Evaluating the cost-utility of intravesical Bacillus Calmette-Guérin versus radical cystectomy in patients with high-risk non-muscle-invasive bladder cancer in the UK

Abstract

Aims

Approximately 75% of bladder cancer (BC) cases present as non-muscle-invasive BC (NMIBC). In patients with high-risk NMIBC, the mainstay treatment is intravesical Bacillus Calmette-Guérin (BCG), with immediate radical cystectomy (RC) as an alternative treatment option. The aim of the present study was to evaluate the cost-utility of BCG versus RC in patients with high-risk NMIBC from the UK healthcare payer perspective.

Materials and methods

A six-state Markov model was developed that covered controlled disease, recurrence, progression to muscle-invasive BC, metastatic disease, and death. The model included adverse events of BCG and RC and monitoring and palliative care. Drug costs were obtained from the British National Formulary. Intravesical delivery, RC, and monitoring costs were sourced from the National Tariff Payment System and the literature. Utility data were obtained from the literature. Analyses were run over a 30-year time horizon, with future costs and effects discounted at 3.5% per annum. One-way and probabilistic sensitivity analyses were performed.

Results

The base case analysis comparing BCG with RC showed that BCG would increase life expectancy by 0.88 years versus RC, from 7.74 to 8.62 years. BCG resulted in an increase of 0.76 quality-adjusted life years (QALYs) versus RC, from 5.63 to 6.39 QALYs. Patients incurred lower lifetime costs if treated with BCG (£47,753) than with RC (£64,264). Cost savings were mainly driven by the lower cost of BCG versus RC, and palliative care costs. Sensitivity analyses showed that results were robust to assumptions.

Limitations

The evidence base informing efficacy estimates of BCG is heterogeneous as different BCG administration schedules were reported in the literature, while incidence and cost data on some BCG-associated adverse events were sparse.

Conclusions

Intravesical BCG led to increased QALYs and reduced costs versus RC for patients with high-risk NMIBC from the UK healthcare payer perspective.

PLAIN LANGUAGE SUMMARY

Intravesical Bacillus Calmette-Guérin (BCG) is a recommended immunotherapy option for patients with high-risk non-muscle invasive bladder cancer (NMIBC) who have undergone transurethral resection of bladder tumor (TURBT). BCG is known for its high efficacy and good safety profile. However, current evidence on its effectiveness against other comparators such as radical cystectomy (RC) is limited, mainly because different BCG schedules are used, particularly with regard to maintenance therapy. Given this lack of evidence, we conducted the first cost-utility analysis which considers adequate BCG therapy relative to RC for the UK. We developed a Markov model that captured the effects of the intravesical BCG and RC on NMIBC, in addition to incidences of adverse events associated with either treatment. Costs of the two treatments, their administration, maintenance, and of treatments for adverse events were modelled alongside the quality-of-life effects of NMIBC, adverse events, and palliative care. We used published clinical data and UK cost data to inform our model. Our results show that BCG was associated with higher quality-adjusted life expectancy than RC and lower total costs from the healthcare system perspective. These results imply that adequate BCG immunotherapy is likely valuable for the National Healthcare System in terms of its effects on patients and indeed cost-saving relative to RC.

Keywords:

• High-risk non-muscle-invasive bladder cancer (NMIBC)
• Bacillus Calmette-Guérin (BCG)
• cost-utility
• cost-effectiveness
• National Health Service (NHS)
• radical cystectomy (RC)

JEL CLASSIFICATION CODES:

• C63
• D61
• I11

Introduction

Bladder cancer was the 12th most common cancer worldwide in 2020.¹ Non-muscle-invasive bladder cancer (NMIBC) accounts for between 70–75% of bladder cancers at the time of initial diagnosis,² and can be further classified into low-, intermediate-, high-, or very high-risk disease.³ Treatment guidelines for high-risk NMIBC recommend patients undergo transurethral resection of bladder tumor (TURBT), followed by intravesical immunotherapy with Bacillus Calmette-Guerin (BCG).^3,4 BCG is a live attenuated strain of Mycobacterium bovis that is administered into the bladder using a catheter and left in the bladder for 2 h, with weekly treatments for the first 6 weeks and, ideally, subsequent maintenance therapy, and with follow-up cystoscopies for the remainder of patient lifetimes.⁵ As an alternative to intravesical BCG, immediate radical cystectomy (RC) may be considered.^3,5,6 This surgical procedure involves the complete removal of the bladder and surrounding tissues, followed by the creation of an ileal conduit (urostomy) or (orthotopic) neobladder to allow storing and passing of urine.⁷

