Nephrol Dial Transplant (2011) 26: 1729–1739doi: 10.1093/ndt/gfq627Advance Access publication 20 October 2010

Screening for renal cancer in recipients of kidney transplants

Germaine Wong1,2,3, Kirsten Howard2, Angela C. Webster1,2,3, Jeremy R. Chapman3

and Jonathan C. Craig1,2

1Centre for Kidney Research, The Children’s Hospital at Westmead, NSW, Australia, 2Sydney School of Public Health, University ofSydney, Sydney, Australia and 3Centre for Transplant and Renal Research, Westmead Hospital, NSW, Australia

Correspondence and offprint requests to: Germaine Wong; E-mail: germainw@chw.edu.au

AbstractBackground. Renal cancer is the most common solidorgan cancer in the kidney transplant population with anexcess risk ~5-fold greater than the general population.It is uncertain whether routine screening for renal canceris cost-effective. The aim of our study is to estimate thecosts and health benefits of ultrasonographic (US) screen-ing for renal cancer in the kidney transplant population.Methods. A Markov model was developed to compare thecosts and benefits in a cohort of kidney transplant recipi-ents (n = 1000, aged 18–69 years), who underwent annualand biennial US screening for renal cancer, compared witha cohort that did not.Results. For recipients of kidney transplants aged 18–69 years, the incremental cost-effectiveness ratio (ICER)for routine US screening ranged from $252 100/LYS forbiennial screening to $320 988/LYS for annual screening.A total of two and one cancer deaths were averted in theannually and biennially screened population, with a rela-tive cancer-specif ic mortality reduction by 25% and12.5%, respectively. Using a series of sensitivity analyses,the ICER was most sensitive to the costs and test specifi-city of ultrasonography, prevalence of disease, and the riskof graft failure in the screened population.Conclusions. Routine screening for renal cancer may reducethe risk of cancer-related deaths in recipients of kidney trans-plants. Uncertainties, however, exist in the model’s influentialvariables including the risk of graft failure among thosewho re-ceived contrast-enhanced diagnostic computer tomography.Given the available evidence, routine screening for renal cancersmay not be cost-effective for recipients of kidney transplants.

Keywords: cost-effectiveness; kidney transplantation; renal cancer;screening

Introduction

Compared with the general population, cancer risk is in-creased by at least 2-fold in kidney transplant recipients

and is their second most common cause of death [1,2].Renal cancer is the most common solid organ cancer inpeople with end-stage kidney disease (ESKD) with an ex-cess risk of at least 5–10-fold greater than the age- andgender-matched general population, and a significantcause of mortality and morbidity [3,4]. At least one in fourdies in the first year after cancer diagnosis, with a 5-yearcancer survival rate of <20% [5].

Early detection and treatment of renal cancer throughroutine screening of kidney transplant recipients are aplausible means to prevent the development of advanced-stage cancer which has an unfavourable prognosis. Generalpopulation screening for cancers such as breast, colorectaland cervical cancer is now standard practice and is effect-ive in reducing cancer-specific mortality [6–10]; however,screening for renal cancer is not recommended in the gen-eral population. In the ESKD population, the benefits ofroutine screening for renal cancer are even less certain.This is predominantly because their kidneys are structural-ly different (as a result of congenital abnormality and/orintrinsic renal disease), thus adding to the complexity oftest result interpretation. Although there is a body of litera-ture assessing the incidence/prevalence of renal cancer inrecipients of kidney transplants [11–16], few studies haveestimated the test performance characteristics (sensitivityand specificity) of ultrasonographic screening and thetreatment effectiveness for renal cancer in the ESKD andtransplant populations [17–20].

The current recommendations for screening for renal can-cer in the kidney transplant population are contradictory. TheAmerican Society of Transplantation and the Kidney DiseaseImproving Global Outcomes (KDIGO) Clinical PracticeGuidelines for the Care of the Kidney Transplant Recipientfound no evidence to support routine screening for renal can-cers in recipients of kidney transplants and did not recom-mend screening using ultrasonography in average-risktransplant recipients [21,22]; guidelines from the EuropeanAssociation of Urology recommended annual screening ofthe native and the graft kidneys using ultrasonography forpeople who are at high risk but did not specify or definethe high-risk populations [23]. The most reliable method toascertain the benefits and harms of screening would ideally

© The Author 2010. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved.For Permissions, please e-mail: journals.permissions@oxfordjournals.org

come from randomized controlled trials (RCT) of screeningcompared with no screening. However, given the paucity oftrial-based data and the low feasibility of adequately poweredtrials, estimates of costs and effects of screening derived fromanalytical modelling are the most effective method to informdecisionmakers about the possible benefits, harms, costs anduncertainties of screening for renal cancer in this setting. Theaims of this study are firstly to estimate the health benefitsand costs of routine annual and biennial screening usingultrasonography of the native kidneys in recipients of kidneytransplants, and secondly to determine the influential and un-certain variables in screening for renal cancer to help informfuture research priorities.

Materials and methods

Using a broad healthcare funder perspective, a Markov model was devel-oped to simulate the lifetime costs and health outcomes of a hypotheticalcohort of kidney transplant recipients.

Structure of the model

The structure of the model is outlined in Figure 1, and details of the struc-ture of the model are provided in Appendix 1. In short, we first structuredthe model to include all the potential consequences of the natural historyof renal cancer and transplantation (Table 1). We then populated themodel using the best available evidence and finally assessed the uncer-tainties around model assumptions and parameter estimates using one-way and multiway sensitivity analyses. In the absence of randomized con-trolled trial evidence of screening for renal cancer, data of health benefits,harms and costs were extrapolated from published observational studies inthe transplant and dialysis population, and where not available, using pub-lished data from the general population (Table 2).

