indicative of moderate,

substantial and considerable amount of heterogeneity,respectively [26]. We assessed the risk of bias of theincluded randomized studies using the Cochrane Colla-boration’s tool for assessing risk of bias (89). The risk ofpublication bias was not assessed with a funnel plot due tothe small number of trials (n= 3) addressing the primaryoutcome [26].

Subgroup analysis and meta-regression were pre-plannedbut were not performed due to the small number of studiesincluded.

The analysis was conducted using RevMan (version 5.3).Data are presented as mean± SD, mean and 95% CI ormean and range, as appropriate. Unless otherwise stated, Pvalues < 0.05 were considered statistically significant.

Results

Study selection

The initial literature search identified 7664 citations, 6025of which were retained after duplicate removal (Fig. 1).Through the initial title and abstract screen, 5967 articleswere excluded that were clearly referring to animal studies,different interventions or study populations and the full textarticles of the remaining 58 citations were retrieved. Onearticle was excluded due to unavailability and 49 wereexcluded after review, leaving 8 articles for inclusion, 1 ofwhich being the follow-up to an earlier study with newoutcomes. Among the remaining eight studies, three studiesincluded data on BMD measured by DXA, two studies hadBMC by DXA, three studies reported BMD or BMC bySPA, one study utilized DPA for bone densitometric end-points, five articles published values of bone biomarkersand three studies reported fractures (one being a 10 yearfollow-up report) (Fig. 1).

Study characteristics

Five out of eight included studies were randomized, pla-cebo-controlled, with four being double blind. Among theremaining three studies, two were randomized controlledand one was controlled. Studies reporting BMD by DXAwere all randomized, double blind, and placebo-controlled.

Baseline demographic and clinical characteristics ofstudy participants are detailed in Table 1. A total of272 subjects were included in 7 studies, 22 of whomwithdrew during follow-up. There were 210 subjects instudies reporting BMD and 142 subjects in studies reportingBMC. There were 233 participants in trials reporting dataon bone biomarkers and 239 were part of studies reportingon fracture outcomes.

Endocrine (2018) 59:39–49 41

The mean age of study participants ranged from 60.7 to71 years and their mean body mass index (BMI) from 24.4to 25.3 kg/m2 in individual studies. All included studiesinvolved women. Three studies reported on subjects’interval since menopause, which ranged from 5 to 17 years[17, 23, 27]. In three trials, baseline IGF-1 levels weredescribed as being normal for age [16, 17, 23].

One trial was conducted in women with osteopenia [24].The remaining six studies included participants withosteoporosis, 5 of which had at least one vertebral fracture[17, 23, 27–29] and one study having 56% of women withan osteoporotic fracture [16].

In addition to the GH or placebo intervention, six studiesmaintained participants on calcium, with daily doses ran-ging between 500 and 1200 mg [16, 17, 23, 24, 28, 29]. Onestudy reported the use of vitamin D at a dose of 400 IU daily[16]. With the exception of one trial [17], participants in allstudies received concurrent therapies for osteoporosis

(along with GH or placebo) including pamidronate, sexsteroid replacement therapy and salmon calcitonin. Therewas a wide range of GH treatment regimens initiated inincluded trials. The two most recent studies used dailydosing with GH (0.3–0.83 mg/d) [16, 17]. Two studies hadpatients on GH administered three times weekly [23, 28].The remaining three studies used cycles of 7 days or2 months of GH therapy [24, 27, 29]. Two studies usedweight-based dosing regimens [23, 24], while the rest usedfixed dosing. All included trials used recombinant humanGH, except for the two oldest studies that prescribed humanGH [28, 29]. Treatment duration ranged between 6 and24 months. One extension study examined fracture dataafter 10 years [25].

Risk of bias assessment

Results for risk of bias assessment, by domain and study,for the primary outcome BMD are shown in Fig. 2. Allstudies were considered at an unclear risk for reporting bias(no protocol registration available) and at low risk fordetection bias. Results for other types of bias were mixed,ranging namely between low and unclear risk of bias(Fig. 2).

Efficacy endpoints

Effect of GH therapy vs. comparator on bone densitometricendpoints

At the lumbar spine (LS), the total hip (TH), and thefemoral neck (FN), there was no significant difference inBMD between women treated with GH (n= 93) for aduration ranging between 12 and 24 months and women onplacebo (n= 71) (Fig. 3a, b, c). There was no evidence ofheterogeneity among subgroups for the three comparisons.