Intravesical BCG is the mainstay of NMIBC treatment.^3,8,9 Despite its acknowledged success, including relative to comparators such as mitomycin-C for intermediate-risk NMIBC,¹⁰ the evidence base supporting BCG is of mixed quality and often inconsistent. An important challenge is the use of different BCG schedules, specifically for BCG maintenance therapy. Differences in treatment schedules, as well as high rates of discontinuation, also make it difficult to assess if treatment failures are due to a lack of BCG efficiency or inadequate therapy, including insufficient TURBT quality, lack of repeat TURBT in T1 tumors, and absence of detrusor muscle in resected Ta high-grade specimen.^3,11,12 Definitions of “adequate BCG therapy” have been provided^13,14 and adequate BCG therapy has been shown to have excellent real-world outcomes.¹⁵ However, these definitions were too recent to be in widespread use in the literature at the time of writing.

Demonstrating the value of BCG to healthcare payers is therefore as challenging as it is important. The aim of the present study was the development of a cost-utility model for the evaluation of the clinical, cost, and quality of life effects of adequate BCG therapy in the treatment of high-risk NMIBC, from the perspective of the English National Health Service (NHS).

Methods

Markov model overview

A Markov model was used to evaluate the decision problem.¹⁶ The conceptualization of the model was informed by the European Association of Urology (EAU) guidelines on NMIBC³ and muscle-invasive bladder cancer (MIBC),¹⁷ respectively, and by prior health economic models.^18–20 Based on these recommendations, adequate BCG was the intervention of interest in the model. The comparator was immediate RC. As recommended by the EAU, all patients were assumed to have undergone TURBT initially to confirm the diagnosis of NMIBC and completely remove visible lesions,³ and the analysis focused on treatment with either BCG or RC after the initial TURBT.

Costs were evaluated from the perspective of NHS England (and ultimately therefore the perspective of the Department of Health and Social Care) and presented in 2022 Great British Pounds (£), using a 30-year time horizon and a 3-monthly cycle length for a population with a mean age of 70 years. Future costs and effects were discounted at 3.5% per annum.²¹

Markov model structure

The Markov model (Figure 1) was developed from treatment recommendations in EAU guidelines and informed by earlier models of NMIBC. Patients entered the Markov model in a controlled disease state in the first model cycle, in which their assigned treatment began. Patients could stay in the controlled disease state, transition to a post-cystectomy state following immediate RC, or transition to either a progression or a recurrence state. A death state was also included as an absorbing state, reachable from all other states in the model.

Figure 1. Markov model schematic. Regular Markov states are colored in blue and tunnel states are in grey. The “Death” state is accessible from every other state in the model and is itself an absorbing state. Immediate radical cystectomy is indicated by the arrow pointing from “Controlled disease” to “Post-cystectomy”. Abbreviations. CHT, Chemohyperthermia; MIBC, Muscle-Invasive Bladder Cancer.

Progression was defined as transitioning from NMIBC to MIBC.^18,20 This state was a one-cycle tunnel state, during which all patients, in line with EAU guidelines on MIBC, were assumed to undergo RC.¹⁷ Recurrence was understood to be the same or lower-stage disease as before the initial TURBT.²⁰ After cystectomy, patients entered a post-cystectomy state.

Model assumptions

Several key assumptions underpinned the model. No difference in efficacy or safety by BCG strain was assumed, in line with recent EAU guidelines,³ which implied that studies using different strains were used to inform different data points in the model. Failure of BCG within 6 months of initiating treatment was labelled as “BCG-unresponsive”.^13,14 Any recurrence beyond this point was not considered to indicate BCG-unresponsiveness and a second round of BCG – specifically, induction BCG – was possible. A proportion of patients eligible for RC after progression was assumed to refuse RC or not undergo surgery for other reasons,²² and it was assumed that these patients would receive chemohyperthermia with mitomycin C.

Model outcomes

Costs, life expectancy (in life years), quality-adjusted life expectancy (in quality-adjusted life years [QALYs]), and incremental cost-utility ratios (ICURs) were the main outcomes of the model. Net monetary benefit (NMB) was calculated using willingness-to-pay (WTP) thresholds of £20,000 and 30,000 per QALY gained, respectively.²³

Clinical data

The risks of recurrence, progression, and death in high-risk patients treated with adequate BCG were obtained from Matulay et al.¹⁵ Matulay et al. reported findings from a real-world US study, which included patients treated between 2004 and 2018 with adequate BCG as defined by the Food and Drug Administration (FDA) and the International Bladder Cancer Group.^13,14 One-year estimates for recurrence-free, progression-free, and overall survival from this study for high-risk patients, respectively, were converted to 3-month transition probabilities, assuming a constant rate of events (Table 1).²⁴

Table 1. Transition probabilities.