Sensitivity analyses

Uncertainties existed in themodel’s parameter estimates and structure. Usinga series of one- and two-way sensitivity analyses, we tested the robustness of

the results to the uncertainties surrounding the model’s estimates. The base-line participation rate of screening renal cancers was assumed to be 70%.Given that transplant recipients were subjected to intense monitoring bytheir treating physicians, a higher than the expected participation rate inscreening trials [24,25] was assumed in recipients of kidney transplants.The expected participation rate was then varied between low and high va-lues of 40–100%. Other variables such as the prevalence of disease (in thehigh-risk population, such as those with prior history of renal cancer, ac-quired cystic disease and analgesic nephropathy), the probability of graftfailure, the estimates of screening test accuracy (test specificity and sensi-tivity), discount rates for costs and benefits, the cancer stage-specific dis-tribution, and the probability of survival with renal cancer were alsoextensively assessed in the sensitivity analysis.

Scenario analyses

We modelled two different scenarios for screening renal cancer in the trans-plant population. The first scenario (base-case analysis) assumed recipientswith failed kidney transplant who returned to dialysis and were continued tobe screened either annually or biennially. The second scenario assumed alltransplant recipients with failed kidney transplants who returned to dialysisand did not continue routine screening for renal cancer.

Model outcomes

The model’s outcomes include the average and incremental health out-comes (in life-years saved) and costs of screening for renal cancers inrenal transplant recipients, the number of renal cancers averted fromearly detection, and the absolute and relative cancer-specific mortalityreduction from screening for renal cancers. The incremental costs andbenefits were calculated according to this formula:

ICER =CostNew − CostComparator

EffectivenessNew − EffectivenessComparator

Future costs and outcomes were discounted at an annual rate of 5%.The discount rate was also tested in the sensitivity analyses within theaccepted rates by most funding agencies [26–28]. In this model, weused a half-cycle correction to minimize overestimation of the expectedoutcomes at the end of the analysis. TreeAge Pro 2008 and Excel wereused to develop the decision-analytic model [29].

Fig. 1. Simplified structure of the model (showing the annual screening arm).

1730 G. Wong et al.

Results

Base-case analysis

Tables 3 and 4 show the results of the base-case analysisof the no screening, annual screening or biennial screen-

ing arms. Assuming a screening participation rate of70%, the incremental benefit of annual screening com-pared with no screening was 0.004 life-years saved(LYS), and the incremental benefit of biennial screeningcompared with no screening was 0.003 LYS. The incre-mental cost of annual screening compared with no

Table 1. Clinical inputs into the model

Input parameters

Base-case values(ranges used insensitivity analyses)

References

Transplant anddialysispopulation

Generalpopulation

Annual age-specific all-cause mortality rate ofpeople on dialysis

Ages [31]25–44 0.064845–64 0.109665–74 0.1494>75 0.2093

Annual age-specific all-cause mortality rate ofrecipients of kidney transplants

Ages [32]25–44 0.00945–64 0.02765–74 0.062≥75 0.10

Annual age-specific prevalence of renal cancersin recipients of kidney transplants

Ages [31,32]<45 0.000645–54 0.000655–74 0.0005≥75 0.0018

Annual age-specific prevalence of renal cancerin people on dialysis

Ages [31,32]25 0.0006745 0.0017655 0.00545

Annual stage-specific mortality rates of renal cancerin recipients of kidney transplant (screened arm)

Stages [37–39]I 0.04II–III 0.2IV 0.25

Annual stage-specific mortality rates of renal cancerin recipients of kidney transplants (unscreened arm)

Stages [31,32]I 0.04II–III 0.25IV 0.33

Annual stage-specific mortality rates of renal cancer inpeople on dialysis (screened arm)

Stages [18,40,41]I 0.14II–III 0.19IV 0.47

Annual stage-specific mortality rate of renal cancer inpeople on dialysis (unscreened arm)

Stages [31,32]I 0.20II–III 0.24IV 0.50

Stage-specific distribution of renal cancer in people ondialysis (screened arm)

Stages [18]I 0.612II 0.063III 0.042IV 0.274

Stage-specific distribution of renal cancer (unscreened arm) Stages [31,32]I 0.230II 0.089III 0.141IV 0.540

Sensitivity of ultrasonography 0.60 (0.50–0.95) [33,34,37,42,43]Specificity of ultrasonography 0.90 (0.70–0.99) [33,34,37,42,43]Participation rate of screening 0.70 (0.40–1.0)Annual probability of graftfailure and return to dialysis

0.030 (0.022–0.040) [31,32]

Annual discount rate for health benefits 0.05 (0.03–0.06) [26]Annual discount date for costs 0.05 (0.03–0.06) [26]Annual probability of disease recurrence Stages [19]

I 0.04II 0.10III 0.48

Screening for renal cancer in recipients of kidney transplants 1731

screening was $1300. The incremental cost of biennialscreening compared with no screening was $900. Theincremental costs-effectiveness ratio (ICER) of annualscreening compared with no screening was $320 988/LYS, and the ICER of biennial screening compared withno screening was $252 100/LYS. Over the entire screen-ing period, there were six deaths from renal cancer per1000 transplant recipients in the annual screening arm,compared with seven and eight deaths from renal cancerper 1000 transplant recipients in the biennial screeningand no screening arms, respectively. Compared with noscreening, the relative risk reduction of death from renalcancer for annual screening was 25%, and 12.5% forbiennial screening, with an absolute risk reduction ofdeath from renal cancer of only 0.2% for annual screen-ing and 0.1% for biennial screening.