DPA was only used in the measurement of BMD in onetrial with no significant difference observed between treat-ment groups [27].

There was no difference in BMC measured by DXA atthe LS and the FN between women treated with GH andplacebo groups (Figs. 3d, e). Similarly, no difference wasobserved in BMC measured by SPA at the distal forearm(Fig. 3f).

Effect of GH therapy vs. comparator on bone biomarkers

The effect of GH therapy on bone biomarkers is summar-ized in Fig. 4.

We found a significant increase in the bone formationmarker procollagen type I carboxy-terminal propeptide(PICP) at the end of the study period (WMD= 14.03, 95%CI [2.68, 25.38], P= 0.02) in the GH group as compared to

Fig. 1 Flow diagram of study selection

42 Endocrine (2018) 59:39–49

Tab

le1

Characteristicsof

subjectsin

stud

iesinclud

edin

themeta-analysis

Participants

Osteoporosis/Osteopenia

Interventio

nOutcome

Firstauthor

(year)

Study

design

Age

(years)

Yearssince

menopause

BMI(kg/

m2)

No.

subjects/

No.

with

draw

n

Yrs

since

diagnosis

Prior

therapy

Concurrent

therapy

Num

ber/site

offractures

Disease

definitio

nBaseline

T-score

Approximate

meantarget

GH

dose

[IU/d

(mg/

d)],regimen

Control

Duration

of therapy

(m)

Outcome

assessed

Funding

Aloia

etal.

[28]

R,C

64.3±7.3

NA

25.2±4.5

25/0

NA

NA

Ca1000

mg/d,

CT100MRC

U4*/w

All≥1VFx(3.4±2.15)

VFx(Lossof

height>25%)

NA

hGH

6/d(2/d),

3*/w

CT

24BMC

FA(SPA

),Fractures

GH

(National

Pitu

itary

Agency),CT

(Arm

our

Pharm

aceuticals)

Aloia

etal.

[29]

R,C

62.4±6.7

NA

NA

14/0

NA

NA

Ca1000–1200

mg/d,

CT100

MRCU

cyclic

All≥1VFx(3.05±1.85)VFx(Lossof

height>25%)

NA

hGH7/d(2.3/d),

cyclic

(hGH

2m,CT3m,rest

3m

for3cycles)

CT

24BMC

FA(SPA

),Fractures

GH

(National

Pitu

itary

Agency),CT

(Arm

our

Pharm

aceuticals)

Erdtsieck

etal.[23]

R,

DB,

PC

63.1

(55–74)

17.2

(5–30)

25.3

(19–32.5)

23/2

NA

NA

Caelem

ental

500mg/d(if

dietaryCa<

1000

mg/d),

Pam

idronate

150mg/day

All≥1VFx

VFx(clin

ical

assessment)

NA

rhGH

0.0625/

kg,up

to4

(0.02/kg),3*/w

Placebo

6BMC

LS,FN

(Lunar),FA

(SPA

),Bone

markers

EliLilly(partly

)

Gonnelli

etal.[27]

R,PC

61.2±5.6

11.8±7.5

24.4

30/4

0none

CT50

IU/d

cyclic

All≥1VFx

VFx

NA

rhGH12/d

(4/d),

cyclic(rhG

H7d,

CT21d,rest61d

for8cycles)

Placebo

24BMD

(DPA),

Bonemarkers

_

Holloway

etal.[24]

R,

DB,

PC

69.2±6.6

NA

NA

84/12

0Estrogen41%

Ca500mg/d,

HRT41%,±

CT100U/d

0BMD

LST

score<−1

NA

rhGH

0.06/kg/d

(0.02mg/kg/d),

cyclic

(rhG

H7

d,CT5d,rest

44dCafor12

cycles)

Placebo

24BMD

LS,TF,

FN

(Hologic),

Bonemarkers

Medications

provided

bysuppliers

Saafet

al.