Clinical parameter	Risk	Data source
Baseline 3-monthly risks: BCG
Controlled disease -> MIBC	1.02%	^15
Controlled disease -> 1st recurrence	5.43%
Controlled disease -> Death	0.25%
Efficacy of repeat vs first BCG course	RR repeat vs first
MIBC	2.43	^26
Recurrence	0.76
Death	1.00
Additional, non-treatment-specific data
Increase in mortality after 1st recurrence	HR: 1.00	Assumption
Absolute survival benefit with NAC	Mean: 0.022	^17
CHT -> MIBC	0.82%	^29
CHT -> Metastatic disease	1.64%
CHT -> Death	1.63%
Not undergoing RC despite eligibility	61.82%	^22
Perioperative death with RC	2.80%	^78
Post-cystectomy -> Metastatic disease	0.88%	^27
Post-cystectomy -> Death	1.88%
Metastatic disease -> Death	9.36%	^28

Abbreviations. BCG, Bacillus Calmette-Guérin; CHT, chemohyperthermia; HR, Hazard ratio; MIBC, muscle-invasive bladder cancer; NAC, neoadjuvant chemotherapy; NMIBC, non-muscle invasive bladder cancer; RC, radical cystectomy; RR, relative risk.

The risk of disease following a first recurrence was affected only by the treatment given for the repeat recurrence. This was based on a finding by Simon et al.,²⁵ who observed no association between the number of recurrences and risk of progression in low- and intermediate-risk patients; in the present analysis, the progression risk following the first recurrence in high-risk disease was also assumed to depend only on the treatment given.

The efficacy of repeat BCG was calculated by applying risk ratios for second induction BCG versus first induction BCG from Daniels et al.,²⁶ which were not statistically significant. No difference was assumed for the effect on mortality. The risk of death following repeat recurrence was assumed to equal that following the first recurrence.

The risk of developing metastatic disease after cystectomy and the risk of death following RC were obtained, in the absence of UK data, from a cohort of patients with high-risk NMIBC who had undergone early RC at a French hospital between 2001 and 2020.²⁷ Over a median follow-up of 65 months, thirty-two patients developed metastases. Five-year overall survival was reported as 68.4%. The study did not report mortality specific to patients with metastatic disease so the corresponding transition probability was calculated from a UK cohort comprising 216 patients with locally advanced or metastatic disease, which reported median overall survival of 16.2 months and 1-year overall survival of 67.5%.²⁸

Transition probabilities following chemohyperthermia were derived from a study of hyperthermic-intravesical chemotherapy (HIVEC) in patients with high-risk NMIBC, which had a median follow-up of 18 months and reported rates of progression to muscle infiltration, metastasis, and death.²⁹

Mortality

For RC, the literature showed that the use of neoadjuvant chemotherapy was associated with an overall survival benefit of 5–8% over 5 years relative to not using neoadjuvant chemotherapy.^17,30 In the base case, the midpoint of this interval (6.5%) was used to model the mortality benefit of neoadjuvant chemotherapy.

Background mortality was modelled using published lifetables for the UK (2018–2020).³¹ General population mortality was adjusted using net survival for bladder cancer in patients aged 65–74 years from Public Health England.³² As net survival estimates were only available for up to 5 years after diagnosis, the last observation (for 5 years) was carried forward and assumed to apply to the remaining years.

Health-related quality of life data

Health-related quality of life data including utility and disutility values (Table 2) were sourced primarily from the BOXIT trial,³³ which provided a recent UK source for the target population (data in this trial pertained predominantly to high-risk patients).

Table 2. Utility and disutility values.