One-way sensitivity analysis

Annual screening. The model was most sensitive to changesin the following variables for annual screening: prevalenceof renal cancer, probability of graft failure and return todialysis, and the cost and test specificity of ultrasonog-raphy. Figure 2A shows the change in the ICER overthe plausible range of estimates tested in the sensitivityanalyses for annual screening compared with no screen-ing. There were substantial uncertainties surrounding eachof these variables on the overall ICER. If a willingness-to-pay (or the cost-effectiveness) threshold was set at therecommended ratio of $100 000/LYS [30], annual screen-ing for renal cancer does not appear to be good value formoney, unless the annual probability of renal graft failurewas <2% (the average rate of graft failure is between 2%

Table 2. Cost data for the model

VariablesAverage costs per input $A(ranges used sensitivity analyses) References

Imaging tests CT [31,32,44–46]Chest and abdomen 250Pelvis 300Ultrasound 120

Pathology tests General blood test 250 [45,46]Interventions Kidney transplantation—surgical costs

(first year)25 025 (20 145–45 500) [45,46]

Kidney transplantation—drugs and management costs (first year) 11 222 (10 567–36 725)Kidney transplantation—drugs and management costs (subsequent years) 12 564 (11 500–38 000) [45–47]Kidney transplantation (subsequent to the diagnoses of renal cancer) 11 219 [31,32,44–46]Dialysis—haemodialysis (first year) 52 720 [31,32,44–46]Dialysis—haemodialysis (subsequent years) 48 708 [31,32,44–46]Dialysis—peritoneal dialysis (first year) 63 590 [31,32,44–46]Dialysis—peritoneal dialysis (subsequent years) 58 318 [31,32,44–46]Total nephrectomy (laparoscopic) 11 966 [46,48]Cytokine therapy—IFN-alpha-2a (3 MIU)3 injections for 12 weeks

1834 [49]

Sorafenib (12 weeks) 6457 [44,49]Sunitinib (20 weeks) 6894 [44,49]Everolimus 1578 (per month) [44,49]

Others GP consultation 35 [45]Specialists consultation 126 [45]

Total costs of treatmentaccording to stage

Localized disease 11 219 [45,46,49,50]Regional disease 12 062 [45,46,49,50]Distant disease 22 574 [45,46,49,50]Palliation (inclusive costs of dying from renalcancer and other causes)

13 650 [45,46,49–51]

Costs for diagnosingrelapses

Imaging 250 [45]Pathology 200 [45]Specialists consultation (×4) 300 [45]

Costs for treatingrelapses

Sorafenib (12 weeks) 6457 [46,49]Chemotherapy (12 weeks) 2000 [46,49]Sunitinib (20 weeks) 6894 [46,49]Everolimus 1578 per month [46,49]

Table 3. Total and incremental costs and benefits of screening renal cell carcinoma

Screening strategiesTotal benefits(LYS)

Total costs($)

Incremental benefits(LYS)

Incrementalcost ($)

Incremental cost-effectivenessratio ($/LYS)

No screen 13.64193 301 700Annual 13.64598 303 000 0.00405 1300 320 988Biennial 13.64550 302 600 0.00357 900 252 100

All screening strategies (annual or biennial) were compared with no screening.

1732 G. Wong et al.

and 4% [31,32]) and the disease prevalence rate was atleast five times greater than the expected prevalence ratein the transplant population.

Biennial screening. A similar range of influential variableswas found in the biennial screening arm. Figure 2B showsthe change in ICER over the plausible range of estimatestested in the sensitivity analyses for biennial screening com-pared with no screening. Similar to the annual screeningarm, uncertainties exist surrounding each of the influentialvariables on the overall ICER. Again, using a willingness-to-pay threshold of $100 000/LYS, screening renal cancerbiennially may not be cost-effective unless the diseaseprevalence is at least 3.5 times greater than the expected dis-ease prevalence in the transplant population and the biennialprobability of renal graft failure is <5.5%.

Two-way sensitivity analyses

A series of two-way sensitivity analyses were performedto assess the combined effects between the influentialvariables within the model. Figure 3 shows the effectof varying disease prevalence in combination with otherinfluential variables in the model; these results suggestthat disease prevalence was the single most importantvariable that determines the overall cost-effectiveness ofscreening renal cancer in recipients of kidney trans-plants. For example, even with the 100% test specificity,screening for renal cancer was not cost-effective unlessthe relative disease prevalence was at least 2.5 timesgreater than the expected age-specific disease prevalenceof renal cancer in the transplant population.

Screening in the high-risk population. Table 5 summarizesthe costs and benefits of annual and biennial screening forrenal cancer using ultrasonography in the high-risk kidneytransplant population. For example, among people with afamily history of renal cancer, biennial screening saves~5.7 days of life compared with no screening, with theoverall ICER of <$70 000/LYS.

Scenario analysis

Scenario analysis was conducted to assess the health out-comes if routine screening was not continued after graftfailure and return to dialysis. The total costs of annualand biennial screening were $302 700 and $302 400.The total benefits (in life-years saved) were 13.6474 and13.6454 LYS. The incremental benefits of annual screen-ing compared with no screening were 0.0056 LYS, and theincremental benefits of screening biennially comparedwith no screening were 0.0035 LYS. The incremental costs

of annual screening compared with no screening were$1000, and the incremental costs of biennial screeningcompared with no screening were $700. The ICER for an-nual screening compared with no screening was $178 571/LYS, and the ICER for biennial screening compared withno screening was $200 000/LYS.

Discussion

Despite the increased risk of renal cancer and the improvedlife expectancy after transplantation relative to remaining ondialysis, routine screening (annual and biennial) using ultra-sonography in this population does not appear good valuefor money using the current data. Annual and biennialscreening for renal cancer achieved very small gains in lifeexpectancy, with <1.5 days of life saved, and at relativelyhigh costs. At best, compared with no screening, the abso-lute gain in survival is only two deaths from renal canceravoided per 1000 recipients if screened annually for62 years, and one death from renal cancer avoided per1000 recipients if screened biennially over the same timeframe, with ICERs of >$300 000/LYS for annual screeningand >$200 000/LYS for biennial screening. If the screeninginterval was reduced from 2 to 1 year, there was an increasein the overall and marginal costs of screening. However, thehealth benefits achieved (total and incremental) throughmore frequent screening are trivial, with the marginalcost-effectiveness ratio of annual screening compared withbiennial screening exceeding $800 000/LYS.

