Effects of Bladder Cancer on UK Healthcare Costs and Patient Health-Related Quality of Life: Evidence From the BOXIT Trial

Background Limited evidence exists regarding the cost and health-related quality of life (HRQoL) effects of non–muscle-invasive bladder cancer (NMIBC) recurrence and progression to muscle-invasive bladder cancer (MIBC). We examined these effects using evidence from a recent randomized control trial. Material and Methods The costs and HRQoL associated with bladder cancer were assessed using data from the BOXIT trial (bladder COX-2 inhibition trial; n = 472). The cost and HRQoL effects from clinical events were estimated using generalized estimating equations. The costs were derived from the recorded resource usage and UK unit costs. HRQoL was assessed using the EQ-5D-3L and reported UK preference tariffs. The events were categorized using the TMN classification. Results Cases of grade 3 recurrence and progression were associated with statistically significant HRQoL decrements (−0.08; 95% confidence interval [CI], −0.13 to −0.03; and −0.10; 95% CI, −0.17 to −0.03, respectively). The 3-year average cost per NMIBC patient was estimated at £8735 (95% CI, 8325-9145). Cases of grade 1, 2, and 3 recurrence were associated with annual cost effects of £1218 (95% CI, 403-2033), £1677 (95% CI, 920-2433), and £3957 (95% CI, 2332-5583), respectively. Progression to MIBC was associated with an average increase in costs of £5407 (95% CI, 2663-8152). Conclusion Evidence from the BOXIT trial suggests that patients with NMIBC will both experience decrements in HRQoL and incur significant costs, especially in the event of a grade 3 recurrence or a progression to MIBC.


Introduction
Bladder cancer is the 9th most common cancer and ranks 13th in terms of cancer-associated mortality worldwide. 1 In the United Kingdom, bladder cancer accounts for 3% of all new cancer cases, with an estimated 10,171 new cases diagnosed in 2015. 2 Clinically, the lesions will be stratified using the TMN classification, with nonemuscle-invasive bladder cancer (NMIBC) classified as stage Tis, Ta, and T1 and muscle-invasive bladder cancer (MIBC) as stage T2, T3, and T4. This distinction is important because the involvement of cancer invading the muscle carries a significantly worse prognosis and requires radical cystectomy, radical chemotherapy, or radical radiotherapy, with or without neoadjuvant chemotherapy. NMIBC has had more favorable survival rates but recurs frequently and has been associated with repeated outpatient visits, cytologic and cystoscopic monitoring, and adjuvant intravesical treatment regimens after transurethral resection.
In the European Union, it has been estimated that the treatment of bladder cancer costs V4.9 billion, representing 5% of the total healthcare cancer costs. 3 In the United States, bladder cancer has been the most costly cancer to treat on a per patient basis. 4,5 Having estimates of the cost and health-related quality of life (HRQoL) effects of clinical events related to bladder cancer is important as a means of understanding its burden, informing resource allocation decisions, and aiding further research. However, current evidence on such effects has been limited in several ways. First, the distinction between NMIBC recurrence and progression to MIBC has often been overlooked. [5][6][7][8] Second, HRQoL studies have predominantly focused on treatment-specific effects [6][7][8][9] and have not sought to understand the HRQoL effects of specific clinical events such as recurrence and progression. Third, systematic reviews have repeatedly criticized the internal validity of HRQoL analyses, commonly citing the use of retrospective or cross-sectional designs, nonvalidated instruments, short time horizons, and failure to adjust for confounders. [7][8][9][10][11] Finally, a paucity of UK-specific cost analyses have been reported.
The present study aimed to estimate the expected costs and HRQoL of patients with a diagnosis of NMIBC and evaluate the effects associated with NMIBC recurrence and progression to MIBC. Our study used evidence from a recent randomized controlled trial of patients with intermediate-and high-risk bladder cancer, the BOXIT trial (bladder COX-2 [cyclo-oxygenase-2] inhibition trial).