States and events	Value
State utilities
Controlled disease^33	0.85
Post-recurrence^33	0.76
Post-cystectomy^33	0.75
Post-chemohyperthermia [assumption]	0.75
Metastatic disease^34	0.62
Event disutilities
NMIBC recurrence^33	−0.083
Radical cystectomy^33	−0.168
BCG administration (per cycle with at least one administration)^48	−0.02
Adverse event disutilities
(Fat) Tissue necrosis of the surrounding area^38	−0.37
Abscess^38	−0.26
Allergic skin rash^37	−0.04
Anastomotic stricture (assumption based on^38)	−0.15
Arthritis, arthralgia^47	−0.05
Bacterial cystitis (assumption based on^38)	−0.17
BCG infection of implants and surrounding tissue^38	−0.26
Bladder perforation^46	−0.48
Bowel obstruction^38	−0.35
Conduit stones (assumption based on^42)	−0.09
Conduit stricture^38	−0.15
Contact dermatitis^37	−0.04
Cystitis (allergic) (assumption based on^38)	−0.17
Deep vein thrombosis^38	−0.23
Epididymo-orchitis (assumption based on^38)	−0.17
Frequency (assumption as no disutility data were identified)	−0.07
General malaise^39	−0.04
Generalized exanthema (assumption: as for allergic skin rash)	−0.04
Granulomatous pulmonary/hepatic reactions^38	−0.24
High fever (>38.5 °C)^38	−0.37
Metabolic acidosis^38	−0.37
Mucositis^41	−0.15
Myocardial infarction^36	−0.21
Necrotizing cystitis (assumption: as for tissue necrosis of the surrounding area)	−0.37
Neutropenic fever^41	−0.15
Neutropenic sepsis^38	−0.43
Palmar-plantar-erythema^41	−0.12
Parastomal hernia^35	−0.09
Pneumonia^38	−0.05
Pulmonary embolism^38	−0.28
Renal replacement therapy (mean values based on,^43 weighted by the number of people with each modality reported by^44	−0.22
Sepsis^38	−0.43
Severe BCG infection^38	−0.37
Stomal stenosis (assumption based on^38)	−0.15
Symptomatic granulomatous prostatitis (assumption based on^38)	−0.17
Transfusion, platelets^45	−0.10
Transfusion, red blood cells (assumption: as for platelet infusion)	−0.10
Upper tract urolithiasis (assumption based on^42)	−0.09
Urinary incontinence^38	−0.14
Urticaria^40	−0.13
Vascular infection^38	−0.26
Vesical fistula^38	−0.22
Wound infection, deep^38	−0.26
Wound infection, superficial^38	−0.26

Abbreviations. BCG, Bacillus Calmette-Guérin; NMIBC, non-muscle invasive bladder cancer.

Several values for post-cystectomy states were available from the literature but were either implausibly high or based on unclear evidence. Given these concerns, the post-cystectomy value was assumed to be equal to the post-recurrence value reported in the BOXIT trial.³³ This value was closely aligned with values used in prior economic models.^6,18

The health state utility value for metastatic disease was sourced from Kulkarni et al.,³⁴ as the value for metastatic disease responsive to chemotherapy, which was originally derived from the breast cancer literature.

The effects of adverse events on quality of life were captured by applying a one-off disutility for the entire duration of the cycle in which the event occurred. Disutilities came from a range of articles, including for conditions other than bladder cancer (Table 2).^35–47 This is in line with previous models in the field that also found a sparsity of relevant and directly applicable (dis-) utility data for adverse events associated with bladder cancer treatment.^18,48 To capture the quality-of-life impact of intravesical administration events, a utility decrement of 0.02⁴⁸ was applied in each cycle during which BCG was administered as either first-line therapy or for repeat induction.

Adverse events

The selection of adverse events for BCG was based on available data and the summary of product characteristics for intravesical BCG.⁴⁹ If cumulative frequencies rather than incidence rates were reported, they were converted to risks adjusted for the 3-month cycle length.²⁴

For RC and neoadjuvant chemotherapy, adverse events and their probabilities came from a model by Magee et al.⁵⁰ and the sources therein. A full overview of the adverse events included for each treatment option, as well as their respective probability estimates, can be found in the supplementary material (Tables S1–S3). Note that adverse events for BCG and RC were included regardless of their frequency, i.e. no minimum frequency was specified for inclusion, to allow capturing rare events.

Cost and resource use

The main source of costs for delivery of intravesical BCG and cystectomy was the 2022/2023 National Tariff Payment System.⁵¹ Costs of medications were captured from the NHS England perspective and, where necessary, inflated to 2022 values using the consumer price index for medical services.⁵²

BCG

BCG dosing was based on the SWOG schedule, which is currently recommended in the BCG treatment guidelines.⁵³ This schedule involves six once-weekly instillations as induction therapy. While the SWOG schedule specifies 3-year maintenance therapy for high-risk patients, the base case was modeled as 1-year maintenance therapy, given that in clinical practice maintenance therapy is usually given from 1 year up to 3 years and many patients fall on the lower end of this time span.^6,15 Maintenance therapy included three instillations each at 3, 6, and 12 months. A scenario analysis using a 3-year maintenance treatment, with an additional three instillations each at 18, 24, 30, and 36 months, was conducted as a post-hoc scenario analysis.