[17]

R,

DB,

PC

67.8±4.4

(60–74)

≥5

24.6±3.2

(20.2–33.2)16/4

NA

Ca/MV

Ca500mg/d

All≥1VFx,

2hipFx,

1shoulder

Fx,

7wristFx

BMD<1g/cm

2in

ultradistalradius

&presence

of1–4

VFx(Lossof

height>15%)

−2.8±

0.72

(TB)rhGH

1.5(±

0.3)/d

(0.5/d)

Range

0.9–2.7/d

(0.3–0.9/d)

Placebo

12BMD

LS,FN

(Lunar),Bone

markers

SwedishMedical

Research

Council+

foundatio

ns

Landin-

Wilh

elmsen

etal.[16]

R,

DB,

PC

60.7±5.7

NA

NA

80/0

1.9±2.3

HRT

Ca750mg/d,

D400U/d,HRT

56%

with

VFx

BMDLSTscore<

−2.5

−2.7(LS)rhGH

1/d(0.33/

d)or

2.5/d

(0.83/d)

Placebo

18BMD/BMC

LS,FN

(Lunar),Bone

markers,

Fractures

Pharm

acia

Upjohn&

Goteborg

University

Krantzetal.

[25]

C71

±6

NA

24.6±3.2

200/6

11.9±2.3

HRT(41%

stopped),

bisphosphonates

(23%

),teriparatid

e(3%)

Ca750mg/d,

D400U/d,HRT

56%

with

VFx

BMDLSTscore<

−2.5

−2.7(LS)NA

Control

PopulationNA

Fractures

_

Dataareshow

nas

mean±

SD

ormean(range)ifno

totherw

isespecified

Rrand

omized,D

Bdo

uble-blin

d,PCplacebo-controlled,

Ccontrolled,

NAno

tavailable,BMIbo

dymassindex,

Cacalcium,M

Vmultiv

itamin,H

RTho

rmon

ereplacem

enttherapy

,CTcalcito

nin,

MRCmedicalresearch

coun

cil,VFxvertebralfracture,B

MDbo

nemineraldensity

,LSlumbarspine,FNfemoralneck,T

Btotalbo

dy,G

Hgrow

thho

rmon

e,hG

Hhu

man

grow

thho

rmon

e,rhGH

recombinant

human

grow

thho

rmon

e,Dday,W

week,M

mon

th,B

MCbo

nemineralcontent,FAforearm,SPAsing

le-pho

tonabsorptio

metry,L

Slumbarspine,FNfemoralneck,D

PAdu

al-pho

ton

absorptio

metry,TFtotalfemur

Endocrine (2018) 59:39–49 43

placebo. A trend in favor of GH was observed among therest of the bone biomarkers, but did not attain statisticalsignificance.

Effect of GH therapy vs. comparator on fracture risk

A statistically significant 37% decrease in the risk of allfractures (combined) was observed with GH treatment ascompared to control at the latest follow-up period (RR=0.63, 95% CI [0.46, 0.87], P= 0.004), with no hetero-geneity among included studies (Fig. 5a). This significanteffect was not maintained when the trial by Krantz et al.with long-term follow-up, was removed in a sensitivityanalysis [25]. No significant difference in the occurrence ofvertebral fractures was observed (Fig. 5b).

Adverse events and mortality

Adverse events related to GH therapy included well-recognized side effects occurring as a consequence offluid retention, including peripheral edema, myalgias,arthralgias, and carpal tunnel syndrome. The frequency andseverity of events were not systematically reported. Whendescribed, symptoms were usually transient and mostlyrelieved by decreasing the GH dose. In five out of sevenstudies, changes in glucose metabolism were activelyexamined (as fasting blood sugar in five studies and asglycosylated hemoglobin, HbA1c, in one study) in the studypopulation with no significant changes observed.

No mortality was reported during the study period inincluded studies; however, most studies were of shortduration. In the 10-year follow-up on the Landin-Wilhelmsen trial, 8% of women on GH died, as comparedto 12% in the control population. No death could be directlyattributed to GH [25].

Discussion

In the present systematic review and meta-analysis, weincluded a limited number of studies addressing our out-comes of interest. Combining the results of three rando-mized trials examining the effect of GH on LS and FNBMD, we observed no significant difference with GHtherapy as compared to placebo. Four trials examined theeffect of GH on fracture risk, one of which had a 10-yearsfollow-up extension. A significant decrease in the risk offracture was only observed when the findings of the long-term study on fracture risk were included in the meta-analysis. A trend towards an increase in bone biomarkerswas observed when combining endpoints of four trialsaddressing this outcome. The trend reached statistical sig-nificance only for the marker PICP.