Using decision-analytic modelling, we have providedthe first step in formulating a decision about whether aprogramme of screening for renal cancer is effective andfeasible in the kidney transplant population. In addition,we have also identified a list of important and influentialfactors that have the greatest impact on the cost-effective-ness ratio. In contrast to many well-established screeningprogrammes, such as colorectal, breast and cervical cancerscreening, the optimal screening strategies and the mortal-ity benefits of routine screening for renal cancer have notyet been established in the general or transplant popula-tion. In particular, the screening test accuracy, the naturalhistory of renal cancer, the extent of benefits of detectingsmall and incidental lesions, and the effectiveness of treat-ment using newer therapies such as the mTOR inhibitorsand the tyrosine kinase inhibitors are all uncertain.

The accuracy of ultrasonography is an important deter-minant of screening efficiency, but is uncertain in recipi-ents of kidney transplants. Not only is ultrasonographyoperator-dependent, but its performance also varies withthe size and morphology of the patient, the kidneys andthe tumour [33,34]. In the general population, observational

Table 4. Marginal costs and benefits of screening renal cell carcinoma (expanding from biennial to annual screening)

Screeningstrategies

Total benefits(LYS)

Total costs($)

Average cost-effectivenessratio ($/LYS)

Marginalcosts ($)

Marginalbenefits (LYS)

Marginal cost-effectivenessratio ($/LYS)

Annual 13.64598 303 000 22 204 400 0.000488 833 333Biennial 13.64550 302 600 22 175

Screening for renal cancer in recipients of kidney transplants 1733

A)

Co

mp

arin

g a

nn

ual

scr

een

ing

wit

h n

o s

cree

nin

g

B)

Co

mp

arin

g b

ien

nia

l scr

een

ing

wit

h n

o s

cree

nin

g

050

000

1000

0015

0000

2000

0025

0000

3000

0035

0000

4000

0045

0000

5000

00

ICE

R (

$/L

YS

)

Rel

ativ

e d

isea

sep

reva

len

ce

Co

sts

of

ult

raso

un

d

Tes

t sp

ecif

icit

y

Gra

ft F

ailu

re

050

000

1000

0015

0000

2000

0025

0000

3000

0035

0000

4000

0045

0000

ICE

R (

$/L

YS

)

Rel

ativ

e d

isea

se

pre

vale

nce

Co

sts

of

ult

raso

un

d

Tes

t sp

ecif

icit

y

Gra

ft f

ailu

re

An

nu

al s

cree

nin

g c

om

par

ed w

ith

no

scr

een

ing

B

ien

nia

l scr

een

ing

co

mp

ared

wit

h n

o s

cree

nin

g

Infl

uen

tial

var

iab

les

Poi

nt e

stim

ates

(pl

ausi

ble

rang

e)

Poi

nt e

stim

ates

(pl

ausi

ble

rang

e)

Pro

babi

lity

of g

raft

failu

re

0.03

1 (b

etw

een

0.02

and

0.0

4)

0.05

8 (b

etw

een

0.05

7 an

d 0.

062)

Tes

t spe

cific

ity o

f ultr

asou

nd

0.90

(be

twee

n 1.

0 an

d 0.

5)

0.90

(be

twee

n 1.

0 an

d 0.

5)

Cos

ts o

f ultr

asou

nd

$120

(be

twee

n $6

0 an

d $2

00)

$120

(be

twee

n $6

0 an

d $2

00)

Rel

ativ

e di

seas

e pr

eval

ence

1.

0 (b

etw

een

5.0

and

0.8)

1.

0 (b

etw

een

5.0

and

0.8)

*The

bla

ck v

ertic

al li

nes

repr

esen

t the

ICE

R o

f ann

ual a

nd b

ienn

ial s

cree

ning

(co

mpa

red

with

no

scre

enin

g) a

t bas

e-ca

se, $

320,

988/

LYS

for

annu

al s

cree

ning

and

$25

2,10

0/LY

S fo

rbi

enni

al s

cree

ning

Fig.2.

One-w

aysensitivity

analyses

assessingtheinfluentialvariablesin

themodel.The

blackvertical

lines

representtheICER

ofannual

andbiennial

screening(com

paredwith

noscreening)

atbase-case,

$320

988/LY

Sforannual

screeningand$252

100/LY

Sforbiennial

screening.

1734 G. Wong et al.

Fig. 3. Two-way sensitivity analysis between influential variables in the biennial screening model.

Table 5. Total and incremental costs and benefits of screening renal cell carcinoma in the high-risk kidney transplant population

High-risk populations Screening strategiesTotal benefits(LYS)

Total costs($)

Incrementalbenefits (LYS)

Incrementalcosts ($)

Incrementalcost-effectivenessratio (ICER) ($/LYS)

Prior history of renal cancer No screening 13.57948 300 008Annual 13.59719 301 568 0.01771 1560 88 085Biennial 13.59518 301 018 0.01570 1010 64 331

Familial or hereditary disease suchas tuberous sclerosis

No screening 13.59883 300 530Annual 13.61278 301 931 0.01395 1401 100 430Biennial 13.61077 301 460 0.01197 930 77 694

History of analgesic nephropathy No screening 13.58538 300 167Annual 13.60193 301 679 0.0165 1512 91 636Biennial 13.59993 301 153 0.0145 986 68 000

Acquired cystic kidney disease No screening 13.55289 299 291Annual 13.57571 301 069 0.02282 1778 77 914Biennial 13.57368 300 411 0.02079 1120 53 872

All screening strategies (annual or biennial) were compared with no screening.

