Economic evaluation

Parts of this chapter are reproduced from Noble et al.29 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

Introduction

The objective of this chapter is to report the within-trial economic evaluation of the UNBLOCS trial, conducted from randomisation to 12-month follow-up, from two perspectives: an NHS secondary care perspective and a wider NHS perspective. This economic evaluation is also reported in Noble et al.29 A cost–consequences analysis of ThuVARP versus TURP in men with BPO in relation to QALYs, the IPSS and Qmax is reported. Furthermore, cost-effectiveness analyses of ThuVARP versus TURP from the two perspectives in relation to QALYs are reported. Costs are derived from resources used by patients in relation to the interventions and the use of NHS services in relation to the treatment of LUTS, or urinary retention secondary to BPO. QALYs are determined from the EuroQol-5-Dimensions, five-level version (EQ-5D-5L), questionnaire and the cross-walk valuation set.30

Methods

Measurement and valuation of relevant resource use

Resource use was collected from randomisation for 12 months. Data on resource use for this analysis came from three main sources: trial CRFs, hospital patient-linked information costing systems (PLICS) and patient-completed questionnaires.

Case report forms were completed by UNBLOCS nurses from the treating hospital and captured the operation, including time (which was calculated from time of start of resection until the time the participant entered recovery) and disposable equipment; and the postoperative stay, including complications, returns to theatre and time spent on different wards (calculated as the number of days from the time leaving recovery). The complete details of the resources collected are in Table 30.

Informatics or similar departments in all sites were contacted and asked if they would be able to provide electronic information from PLICS for inpatient stays and day cases in the form of International Classification of Diseases, Tenth Revision (ICD-10);38 OPCS-4;6 and Healthcare Resource Group (HRG) codes; admission and discharge dates; and, for outpatient visits and procedures, service codes, HRG/currency codes and attendance dates. For all centres for which this was possible, a Microsoft Excel spreadsheet containing a list of study identifiers (IDs) and the dates of randomisation and 12-month follow-up were sent to the on-site study research nurses. The nurses added NHS and hospital numbers to the study IDs, which were then forwarded to the relevant contact. Data on all inpatient stays and outpatient attendances between the two given dates were then added to this spreadsheet and returned to the research nurse, who anonymised the data before transferring them to the co-ordinating centre. For one centre for which this was not possible, the local principal investigator manually extracted data on all inpatient stays and outpatient visits from the hospital systems.

All inpatient admissions occurring on the randomisation date were deleted, as were events that, according to clinician opinion, could not be related to the treatment of LUTS or urinary retention secondary to BPO. When HRG codes were not supplied, the ICD-1038 and OPCS-46 were used to map the event to a relevant HRG code.

The resource use questionnaire was administered to participants at 3- and 12-month follow-up. Either it was posted or nurses provided it to the participant at clinic to be completed in their own time and then returned by post. The questionnaires were used to collect information on NHS and private community-based health-care use (e.g. GP visits, district nurse visits), other NHS hospital health-care use, medications, Personal Social Services resource use in addition to travel, time off work/usual activities and any other expenses resulting from their treatment.

All resource use and the unit costs used to value the resources are given in Table 30. The 2016/17 costs excluding VAT (value-added tax) were used to value the resource use.

Any costs given in an earlier year were upgraded using the Hospital and Community Health Services Index.

A form of microcosting was used to cost the operation and the initial hospital stay. The trust finance department of one of the participating hospitals and equipment manufacturers were approached to obtain values for the initial NHS resources used. Any equipment that differed between the two procedures was costed. For ThuVARP this included the cost of the laser itself and for TURP this included the cost of the generator. For both procedures, the costs of any consumables and working elements required for the operations were included. To obtain a cost per procedure for the capital equipment, an ‘annual equivalent cost’, which took into account capital charges/depreciation over the useful life of the equipment,39 was calculated for the two operations. The annual maintenance costs for the laser machine and the generator were then added to these respective costs. These costs were divided by the number of TURP procedures (n = 260) carried out in a single operating theatre in 1 year. If one procedure had been converted to another, then both capital equipment costs were assigned to the participant. The cost per minute of a medi room (a room in which patients are admitted to prior to surgery and return to following surgery) was used to value time in recovery.