BOXIT Trial
The BOXIT trial (ISRCTN registry no. ISRCTN84681538; Cancer Research UK no. CRUK/07/004) was a randomized phase III placebo-controlled trial that evaluated the addition of celecoxib to standard treatment for patients with NMIBC and an intermediate or a high risk of recurrence. From 2007 to 2012, 472 patients with transitional cell carcinoma NMIBC were recruited. The patients had a mean age of 65.9 years, and most patients were men (79%). The median follow-up at the point of analysis was 44 months (interquartile range, 36-57 months). The trial found no clear treatment benefit for celecoxib, with no significant differences in the interval to the first recurrence of bladder cancer (ie, NMIBC or MIBC) between patients randomized to either celecoxib or placebo for 2 years. Further details of the study design, treatment schedules, patients, and clinical results from the trial have been previously reported. 12

Clinical Events
At trial entry, cases of intermediate-and high-risk NMIBC were defined according to the clinicopathologic features outlined by the European Association of Urology 2002 guidelines. 13 The clinical events of interest during the trial were NMIBC recurrence and progression to MIBC. The grade and stage of NMIBC and MIBC were classified using the World Health Organization TNM classification. 14 Patients could have experienced > 1 recurrence of NMIBC during the follow-up period. Disease progression was defined as the development of MIBC (stage ! pT2). Intermediateand high-risk patients were recommended to undergo single adjuvant intravesical mitomycin C. The intermediate-risk patients were recommended to undergo 6 cycles of once-weekly adjuvant intravesical mitomycin C, and high-risk patients were recommended to undergo induction bacillus Calmette-Guérin with maintenance therapy for 3 years in accordance with international guidelines. 15,16 Surveillance cystoscopy was performed at 3-month intervals for the first 2 years and then every 6 months for the third and fourth years. In the present report, we focused on the first 3 years of follow-up.

HRQoL, Resource Use, and Cost Data
HRQoL was measured using the EQ-5D-3L, a generic, preference-based measure encompassing 5 dimensions of health (ie, mobility, self-care, usual activities, pain or discomfort, anxiety or depression) and an overall health rating, measured using a visual analog scale. 17 The HRQoL values were generated using reported UK preference "tariffs" for the 243 health states described in the EQ-5D-3L. 18 The values range from 1.0 (perfect health status) to À0.594, with 0 indicating death and negative values reflecting health states considered to be worse than death. 19 The 346 high-risk patients in the trial completed scheduled EQ-5D self-assessments at baseline (trial entry) and at 2, 3, 6, 12, 24, and 36 months of followup. The 126 intermediate-risk patients completed scheduled EQ-5D self-assessments at baseline and at 12, 24, and 36 months of follow-up.
The cost analysis used resource use data from questionnaires collected during the trial and took the perspective of the National Health Service and personal social services. The relevant resources used were those related to the diagnosis, treatment, and 3-year follow-up data of the patients included in the BOXIT trial. These included endoscopic investigations, together with the primary, secondary, and palliative care data, and the therapeutic procedures used, including radical cystectomy, chemotherapy, radical radiotherapy, immunotherapy regimens, and intravesical therapy regimens. Missing information relating to the quantity or specific type of treatment administered after a clinical event was assumed to follow usual practice. The unit costs were obtained from a variety of sources (Supplemental Table 1 available in the online version) and inflated to 2017 prices. 20 The costs of inpatient visits were determined using a fixed component relating to the first 2 days of stay, with a marginal component related to any additional days. Care was assumed to have been elective, unless stated otherwise. The total costs were aggregated into years after baseline, with each year estimated by multiplying the number of resources consumed during that period by their respective unit costs and summating.
The HRQoL analysis set consisted of the 316 high-risk patients who had fully completed ! 1 EQ-5D questionnaire(s) during the trial. The focus on high-risk patients was to use the most EQ-5D data available and provide the most interpretable estimates of effects, given the small number of MIBC and grade 3 NMIBC events in the intermediate-risk patients and the different EQ-5D follow-up schedules for the 2 risk groups. An analysis that included both risk groups with annual EQ-5D follow-up data was performed as a secondary analysis.

Analysis Methods
The standard approach used to analyze HRQoL and cost data from clinical trials has been to compare these between treatment arms over time to calculate the quality-adjusted life-years (QALYs) and total costs for each patient in the trial. 21 For trials showing no clinically or statistically significant benefit from a new treatment, this method will have little value. However, such trials offer a means of estimating the costs and HRQoL associated with a disease. This can include an exploration of how the HRQoL and costs vary between patients and how the patient characteristics and the clinical events they experience could explain some of this variation. 22,23 Such analyses can provide valuable information to those assessing the potential value of other new treatments for similar patients. 24 In the present study, 2 forms of analysis were conducted for both costs and HRQoL. The first analysis was descriptive, with the mean EQ-5D scores calculated at each follow-up period of interest, and the mean costs calculated annually. The patients were grouped in accordance with the types of events experienced during the 3-year follow-up period. The costs were categorized into resource-related groups for comparison. The second analysis was used to establish the effects of an event (ie, NMIBC, MIBC) on each outcome measure. The patients' clinical events were linked to their closest post-event assessment. If multiple NMIBC recurrences had occurred between the EQ-5D or cost assessments, the recurrence with the highest grade was recorded. The effects of the events on the HRQoL and costs were computed using repeated measures regression, controlling for the relevant baseline covariates chosen on the basis of clinical relevance. These included: baseline HRQoL, randomized treatment, history of bladder cancer, patient characteristics (ie, age, body mass index, gender, diabetes), follow-up year, risk group, and interaction terms, as appropriate.
To evaluate the HRQoL and costs, separate generalized estimating equation models were implemented in accordance with reporting guidelines. 25,26 The model fit, comparison, and selection of the working correlation structure was performed using the quasilikelihood information criterion. 27,28 Dependent variables of the annual costs and EQ-5D scores were assumed to follow the gamma and normal distributions, respectively.