The absence of a significant change in BMC and BMDmay be the result of the short duration of GH therapy inincluded studies. Growth hormone replacement has beenfound to increase bone turnover and expand the boneremodeling space, a process that results in an initialdecrease or no change in BMC and BMD among patientswith organic GHD, followed by a subsequent time-relatedincrease in bone densitometric endpoints. This is illustratedin the observed non-significant increase in both bone for-mation and bone resorption biomarkers in this review and inthe published literature, reflecting a state of increased boneturnover [30]. In our previous meta-analysis on GHR inGHD, a significant increase in BMC and BMD was onlyobserved after a minimum of one year of therapy [8]. In thestudy by Saaf et al. no significant change in BMD wasobserved during the 12-month trial duration. It was only inpatients who were maintained on a second additional yearof GH therapy that a significant increase from baseline wasobserved in LS BMD [17]. Whether older patients treatedwith GH, such as our patient population, may require alonger duration of therapy in order to result in significantincreases in bone mass remains unanswered but may bepostulated. The reason being that remodeling efficiencypresumably decreases with aging, and as such, agents suchas GH, that work through expansion of new remodelingspace may require a longer treatment duration in order todemonstrate efficacy [24, 31].

The observed limited efficacy of GH therapy on bonedensity may also be secondary to the fact that eligible

Fig. 2 Risk of bias assessment across studies evaluating bone mineraldensity (BMD). Light grey shading and (+) sign represent a low riskof bias; White shading and (?) sign represent an unclear risk of bias;Dark grey shading and a (−) sign represent a high risk of bias

44 Endocrine (2018) 59:39–49

Fig. 3 Growth hormone (GH) effect vs. control on bone densitometric endpoints (BMD and BMC). Effect of GH vs. control on a BMD at thelumbar spine b BMD at the total hip. c BMD at the femoral neck. d BMC at the lumbar spine. e BMC at the femoral neck. f BMC at the forearm

Endocrine (2018) 59:39–49 45

studies included only women. A differential effect of GHreplacement in men and women has been described inGHD. Indeed, in our previous meta-analysis, in gender-related subgroup analysis, the increase in BMD was onlysignificant in men [8]. In an uncontrolled trial in 29 menwith idiopathic osteoporosis, intermittent or continuoustreatment with GH over a period of 24 months resulted inincreased BMC and BMD [32]. One possible explanationfor this dichotomous effect is the need for women on oralestrogen to receive higher treatment doses than men,because of an inhibitory effect of oral estrogen on GH-induced IGF-1 synthesis in the liver [33]. In the two largestincluded trials, GH was administered concomitantly withestrogen replacement therapy, either in the whole study

population or in 41% of participants [16, 24]. Both oral andtransdermal estrogen replacement were used in one of thesetwo studies. The need for a higher dose of GH in that casewas not specifically addressed and might have affectedobserved results.

The majority of included studies in this review (5/7)involved concomitant treatment with GH and an anti-resorptive agent. It has been postulated that the use of anti-resorptive agents, such as bisphosphonates, may counteractthe mechanism of action of GH through blunting of theexpected increase in bone turnover. In the included trial byErdtsieck et al. treatment with a combination of GH andpamidronate was less effective on BMC than therapy withthe bisphosphonate alone [23]. An analogy to this approach

Fig. 4 Growth hormone (GH) effect vs. control on bone turnover biomarkers. Effect of GH vs. control on a Osteocalcin. b PICP (procollagen typeI carboxy-terminal propeptide). c Total urinary pyridinolines. d CTX (C-telopeptide cross-link of type I collagen)

46 Endocrine (2018) 59:39–49

can be made regarding the combination of parathyroidhormone (PTH), another anabolic agent, with bispho-sphonates in osteoporosis. In a randomized trial in men withosteoporosis, BMD increased significantly more in patientstreated with PTH alone than in those treated with alen-dronate alone or with a combination of both agents, aneffect that was potentially attributable to attenuation ofPTH-induced increase in bone formation by concurrentbisphosphonate therapy [34].