The NHS Reference Costs 2016/1731 were used to value inpatient stays and outpatient visits and procedures. The elective inpatient unit cost related to the relevant HRG code was used. For outpatient procedures, the unit cost relating to the service code and HRG code for each procedure was used. A consultant-led unit cost relating to the relevant service and currency code was used for outpatient appointments. When the outpatient currency code was missing, a value based on the unit cost for that service code for total outpatient attendances was applied.

Outcome data collection and valuation

The primary economic outcome in the UNBLOCS trial is the QALY as recommended by NICE. The EQ-5D-5L was given to participants at baseline and at the 3- and 12-month follow-up to either complete at clinic or take home and return by post. At the 6-week follow-up, the EQ-5D-5L was administered by post. The participants’ self-reported EQ-5D-5L values at baseline and at the 6-week and 3- and 12-month follow-up were transformed into utility scores using the cross-walk valuation set as currently recommended by NICE. QALYs for each participant are calculated from the utility scores using the area under the curve approach, and this takes into account any deaths that have occurred during the study.40 The co-primary outcomes of the effectiveness analysis, IPSS and Qmax, are also evaluated in the cost–consequences analysis.

Data cleaning and missing costs and outcomes

The health economists were blinded to the randomisation allocation until all of the data cleaning and valuation of resources had taken place. Simple imputation was used for a few missing data items occurring during the operation. If the start of resection time was missing (n = 35), then the anaesthetic start time was used. If the time leaving recovery ward or discharge time (for day cases) was missing (n = 52), then a 3-hour duration was used, based on information given by one of the hospitals. If the type of laser fibre used was missing (n = 5), then a reusable fibre was assumed; if the number of laser fibres was missing (n = 14), then one was assumed. In relation to the missing items from the resource use questionnaires, if the questionnaire had been returned and an item was missing, then it was assumed that no resource had been used. In relation to medications, if a dose was missing, then the usual dose was obtained from the British National Formulary.41 In relation to incontinence pads/pads costs, the cost of the most commonly used pad was imputed.

Multiple imputation by chained equations was used. The model included baseline, 6-week and 3- and 12-month utility variables, trial arm, baseline diagnosis of LUTS or urinary retention, baseline comorbidities, age and centre. The typical approach is that the number of imputations should surpass the percentage of incomplete cases; thus, 54 individual imputations were conducted and combined using Rubin’s rules27 in Stata 15.1 in relation to both the secondary care and the NHS perspective. The randomisation seed of 12678 was employed to create reproducible imputations.

Analysis

The economic analyses are conducted using an intention-to-treat approach, that is, analysing participants in the arm to which they were randomised, irrespective of any post-randomisation changes. As the trial period does not fall beyond 1 year, no costs and effects are discounted. No modelling has been specified within this evaluation, as the work is seen as a definitive trial, and experience has shown that most of the uncertainty in relation to cost differences are captured within the first 12 months, the duration of this trial.

The cost of each item of resource used during the 12 months of follow-up is evaluated as the resource use (e.g. number of minutes in theatre) multiplied by its unit cost. The total cost for each resource category (e.g. theatre equipment, subsequent inpatient stays), the NHS secondary care and NHS perspective for each individual patient was calculated as the sum of the cost of resource use items. The mean resource use and costs were estimated and presented by trial arm for each resource use category (e.g. outpatient visits, equipment costs). The mean EQ-5D-5L domain scores by time point were estimated and presented by arm. The method of seemingly unrelated regressions, which accounts for the correlation between costs and QALYs, was used to estimate adjusted mean costs and QALYs and the differences in adjusted mean costs, QALYs (and their associated 95% CIs) between the trial arms in relation to secondary care NHS costs and NHS costs. Costs and QALYs were adjusted for the stratification variables used in the randomisation process [centre and patient eligibility classification at baseline (bothersome LUTS or urinary retention)].