Intravesical BCG was assumed to be adequate, as per definitions advanced by the FDA¹³ and Kamat et al.,¹⁴ and in line with efficacy data from Matulay et al.¹⁵ that informed transition probabilities in the BCG arm of the model.

Costs for BCG were modelled based on the 2015 National Institute for Health and Care Excellence (NICE) guideline,⁶ including costs for delivery and drug acquisition. Drug costs came from the British National Formulary (BNF), and, in the base case analysis, BCG was priced at £150 per vial.

Per-instillation intravesical delivery costs for BCG were £163 per delivery.⁵¹ In line with the 2015 NICE guideline, the model incorporated a calculation of the average weighted delivery cost by specifying the proportion of patients undergoing outpatient versus combined day case/ordinary elective spell administration procedures. The proportions of patients undergoing such procedures were determined based on those outlined in the NICE guideline⁶ (57% day case).

Radical cystectomy

One RC was modelled for each patient transitioning into the post-cystectomy state. Additionally, and in line with NICE guidance⁶ and guidelines on the management of bladder cancer from the West Midlands Expert Advisory Group for Urological Cancer,⁵⁴ follow-up after RC was modelled as three consultations with a urologist, one cystoscopy per year, two blood tests (to test for metabolic acidosis, B12, and folate deficiency) per year, one glomerular filtration rate test per year, and three CT scans (guidance specifies to have CT scans 6, 12, and 24 months after RC⁵⁴ or annually as per the NICE guidance⁶).

RC was costed in line with the NICE 2015 guideline,⁶ with 2022/23 National Tariff costs by complication score weighted by 2019/2020 activity data from the NHS 2019–20 national schedule of costs (the latest version of these data to be made publicly available based on pre-COVID-19 data).⁵¹ This yielded a cost of £10,883.19. Costs also included follow-up procedures associated with RC and their respective costs (per procedure). These procedures comprised cystoscopy (£251), CT scan (£83), glomerular filtration rate (GFR) testing (£257), blood test (£3.26), and follow-up with a urologist/consultant (£90).

Neoadjuvant chemotherapy

In the base case, it was assumed that all patients were eligible for cisplatin-based neoadjuvant chemotherapy. Based on a previous study,⁵⁵ 16.5% of patients were modelled to receive cisplatin-based neoadjuvant chemotherapy, of whom 90% received gemcitabine plus cisplatin and 10% received accelerated M-VAC.⁶

A single cycle of gemcitabine plus cisplatin included 1000 mg of gemcitabine per square meter of body surface area (BSA) on days 1, 8, and 15, and 70 mg/m² of cisplatin on day 2.^6,56 Three cycles were modelled for neoadjuvant chemotherapy.⁶

A single cycle of accelerated M-VAC included 30 mg/m² methotrexate, 3 mg/m² vinblastine, 30 mg/m² adriamycin/doxorubicin, and 70 mg/m² cisplatin, all given on day 2 of a cycle.⁶ As for gemcitabine plus cisplatin, neoadjuvant chemotherapy with accelerated M-VAC included three cycles. Granulocyte colony-stimulating factors (G-CSF) were assumed to be used in 25.2% of patients undergoing neoadjuvant chemotherapy, to prevent febrile neutropenia.^55,57 For dosing based on BSA, a mean BSA of 1.71 m² for women and 1.79 m² for men was used, based on retrospective data for adult cancer patients in the UK.⁵⁸ The mean weights in kilogram for women (71.8 kg) and men (85.9 kg) as reported by the Office for National Statistics for 2019 to 2020 were used for weight-based dosing.⁵⁹

Palliative care

Some patients dying from the metastatic disease were assumed to have received palliative care. Although no data for the UK were identified, a US study reported that 4.1% of patients with advanced bladder cancer (T4, N+, M+) received palliative care,⁶⁰ and this value was used in the base case analysis.

In line with the NICE guidance on bladder cancer,⁶ palliative care was assumed to consist of 3.1 admissions, on average, each with an average length of stay of 11.4 d. The entire cost of palliative care was applied to 4.1% of patients upon transitioning from metastatic disease to death.

Costs of palliative care were calculated based on a daily cost obtained from the Unit Costs of Health and Social Care 2021,⁶¹ which reported an average inpatient specialist palliative care cost of £419.98 per bed day.

