Cost-Effectiveness of Robot-Assisted Radical Cystectomy vs Open Radical Cystectomy for Patients With Bladder Cancer

This economic evaluation of a randomized clinical trial compares the cost-effectiveness of intracorporeal robot-assisted radical cystectomy with that of open radical cystectomy for patients with bladder cancer, including the consideration of prespecified patient subgroups.


Introduction
Each year more than 550 000 new cases of bladder cancer are diagnosed worldwide. 1Around one-third of bladder cancers require radical treatment, including radical cystectomy with pelvic lymphadenectomy. 2 An estimated 33 429 radical cystectomy operations were performed in the US from 2008 to 2011, 3 with most patients developing 1 or more complications and 20% to 30% readmitted post discharge. 4,5While reductions in morbidity from radical cystectomy have been achieved through robot-assisted radical cystectomy (RARC), 6,7 the cost-effectiveness of these approaches is unclear.
A systematic review of economic studies of RARC and open radical cystectomy (ORC) 8 highlighted that the costs of the robot-assisted procedure are likely to be higher than open comparators, despite savings from reductions in complications.However, changes associated with patient throughput and shorter surgery times were highlighted as important considerations, as well as the need for prospectively collected quality of life information that could be used to generate quality-adjusted life-years (QALYs).A prospectively designed cost-effectiveness analysis that was integrated into a randomized comparison of total intracorporeal RARC (iRARC) and ORC provides this information.The objective of this study was to compare the cost-effectiveness of iRARC with that of ORC for patients with bladder cancer, including the consideration of prespecified patient subgroups.

Patients and the iROC Study
The cost-effectiveness analysis was integrated into the iROC (Intracorporeal Robot-Assisted Radical Cystectomy vs Open Radical Cystectomy) study, which was a multicenter, unblinded, randomized clinical trial that recruited patients from 9 centers in the United Kingdom from March 20, 2017, to January 29, 2020. 6The primary objective of the trial was to investigate the effects of the different forms of surgery on patient recovery.Patients were eligible to be recruited to the trial if they were adults with nonmetastatic urothelial, squamous, adenocarcinoma, or variant bladder cancer.Of 338 patients randomized (169 in each group), 317 underwent radical cystectomy, with those in the robot-assisted group spending a mean of 2.2 (95% CI, 0.50-3.85)days longer alive or out of hospital.Statistically significant differences in health-related quality of life (HRQOL) and disability were also identified at 5 weeks using the European Quality of Life 5-Dimension 5-Level instrument (EQ-5D-5L) 9 and the World Health Organization Disability Assessment Schedule, version 2.0. 10 The trial received ethical approval from the Newcastle and North Tyneside Research Ethics Committee, and all patients gave written informed consent (for details, see Catto et al 6 ).Reporting aligned with Consolidated Health Economic Evaluation Reporting Standards (CHEERS) and is detailed in eMethods in Supplement 1.

Cost-Effectiveness Analysis
The cost-effectiveness analysis used QALYs based on patient EQ-5D-5L responses and was undertaken from the perspective of the United Kingdom National Health Service. 11When analyzed as described, EQ-5D-5L scores range from 1.00 (perfect health) through 0 (a health state equivalent to death) to −0.594 (the worst possible health classified by the instrument).The primary analysis used a time horizon of 90 days post surgery so that it aligned with the trial evidence, and secondary analyses were based on projections of patient recovery to 180, 270, and 360 days.The general JAMA Network Open | Surgery approach was consistent with methodological guidelines 12 outlined in the trial protocol 13 and prespecified in a health economics analysis plan (eMethods in Supplement 1).Costs have been converted to US dollars using the purchasing power parity for 2021. 14

Resource Use
The principal differences in resource use were expected to be related to theater equipment, staff mix in the theater, length of theater time, and length of ward stay.Further differences were considered in terms of intensive care, high-dependency care, units of blood transfused, family physician attendances, emergency department attendances, and readmissions.Data were available via case report forms from the iROC trial and relate to the period from admission to the theater suite.Length of stay was calculated as the number of separate days on which a patient was present in hospital.