A concerning characteristic of included studies is the useof different GH doses and treatment regimens, which mighthave diluted the observed results. In our prior meta-analysison GH replacement in GHD, no significant association wasobtained between the treatment dose and the change inBMD. However, a positive association was observedbetween on-therapy IGF-1 level, a reflection of GH action,and change in BMD [8]. In all included studies in thepresent meta-analysis, a significant increase from baselinein serum IGF-1 level was observed in parallel to GHadministration, with a quick return to baseline after treat-ment cessation. Whether these reversible changes in IGF-1levels might have mitigated bone densitometric endpoints intrials where GH was administered cyclically is plausible. Intwo trials, GH was given for 1 week every 49–82 days and,as such, the change in IGF-1 was limited to this shorttreatment period [24, 27].

Despite a non-significant change in bone densitometricendpoints, the current meta-analysis showed a significantdecrease in fracture risk with GH therapy, albeit largelydriven by a single study. This differential effect of GH on

bone-related endpoints might reflect a beneficial effect onbone that is not detected by measuring bone density. It isconceivable, but remains speculative, that GH coulddecrease fracture risk by improving aspects of bone quality.However, the impact of GH on bone quality is not clearlydefined and requires further investigation. In a prospectivetrial studying the link between osteoporotic fractures andserum IGF-1 in postmenopausal women, decreased serumIGF-1 levels predicted a higher risk of fractures indepen-dently of BMD, and, as such, systemic IGF-1 was suggestedto carry an important role in maintaining bone strength andbone quality. The mechanism of such an effect remains,however, unclear [14]. The differential effect of GH onfracture risk may also reflect differences between studiesevaluating this outcome, as compared to trials includingbone densitometric endpoints. An important distinguishingtrial is the 10-year follow-up study by Krantz et al. reportinglong-term fracture risk with GH [25], suggesting the needfor extended follow-up in order to observe clinically sig-nificant results in osteoporosis trials. This trial, however,has significant caveats; the actual randomized trial extendedfor only 18 months [16]. The actual Krantz study comparedthe initial GH treatment group to a control group of age-matched postmenopausal women. The two groups differedin their BMI, level of physical activity and their use ofosteoporosis therapies. In the GH group, at the end of fol-low-up, 59% of women were on sex steroid replacement,23% were started on bisphosphonates and 3% were onteriparatide. Meanwhile, in the control group, at the end ofthe study, 8% were on estrogen and the use of other

Fig. 5 Growth hormone (GH) effect vs. control on fracture risk. Effect of GH vs. control on a fracture risk, b vertebral fracture risk

Endocrine (2018) 59:39–49 47

osteoporosis therapies reached 4% [25]. The groups werethus not matched with regards to their disease status andtherapy, which might have affected and interacted with anypotential effect of GH treatment on bone.

Any potential use of GH as a therapeutic agent has totake into account its potential safety. The safety of GHreplacement has been recently appraised by internationalprofessional societies involved with such treatment [35]. Inthe consequent position paper, aggregate evidence did notsupport an association between GH and all-cause mortality,and data on cancer risk were reassuring. The reviewed lit-erature again recognized that the most common side effectsrelated to GH were musculoskeletal symptoms associatedwith fluid retention, potential for exacerbation of obstructivesleep apnea, and increased risk of glucose intolerance anddiabetes mellitus in individuals at higher risk [35]. Thesafety of GH treatment in the absence of organic deficiency,however, cannot be extrapolated from studies of GHreplacement. More data on the safety of GH therapy in thispopulation are clearly needed.

In conclusion, GH treatment, based on available evi-dence, has not been shown to be effective in improvingBMD in women with age-related bone loss. It may howeverdecrease fracture risk, without observed major adverseevents. Existing literature suffers from significant caveats.Randomized controlled trials of sufficient size and durationmay help elucidate the effects of a stable extended regimenof recombinant GH therapy in men and women with age-related osteoporosis on bone density, bone micro-architecture and long-term fracture risk.