Screening for renal cancer in recipients of kidney transplants 1735

studies have reported lower test sensitivity and specifi-city among smaller tumours (lesions <3 cm) even withmodern techniques of harmonic imaging and ultrasoundcontrast [47,48]. The difficulties associated with ultra-sonographic screening in people with CKD include theeffect of multi-cystic diseases and small scarred nativekidneys on the overall test accuracy and poor reliabilityin differentiating small hyperechoic renal cancers fromlesions such as adenomas and angiomyolipomas. Infor-mation and evidence to address these issues are lacking.The detection of an equivocal solid lesion is followed by areference standard such as the contrast-enhanced computertomography. Unnecessary use of the intravenous contrast-enhanced imaging can lead to unwanted complications ofcontrast-induced nephropathy and increased risk ofgraft failure, particularly among those with reducedrenal function (eGFR <30 mL/min) [35], and is one ofthe major influential factors on the overall cost-effective-ness ratio in our model.

In general, malignant lesions are more aggressive inpeople with kidney transplants compared with cancers inthe general population. The extent of the aggressiveness isdependent upon the immunosuppressive state of the recipi-ents. Previous studies have reported substantial variabilityin the natural growth rates of small and incidentally detectedrenal cancers in the general population, varying between0.08 and 17.34 cm³ annually, and are dependent upon thehistological grading of the tumour [36]. The impact oflong-term immunosuppression upon the growth rates ofthese small tumours is uncertain and will be difficult tomeasure because recipients are likely to be treated surgical-ly, regardless of their tumour size, immediately after diagno-sis. In the absence of data from recipients of kidneytransplants, we assumed that the rates of cancer growthare comparablewith that in the general population. The ben-efits and cost-effectiveness of routine screening may bemore favourable if we had confirmed accurate data of theexpected increased rates of cancer development (due tothe effect of long-term immunosuppression use) in thetransplant population.

As expected, disease prevalence is the most influentialdeterminant on cost-effectiveness in this setting. The great-er the disease burden within the target population, thegreater the number of diseased individuals who may bene-fit from the early detection and effective treatment of thedisease share the costs of screening. Even though our re-sults may not support a policy of population screening forrenal cancer among the average-risk kidney transplant re-cipients, routine biennial screening may be a good use ofour healthcare dollars if we target screening among thehigher-risk population, such as those with a history of ac-quired cystic disease, a family history of renal cancer or afamilial syndrome of kidney cancer such as von Hippel–Lindau syndrome. The findings from our analyses hadshown that if the relative prevalence of disease is fourtimes greater than the expected (adjusted for age and gen-der), the ICER for biennial screening improves from abaseline of >$200 000/LYS to <$100 000/LYS.

The impact of shortened life expectancy on theoverall benefits of population cancer screening inpeople with CKD, in particular among those on dia-

lysis, is a concern for some clinicians and decisionmakers because the average expected survival for someon dialysis (or with end-stage kidney disease) is shorterthan the time lived to develop cancer [31,32]. The impactof life expectancy on the benefits, costs and harms ofscreening renal cancer among those on dialysis has beenassessed extensively in our scenario analyses. Findingsfrom our model suggest that limiting routine screening tothose with longer life expectancy will improve the overallcost-effectiveness ratio of screening, but the incrementalgains are small, saving less than half a day of life in the en-tire screening period.

There are a number of other potential limitations in thestudy. First, the estimates of survival benefits and prob-abilities of a favourable stage shift from routine screeningwere extrapolated from prospective and retrospective ob-servational studies in the transplant and non-transplantpopulations, and are subjected to potential selection,measurement and analytical bias. For some, a carefullydesigned and well-conducted randomized controlled trialof screening for renal cancer may seem unreasonable andimpractical, but the decision to implement a population-based screening programme on asymptomatic individualrelies on convincing evidence of screening benefits andtreatment effectiveness. Well-conducted primary studieson the benefits/harms of early detection and treatment ef-fectiveness of renal cancer in the transplant populationshould undoubtedly be future research priorities. Pro-spective studies assessing the screening test accuracy ofultrasonographic screening and the treatment benefits ofrenal cancer in all transplant recipients may not be feas-ible, but a study that limits to the high-risk populationsuch as those with a family history of renal cancer andwith underlying acquired cystic disease is potentiallyachievable. Second, we have not taken into account theimplication and potential harms and costs of over-detec-tion and inconsequential disease in our model. Screeningcan sometimes lead to the detection of inconsequentialdisease, a disease state that is detected by screening thatwould not contribute to poor outcomes if it remained un-detected. Early detection of disease that would not even-tually present clinically in individuals can lead to harmsfor the individual (and potentially unnecessary costs) be-cause it would lead to treatment, often invasive, and of nosignificant clinical benefits. Third, we have only consid-ered screening in the native kidneys and not in the trans-planted kidneys. Approximately 10–13% of renal cancersin the kidney transplant population occur in the renal graft[31]. The test accuracy of the ultrasonography and theoverall cost-effectiveness of screening may be differentif we had included cancers in the transplanted kidneys.Given the paucity of data of benefits and harms ofscreening renal cancer in the transplant population, a de-terministic rather than a probabilistic model was used tohandle the uncertainties within the model’s parameter es-timates. A probabilistic model has the advantage of esti-mating the joint effects of all the parameter’s uncertaintiesover a designated time frame, and allowing the estimationof uncertainties surrounding the mean cost-effectivenessratios. Finally, outcomes are measures in life-years savedrather than adjusted life-years, which may better estimate

1736 G. Wong et al.

the overall survival and quality of life of people with cancerand kidney disease.

Conclusions

In summary, screening for renal cancer in the kidney trans-plant population may not provide valuable benefits to pa-tients, but targeted screening among those with underlyingrisk factors may be a better use of the limited healthcareresource. Using decision-analytic modelling, we have inte-grated the best available evidence to inform policy makersand clinicians about the effectiveness and uncertainties ofroutine screening for renal cancer in transplant recipients.Better information about the screening test characteristicsof ultrasonography for renal cancers, and the clinicalhealth outcomes and benefits after the implementation ofroutine screening are needed before it can be routinely re-commended in recipients of kidney transplants.