Unit Costs
All unit costs are based on fully absorbed accounting principles, or market prices if available; no hospital charges were used.Robot costs were based on the purchase and maintenance price for a surgical robot (Da Vinci X; Intuitive), a simulator, instruments, and staff time for training.
Nonrecurrent costs were annuitized using a discount rate of 3.5% in advance over 10 years and zero reuse value. 11Capital costs were allocated across 206 patients per annum, based on a study by Lam et al, 15 and surgeon training costs were allocated across 40 patients per annum.The resultant cost for iRARC was £2638 (US $3807) per patient (eTable 1 in Supplement 1).Equipment costs for ORC were taken from a recent United Kingdom-based study that identified and priced each individual component of theater equipment used, which produced a cost of £1514 (US $2185) per patient. 16st per theater minute and cost per ward day were calculated in consultation with business managers at one of the larger recruiting sites.Theater costs were based on staffing, equipment, and consumables in urology theaters.Both theater and ward costs were then adjusted, pro rata, to match national average costs (eTable 2 in Supplement 1).All other unit costs were taken from publicly available sources (eTable 2 in Supplement 1).All costs were at 2020-2021 price levels, with adjustment to that level if required, using the National Health Service Cost Inflation Index. 17

Outcomes
We calculated QALYs using linear interpolation of EQ-5D-3L United Kingdom tariff values.Tariff values were calculated using the van Hout crosswalk tariff. 18The EQ-5D-5L was completed at baseline and 5 weeks and 90 days post surgery.However, this was not considered to be adequate for incorporating potential short-term differences in HRQOL, and so EQ-5D-5L values at 5 days post surgery were imputed using quantified activity levels recorded at that time.Specifically, using the aldvmm package in R, version 4.2.2 (R Project for Statistical Computing), an adjusted limited dependent variable mixture regression model was fitted between the week 5 EQ-5D-5L and week 4 activity data (mean across days 4, 5, and 6), together with appropriate covariates.The results were then used to estimate the day 5 EQ-5D-3L score using the mean activity data across the 3 days.

Statistical Analysis
Data were analyzed from January 13, 2022, to March 10, 2023.Mean resource use was estimated and compared using unpaired, 2-tailed t tests with unequal variances for continuous variables, negative binomial regression for count data, and χ 2 tests for event rates.A 5% level of statistical significance was used with 2-sided hypothesis tests (P < .05).Costs and QALYs were compared using seemingly unrelated regression models to account for correlation between costs and QALYs. 19For cost, the sole independent variable was treatment group, while for QALYs, treatment group, age, sex, and baseline EQ-5D-5L utility score were used as independent variables.Subgroup analyses incorporated an interaction term into the same regression specification.The analysis was based on patients for whom the primary outcome of the clinical trial was available (n = 305).

JAMA Network Open | Surgery
Missing data for EQ-5D-5L utility score (at baseline, 5 days, 5 weeks, and 12 weeks) were based on 20 imputed data sets.The imputed data sets were established in chain regressions with covariates of age, sex, and group.All analyses, unless otherwise stated, were undertaken in Stata, version 17 (StataCorp LLC).
Incremental cost-effectiveness ratios (ICERs) were calculated using the coefficients of the treatment variables of the seemingly unrelated regressions.Five thousand mean values for incremental costs and QALYs were bootstrapped and plotted on the cost-effectiveness plane to give a cost-effectiveness acceptability curve.The main conclusions of the analysis are based on a United Kingdom funding threshold of £20 000 (US $28 8960) per QALY gained. 11eplanned subgroup analyses included the following: chemotherapy vs no chemotherapy, stage T2 or less vs T3 or greater, age younger than 70 years vs 70 years or older, performance status 0 vs 1 or greater on the ECOG Performance Status Scale (score range, 0 [fully active] to 5 [dead]), male vs female sex, and type of diversion, consisting of ileal conduit vs neobladder or other reconstruction.A further post hoc analysis was undertaken to assess where body mass index (BMI; calculated as weight in kilograms divided by height in meters squared) may be a potential effect modifier.
Sources of methodological and cost uncertainty were identified and explored through deterministic sensitivity analyses.Sources relating to methodological uncertainty included: • complete case analysis (cases with complete EQ-5D-5L data at baseline and 5 weeks and 90 days post surgery); • omission of the day 5 utility imputation; • use of last observation carried forward for missing data; • scoring the EQ-5D-5L using an alternative algorithm 20 ; and • extrapolation of day 90 results assuming convergence of mean utilities at 180, 270, and 360 days post surgery.
Sources relating to cost uncertainty are detailed in eTable 3 in Supplement 1, but briefly included: • lower theater cost per minute for iRARC; • lower ORC equipment costs; • lower cost of a day on a ward; • alternative life span and throughput of the surgical robot; and • alternative hospital costs.

Results
Data were available for 305 patients, all of whom were included in the analysis.Race and ethnicity data were not collected as there is no robust evidence that race or ethnicity is related to prognosis.Mean (SD) age was 68.3 (8.1) years; 241 (79.0%) were men and 64 (21.0%) were women.

