Cost-effectiveness of surveillance schedules in older adults with non-muscle-invasive bladder cancer.

OBJECTIVE: To estimate the cost-effectiveness of surveillance schedules for non-muscle-invasive bladder cancer (NMIBC) amongst older adults. PATIENTS AND METHODS: We developed a MIcrosimulation SCreening ANalysis (MISCAN) microsimulation model to compare the cost-effectiveness of various surveillance schedules (every 3 months to every 24 months, for 2, 5 or 10 years or lifetime) for older adults (aged 65-85 years) with NMIBC. For each surveillance schedule we calculated total costs per patient and the number of quality adjusted life-years (QALYs) gained. Incremental cost-effectiveness ratios (ICERs), as incremental costs per QALY gained, were calculated using a 3% discount.
RESULTS: As age increased, the number of QALYs gained per patient decreased substantially. Surveillance of patients aged 65 years resulted in 2-7 QALYs gained, whereas surveillance at age 85 years led to <1 QALY gained. The total costs of the surveillance schedules also decreased as age increased. The ICER of 6-monthly surveillance at age 65 years for lifetime was $4999 (American dollars)/QALY gained. Amongst patients aged >75 years, the incremental yield of QALY gains for any increase in surveillance frequency and/or duration was quite modest (<2 QALYs gained).
CONCLUSION: With increasing age, surveillance for recurrences leads to substantially fewer QALYs gained. These data support age-specific surveillance recommendations for patients treated for NMIBC.

Introduction

About 75% of all bladder cancer diagnoses are non‐muscle‐invasive bladder cancer (NMIBC) 1. NMIBC is generally treated surgically by transurethral resection (TUR). The 10‐year cancer‐specific survival is relatively high, ranging from 70% to 85% for high‐grade to 88–98% for low‐grade cancers 2, 3. However, the probability of recurrence is high (up to 75%) and 10–20% can progress to muscle‐invasive bladder cancer (MIBC), which is potentially lethal 2, 4, 5, 6. Therefore, surveillance by cystoscopy is offered to patients with a history of NMIBC to facilitate early diagnosis and treatment before progression.

All international guidelines on the management of NMIBC recommend cystoscopy at regular intervals, generally based on the patient's risk, but the recommended intervals differ 7. The AUA and the European Association of Urology (EAU) recommend low‐risk patients to have two cystoscopies in the first year, followed by annual cystoscopy for 5 years. The EAU recommends discontinuing surveillance after 5 years, whereas the AUA recommends shared decision‐making to stop or continue annual surveillance 8, 9. The National Institute for Health and Care Excellence (NICE) guideline recommends surveillance only at 3 and 12 months for low‐risk patients, discharging low‐risk patients to primary care if recurrence‐free at 12 months 10. For high‐grade patients, various guidelines generally recommend surveillance every 3 months for the first 2 years, followed by every 6 months for 2 or 3 years, followed by annual and even lifelong surveillance.

Surveillance cystoscopy is an invasive procedure. Patients may experience anxiety and pain, or complications such as painful urination and UTI, which influence quality of life (QoL) 11, 12. Furthermore, bladder cancer is largely a disease of older adults, and has the highest median patient age at diagnosis (73 years) of all cancer sites 13, 14. Despite relatively stable age‐adjusted incidence rates, the overall burden of bladder cancer has increased dramatically with the ageing of the population, increasing from stable figures of ~50 000 cases/year in the USA in the 1990s to >80 000 cases/year projected for 2018 15, 16. Over this same period, the median age at diagnosis of bladder cancer has increased from 67 to 73 years.

Given substantial competing causes of death in this predominantly elderly population, the intensity and duration of the surveillance programme should be evaluated for different age groups. Surveillance programmes should be optimised in terms of frequency of cystoscopy and duration of the surveillance, to simultaneously maximise the life‐years gained and minimise the harms and the costs.

The present study objective was to determine the most cost‐effective surveillance protocol for older adults with NMIBC. We evaluated surveillance strategies of varying frequency and duration.