Costs associated with adverse events and additional costs

Costs associated with the treatment of adverse events and additional costs came from a variety of documents, primarily the 2022–2023 NHS tariff.⁵¹ Where multiple tariff codes were available for different degrees of comorbidity, average costs across these comorbidity categories were calculated as averages weighted by activity in each category as per activity data from the NHS 2019–20 national schedule of costs. A full description of these costs can be found in Table S4.

Sensitivity analysis

One-way and probabilistic sensitivity analyses (PSA) were conducted to evaluate the robustness of the model and the base-case results. Key model parameters were sampled during PSA to investigate the effects of uncertainty surrounding the model inputs; PSA consisted of 10,000 Monte Carlo iterations. For the one-way sensitivity analyses, individual model parameters were altered to investigate the magnitude of their effect on model outcomes, and a tornado diagram was used to summarize NMB outcomes.

Results

Base case

The base case analysis comparing BCG with RC showed that BCG would increase life expectancy by 0.88 years versus RC, from 7.74 to 8.62 years (Table 3). BCG also resulted in an increase of 0.76 QALYs versus RC, from 5.63 to 6.39 QALYs, yielding an NMB of £31,748 for a WTP threshold of £20,000 and of £39,366 for a WTP of £30,000 per QALY gained. Patients treated with BCG incurred lower costs than those treated with RC, with costs of £64,264 relative to £47,753 over patient lifetimes. Cost savings were mainly driven by the lower cost of BCG versus RC, with palliative care also leading to further cost savings (£28,638 for RC and £18,713 for BCG). A full breakdown of costs for each treatment can be found in Table 3. Based on these results, BCG was the dominant intervention, improving quality-adjusted life expectancy and reducing costs versus RC.

Table 3. Base case results and cost breakdown.

	Life expectancy, years	Quality-adjusted life expectancy, QALYs	Costs, £	Incremental cost-utility ratio, £per QALY gained	Net monetary benefit, £ (for a WTP threshold of £20,000 per QALY gained)
Radical cystectomy	7.74	5.63	64,264 (RC + NAC: 12,850; Chemotherapy for metastases: 10,883; Palliative care: 28,638; Monitoring: 9797; Adverse events: 2096)	–	–
BCG	8.62	6.39	47,753 (BCG: 5386; RC + NAC: 5608; Chemotherapy for metastases: 7842; Palliative care: 18,713; Monitoring: 8573; Adverse events: 1631)	–	–
Incremental	0.88	0.76	−16,512	BCG is more effective and less costly	31,748

Abbreviations. BCG, Bacillus Calmette-Guérin; NAC, Neoadjuvant Chemotherapy; QALYs, Quality-Adjusted Life Expectancy; RC, Radical Cystectomy; WTP, Willingness-to-Pay Threshold.

Probabilistic sensitivity analyses

In most PSA iterations, ICURs fell within the south-eastern quadrant of the cost-utility plane, indicating that the use of BCG resulted in an increase of QALYs and cost savings (Figure 2). At all WTP thresholds between GBPP 20,000 and £100,000, there was a 100% likelihood of adequate BCG therapy being considered a cost-effective strategy.

Figure 2. Cost-utility scatterplot. Blue circles represent combinations of incremental costs and quality-adjusted life-years from 10,000 Monte Carlo iterations. The orange diamond represents the mean value across these 10,000 iterations.

One-way sensitivity analysis

For the one-way sensitivity analyses (OWSA), the ten parameters with the largest impact on the final NMB are presented in Figure 3. The relative risk for a first recurrence following repeat induction relative to a first course of BCG was the model parameter associated with the greatest variation in NMB when compared with the original base case results. Specifically, when the risk of a recurrence following repeat induction of BCG was three times higher than after a first course, the NMB decreased by more than £8000 while a relative risk of 0 was associated with an NMB increase of more than £10,300. Other relatively influential parameters were the relative risk for muscle-invasive disease following a repeat versus first induction with BCG and the probability of transitioning from the post-cystectomy state to the metastasis state. Notably, the NMB remained positive, favoring BCG over RC, in all one-way sensitivity analyses.

Figure 3. Tornado diagram for one-way sensitivity analyses. The ten most influential drivers of cost-utility results are presented. Abbreviations. CHT, Chemohyperthermia; MIBC, Muscle-Invasive Bladder Cancer; NMB, Net Monetary Benefit; RC, Radical Cystectomy; RR, Relative Risk; TP, Transition Probability.