Patient Characteristics and Events
Patients experiencing disease recurrence and progression had characteristics similar to those of the patients without disease recurrence and progression. However, modest differences in the rates of diabetes and a history of NMIBC were noted (Table 1). We assessed whether systematic differences were present between patients with and without missing EQ-5D data at different follow-up points and found that the differences were small (Supplemental Tables 2 and 3 available in the online version). This finding supported the assumption from our complete case analysis that the occurrence of missing data was completely at random.

HRQoL Analysis
The completion rate of the EQ-5D during the 3-year period was 79% (range, 58%-84%) across the points of follow-up. The completion rates after a NMIBC recurrence and progression to MIBC were 60% and 38%, respectively. An overview of the observed mean EQ-5D index scores for high-risk patients and the proportion of events occurring between each EQ-5D follow-up period are presented in Figure 1. Full details of the HRQoL descriptive results are provided in Supplemental Tables 5 and 6 (available in the online version).
The set of subgroups comprising patients who had experienced ! 1 of the specified clinical events during the 3-year follow-up period or had experienced no event is presented in Figure 1. These findings suggest that NMIBC recurrence and MIBC progression could be associated with deterioration in HRQoL at specific points. The variation in HRQoL at specific follow-up points was largely driven by the events experienced by the patients. In contrast, the variation in HRQoL between the followup points was related to the underlying within-patient variations, the nonuniform distribution of events over time, and the sampling error, exacerbated by partitioning modestly sized subgroups. A comparison of the EQ-5D dimensions stratified by the event-related subgroup found greater proportions of patients reporting problems with pain or discomfort and undertaking usual activities when experiencing a grade 3 recurrence or MIBC progression compared with no events during the 3-year follow-up period (Supplemental Figure 1 available in the online version). The statistically significant effects of clinical events on HRQoL in terms of the estimated decrements and mean health-state values are listed in Table 2. Progression to MIBC and NMIBC grade 3 recurrence were associated with predicted mean decrements in HRQoL of À0.10 (95% confidence interval, À0.17 to À0.03) and À0.08 (95% confidence interval, À0.13 to À0.03), respectively (P < .01). In contrast, the recurrence of NMIBC grade 1 and grade 2 was associated with positive, but statistically insignificant (P > .1), increments in HRQoL compared with patients without cancer.
The secondary analysis showed that introducing an interaction term into the regression analysis revealed that patients with grade 3 NMIBC recurrence in the first year experienced larger decrements in HRQoL (À0.11) compared with those with recurrence in subsequent years (À0.04). The small numbers precluded the same analysis for MIBC progression. Including both high-and intermediate-risk patients in the analysis using only the annual EQ-5D assessment data generated NMIBC recurrence estimates closer to 0 for all grades, with only MIBC events resulting in a statistically significant decrement in HRQoL (P < .05). Irrespective of the bladder cancer grade or stage, radical cystectomy was associated with a À0.17 decrement in HRQoL. All regression results and primary varianceecovariance matrices are presented in Supplemental Tables 7 to 13 (available in the online version).

Cost Analysis
The mean costs per patient for each type of care (Supplemental Table 1 available in the online version), annually and in total, are reported in Figure 2. The mean cost of treatment for a patient with

Edward Cox et al
Clinical Genitourinary Cancer August 2020 -e423 NMIBC was £4854 in the first year, with a total cost of £8735 over three years. These results suggest that the costs decline over time, with mean costs of £1496 in year 3. Endoscopic surveillance was the principal cost driver, accounting for > 52% of the total costs and representing a high proportion in years 2 (£1384 of £2386) and 3 (£835 of £1496). These estimates resulted in a 3-year total cost for the UK NMIBC bladder cancer cohort diagnosed in 2015 at £66.14 million, assuming that 74.5% of the 10,171 UK bladder cancer cases were NMIBC. 2,29 The effect of the clinical events on annual costs is shown in Figure 3, which indicated that MIBC progression and all grades of NMIBC recurrence led to increased costs. Higher grades of NMIBC were associated with higher costs, with grade 3 recurrence events necessitating more intensive therapy and closer surveillance. Progression to MIBC was associated with the greatest cost increment, with a £5407 increase in the expected annual cost per patient, again reflecting the more intensive therapy. Additionally, the treatment and surveillance of high-risk patients were associated with a £1968 increase in mean costs in the first year, although the costs had declined to £457 and £74 in years 2 and 3, respectively. The predicted mean costs per patient by year, event status, and risk group are shown in Table 3.

