Primary Economic Evaluation: Methods

The clinical and economic reviews considered robotic surgery in four indications. When the protocol was written, a decision was made to select one of the four indications for a primary economic evaluation. The selection of the indication was to be made in consultation with the clinical experts for this report and was to consider incremental clinical evidence and the potential clinical and economic impact of robotic surgery based on the relative size of the eligible patient populations and utilization. While the clinical evidence on robotic prostatectomy did not suggest the greatest relative impact on patient outcome, and other indications also had sizable eligible patient populations, prostatectomy is the most frequently performed robotic surgical procedure in Canada (62% of all robotic procedures in 2010 [Danny Minogue, Minogue Medical Inc.,

Montreal, Quebec, Canada: personal communication, December 31, 2010]), and is performed at more Canadian centres (10 of 11) than any other robotic procedure. Given the frequency of use of robotic technology in prostatectomy in Canada, an economic evaluation of robotic surgery in this indication was considered to be appropriate.

5.3.1 Type of economic evaluation

The results obtained from the clinical review and meta-analyses did not show meaningful differences between RARP and ORP, or RARP and LRP, in mortality, general health-related quality of life, or return to normal activities. Differences were seen in urinary function at 12 months, sexual function at 12 months, and in positive margin rates in pT2 stage disease, in comparisons between RARP and ORP. The difference in complication rates between RARP and ORP approached statistical significance and was statistically significant when only procedures conducted after the learning curve were considered.

Sexual and urinary function are aspects of disease-specific quality of life (QOL), but data on the relative impact of surgical approaches on general health-related QOL are limited, and some clinicians have questioned whether observed differences between RARP and ORP are clinically meaningful.¹⁵³ One short-term observational study⁵³ using the 12-item Short Form Health Survey (SF-12) suggests that there is little difference between RARP and ORP, with physical component scores returning to baseline levels within six and seven weeks, respectively, and mental

component scores exceeding baseline levels similarly in the two groups during follow-up. An abstract for a cost-utility analysis¹⁴⁰ that was described in the economic review reported no difference in QALYs after one year compared with open surgery. Observational studies in radical prostatectomy show that mean SF-12 and SF-36 scores approach or reach baseline levels within a year and remain at these levels for up to three and four years, even with sexual function and urinary function remaining low post-surgery.^154-156 A Canadian study that looked at utility and QOL in ORP patients using the Patient-Oriented Prostate Utility Scale (PORPUS) reported that QOL and utility values changed similarly over time, and that utility values approached baseline levels at 18 months to 30 months (baseline 0.94, 18 to 30 months 0.90, clinically important difference in PORPUS 0.05).¹⁵⁷ A second Canadian study in ORP patients who were

administered the PORPUS-U¹⁵⁸ reported a mean baseline score of 0.97, and a mean one-year score of 0.94. Although surgery for prostate cancer may have impacts on urinary function and sexual function, it also simultaneously results in improvements in other QOL domains.¹⁵⁹ A Canadian study¹⁶⁰ assessed utility decrements attributed to sexual function and urinary function using four instruments (PORPUS-URS, PORPUS-USG, Health Utilities Index [HUI], and Quality of Well-Being Scale [QWB]). Estimating between-group differences (robotic surgery compared with open surgery) in QALYs using the utility estimates from this study and the sexual function and urinary function results from the clinical section of this report resulted in estimates of 0.01 to 0.02 QALYs, depending on the instrument. However, these estimates are based on observational data in which baseline sexual and urinary function were often not reported,^43,57,59 and where there were imbalances in several studies in terms of age,^43,57 follow-up,^39,59 and disease progression.

Higher rates of non-localized prostate cancer among the open surgery groups were seen in most studies, with these differences being statistically significant in three studies.^45,54,59 In addition, none of the studies included in the analysis of the sexual function outcome controlled for the effects of medication for erectile dysfunction, which can differ by treatment group.¹⁶¹ Data on comorbidity were generally lacking. Estimating between-group differences in urinary and sexual function beyond one year is difficult, given the lack of longer-term data on these outcomes in robotic prostatectomy, a decline in sexual function with age, and the use of medication and aids for erectile dysfunction.