Acknowledgements The authors would like to thank Dr. Elie Aklfor his contribution to the methodology of the meta-analysis, Ms AidaFarha for her contribution to the systematic search and all studyauthors who generously provided data from their individual studiesand answered study-related queries. Drs Barake, Arabi, El-HajjFuleihan, Tritos and Klibanski participated in study design; Drs Bar-ake, Nakhoul and El Ghandour participated in article screening; DrsBarake and Nakhoul participated in data extraction and analysis; DrsBarake and Tritos participated in manuscript writing; All co-authorshad the opportunity to revise, view and approve of the latest version ofthe manuscript. Research reported in this publication was supported bythe Fogarty International Center and Office of Dietary Supplements ofthe National Institutes of Health under Award Number D43TW009118. This funding supported the Scholars in Health ResearchProgram (SHARP), American University of Beirut, Beirut, Lebanon.The content is solely the responsibility of the authors and does notnecessarily represent the official views of the National Institutes ofHealth.

Compliance with ethical standards

Conflict of interest Nicholas A Tritos has received institution-directed research support from Ipsen, Novo Nordisk, Novartis andPfizer for work unrelated to this manuscript. Anne Klibanski hasreceived research funding from Ipsen and consultancy from Chiasma.The remaining authors declare that they have no competing interests.

References

1. A. Giustina, J.D. Veldhuis, Pathophysiology of the neuroregula-tion of growth hormone secretion in experimental animals and thehuman. Endocr. Rev. 19(6), 717–797 (1998). https://doi.org/10.1210/edrv.19.6.0353

2. C. Ohlsson, B.A. Bengtsson, O.G. Isaksson, T.T. Andreassen, M.C. Slootweg, Growth hormone and bone. Endocr. Rev. 19(1),55–79 (1998). https://doi.org/10.1210/edrv.19.1.0324

3. P.V. Carroll, E.R. Christ, B.A. Bengtsson, L. Carlsson, J.S.Christiansen, D. Clemmons, R. Hintz, K. Ho, Z. Laron, P. Sizo-nenko, P.H. Sonksen, T. Tanaka, M. Thorne, Growth hormonedeficiency in adulthood and the effects of growth hormonereplacement: a review. Growth Hormone Research Society Sci-entific Committee. J. Clin. Endocrinol. Metab. 83(2), 382–395(1998). https://doi.org/10.1210/jcem.83.2.4594

4. G. Mazziotti, A. Bianchi, S. Bonadonna, M. Nuzzo, V. Cimino,A. Fusco, L. De Marinis, A. Giustina, Increased prevalence ofradiological spinal deformities in adult patients with GH defi-ciency: influence of GH replacement therapy. J. Bone Miner. Res.21(4), 520–528 (2006). https://doi.org/10.1359/jbmr.060112

5. G. Mazziotti, M. Doga, S. Frara, F. Maffezzoni, T. Porcelli, L.Cerri, R. Maroldi, A. Giustina, Incidence of morphometric ver-tebral fractures in adult patients with growth hormone deficiency.Endocrine 52(1), 103–110 (2016). https://doi.org/10.1007/s12020-015-0738-z

6. T. Rosen, L. Wilhelmsen, K. Landin-Wilhelmsen, G. Lappas, B.A. Bengtsson, Increased fracture frequency in adult patients withhypopituitarism and GH deficiency. Eur. J. Endocrinol. 137(3),240–245 (1997)

7. C. Wuster, R. Abs, B.A. Bengtsson, H. Bennmarker, U. Feldt-Rasmussen, E. Hernberg-Stahl, J.P. Monson, B. Westberg, P.Wilton, K.S. Group, K.I.B.P. the; Upjohn International Metabolic,D., The influence of growth hormone deficiency, growth hormonereplacement therapy, and other aspects of hypopituitarism onfracture rate and bone mineral density. J. Bone Miner. Res. 16(2),398–405 (2001). https://doi.org/10.1359/jbmr.2001.16.2.398

8. M. Barake, A. Klibanski, N.A. Tritos, Effects of recombinant humangrowth hormone therapy on bone mineral density in adults withgrowth hormone deficiency: a meta-analysis. J. Clin. Endocrinol.Metab. 99(3), 852–860 (2014). https://doi.org/10.1210/jc.2013-3921

9. D.R. Clemmons, J.J. Van Wyk, Factors controlling blood con-centration of somatomedin C. Clin. Endocrinol. Metab. 13(1),113–143 (1984)

10. A. Iranmanesh, G. Lizarralde, J.D. Veldhuis, Age and relativeadiposity are specific negative determinants of the frequency andamplitude of growth hormone (GH) secretory bursts and the half-life of endogenous GH in healthy men. J. Clin. Endocrinol. Metab.73(5), 1081–1088 (1991). https://doi.org/10.1210/jcem-73-5-1081