Patients and Methods

MIcrosimulation SCreening ANalysis (MISCAN) model

We developed a microsimulation model using the MISCAN framework to simulate the impact of surveillance on survival in a patient with NMIBC. MISCAN is a microsimulation model developed for the evaluation of screening and has been used previously to estimate the harms, benefits and cost‐effectiveness of breast, colorectal, cervical and prostate cancer screening 17, 18, 19, 20. In this model, individual life histories of patients are simulated by modelling the transitions between possible health states (Fig. 1). All patients start in the no bladder cancer state after having been diagnosed and treated for bladder cancer. Some patients will have a NMIBC recurrence, which will be treated by TUR, or will progress to the MIBC state. An assumption in the model is that progression can only take place after recurrence. MIBC is assumed to be treated with cystectomy, which ends surveillance. In each subsequent state, there is an increasing risk of death from bladder cancer, therefore preventing progression to the MIBC state by surveillance using cystoscopy will decrease the risk of death and will increase the number of quality adjusted life‐years (QALYs).

Figure 1

Possible transitions in the MISCAN model.

Model inputs were drawn from a previously developed Markov model with time steps of 3 months 21. This semi‐Markov model was based on the Dutch Cost‐Effectiveness of Follow‐up of Urinary Bladder Cancer (CEFUB) trial, which included 448 patients with non‐muscle‐invasive urothelial cancer (NMI‐UC) 22. In that trial, the efficacy of microsatellite analysis on voided urine for detecting tumour recurrences in the follow‐up of patients was evaluated. Patients included had a primary or recurrent NMI‐UC [pTa (85%), pT1 (15%), Grade 1 (43%) or Grade 2 (57%)] of the urinary bladder, based on histopathological examination of the surgically removed tumour. About half of the cohort (118/228) was newly diagnosed, whereas the remainder was enrolled at the time of recurrence (45 at first recurrence, 30 at second recurrence, 16 at third recurrence, 19 at the fourth or more recurrence). The probability of bladder cancer death by time after progression is calibrated to data from the Dutch Cancer Registry 21. The probability of other causes of death is based on the USA life table for 2010 for the general population. The parameters used in the MISCAN model are described in Table 1 21, 23.

Table 1

The parameters used in the MISCAN model, based on de Bekker‐Grob et al. 21 and Zhang et al. 23

Parameter	Value
Recurrence time, years, mean	4.18 (exponential distribution)
Time until progression, years, mean	1.46 (exponential distribution)
Time until death after progression, years, mean	2.57
Sensitivity cystoscopy, %	98
Specificity cystoscopy, %	88
Costs of cystoscopy, $	168
Costs of TUR (after recurrence), $	1409
Costs of cystectomy (after progression), $	7997
Disutility of cystoscopy	0.025 for 1 month
Disutility of TUR	0.03 for 1 month
Utility in NMIBC	0.94
Utility in MIBC	0.80 for lifetime

$, American dollars.

Screening strategies

We used MISCAN to simulate five cohorts of 1 million patients aged 65, 70, 75, 80 and 85 years at diagnosis and start of surveillance. To each cohort we applied 16 surveillance protocols, which comprised all combinations of frequencies of 3 months, 6 months, 12 months or 24 months and surveillance durations of 2, 5 and 10 years, or lifetime: 80 surveillance protocols with variable age of diagnosis, intensity and duration were investigated, assuming 100% attendance to surveillance.

Cost‐effectiveness

For each surveillance protocol, we calculated the number of cystoscopies, recurrences, progressions, and QALYs gained for a lifetime horizon. The utility estimates to calculate QALYs were based on Zhang et al. 23. The total costs of each surveillance protocol were calculated using cystoscopy and intervention costs (TUR and cystectomy) from the literature 21. We used incremental cost‐effectiveness ratios (ICERs) as the ratio of incremental costs to incremental life‐years gained. An annual discount of 3% was used, and the threshold for the ICER was set on $100 000 (American dollars).