Positive margin rates in pT2 stage disease after prostatectomy are predictors of disease recurrence in general; however, their impact in pT2 disease is less clear.^162,163 Given the low positive margin rates in pT2 stage disease and the estimated differences in these rates in ORP and RARP, the impact of RARP on overall disease recurrence will be small (0.71% over five years, assuming a large difference in recurrence rates between positive and negative margins in pT2 disease¹⁶⁴). In Drouin et al.’s study,⁷⁹ 83% of patients had pT2 stage disease (the remainder being pT3), and the PSA recurrence rates among the three surgical approaches were the same at five years. The difference in complication rates in the RARP-ORP meta-analysis approached statistical significance, and attained statistical significance when only post–learning curve cases were considered, but a large proportion of these complications are minor and are often accounted for by transfusions of low blood volume. Based on the clinical data reviewed for this report, an estimated 25% of all complications in prostate surgery are major. Based on the complication rates in the clinical section of this report, the marginal difference in major complications between RARP and ORP would therefore be less than 1% for all cases, and 1.2% in cases that occur after the learning curve. The long-term impact of these possible differences is unclear. One study of more than 1,100 patients⁵⁵ that looked at readmissions and post-study visits for complications found no differences between patients who had undergone RARP and ORP.

Because clinically important between-group differences in survival, general QOL, morbidity, and potential disease recurrence could not be shown, a cost-minimization analysis was conducted. The results of this economic evaluation of robotic prostatectomy are presented in terms of average per-patient total and incremental costs for RARP compared with ORP and RARP compared with LRP.

5.3.2 Target population

The target population in this analysis is males with a diagnosis of prostate cancer for whom prostatectomy is the recommended therapy. The average age of patients in the clinical studies that were reviewed for this report is 61 years.

5.3.3 Comparators

RARP was compared with ORP and with LRP.

5.3.4 Perspective

Analyses were conducted from the perspective of the publicly funded health care system.

5.3.5 Effectiveness

Effectiveness in major patient outcomes is assumed to be equivalent between comparators.

5.3.6 Time horizon

Because the expected outcomes and treatment of patients could not be shown to differ after hospital discharge, the time horizon for this analysis is the length of hospitalization. The useful life of the robotic equipment was assumed to be seven years in the base case.

5.3.7 Modelling

Analyses were conducted in Microsoft Excel 2010, version 14.0, and in TreeAge Pro Suite 2009, version 1.0.2. Because analysis of the clinical data was conducted separately for RARP

compared with ORP and for RARP compared with LRP, separate models were used for the RARP with ORP and RARP with LRP comparisons. Simple decision-analytic models (two treatment arms with no subsequent decision nodes) were constructed to compare costs by treatment group, and to conduct probabilistic sensitivity analyses on the incremental cost estimates. An internal validation of the models was conducted by varying model parameters to extreme values and assessing the feasibility of the resulting cost estimates.

5.3.8 Resource use and costs

Follow-up of patients post-discharge was assumed not to differ by surgical approach.

a) Surgical equipment and supplies

The da Vinci Si Surgical System is distributed in Canada through Minogue Medical Inc., and this distributor quoted costs of the system and its operation in US dollars (Danny Minogue, Minogue Medical Inc., Montreal, Quebec, Canada: personal communication, December 31, 2010). US prices were converted to Canadian prices using the average exchange rate in the previous year (US$1 is C$1.016, April 2010 to March 2011;¹⁶⁵ Table 16).

Table 16: Capital and Operating Costs of da Vinci Surgical System*

Item US Dollars Canadian Dollars

da Vinci Si Surgical System 2,600,000 2,643,680

Start-up reusable equipment and accessories 200,000 203,360

Disposables and consumables (per procedure) 2,500 2,542

Training of surgeons† (each) 6,000 6,101

Training other personnel Nursing and CPD in-service at no charge

Annual maintenance (after first year warranty) 175,000 177,940

CPD = continuing professional development.