A BCG maintenance period of 3 years was associated with cost savings of £29,000 relative to RC and NMBs of £62,453 and £79,179 at willingness-to-pay thresholds of £20,000 and £30,000 per QALY gained, respectively.

Discussion

The present study sought to determine the cost-utility of adequate BCG therapy in comparison to RC for treating patients with high-risk NMIBC. To our knowledge, this is the first study that attempts to evaluate the cost-utility of BCG in a patient population where only adequate BCG therapy has been administered.

The analysis showed that adequate BCG is cost-effective compared to RC in patients with high-risk NMIBC. Specifically, BCG resulted in 0.76 additional QALYs versus RC, while reducing costs over patient lifetimes. Cost savings with BCG versus RC were driven by factors such as acquisition and palliative care costs. The PSA demonstrated that, at all WTP thresholds between £20,000 and £100,000, adequate BCG therapy was highly likely to be considered cost-effective, with most PSA iterations suggesting that BCG would lead to cost-savings and QALY gains when compared to RC. The OWSA showed that patients not undergoing RC despite their eligibility had the greatest impact on the final NMB.

There were several limitations to this study, with the first being that there were limited relevant health outcomes data available in the literature, particularly concerning adequate BCG therapy. Given these limitations, assumptions were made concerning the applicability of similar data, considering the treatment setting and the risk group of patients. For example, where UK study data were unavailable, results from US studies informed proxy clinical inputs for the economic model, and, in the case of progression following prior recurrence, no data specific to high-risk NMIBC could be identified so data from intermediate-risk NMIBC was assumed to apply. This is particularly relevant for the clinical efficacy data, which were taken from a single-center US study¹⁵ owing to the current sparsity of data on the efficacy of adequate BCG therapy for high-risk NMIBC. It should be noted that the real-world recurrence and progression data used for the analysis are closely aligned with those for standard frequency BCG administration from the European multi-center NIMBUS trial in patients with high-grade NMIBC.⁶² While there are some differences between adequate therapy and the reference therapy used in the NIMBUS trial, the comparability of results, which point to the excellent efficacy of induction followed by maintenance BCG therapy, suggest that the clinical efficacy estimates informing the present analysis can be considered robust. However, the use of UK-specific data, once available, would likely be preferable.

Similar to clinical efficacy data, if data for the risks surrounding the development of adverse events were unavailable (e.g. vascular infections from BCG therapy), assumptions and calculations were made. These calculations were based on data such as the number of cases reported in the literature, as well as incidence categories outlined in the summary of product characteristics.⁴⁹ Although these calculations were based on data on the intervention in question, a risk of incorrect estimation remained for the final figures obtained, which affects the generalizability of the results.

A second limitation was that several assumptions were made regarding the incorporation of adverse events, due to the limitations of the Markov model design. An example of this was incorporating multiple CT scans and follow-ups with a consultant during the first year after treatment with RC. Guidance for CT scans states that patients should receive scans at 6, 12, and 24 months after RC, or three annually according to NICE guidelines.⁶ However, the Markov model structure could not fully account for time-in-state. As a result, an informed assumption was made to include three CT scans and three consultant follow-ups for the first year following RC, based on the existing literature. The specific assumptions were made to best reflect the most important aspects of patient experiences concerning treatment with RC, and subsequent health outcomes and costs. However, these figures also potentially represent an underestimation of the true number of scans and follow-ups (and therefore costs) that RC patients may experience in their first-year post-treatment. Consequently, the final cost-utility results may in turn represent an underestimation of the final cost savings associated with BCG.

A potential further limitation is that data on the clinical outcomes of adequate BCG therapy were taken from a study conducted at a high-volume reference center in the US, which may not reflect standard treatment.¹⁵ However, no other data source of comparable quality and sample size could be identified so the data by Matulay et al.¹⁵ were used despite concerns about their potential generalizability. It is noteworthy that the European NIMBUS trial reported comparable results,⁶³ which may also not be considered fully generalizable to standard care given the clinical trial setting but should alleviate concerns regarding country-specific differences.