Acknowledgements. We thank Simon Gruenewald for reading and hiscomments on the manuscript. We thank the ANZDATA Registry for pro-viding the relevant data for the model. Partial analysis of this report waspresented at the 45th Annual Scientific Meeting of the Australian andNew Zealand Society of Nephrology.

Conflict of interest statement. G.W. is partially supported by the Screeningand Diagnostic Test Evaluation Program (STEP) grant.

Appendix 1

Structure of the model

The model compared the health outcomes, in life-yearssaved (LYS), and healthcare costs, of no screening, annualscreening and biennial screening in a hypothetical cohortof transplant recipients (n = 1000) aged 18–69 years. Simi-lar to other established population screening programmes[1], an upper age limit was set at 70 years because of theexpected diminishing marginal benefits from screeningolder populations [2]. Individuals’ progression throughthe model was based upon age-specific transition probabil-ities (annual and biennial) through mutually exclusivehealth states of renal cancer over time. The age range in-cluded in the model was chosen to reflect the actual agerange of recipients of kidney transplants in Australia andNew Zealand [3]. The entire lifetime of the cohort of trans-plant recipients was modelled, whereby each individualwas at risk of allograft graft failure and subsequently re-turned to dialysis at the end of each cycle (annual andbiennial). The model assumed the population cohort wastransplanted once only: all recipients with failed first trans-plant were managed subsequently on dialysis until death.

The model was constructed with three arms to assess therelative costs and benefits of annual or biennial screeningcompared with no screening. Those who actually partici-pated in routine screening could test positive or negativeon ultrasonography for renal cancer, based on the preva-lence of renal cancer and diagnostic characteristics ofthe ultrasonography test. A positive screen was defined as

the presence of complex cystic (defined according to theBosniak classification) or typical solid lesions on the nativekidneys [4]. Transplant recipients with cancer who did notparticipate in screening could have their disease diagnosedclinically, or remain undiagnosed. All recipients with posi-tive screening on ultrasound or on clinical suspicion of can-cer were investigated further using contrast-enhancedcomputer tomography (CT) scans.

Transplant recipients diagnosed with renal cancer in theirnative kidneys were defined to have been treated based onthe staging of the initial diagnosis. Recipients with earlystage cancer were treated with curative surgical intervention(open or laparoscopic total nephrectomy). Recipients withaggressive disease associated with local lymph node exten-sion and systemic metastases received cytokine therapysuch as interferon-alpha-2a or newer oral tyrosine kinase in-hibitors (sorafenib and sunitinib), and maintained on com-bination immunosuppressive therapy including mammaliantarget of rapamycin (mTOR) inhibitors (such as everolimusand sirolimus) and steroid therapy. Individuals with poorprognostic features did not receive any curative treatment,but instead received palliative chemotherapy and supportivetherapy for symptomatic and pain control.

Recipients with early-stage disease who received treat-ment for their cancer could survive with cure, relapse aftertreatment, die from renal cancer or die from other causes.Recipients who had recurrent disease diagnosed had achance of developing metastatic disease or could respondto treatment. Individuals with metastatic disease could sur-vive that year with cancer, die from renal cancer or diefrom other causes. The standard maintenance immunosup-pression for transplant recipients was changed from calci-neurin inhibitor-based therapy to mTOR inhibitor-basedimmunosuppressive regimen after the diagnosis and/ortreatment of renal cancer. Recipients in the annual screen-ing arm with no cancer who remained alive at the end of1 year cycled back to the annual screening decision node.

The biennial screening arm was similar to the annualscreening arm but differed only in the cycle length (2 yearsinstead of 1 year) and the transition probabilities (biennialprobabilities rather than annual). In the annual screeningarm, the cumulative benefits and costs were calculatedafter 62 annual cycles, and in the biennial screening arm,the cumulative benefits and costs were calculated after 31biennial cycles, by which time all patients were deceased.

Input parameters for the model

Clinical data. A comprehensive literature search was con-ducted to obtain the age- and cancer-stage-dependent tran-sition probabilities from relevant published literature in thetransplant and non-transplant populations (Table 1). Theincluded variables were age-specific participation rate, testsensitivity and specificity of US screening, stage-specificdistribution of cancer in the screened (annual and biennial)and unscreened arms, probability of recurrence, and prob-ability of complications from diagnostic investigations.The age-specific prevalence of renal cancer in transplantrecipients and those who returned to dialysis after graftfailure, the age- and stage-specific probability (annualand biennial) of survival and death from renal cancer,

Screening for renal cancer in recipients of kidney transplants 1737

and the age-specific mortality rates of death from othercauses were sourced from the Australian and New ZealandDialysis and Transplant (ANZDATA) Registry 1995–2007.Given that screening for renal cancer is not routinely re-commended for recipients of kidney transplant, we as-sumed that the rates of opportunistic screening amongtransplant recipients were minimal.

Cost data. Only direct healthcare costs were included inthe analysis (Table 2). Unit costs of screening for renal can-cer were obtained from the Australian Diagnosis RelatedGroups (AR-DRG) [5], the Medicare Benefits Schedule(MBS) [6], the Cancer Institute of NSW (CI NSW) [7]and the published data from trials of the intervention withmTOR and tyrosine kinase inhibitors for recipients withrenal cancer [8–13]. The average Australian costs were thenassigned to each of the disease health states, and all costswere subsequently updated to 2008 using the Medicarecomponent of the Consumer Price Index (CPI) [14]. If thecost data for certain interventions and specified health stateswere unavailable in Australian dollars, costs from other highincome countries with similar healthcare systems such asCanada and the UK were extrapolated and converted tothe 2008 Australian dollar using purchasing power parities.