Discussion
Published economic evaluations of treatments for bladder cancer have lacked robust estimates of clinical effects on HRQoL and costs. 30,31 Furthermore, clinicians should understand the consequences of clinical events on patients' well-being and the health service costs. The present study has provided new evidence on the costs and HRQoL associated with NMIBC occurrence, recurrence, and progression to MIBC, supporting future clinical and economic evaluations. Our findings suggest that NMIBC will have an average cost of £8735 during a 3-year period, with cases of grade 1, 2, and 3 NMIBC recurrences and progression to MIBC associated with £1218, £1677, £3957, and £5407 increases in annual costs, respectively. In addition, grade 3 recurrence and progression to MIBC were associated with statistically significant decrements in HRQoL (À0.08 and À0.10, respectively).
Singer et al 32 reported that patients with bladder cancer, whether muscle invasive or not, will experience significant and clinically relevant deteriorations in HRQoL. Little evidence has contradicted the idea that patients with MIBC will experience a significant health burden; however, the same cannot necessarily be said for those with NMIBC. The commonly reported NMIBC morbidities have included mental health effects at diagnosis, physical discomfort, sexual problems, and urinary symptoms. [33][34][35] However, these have rarely translated into reductions in longer term health outcomes and, in some cases, have not been recorded at all. 9,36 It has been suggested that patients might become "accustomed" to NMIBC and its related management, accepting recurrences as a part of their lives. 10 The evidence presented from the BOXIT trial has offered some additional support for this view, but suggests that not all cases of NMBIC recurrence should be considered equal. Based on the recommended NMIBC surveillance guidelines, our results suggest that the negative effect of a NMIBC recurrence on HRQoL will be concentrated within the high-grade strata (grade 3), especially in the first year after the diagnosis. Furthermore, no evidence of negative HRQoL outcomes from grade 1 or 2 NMIBC recurrences was found. This might, at least in part, be explained by the low rates of radical surgery observed for grade 1 and 2 NMIBC recurrences. The results from supplementary analyses have supported these findings, with the use of cystectomy a large and significant predictor of HRQoL status. In addition, the patient groups with the highest rates of radical surgery (for grade 3 recurrence and progression) were most likely to report related problems with pain or discomfort and undertaking usual activities. A fuller understanding of the mechanisms behind these findings requires further prospective research.
Sangar et al, 37 estimated that the UK cost in 2001 to 2002 for the diagnosis, treatment, and 5-year follow-up of each bladder cancer case was £55.39 million, at a mean cost of £8349.20. Allowing for inflation and the different follow-up periods, their results are similar to those from the present study. To put this into context, it would be less costly per patient to treat stage 2 colon, rectal, and nonesmall-cell lung cancer in the United Kingdom. 38 The results from our analysis compliment those from the earlier study, showing the prominent role of endoscopic surveillance in driving the costs, which has remained the primary target for innovation in bladder cancer management. 5,39,40 Optimizing surveillance has also remained a research priority. Less costly and noninvasive urinary biomarkers represent an attractive option; however, to date, no commercially available test has the diagnostic accuracy to replace cystoscopy because patients and physicians require a test with high sensitivity before widespread acceptance. [41][42][43] Similar to others, we found that progression to MIBC will be associated with higher costs for intermediate-and high-risk patients. 44 The relatively large sample size, prospective study design, and the use of a validated HRQoL instrument represent the strengths of the present study. To the best of our knowledge, this is the first study to estimate both the mean and the marginal HRQoL and the cost effects across multiple grades and stages of bladder cancer. However, the present study had several important limitations. Despite the BOXIT protocol remaining representative of current UK guidelines (other than celecoxib treatment), differences between the BOXIT trial and current clinical practice have occurred (eg, the European Association of Urology now recommends bacillus Calmette-Guérin instillations for intermediate-risk patients and have revised the definitions of risk 45 ). In addition, the trial's exclusion criteria could have limited the generalizability of our study, with results applicable to a cohort healthier than what might be observed in clinical practice. With respect to HRQoL, the EQ-5D is a generic measure of health outcomes suitable to assess the value of healthcare interventions across different disease areas. The EQ-5D is the preferred instrument of the National Institute for Health and Care Excellence for cost-effectiveness analysis. Although the measure showed important differences between the patient subgroups, further research could assess whether the EQ-5D is sufficiently sensitive to detect important clinical changes in patients with bladder cancer. Regarding the study findings, the true negative repercussions of MIBC might differ from those reported because the number of patients who progressed to MIBC was relatively small because BOXIT trial was powered to investigate the interval to the first recurrence. This, coupled with the low postprogression EQ-5D response rate, resulted in uncertain estimates and might lead to Bladder Cancer Cost and Health-related Effects e424 -Clinical Genitourinary Cancer August 2020 overestimates of the HRQoL because patients with relatively poor health outcomes after the development of MIBC might be less likely to complete the EQ-5D. Moreover, the increasingly protracted EQ-5D follow-up periods meant that the clinical events in the study became progressively distant from completion of the EQ-5D. Whether improvements in the reported postevent HRQoL outcomes over time stemmed from the true underlying dynamics of bladder cancer or had resulted only from time-related disparities between the event and the follow-up evaluation remains to be determined.
The costs could have been underestimated for several reasons. First, our analysis of the effect of the events on the annual costs neglected the potential dynamics and spillover effects between the evaluation periods. Bladder cancer events will inevitably prompt immediate resource use; however, the costs incurred from stricter surveillance and the greater risk of related events will be realized further into the future. Understanding these dynamics requires a more detailed collection of the resource use data and remains a potential avenue for further research. Second, the assumption that the treatments were elective could, again, have underrepresented the costs. Third, the protracted and persistent nature of bladder cancer has far broader cost effects than those incurred only by the NHS within 3 years. A wider perspective would give a more comprehensive account of the earnings, productivity, and time lost by patients with bladder cancer and their informal caregivers.