QALYs and Cost-Effectiveness
The profile of HRQOL as measured by EQ-5D-5L is shown in Figure 2. The QALYs associated with these profiles, calculated as the areas under the curves, produce an incremental benefit associated with iRARC of 0.01124 QALYs (95% CI, 0.00391-0.01857QALYs; P = .003).The ICER is £100 008 (US $144 312) per QALY gained (Table 2), with a 16.3% chance that iRARC is cost-effective at £20 000 (US $28 860) per QALY gained.The cost-effectiveness plane and associated cost-effectiveness acceptability curves are shown in eFigures 1 and 2 in Supplement 1, respectively.

Sensitivity Analysis
Sensitivity analyses relating to methodological uncertainties (Table 2) revealed changes associated with the ICER of alternative approaches to imputation.The use of a complete case analysis yielded the greatest changes, suggesting that iRARC reduces cost and improves health (dominant), increasing the probability that iRARC is cost-effective to 63.3%.Use of the last observation carried forward as an imputation method had a less dramatic effect, showing a 32.0% chance of iRARC being cost-effective.The extrapolation of health benefits to 1 year after surgery (90 days of patient follow-up plus 270 days of extrapolation) was associated with a 19.9% chance of being cost-effective (Table 2).
Changes in unit costs reflected iRARC surgery times being shorter than those observed in iROC, with the greatest changes to the ICER; a 10% reduction showed that iRARC was dominant and had a 61.9% chance of being cost-effective at £20 000 (US $28 860) per QALY gained.The estimated changes in robot life expectancy and throughput yielded smaller changes to the ICER.Lower ward costs that may better reflect marginal cost savings yielded only small changes to the ICER.The results

Subgroup Analysis
The subgroup analyses revealed large changes to cost-effectiveness when different age groups, tumor stages, and performance status were evaluated (Table 3).The probabilities of iRARC being cost-effective at £20 000 (US $28 860) per QALY gained were 82.2% among patients 70 years or older, 77.6% among those with large tumors, and 84.7% among those with a worse performance status.There were moderate improvements in cost-effectiveness among patients undergoing an ileal conduit diversion, with an ICER of £58 101 (US $83 840) per QALY gained and probability of being cost-effective of 32.9%, and those with a BMI of 25 or above, with an ICER of £66 656 (US $96 185) per QALY gained and probability of being cost-effective of 34.1%.
An exploratory post hoc analysis was undertaken to examine individual cost components and how they differed between patient subgroups (eTable 5 in Supplement 1).Our findings suggest that the improved cost-effectiveness of iRARC for more elderly patients was associated with larger reductions in length of stay and readmissions.Reductions in length of stay were also present for patients with larger tumors, while for performance status, there appeared to be no clear association.

Discussion
Our primary economic analysis suggests that iRARC has greater costs per procedure than ORC.There are clear differences in resources used by each surgical approach, with the higher equipment costs of iRARC being partly offset by savings in other ward and critical care costs.This higher cost per patient (£1124 [US $1622]; P = .20)was associated with greater health benefits (0.01124 QALYs; P = .003);however, the resultant ICER of £100 008 (US $144 312) was above normal funding thresholds in the United Kingdom.
Our sensitivity analysis highlighted several practical issues.First, reducing robotic surgical times has a large effect on cost-effectiveness.Second, more general changes in the level of health service costs (excluding equipment prices) had relatively small effects due to the counteracting effects of iRARC producing higher theater costs but lower ward costs.However, it should be recognized that these conclusions are based on changes in one cost component at a time.For example, if it is possible to reduce the additional cost of iRARC by £899 (US $1297) per patient by a combination of changes to working practices and/or price, then iRARC becomes cost-effective at a threshold of £20 000 (US $28 860) per QALY gained.When analyzed as described, EQ-5D-5L scores range from 1.00 (perfect health) through 0 (a health state equivalent to death) to −0.594 (the worst possible health classified by the instrument).iRARC indicates robot-assisted radical cystectomy with intracorporeal urinary diversion; ORC, open radical cystectomy.Subgroup analyses revealed large changes in cost-effectiveness; iRARC was cost-effective in patients 70 years or older, with tumor stages of T3 or greater, or with a performance status of 1 or above.Our exploratory analysis of these findings suggests that some of these differences are due to younger patients being able to tolerate open surgery better, and so the length of hospital stay and readmission benefits of iRARC are much greater in the older patient group.However, given the post hoc nature of these analyses and the correlations between subgroup membership, we do not believe that it is appropriate to tease apart these differences further.
Nonetheless, the subgroup analysis suggests that even if payers consider iRARC not to be costeffective for all patients within the iROC study, subgroups can be identified for whom iRARC is costeffective.Other subgroups relating to type of diversion and BMI may also be cost-effective, depending on the funding thresholds adopted by various countries or health plans.