*Danny Minogue, Minogue Medical Inc., Montreal, Quebec, Canada: personal communication, December 31, 2010.

†Cost of training first four surgeons is included in the purchase price of the robot. Experience of Canadian centres suggests that after the first year, one new surgeon will be trained at each centre each year.

The undiscounted annual and cumulative costs to a centre for acquiring and operating this technology are shown in Appendix 18. Some costs of this technology are fixed (acquisition costs of robot), but others recur annually or vary by the number of procedures that are performed (maintenance contract, disposable and consumable equipment).

Initial capital expenditures were annuitized using the method described by Richardson and Gafni.¹⁶⁶ A discount of 5% was used. The base case assumptions for this estimation were that the useful life of the equipment is seven years, and that it has no residual value at the end of its use.

The assumption about the useful life of equipment was based on convention in other studies of this technology (five or seven years) and on the fact that two Canadian centres have been operating their robotic equipment for seven years. Longer durations of use are possible, but technological change may limit the useful life of equipment. Based on the experience of Canadian centres, it was assumed that one new surgeon would receive the mandatory robotic training course provided by Intuitive Surgical, Inc., at each centre each year, after the first year.

No other training costs were considered. Expenditures on training and maintenance over the life of the robot were discounted at 5%. Assuming an average caseload of 130 procedures per centre per year (the average number of procedures performed at 11 Canadian centres in 2010 [range 25 to 268]), the total cost of the robotic equipment in the base case was estimated to be C$7,427 per procedure.

The costs of supplies for laparoscopic prostatectomy and open radical prostatectomy were obtained from the literature¹²⁹ and were estimated to be C$831 and C$212 per procedure,

respectively. Laparoscopic equipment was assumed to be disposable, and therefore there were no associated maintenance costs.

b) Hospital costs

No reliable national Canadian data were available on length of hospital stay among patients undergoing robotic prostatectomy (alone or in comparison with open prostatectomy or

laparoscopic prostatectomy). As a result, comparative data on lengths of hospital stay that were obtained from the clinical review of this report were used. The per diem hospital costs were estimated from special tabulations obtained from the Canadian Institute for Health Information’s Discharge Abstract Database for 2009-2010 (Sources: Canadian Institute for Health Information, Ottawa, Ontario, Canada. Discharge Abstract Database). Resource intensity weights were

multiplied by the average cost per weighted case (CPWC) in Canadian hospitals. The CPWC for 2009-2010 data was unavailable when this report was written, and a 2008-2009 estimate,

adjusted +3.5% to account for observed annual growth rates in CPWCs, was used. The estimated hospitalization cost was then divided by the average length of hospital stay that was estimated for radical prostatectomy patients, to provide a per diem cost. The estimated hospital per diem cost for prostatectomy was $2,353. The per diem costs were then multiplied by the average lengths of hospital stay estimated for the three surgical approaches in the meta-analyses.

c) Professional fees

Procedural surgical and anesthesia fees were obtained from the fee schedules of the four provinces performing robotic prostatectomy (British Columbia, Ontario, Quebec, and Alberta).^167-170 This represents a range of fee scale (low to high) seen in Canada and, taken together, 86% of the Canadian population. Surgeons who perform robotic prostatectomy bill the respective provinces for a laparoscopic procedure, because there are no unique billing codes for robotic prostatectomy. Using a weighted (by population) average, surgeon fees for open

prostatectomy and for laparoscopic or robotic radical prostatectomy were estimated to be $1,022 and $1,381, respectively. The fees for anesthesia have a time component in three provinces:

British Columbia, Ontario, and Quebec. Accounting for differences in operative times as reported in the meta-analyses, the weighted average fees for anesthesia in open surgery, laparoscopic surgery, and robotic surgery were $470, $615, and $581, respectively.

d) Transfusions

The probabilities of transfusion for each surgical approach were obtained from the results of the meta-analysis. In the comparison between RARP and ORP, these probabilities were 2.9% and 14.5%, respectively, and in the comparison between RARP and LRP, these probabilities were 2.5% and 4.6%, respectively. The number of red blood cell units transfused at each transfusion was estimated from the data on blood loss, which were obtained from the meta-analysis. It was assumed that up to 450 mL of lost blood would result in a transfusion with one unit of red blood cells. The cost of a unit of red blood cells in Canada was estimated from the literature to be