This study has highlighted limitations in the BCG outcomes evidence base, particularly regarding adequate BCG therapy. Although the clinical effectiveness and safety of BCG have been well-established, the clinical literature on the effectiveness and safety of BCG is heterogeneous in terms of the BCG dosing schedules used, the populations under investigation and the outcomes reported. These differences can be linked to shortcomings in therapy based on how BCG is used in clinical practice, in particular the inconsistent use and scheduling of maintenance therapy. While existing clinical guidelines describe what constitutes adequate BCG therapy, clinical practice has been found to deviate frequently from guideline recommendations.^11,64,65 This non-adherence to guidelines can be considered one of the major residual problems in NMIBC care and is compounded by challenges in completing therapy, including a shortage of BCG and, recently, treatment interruptions due to the COVID-19 pandemic.^66–68 The BCG shortage has led to rationing or foregoing treatment with BCG and investigations of reduced BCG dosing. Additionally, alternatives to BCG are now being considered due to supply shortages, which is already reflected in recommendations to consider using intravesical chemotherapy in intermediate-risk NMIBC patients.^3,69

A further limitation is the lack of comparative data for BCG relative to RC. Such data will likely be difficult to generate, however, as patient numbers available for RC are generally small – while NICE guidance and the NHS specify that immediate RC is an alternative treatment option for high-risk NMIBC,^5,6 clinicians and patients may prefer to not use RC early in the disease course and consider RC only for muscle-invasive disease. In a recent trial of photodynamic diagnosis-guided TURBT in patients with intermediate- and high-risk NMIBC from the UK, less than 5% of all patients underwent immediate RC.⁷⁰ Similar findings were obtained by the BRAVO-Feasibility study, which compared patients with high-risk NMIBC treated with maintenance BCG and RC in the UK.⁷¹ The study authors found recruitment to be challenging – many patients expressed a treatment preference, in most cases in favor of BCG over RC, and 20% of patients assigned to RC withdrew from the study as they favored BCG (conversely, only 8% of patients assigned to BCG withdrew due to a preference for RC). Note, however, that the more limited use of immediate RC in the UK does not affect the comparative health economic argument made in the present study, with its focus on value for money.⁷²

Limited evidence is available for the quality-of-life impact of BCG relative to RC. Results on quality of life in the BRAVO-Feasibility were underpowered, but the quality of life outcomes at 12 months were similar for BCG and RC after a rebound in quality of life from 3 months with RC, while treatment decision regret was higher in the BCG relative to the RC arm.⁷¹ Observational studies^73,74 and reviews⁷⁵ also reported that quality of life may recover within 12–24 months after RC but stated that sexual function and body image may not fully recover, or recover more slowly, and remain below values in the general population. Additional evidence to fill current gaps in patient-reported outcomes and treatment perceptions in NMIBC would be highly welcome and might contribute to improved communication with patients regarding their concerns around their disease and potential treatment sequelae.^71,76,77

While our study demonstrates that adequate BCG therapy is a cost-effective treatment option relative to immediate RC for NMIBC in the UK, current real-world practices show that inconsistent use of BCG and non-adherence to treatment guidelines remain important barriers to effective therapy. For the NHS to see maximal utility gains from adequate BCG therapy, more work is required to ensure uniformity of dosing and scheduling, as well as improve patient adherence.

Conclusion

Adequate intravesical BCG led to QALY gains and reduced costs for patients with high-risk NMIBC from the perspective of NHS England when considering RC as a comparator. Further work investigating the effects of adequate BCG treatment on patient health outcomes could be beneficial in strengthening the evidence base for future cost-effectiveness studies.

Transparency

Author contributions

KGH, CH, JP, and RFP conceived and designed the analysis. JP developed the simulation model and conducted the statistical and health economic analyses in collaboration with RFP. WA prepared the first draft of the manuscript, which was revised critically for intellectual content by all authors. All authors approved the final version to be published. All authors agree to be accountable for all aspects of the work.

Reviewer disclosures

A reviewer on this manuscript has disclosed that they have undertaken paid consultancy for a medical device manufacturer that operates within the bladder cancer field. Their technology, however, is of no direct relevance to either BCG treatment or radical cystectomy. The other peer reviewers on this manuscript have no other relevant financial relationships or otherwise to disclose.

Previous presentations

The model was previously presented at ISPOR Europe 2022 (Vienna, November 6 to November 9, 2022) as poster EE563.

Acknowledgements

None reported.

Declaration of funding

medac funded the development of the health economic model and analysis, the preparation of the manuscript, and the article processing charge for the manuscript.

Disclosure statement of financial/other relationships

KGH and CH are full-time employees of medac GmbH, which manufactures and markets a BCG strain. RFP is a full-time employee, director, and shareholder in, and JP and WA are full-time employees of, Covalence Research Ltd, which received consultancy fees from medac to develop the Markov model and analysis and prepare the manuscript. IO has no conflicts of interest to disclose.