11. Z. Zadik, S.A. Chalew, R.J. McCarter Jr., M. Meistas, A.A.Kowarski, The influence of age on the 24-hour integrated con-centration of growth hormone in normal individuals. J ClinEndocrinol Metab 60(3), 513–516 (1985). https://doi.org/10.1210/jcem-60-3-513

12. S. Ljunghall, A.G. Johansson, P. Burman, O. Kampe, E. Lindh, F.A. Karlsson, Low plasma levels of insulin-like growth factor 1(IGF-1) in male patients with idiopathic osteoporosis. J. Intern.Med. 232(1), 59–64 (1992)

13. M. Munoz-Torres, P. Mezquita-Raya, F. Lopez-Rodriguez, E.Torres-Vela, J. de Dios Luna, F. Escobar-Jimenez, The con-tribution of IGF-I to skeletal integrity in postmenopausal women.Clin. Endocrinol. 55(6), 759–766 (2001)

14. P. Garnero, E. Sornay-Rendu, P.D. Delmas, Low serum IGF-1and occurrence of osteoporotic fractures in postmenopausalwomen. Lancet 355(9207), 898–899 (2000)

48 Endocrine (2018) 59:39–49

15. C. Ohlsson, D. Mellstrom, D. Carlzon, E. Orwoll, O. Ljunggren,M.K. Karlsson, L. Vandenput, Older men with low serum IGF-1have an increased risk of incident fractures: the MrOS Swedenstudy. J. Bone Miner. Res. 26(4), 865–872 (2011). https://doi.org/10.1002/jbmr.281

16. K. Landin-Wilhelmsen, A. Nilsson, I. Bosaeus, B.A. Bengtsson,Growth hormone increases bone mineral content in post-menopausal osteoporosis: a randomized placebo-controlled trial. J.Bone Miner. Res. 18(3), 393–405 (2003). https://doi.org/10.1359/jbmr.2003.18.3.393

17. M. Saaf, A. Hilding, M. Thoren, S. Troell, K. Hall, Growth hormonetreatment of osteoporotic postmenopausal women - a one-year pla-cebo-controlled study. Eur. J. Endocrinol. 140(5), 390–399 (1999)

18. S. Yang, L. Cao, S. Cai, J. Yuan, J. Wang, A systematic review ofgrowth hormone for hip fractures. Growth Horm. IGF Res. 22(3-4), 97–101 (2012). https://doi.org/10.1016/j.ghir.2012.03.002

19. H. Liu, D.M. Bravata, I. Olkin, S. Nayak, B. Roberts, A.M.Garber, A.R. Hoffman, Systematic review: the safety and efficacyof growth hormone in the healthy elderly. Ann. Intern. Med. 146(2), 104–115 (2007)

20. H.F. Atkinson, R.F. Moyer, D. Yacoub, D. Coughlin, T.B. Bir-mingham, Effects of recombinant human growth hormone forosteoporosis: systematic review and meta-analysis. Can. J. Aging36(1), 41–54 (2017). https://doi.org/10.1017/S0714980816000696

21. A. Cranney, P. Tugwell, S. Cummings, P. Sambrook, J. Adachi,A.J. Silman, W.J. Gillespie, D.T. Felson, B. Shea, G. Wells,Osteoporosis clinical trials endpoints: candidate variables andclinimetric properties. J. Rheumatol. 24(6), 1222–1229 (1997)

22. COMET: initiative. www.comet-initiative.org. Accessed in Jan-uary 2016

23. R.J. Erdtsieck, H.A. Pols, N.K. Valk, B.M. van Ouwerkerk, S.W.Lamberts, P. Mulder, J.C. Birkenhager, Treatment of post-menopausal osteoporosis with a combination of growth hormoneand pamidronate: a placebo controlled trial. Clin. Endocrinol. 43(5), 557–565 (1995)

24. L. Holloway, L. Kohlmeier, K. Kent, R. Marcus, Skeletal effectsof cyclic recombinant human growth hormone and salmon calci-tonin in osteopenic postmenopausal women. J. Clin. Endocrinol.Metab. 82(4), 1111–1117 (1997). https://doi.org/10.1210/jcem.82.4.3901