Sensitivity analysis

We performed one‐way sensitivity analyses, in which we varied the comorbidity status, disutilities, and costs. Comorbidity status was based on the life tables for the USA population having no comorbidity and severe comorbidity 24. The disutility of cystoscopy and TUR were increased to 0.1 for 1 month and the costs for cystoscopy, TUR and cystectomy were all lowered by 10%.

Results

The discounted QALYs gained per patient vs costs per patient are presented for all simulated surveillance scenarios in Fig. 2. The younger the patient was when entering a surveillance programme, the more QALYs gained. The duration of surveillance increased both QALYs gained and the associated costs. However, the difference in QALYs gained between a surveillance protocol running for 10 years and one running for a lifetime decreased as patients aged and was negligible for patients aged >80 years: e.g., screening every 3 months for 10 years or lifetime both resulted in 0.92 QALY gained amongst 85‐year olds and ~1.7 QALYs amongst individuals aged 80 years. The frequency of a surveillance protocol had a similar effect across starting ages.

Figure 2

The costs and QALYs gained per patient of all evaluated surveillance strategies. Both costs and QALYs are discounted. The colours of the symbols represent the ages at the start of the surveillance strategy and the symbols the different frequencies of surveillance (square every 3 months, diamond every 6 months, triangle every 12 months, and circle every 24 months).

In Table 2 only the cost‐effective strategies within each cohort are presented. Surveillance strategies that allow for more surveillance (in both ways – by more frequent cystoscopy or by longer monitoring) resulted in larger cost increases. The number of diagnosed recurrences increased with the number of cystoscopies. In contrast, the overall number of diagnosed cancer progressions followed a parabolic trend, where it increased with the duration of the surveillance protocol but started to decrease when more than one test per year was performed. This suggests that frequent surveillance can detect recurrences before they progress to muscle invasion. Detecting recurrences has a major effect on the costs of a protocol, as the treatment of progression is over five‐times as expensive as treatment for NMIBC.

Table 2

Effects and costs (in American dollars) of the cost‐effective strategies per age cohort. The number of screens, recurrences and progression, costs and QALYs gained are given per patient over a lifetime, without discount. In the calculation of the ICER a 3% discount rate was used

Start age, years	Frequency, months	Duration, years	Screens, n	Recurrence, n	Progression, n	Costs, $	QALYs	ICER, $
65	24	2	0.91	0.17	0.11	1 264	1.87	920
65	24	5	2.46	0.39	0.26	3 050	3.99	1 123
65	12	5	4.51	0.55	0.19	3 017	3.86	1 145
65	12	10	8.14	0.87	0.30	4 984	5.60	1 450
65	12	Lifetime	14.34	1.19	0.41	7 374	6.60	2 786
65	6	Lifetime	30.01	1.78	0.31	10 050	7.12	4 999
65	3	Lifetime	60.59	2.27	0.20	14 973	7.29	24 758
70	12	2	1.86	0.25	0.08	1 329	1.53	1 112
70	12	5	4.36	0.53	0.18	2 924	2.98	1 445
70	12	10	7.66	0.82	0.28	4 709	4.15	2 177
70	12	Lifetime	11.66	1.05	0.36	6 327	4.62	4 579
70	6	Lifetime	24.49	1.54	0.27	8 429	5.06	4 675
70	3	Lifetime	49.65	1.92	0.17	12 405	5.22	20 195
75	12	2	1.81	0.24	0.08	1 288	1.13	1 410
75	12	5	4.13	0.50	0.17	2 770	2.11	1 932
75	12	10	6.90	0.75	0.26	4 276	2.76	3 237
75	6	10	14.38	1.03	0.18	5 306	3.00	4 147
75	6	Lifetime	19.13	1.27	0.22	6 770	3.26	7 340
75	3	Lifetime	38.94	1.56	0.14	9 841	3.40	19 065
80	12	2	1.71	0.23	0.08	1 217	0.75	1 958
80	12	5	3.74	0.46	0.16	2 517	1.30	2 918
80	6	5	7.76	0.60	0.10	2 990	1.42	3 895
80	6	10	12.10	0.88	0.15	4 494	1.78	5 502
80	6	Lifetime	14.09	0.98	0.17	5 122	1.83	15 572
80	3	Lifetime	28.83	1.19	0.11	7 364	1.94	19 402
85	12	2	1.53	0.20	0.07	1 092	0.41	3 112
85	6	2	3.86	0.31	0.05	1 528	0.53	4 118
85	6	5	6.54	0.51	0.09	2 526	0.75	5 404
85	6	10	9.07	0.67	0.12	3 406	0.85	11 558
85	3	10	18.63	0.80	0.07	4 824	0.92	20 914