Quality-adjusted life-years were adjusted for the stratification variables and baseline utility.42

Cost–consequences analysis

A cost–consequences analysis was conducted in which adjusted secondary care costs and all NHS costs were compared with adjusted QALYs, IPSS and Qmax scores.

Cost-effectiveness analysis

Cost-effectiveness analyses were conducted in which the secondary care NHS costs and all NHS costs were compared with QALYs. Incremental cost-effectiveness ratios were created using seemingly unrelated regressions if neither treatment was dominant (i.e. less expensive and more effective). Seemingly unrelated regression outputs were used to estimate the incremental net monetary benefit (INMB) statistic at the standard NICE willingness to pay threshold of £20,000 per QALY. Cost-effectiveness acceptability curves, which show the probability that ThuVARP is the cost-effective option compared with TURP at different willingness-to-pay-per-QALY thresholds, were created to explore uncertainty.

Sensitivity analyses

A series of one-way sensitivity analyses were conducted on the NHS secondary perspective analysis (as this contained the most complete data and the greatest cost drivers) to test the robustness of different parameter estimates and assumptions made in relation to resource use and costs. The sensitivity analyses included a complete-case analysis; the exclusion of prostate cancer-related hospital resource use; the addition of TURP capital costs to those patients randomised to ThuVARP to account for TURP equipment that needed to be available because of laser equipment failure; the application of the average times of theatre for each arm from the last 25% of cases in each centre to all other respective cases in order to examine the learning curve; the exclusion of post-recovery ward costs for those participants who underwent a ThuVARP procedure in the five centres where day-case TURP procedures were not conducted to examine the cost implication of being able to conduct the ThuVARP procedure as a day case; the exclusion of the capital equipment costs from both operations to reflect the fact that manufacturers often give capital equipment free provided that enough consumables are purchased; and adjustment for the number of people who would have had the operation in 1 year.

Results

Complete resource use and cost data for the secondary care perspective were obtained for 385 participants (95% of those patients randomised who did not withdraw their data). In the case of the NHS perspective, the completeness of the data declined to 47%. The resource use questionnaires administered at 3 and 6 months were not completed adequately to allow a patient or a societal perspective to be conducted. The number of complete cases from a patient perspective would have been 127. The number of complete cases and the means and costs of these resource use items are given in Appendix 1. EQ-5D-5L data were complete for 89% of participants at baseline, 82% of participants at 6 weeks, 78% of participants at 3 months and 81% of participants at 12 months; however, because there was a great deal of intermittent missingness, complete QALY data were obtained for only 212 (52%) participants. The seemingly unrelated regression analyses were conducted on all randomised participants who did not withdraw consent for their data to be used (n = 407).

In terms of resources used (Table 31), the ThuVARP procedure took, on average, 21 minutes longer than the TURP procedure. Participants in the TURP arm spent longer in recovery (14 minutes) and they spent more time in high dependency units/intensive therapy units, and they had slightly more inpatient stays and outpatient visits and slightly fewer day cases. Participants in the TURP arm also had slightly more community-based health service contacts and medications.

In the secondary care analysis, the total adjusted mean costs in the ThuVARP arm were slightly higher (£4252) than those in the TURP arm (£4244), a cost difference of just £9 (see Table 33). The higher cost of the time in theatre has, to some extent, been offset by the higher cost of disposable equipment in the TURP arm and the greater use in the TURP arm of specific resources outlined above. In the NHS analysis, this cost difference had decreased slightly to £4 as a result of the slightly greater community care NHS use among participants in the TURP arm.

The EQ-5D-5L domain scores (Table 32) show an improvement (i.e. the mean score decreases) between baseline and 12 months in both arms, with the exception in the ThuVARP arm of mobility and self-care.