Reference list

1. Australian Government Department of Health and Aging. Australia’sBowel Cancer Screening Pilot and Beyond. Final Evaluation Report.2005

2. LeBrun CJ, Diehl LF, Abbott KC et al. Life expectancy benefits ofcancer screening in the end-stage renal disease population. Am J Kid-ney Dis 2000; 35: 237–243

3. Australian and New Zealand Dialysis and Transplant Registry. The30th Annual Report. 11-1-0007

4. Bosniak MA. The use of the Bosniak classification system for renalcysts and cystic tumors. J Urol 1997; 157: 1852–1853

5. Australian Government Australia Institute of Health and Welfare.AR-DRG Data Cubes from 2006–2007. 2007.

6. Australian Government Department of Health and Ageing. MedicareBenefits Schedule Book 2006

7. Cancer Institute NSW Standard Cancer Treatment 2006; https://www.treatment.cancerinstitute.org.au/cancerinstitute

8. Bellmunt J, Gonzalez-Larriba JL, Climent MA et al. Sorafenib TAR-GET trial results in Spanish patients. Clinical and Translational On-cology: Official Publication of the Federation of Spanish OncologySocietes and of the National Cancer Institute of Mexico 2007; 9:671–673

9. Bukowski R, Cella D, Gondek K et al. Sorafenib TARGET ClinicalTrial Group. Effects of sorafenib on symptoms and quality of life:results from a large randomized placebo-controlled study in renalcancer. Am J Clin Oncol 2007; 30: 220–227

10. Gao X, Reddy P, Dhanda R et al. Cost-effectiveness of sorafenib ver-sus best supportive care in advanced renal cell cancer. J Clin Oncol2006; 24: 4604

11. Motzer RJ, Hutson TE, Tomczak P et al. Sunitinib versus interferonalfa in metastatic renal-cell carcinoma. N Engl J Med 2007; 356:115–124

12. Motzer RJ, Escudier B, Oudard S et al. Efficacy of everolimus inadvanced renal cell carcinoma: a double-blind, randomised, place-bo-controlled phase III trial. Lancet 2008; 372: 449–456

13. Remak E, Charbonneau C, Negrier S et al. Economic evaluation ofsunitinib malate for the first-line treatment of metastatic renal cellcarcinoma. J Clin Oncol 2008; 26: 3995–4000

14. Quality Statistical Release: Consumer Price Index. Reserve Bank ofAustralia, 2009

References

1. Grulich AE, van Leeuwen MT, Falster MO et al. Incidence of cancersin people with HIV/AIDS compared with immunosuppressed trans-plant recipients: a meta-analysis. Lancet 2007; 370: 59–67

2. Webster AC, Craig JC, Simpson JM et al. Identifying high riskgroups and quantifying absolute risk of cancer after kidney trans-plantation: a cohort study of 15,183 recipients. Am J Transplant2007; 7: 2140–2151

3. Vajdic CM, McDonald SP, McCredie MR et al. Cancer incidence be-fore and after kidney transplantation. JAMA 2006; 296: 2823–2831

4. Miao Y, Everly JJ, Gross TG et al. De novo cancers arising in organtransplant recipients are associated with adverse outcomes comparedwith the general population. Transplantation 2009; 87: 1347–1359

5. Chapman JR, Webster AC. Australian and New Zealand Dialysis andTransplant Registry. The 25th Annual Report. Chapter 10 2002

6. Towler B, Irwig L, Glasziou P et al. A systematic review of the effectsof screening for colorectal cancer using the faecal occult blood test,hemoccult. BMJ 1998; 317: 559–565

7. Andersson I, Janzon L. Reduced breast cancer mortality in womenunder age 50: updated results from the Malmo MammographicScreening Program. J Natl Cancer Inst 1997; 22: 63–67

8. Eddy DM. Screening for breast cancer. Ann Intern Med 1989; 111:389–399

9. American College of Obstetricians and Gynecologists. Cervical cy-tology screening. ACOG Practice Bulletin No. 45

10. ACOG. ACOG Practice Bulletin. Breast Cancer Screening. Number42, April 2003. Int J Gynaecol Obstet 2003; 81: 313–323

11. Einollahi B, Simforoosh N, Lessan-Pezeshki M et al. Genitourinarytumor following kidney transplantation: a multicenter study. Trans-plant Proc 2009; 41: 2848–2849

12. Nobushita T, Hara N, Nakagawa Y et al. Renal cell carcinomas aris-ing from the allograft and bilateral native kidneys. Int J Urol 2008;15: 175–177

13. Schwarz A, Vatandaslar S, Merkel S et al. Renal cell carcinoma intransplant recipients with acquired cystic kidney disease. Clin J AmSoc Nephrol 2007; 2: 750–756

14. Ianhez LE, Lucon M, Nahas WC et al. Renal cell carcinoma in renaltransplant patients. Urology 2007; 69: 462–464

15. Moudouni SM, Lakmichi A, Tligui M et al. Renal cell carcinoma ofnative kidney in renal transplant recipients.BJU Int 2006; 98: 298–302

16. Denton MD, Magee CC, Ovuworie C et al. Prevalence of renal cellcarcinoma in patients with ESRD pre-transplantation: a pathologicanalysis. Kidney Int 2002; 61: 2201–2209

17. Gulanikar AC, Daily PP, Kilambi NK et al. Prospective pretransplantultrasound screening in 206 patients for acquired renal cysts and renalcell carcinoma. Transplantation 1998; 66: 1669–1672

18. Ishikawa I, Honda R, Yamada Y et al. Renal cell carcinoma de-tected by screening shows better patient survival than that detectedfollowing symptoms in dialysis patients. Ther Apher Dial 2004; 8:468–473

19. Parienty RA, Richard F, Pradel J et al. Local recurrence after neph-rectomy for primary renal cancer: computerized tomography recogni-tion. J Urol 1984; 132: 246–249