For example, 12‐monthly surveillance from the age of 65 years is expected to result in 6.60 QALYs gained per patient (without discount), whilst 6‐monthly surveillance from the same age gained an extra 0.52 QALYs at an increased cost of $10 050/patient. The corresponding ICER was $4999 (using a 3% discount). This small effect is achieved by more than doubling the number of surveillance cystoscopies.

The sensitivity analyses showed that the results were robust to the parameters varied (Appendix S1). Only in a few cases, a different strategy was amongst the cost‐effective strategies. Comorbidity status had the largest, although still limited, impact on the results.

Discussion

Our present findings suggest that cystoscopy‐based surveillance for NMIBC is cost‐effective for younger patients. However, we found that, even in the youngest cohort (65 years), lifelong quarterly cystoscopy was associated with high‐costs for modest potential QALY gains. Perhaps more importantly, within the context of the high, and increasing, median age at bladder cancer diagnosis, we found that the incremental yield of QALY gains for any increase in surveillance frequency and/or duration amongst patients aged ≥75 years was modest. Given the changing age structure of the population and the disproportionately high incidence rates in the oldest ages 25, the burden of bladder cancer in the elderly is projected to increase by 54% in the next 15 years 26, underscoring the public health impact of these findings.

In contrast to other urological cancers, in particular prostate cancer 27, age and competing risks have not, to date, been explicitly considered in bladder cancer guidelines. A multidisciplinary stakeholder group of clinicians, patients, payers and patient advocates identified cancer care for the elderly and post‐treatment surveillance as two of the top three cancers comparative effectiveness research priorities 28. Post‐treatment surveillance for NMIBC is uniquely invasive, intensive, and under current recommendations often lifelong. Given the rapidly growing and ageing population of patients with bladder cancer, it is critically important to define age‐based surveillance schedules for this group.

There are two other studies that evaluated different surveillance protocols. Overall, the results were comparable: for younger patients an intensive surveillance strategy was more cost‐effective than for older patients. The analysis of Zhang et al. 23 focused on low‐grade NMIBC, which accounts for about two‐thirds of all NMIBCs. International guidelines for surveillance were simulated and the number of cystoscopies and the QALYs were calculated for men and women separately. The authors concluded that patient‐specific factors such as the presence of comorbidity, or perception of utility loss from cystoscopy, should play an important role in determining the best surveillance protocol. Kent et al. 29 calculated the expected delay in detection of the next tumour based on several surveillance scenarios and concluded that the optimal surveillance schedule is less frequent for low‐risk patients than for high‐risk patients. In our present study, we considered the full spectrum of NMIBC, including the relatively higher risk subset with multiple recurrences, in contrast to Zhang et al. 23, and found relatively small incremental benefits of QALYs gained for more intensive or lengthy surveillance programmes amongst patients aged ≥75 years. As a result of advanced chronological age, patients with bladder cancer have a high burden of multiple chronic conditions 30 and the American Society for Clinical Oncology has called for more explicit consideration of these factors in the guideline development process 31. A previous analysis found that in patients with the highest rates of comorbidity, the optimal surveillance protocol was no surveillance 23. In our present analysis, even for patients with severe comorbidity surveillance could still be cost‐effective.