Participants who had complete QALY data had a similar number of adjusted mean QALYs (0.86, 95% CI 0.83 to 0.89 in the TURP arm and 0.84, 95% CI 0.81 to 0.87 in the ThuVARP arm). For the multiple imputation analysis, this was still the case (0.84 TURP vs. 0.83 ThuVARP) (Table 33); the difference is equivalent to an extra 4 days of perfect health. TURP therefore weakly dominates ThuVARP in that it is very slightly more effective and less costly, but this could be a result of chance. This is also the case for the two primary outcomes.

No incremental cost-effectiveness ratios were created because TURP weakly dominated ThuVARP. The INMB at a threshold of £20,000 per QALY in relation to the secondary care perspective was –£236.24 (95% CI –£891.96 to £419.48) and in relation to the NHS perspective was –£231.57 (95% CI –£892.46 to £429.32). Figure 10 illustrates the cost-effectiveness acceptability curves for the two perspectives and shows that, at the willingness-to-pay threshold of £20,000 per QALY, the probability that ThuVARP was the cost-effective treatment compared with TURP was 24% for the NHS secondary care perspective and 25% for the NHS perspective.

FIGURE 10 Cost-effectiveness acceptability curves from (a) an NHS secondary care perspective; and (b) an NHS perspective. (Figure not reproduced here; see the original article.)

The sensitivity analyses (Table 34) show the initial results to be robust; only when costing ThuVARP as day-case surgery did ThuVARP have a positive INMB (£98.95) at a threshold of £20,000 per QALY.

Discussion

The total mean secondary care costs in the ThuVARP arm were £9 higher than those costs in the TURP arm, which reduced to £4 when all NHS costs were included; however, this could be down to chance. A slightly higher QALY score in the TURP arm meant that weak dominance was shown for both perspectives.

Strengths

The use of PLICS to obtain information on follow-up outpatient visits (including procedures) and inpatient stays meant that these data can be assumed to be complete, and it is likely that more complete information was obtained this way than would have been the case through a medical note review. It was also less burdensome on the research nurses. Detailed data collection during the operation meant that a form of microcosting to establish the difference between the two operations could be used.

Limitations

In relation to the use of PLICS, the lack of diagnosis codes in the outpatients’ information may have meant an overestimate of the number of urological appointments related to the treatment of LUTS or urinary retention secondary to BPO; however, this would have been the same in both trial arms. There may have been an underestimate of uncertainty around the theatre costs resulting from the use of simple imputation methods. Resection time missingness did not vary by arm. Missingness of recovery time is slightly higher for TURP (34 vs. 24), which could have led to a slight overestimate of the costs in the TURP arm. Equipment was costed using a bipolar list of TURP equipment, as this was the most common procedure. Even if the monopolar fixed equipment cost per procedure had been double that of the bipolar equipment (£15.64), this would not have had a significant effect on the results.

There is uncertainty with the NHS analysis. The number of complete cost cases was < 50% and the percentage of complete cases differed by arm; furthermore, the unadjusted NHS costs were lower than the secondary care perspective costs, potentially a reflection of healthier participants completing more of the resource use questions. Although the results using multiple imputation showed that the difference between the two arms diminished because of the higher number of community-based contacts that was reflected in community-based resources use, the results from the NHS perspective need to be treated with caution compared with those from the NHS secondary care perspective. The resource use questions were in a separate questionnaire, and therefore perhaps less likely to have been completed than if they had been in the main trial questionnaire. The need to have additional questions to have a wider-than-NHS perspective may have led to non-completion. Additionally, because of previous research,43 the participants had been given a resource use log to complete throughout the study to act as an aide-memoire, but they may have felt that they were being asked to do the same thing twice. It was therefore not appropriate to conduct an analysis from a patient or societal perspective as originally planned because of the incompleteness of the data.

Conclusions

The results indicate that, from an NHS secondary care perspective, participants in the ThuVARP arm had slightly higher costs and fewer QALYs than those in the TURP arm, but these differences were consistent with chance. At a willingness-to-pay threshold of £20,000 per QALY, there is only a 24% probability that ThuVARP is the cost-effective intervention compared with TURP.