20. Hasegawa Y, Mita K, Matsubara A et al. Multidisciplinary treatmentincluding sorafenib stabilized the bone metastases of renal cell car-cinoma in an immunosuppressed renal transplant recipient. Int J ClinOncol 2009; 14: 465–467

21. Members of the KDIGO Working Group. KDIGO Clinical PracticeGuidelines for the Care of Kidney Transplant Recipients. Am J Trans-plant 2009; 9: S1–S155

22. Kasiske BL, Vazquez MA, Harmon WE et al. Recommendations forthe outpatient surveillance of renal transplant recipients. AmericanSociety of Transplantation. J Am Soc Nephrol 2000; 11: 1910–1917

23. Kälble T, Lucan M, Nicita G et al. Guidelines on renal transplant-ation. European Association of Urology 2006

24. Kronborg O, Jorgensen OD, Fenger C et al. Randomized study ofbiennial screening with a faecal occult blood test: results after ninescreening rounds. Scand J Gastroenterol 2004; 39: 846–851

1738 G. Wong et al.

25. Mandel JS, Bond JH, Church TR et al. Reducing mortality fromcolorectal cancer by screening for fecal occult blood. MinnesotaColon Cancer Control Study. N Engl J Med 1993; 328: 1365–1371

26. Commonwealth Department of Health and Ageing. Guidelines for thePharmaceutical Industry on the Preparation of Submissions to thePharmaceutical Benefits Advisory Committee. Canberra, ACT: Com-monwealth Department of Health and Ageing, 2006

27. Evers S, Goossens M, de Vet H et al. Criteria list for assessment ofmethodological quality of economic evaluations: Consensus onHealth Economic Criteria. Int J Technol Assess Health Care 2005;21: 240–245

28. National Institute for Clinical Excellence. NICE: Guideline Develop-ment Methods. Incorporating Health Economics in Guidelines andAssessing Resource Impact http://www.nice.org.uk/niceMedia/pdf/GDM_Chapter8_0305.pdf. 2004.

29. TreeAge Software, Inc http://treeage.com/products/index.html. 200930. Quinn RR, Naimark DM, Oliver MJ et al. Should hemodialysis pa-

tients with atrial fibrillation undergo systemic anticoagulation? Acost-utility analysis. Am J Kidney Dis 2007; 50: 421–432

31. Riccabona M, Szolar D, Preidler K et al. Renal masses—evaluationby amplitude coded colour Doppler sonography and multiphasic con-trast-enhanced CT. Acta Radiol 1999; 40: 457–461

32. Schmidt T, Hohl C, Haage P et al. Diagnostic accuracy of phase-in-version tissue harmonic imaging versus fundamental B-mode sonog-raphy in the evaluation of focal lesions of the kidney. Am JRoentgenol 2003; 180: 1639–1647

33. Herts BR, Schneider E, Obuchowski N et al. Probability of reducedrenal function after contrast-enhanced CT: a model based on serumcreatinine level, patient age, and estimated glomerular filtration rate.Am J Roentgenol 2009; 193: 494–500

34. Fenton JJ, Weiss NS. Screening computed tomography: will it resultin overdiagnosis of renal carcinoma? Cancer 2004; 100: 986–990

35. Australia and New Zealand Dialysis and Transplant Registry (AN-ZDATA), Special data request (2009). 2005

36. Australian and New Zealand Dialysis and Transplant Registry. The30th Annual Report. 1-2-2007

37. Mihara S, Kuroda K, Yoshioka R et al. Early detection of renalcell carcinoma by ultrasonographic screening—based on the results

of 13 years screening in Japan. Ultrasound Med Biol 1999; 25:1033–1039

38. Neuzillet Y, Lay F, Luccioni A et al. De novo renal cell carcinoma ofnative kidney in renal transplant recipients. Cancer 2005; 103: 251–257

39. Reinberg Y, Matas A, Manivel C et al. Outcome of renal transplant-ation or dialysis in patients with a history of renal cancer. Cancer1992; 70: 1564–1567

40. Kojima Y, Takahara S,Miyake O et al. Renal cell carcinoma in dialysispatients: a single center experience. Int J Urol 2006; 13: 1045–1048

41. Tsui KH, Shvarts O, Smith RB et al. Prognostic indicators for renalcell carcinoma: a multivariate analysis of 643 patients using the re-vised 1997 TNM staging criteria. J Urol 1295: 1090–1095

42. Malaeb BS, Martin DJ, Littooy FN et al. The utility of screeningrenal ultrasonography: identifying renal cell carcinoma in an elderlyasymptomatic population. BJU Int 2005; 95: 977–981

43. Tamai H, Takiguchi Y, Oka M et al. Contrast-enhanced ultrasonog-raphy in the diagnosis of solid renal tumors. J Ultrasound Med 2005;24: 1635–1640

44. Australian Government Department of Health and Ageing. Sceduleof Pharmaceutical Benefits. 10 A.D.

45. Australian Government Department of Health and Ageing. MedicareBenefits Schedule Book. 6-11-2006.

46. Australian Government Australia Institute of Health and Welfare.AR-DRG Data Cubes from 2006–2007. 2007

47. Australian and New Zealand Dialysis and Transplant Registry. The30th Annual Report. 11-1-0007

48. Bensalah K, Raman JD, Bagrodia A et al. Does obesity impact thecosts of partial and radical nephrectomy? J Urol 179: 1714–1717

49. Cancer Institute NSW standard cancer treatment (https://www.treat-ment.cancerinstitute.org.au/cancerinstitute). 2006

50. Australian Institute of Health and Welfare. Australian Hospital Statis-tics (2003–2004). 2004

51. Belasco JB, Danz P, Drill A et al. Supportive care: palliative care inchildren, adolescents, and young adults—model of care, interven-tions, and cost of care: a retrospective review. J Palliat Care 2000;16: 39–46

Received for publication: 18.4.10; Accepted in revised form: 18.9.10

Screening for renal cancer in recipients of kidney transplants 1739