Strong points of the present study are that we included lifetime surveillance and evaluated the results over a lifetime horizon. Furthermore, we applied cost‐effectiveness analysis to determine the optimal surveillance protocol. This study also has several limitations.

We did not explicitly model heterogeneity in NMIBC. Surveillance can be stratified by pathology findings, treatment and recurrence history into low‐, intermediate‐ and high‐risk disease 9. Individuals with high‐grade tumours have a higher risk of recurrence and progression to MIBC and may require a more intensive surveillance protocol 5, 6. However, to reduce the burdens and costs of surveillance, patients at low‐risk of recurrence could follow a less intensive surveillance strategy 32.

Risk‐stratified surveillance has had slow uptake in the USA, with evidence of a mismatch between patient‐level risk and intensity of surveillance 33. The CEFUB cohort, which the model is based on, did not include the highest‐risk minority of cases with Grade 3 tumours. However, it reflects the majority of incident and prevalent cases, including both incident and recurrent cases, and is relevant for the subset where there is lesser consensus in surveillance guidelines. As such, our present analysis is one of the first to apply modelling to examine these areas of uncertainty, especially the questions of age, intensity and duration of surveillance. Whilst the inclusion criteria of the CEFUB trial reflected relatively lower grade, the fact that half of the cohort included recurrent tumours (with about half of those being multiply recurrent cases), the risks of progression in that cohort was nontrivial, higher than might be inferred from inclusion criteria of Grades 1/2 only. The inclusion in the present study population of newly diagnosed and recurrent cases reflects a broader spectrum of risk.

Another limitation is that we assumed that all patients with progression will undergo cystectomy; however, in practice, a proportion of patients will not undergo cystectomy. This will impact survival, QoL and costs, but the magnitude is unclear. Future work will apply the MISCAN model to these questions. Also, we did not consider adjuvant intravesical therapies and whilst these adjuncts may reduce risks of recurrence, implementation in real‐world practice is infrequent 34.

Finally, we did not stratify by gender. Bladder cancer is three‐times more common in men 1, and there is some uncertainty about the potential for varying recurrence and progression patterns by sex 35. Zhang et al. 23 found that women should be screened more intensively. However, currently the guidelines do not stratify recommended surveillance schedules by sex.

Another cost‐effective possibility to reduce the number of cystoscopies might be the utilisation of biomarker tests [e.g. fibroblast growth factor receptor 3 (FGFR3) mutation analysis] in voided urine samples 21, 36. A previous analysis concluded that the effectiveness of surveillance with the FGFR3 urine test was similar to that of surveillance entirely by cystoscopy, and the costs were substantially lower. The utility of such strategies, specifically amongst older patients where we found more modest benefits of regular cystoscopy, warrants further study.

In conclusion, various schedules for surveillance cystoscopy amongst patients with NMIBC are cost‐effective. However, recognising that bladder cancer has the highest median age at diagnosis amongst all cancer sites, our present finding of modest QALY gains amongst patients aged ≥75 years calls into question the utility of intensive or lengthy surveillance schedules amongst this large and growing subset of patients. Future studies should focus on surveillance stratified by age and competing risks, and evaluate the cost and QoL tradeoffs of innovative urine‐based biomarker tests.

Conflicts of Interest

Dr Nielsen has research funding from the National Institutes of Health, not directly related to the submitted work. Dr Nielsen is on the Medical Advisory Board for Grand Rounds and has research funding from the Patient‐Centered Outcomes Research Institute (PCORI). All other authors do not have conflicts of interest.

(non‐)muscle‐invasive bladder cancer

Cost‐Effectiveness of Follow‐up of Urinary Bladder Cancer

fibroblast growth factor receptor 3

incremental cost‐effectiveness ratio

MIcrosimulation SCreening Analysis

non‐muscle‐invasive urothelial cancer

quality adjusted life‐year

quality of life

transurethral resection

Appendix S1. Sensitivity analyses.

Click here for additional data file.