Strengths and Limitations
The biggest strength of this study is that it was integrated into a high-quality randomized clinical trial with patient-level data for all major cost components and HRQOL.As such, it overcomes the weaknesses identified in a previously published systematic review of robot-assisted radical cystectomy for patients with bladder cancer. 8In addition, the subgroup analyses yielded information for payers who may wish to limit coverage for economic or budget impact reasons.
[26][27] The economic evaluation was also undertaken with direct and indirect stakeholder involvement to ensure its relevance.It was designed with the assistance of the lead clinician of the underlying iROC study, while the final design and conduct of the iROC study was overseen by a steering committee that included patient representatives. 13e main limitation relating to the clinical evidence is the use of the 90-day follow-up, which is expected to systematically underestimate the HRQOL benefits of iRARC.However, sensitivity analysis explored this, and with extrapolation to what was considered to be the largest plausible length of morbidity benefit (360 days), the ICER remained high.The main limitation for the costs was the lack of study data on the use and cost of instruments for ORC, which meant relying on a previously published figure.While this figure was the most relevant figure available, other available estimates for ORC and related surgical procedures are much lower.This was explored in the sensitivity analysis.
The generalizability of costs and cost-effectiveness needs to be considered by funders.While our sensitivity analysis showed that variability in costs has a limited effect, this is based on United Kingdom cost structures and surgical practices.As such, the extent to which these economic results can be generalized beyond the United Kingdom is unknown.Similarly, the funding threshold presented may not be relevant to other countries (even if translated to local currency units).
Consequently, local information needs to be taken into account when our results are used to inform policy in other countries.This is perhaps most pertinent to length of stay, as hospitals and/or countries with short lengths of stay may be less likely to be able to deliver the absolute reductions seen here, which will lead to reduced cost-effectiveness.
There were 5 deviations from the health economics analysis plan.Three of these relate to the adoption of alternative methods: the choice of hospitals used to estimate costs, the method of extrapolation beyond the trial follow-up, and the source of equipment costs relating to ORC.All changes were explored with sensitivity analysis and were found to alter the results only minimally.
The other 2 changes relate to the subgroup analysis of BMI and the exploratory analysis relating to the subgroups analysis of age, tumor size, and performance status.These are clearly reported as post hoc analyses with findings interpreted accordingly.

Table 1 .
The Difference in Resource Use for the ORC and iRARC Interventions a Relates to the percentages in the 2 preceding columns.bIndicates unpaired 2-tailed t test with unequal variance.

Table 2 .
Sensitivity Analysis of Incremental Costs, QALYs, and Cost-Effectiveness of iRARC vs ORC higher-and lower-cost hospitals, with the use of United Kingdom upper and lower quartile costs yielding a small change to the ICERs.These results are shown in Table2.
a iRARC dominant indicates that iRARC is more effective and less costly, and in such cases, an ICER is not generally reported due to problems associated with the interpretation of negative ratios.bIndicates the probability that iRARC is cost-effective at £20 000 per QALY gained.JAMA Network Open | SurgeryRobot-Assisted Radical Cystectomy vs Open Radical Cystectomy for Bladder Cancer JAMA Network Open.2023;6(6):e2317255.doi:10.1001/jamanetworkopen.2023.17255(Reprinted) June 30, 2023 6/12 Downloaded From: https://jamanetwork.com/ by a University College London User on 07/02/2023 appear generalizable to

Table 3 .
Subgroup Analyses of Incremental Costs, QALYs, and Cost-Effectiveness of iRARC vs ORC means that iRARC is more effective and less costly, and in such cases, an ICER is not generally reported due to problems associated with the interpretation of negative ratios.Indicates the probability that iRARC is cost-effective at £20 000 per QALY gained.Calculated as weight in kilograms divided by height in meters squared.
a iRARC dominant b c Downloaded From: https://jamanetwork.com/ by a University College London User on 07/02/2023

Open | Surgery SUPPLEMENT 1. eMethods.
Health Economics Analysis Plan eTable 1. Unit Cost of Robot per Patient eTable 2. Unit Costs for the Primary Analysis eTable 3. Unit Costs for Sensitivity Analyses eTable 4. Staff Mix of Surgeons in Theater eTable 5. Exploratory Analysis of Subgroup Differences by QALYs and Individual Components of Resource eFigure 1. Cost-Effectiveness Plane With Bootstrapped Incremental Costs and Quality-Adjusted Life-Years (QALYs) eFigure 2. Cost-Effectiveness Acceptability Curve for iRARC eReferences