25. E. Krantz, P. Trimpou, K. Landin-Wilhelmsen, Effect of growthhormone treatment on fractures and quality of life in post-menopausal osteoporosis: a 10-year follow-up study. J. Clin.Endocrinol. Metab. 100(9), 3251–3259 (2015). https://doi.org/10.1210/jc.2015-1757

26. J.P.T. Higgins, S. Green, Cochrane handbook for systematicreviews of interventions. version 5.1.0. The Cochrane Colla-boration (2011)

27. S. Gonnelli, C. Cepollaro, M. Montomoli, L. Gennari, A. Mon-tagnani, R. Palmieri, C. Gennari, Treatment of post-menopausalosteoporosis with recombinant human growth hormone and sal-mon calcitonin: a placebo controlled study. Clin. Endocrinol. 46(1), 55–61 (1997)

28. J.F. Aloia, A. Vaswani, A. Kapoor, J.K. Yeh, S.H. Cohn, Treat-ment of osteoporosis with calcitonin, with and without growthhormone. Metabolism 34(2), 124–129 (1985)

29. J.F. Aloia, A. Vaswani, P.J. Meunier, C.M. Edouard, M.E. Arlot,J.K. Yeh, S.H. Cohn, Coherence treatment of postmenopausalosteoporosis with growth hormone and calcitonin. Calcif. TissueInt. 40(5), 253–259 (1987)

30. T.B. Hansen, K. Brixen, N. Vahl, J.O. Jorgensen, J.S. Chris-tiansen, L. Mosekilde, C. Hagen, Effects of 12 months of growthhormone (GH) treatment on calciotropic hormones, calciumhomeostasis, and bone metabolism in adults with acquired GHdeficiency: a double blind, randomized, placebo-controlled study.J. Clin. Endocrinol. Metab. 81(9), 3352–3359 (1996). https://doi.org/10.1210/jcem.81.9.8784096

31. R. Marcus, Normal and abnormal bone remodeling in man. Annu.Rev. Med. 38, 129–141 (1987). https://doi.org/10.1146/annurev.me.38.020187.001021

32. P. Gillberg, H. Mallmin, M. Petren-Mallmin, S. Ljunghall, A.G.Nilsson, Two years of treatment with recombinant human growthhormone increases bone mineral density in men with idiopathicosteoporosis. J. Clin. Endocrinol. Metab. 87(11), 4900–4906(2002). https://doi.org/10.1210/jc.2002-020231

33. M.E. Molitch, D.R. Clemmons, S. Malozowski, G.R. Merriam, M.L. Vance, S. Endocrine, Evaluation and treatment of adult growthhormone deficiency: an Endocrine Society clinical practiceguideline. J. Clin. Endocrinol. Metab. 96(6), 1587–1609 (2011).https://doi.org/10.1210/jc.2011-0179

34. J.S. Finkelstein, A. Hayes, J.L. Hunzelman, J.J. Wyland, H. Lee,R.M. Neer, The effects of parathyroid hormone, alendronate, orboth in men with osteoporosis. N. Engl. J. Med. 349(13),1216–1226 (2003). https://doi.org/10.1056/NEJMoa035725

35. D.B. Allen, P. Backeljauw, M. Bidlingmaier, B.M. Biller, M.Boguszewski, P. Burman, G. Butler, K. Chihara, J. Christiansen,S. Cianfarani, P. Clayton, D. Clemmons, P. Cohen, F. Dar-endeliler, C. Deal, D. Dunger, E.M. Erfurth, J.S. Fuqua, A.Grimberg, M. Haymond, C. Higham, K. Ho, A.R. Hoffman, A.Hokken-Koelega, G. Johannsson, A. Juul, J. Kopchick, P. Lee, M.Pollak, S. Radovick, L. Robison, R. Rosenfeld, R.J. Ross, L.Savendahl, P. Saenger, H. Toft Sorensen, K. Stochholm, C.Strasburger, A. Swerdlow, M. Thorner, GH safety workshopposition paper: a critical appraisal of recombinant human GHtherapy in children and adults. Eur. J. Endocrinol. 174(2), P1–9(2016). https://doi.org/10.1530/EJE-15-0873

Endocrine (2018) 59:39–49 49