$429.43.¹⁷¹

The costs that were reported in US dollars were converted to Canadian dollars using the average exchange rate of the year in which the costs were reported.¹⁶⁵ All costs are reported in 2011 Canadian dollars. The costs that were obtained from sources dating before 2011 were inflated using the Canadian Consumer Price Index.¹⁷²

The health care resource use estimates and cost estimates that are used in this analysis are shown in Appendix 19, in Tables A27 and A28, respectively.

5.3.9 Discount rate

To estimate the present value of a procedure using robotic equipment with a specified lifespan, future costs were discounted at 5% per year in the base case. Rates of 0% and 3% were

considered in the sensitivity analysis, as suggested by CADTH Guidelines.¹²⁸

5.3.10 Variability and uncertainty

Sensitivity analyses were conducted on the estimated incremental costs of RARP compared with ORP, and RARP compared with LRP.

One-way and multi-way deterministic sensitivity analyses were conducted for key model parameters to assess the robustness of the base case results. The methods that were used to determine the parameter values for the sensitivity analysis included plausible ranges as determined by the variability of parameter estimates, the literature, guidelines, and expert opinion. For parameters with values that were most uncertain or for which variability was unknown, ranges of ± 50% of the estimated mean value were used.¹⁷³ The parameters that were included in the deterministic sensitivity analyses of the base case were:

Discount and annuitization rate (0% and 3%)¹²⁸

Cost of robotic disposables and consumables (± 25%)¹²⁹ Cost of robotic annual maintenance contract (± 25%)¹²⁹

Cost of all recurring robotic costs (disposables, maintenance, training; ± 25%) Useful life of robot (five years and 10 years)

Useful life of robot by average annual caseload (range of 50 to 500) Break-even number of procedures per year

Donation of robotic equipment

Exclusion of non-robotic equipment and supply costs Cost of non-robotic equipment and supplies (± 50%)

Length of hospital stay (post–learning curve and marginal difference needed for break-even) CPWC (0% to 8%)

Specialist fees (Quebec, Alberta)^169,170 Number of transfusions

Complications (extreme cost scenario)

Exchange rate (US$1 is C$0.85 to C$1.15, current exchange).

The donation of robotic equipment by a party lying outside the definition of the publicly funded health care system was considered as a scenario in the sensitivity analysis, because some

Canadian centres have obtained a surgical robot in this manner.

One study of more than 1,700 patients³⁴ compared complications in RARP and ORP and reported that most of the statistically significant differences occurred among minor

complications, and the only statistically significant difference among major complications was seen with pulmonary embolism (0.1% compared with 1.0% in RARP and ORP, respectively).

According to data from the Canadian Institute for Health Information’s Patient Cost Estimator, the cost of a hospitalization for pulmonary embolism in Canada is $6,010 (2008-2009 data.

Sources: Canadian Institute for Health Information, Ottawa, Ontario, Canada. Discharge Abstract Database). Although it is likely that the impact of complications is captured in the base case model through LOS and transfusions, we assumed an extreme scenario in which all 25% of complications that are estimated to be severe, according to our clinical review, cost three times the average cost of a hospital stay plus professional fees ($34,445 for ORP and $24,726 for RARP) that were used in the base case model (Table 17).

Probabilistic sensitivity analyses using Monte Carlo simulation were conducted to estimate the uncertainty in incremental costs. The probabilities used in the model were assumed to follow a beta distribution. The length of hospital stay was assumed to follow a gamma distribution.

Surgical equipment (including robot) costs were assumed to follow a fixed distribution, and all other health care costs were assumed to follow a gamma distribution, with standard errors being estimated at 50% of the mean. The distributions of estimates that were used in the models are shown in Tables A29 and A30 in Appendix 20. The uncertainty in the incremental total costs for each model was expressed in terms of 95% confidence intervals.