Conclusion
The results from our analysis of the BOXIT trial data suggest that patients with NMIBC will experience decrements in HRQoL, with significant costs imposed in the event of disease recurrence or progression, and the costs increasing with the abnormality and invasiveness of the lesion.

Clinical Practice Points
A need exists to evaluate the costs and HRQoL implications of bladder cancer and its recurrence and progression to assess the disease burden, inform resource allocation decisions, and aid further research. It has been shown that NMIBC will be associated with considerable costs in the United Kingdom and that patients will experience significant decrements in HRQoL with progression to MIBC; however, the effect from NMIBC recurrence is less clear. In our study, we reported both the mean and the marginal UK HRQoL and cost effects across multiple grades and stages of bladder cancer for patients with NMIBC and found significant decrements in HRQoL related to grade 3 recurrence and progression to MIBC; the cost effects increased with the lesion's abnormality and invasiveness. The evidence presented from the BOXIT trial suggests that not all cases of NMBIC recurrence should be considered equal with respect to the effects on patient HRQoL or the consequent healthcare costs. The results from the present study could help to lay the foundation for future related burden of disease studies and costeffectiveness analyses.

Acknowledgments
Grateful thanks are due to the BOXIT Management Group and Trial Steering Committee for the release of data to conduct this research. Thanks also to all the patients who participated in the study; all involved staff at the participating centers and contributing principal investigators; and the trial unit staff at Institute of Cancer Research-Clinical Trials and Statistics Unit involved in the coordination, data management, and analysis of the trial data. The BOXIT was supported by Cancer Research UK (CRUK 07/004). Pfizer US provided the celecoxib and placebo drugs, and education grants were received from Kyowa Hakko UK Limited and Alliance Pharma Plc (formerly, Cambridge Laboratories Ltd). Trial recruitment and on-going patient follow-up were supported within centers by the National Institute for Health Research funded National Cancer Research Network. Central trial management costs were funded by Cancer Research UK (grant C8262/A5669). None of the funders played a role in the study design, data collection, analysis, interpretation, writing of the report, or the decision to submit the report for publication.

EQ-5D dimension scores by event status
No problems Some problems Extreme problems

Edward Cox et al
Clinical Genitourinary Cancer August 2020 -e427 Supplemental

Edward Cox et al
Clinical Genitourinary Cancer August 2020 -e433 Supplemental

Edward Cox et al
Clinical Genitourinary Cancer August 2020 -e435 Supplemental

Edward Cox et al
Clinical Genitourinary Cancer August 2020 -e437 Supplemental   Bladder Cancer Cost and Health-related Effects e442 -Clinical Genitourinary Cancer August 2020