Clinical inputs

Resource use and costs

Quality of life

Quality of life weights for the model were taken from previous cost-effectiveness study of patients (identical in age to our cohort) with rectal cancer.

Probabilistic sensitivity analysis

Probabilistic sensitivity analysis was also conducted to assess the combined parameter uncertainty in the model.

Probabilistic Sensitivity Analysis

Base-case assumptions

The results of 10,000 runs of the PSA are shown using ICER scatterplots and cost-effectiveness acceptability curves (CEAC). The ICER scatter plots show the incremental costs and QALYs associated with each of the 10,000 runs of the PSA along with the mean result. The CEAC graphs show the probability of each strategy being considered cost-effective at the various cost-effectiveness thresholds on the x axis.

Figure 1 presents the probabilistic results of the base case analysis. Of these 10,000 iterations over 75% of them are health improving (to the right of the Y-axis) and over 80% are cost decreasing (below the X-axis) with the majority of iterations being both cost saving and health increasing.

Figure 1

ICER scatterplot base case results. CE: cost effectiveness; ICER: incremental cost-effectiveness ratio; QALYs: quality-adjusted life years

Figure 2 presents the CEAC for the base case results. The probability that the simultaneous approach is the preferred option is 80% at a cost effectiveness ratio of £0 i.e. where the cheapest option is preferred. At £20,000 threshold there is a 86% probability of the simultaneous approach being the preferred option. This remains above 80% beyond values above the £100,000 threshold. For no threshold does a staged approach become the preferred option.

Figure 2

Cost effectiveness acceptability curve base case results.

Clinical evidence review values results

When results from the clinical evidence review are used to inform the economic model there is a greater degree of uncertainty around the results (Figure 3). When these inputs are considered a greater number of iterations show the simultaneous approach as cost saving (94%) but with less than 20% of iterations being health improving.

Figure 3

ICER scatterplot clinical evidence review inputs. CE: cost effectiveness; ICER: incremental cost-effectiveness ratio; QALYs: quality-adjusted life years

The CEAC for this analysis shows uncertainty around the preferred option (Figure 4). At the £20,000 threshold 42% showed the simultaneous approach to be cost effective although as shown by the cost effectiveness plane the majority of these would also be health decreasing. As the threshold increases the probability of the staged approach being the most cost effective option also increases. The threshold at which the CEACs cross and we are indifferent between the two options is £17,000 per QALY.

Figure 4

Cost effectiveness acceptability curve clinical evidence review inputs.

Critical outcomes

Important outcomes

Critical outcomes

Important outcomes

Critical outcomes

Important outcomes

Socioeconomic and demographics variables

The model assumed a uniform age of the cohort of 60 years of age based and 60% male identical to those reported in the only study (Mayo 2013), identified in the accompanying clinical evidence review, that reported overall survival. The uniform age of the model cohort was not varied during either deterministic or probabilistic sensitivity analysis as evidence was not available, given the 1 study identified, to do this in a meaningful way. The proportion of males was varied using a beta distribution based on the absolute numbers reported (Table 16).

Overall and disease-free survival

Survival curves for the economic model were estimated entirely from the accompanying clinical evidence report. The accompanying clinical evidence review identified one study which reported overall survival for this patient group. (Mayo 2013) The study estimated an absolute overall survival from a staged approach at 44.0% after 5 years. From the same study the hazard ratio for a simultaneous compared to a staged approach was 1.08 (0.89-1.32) suggesting decreased overall survival in the simultaneous approach although the 95% confidence interval crossed the point of no difference. The study was a retrospective cohort study involving 1004 patients. Full details of the study are available in the accompanying clinical evidence review.

It was considered by the committee that retrospective cohort studies would always overestimate the overall survival of the staged approach. This is because there will be a subset of patients in the staged approach who would receive the first stage of the resection, with intention for the second stage, but did not receive both either because of disease progression, change in the frailty of the patient making further surgery inappropriate or through death. This sub patient group is likely to be older, have more aggressive disease or greater comorbidities. Their life expectancy is less than that of the rest of the population. The 3 retrospective studies did not or were unable to identify these patients and they were excluded from the analysis increasing the mean survival for the cohort. A prospective cohort study or randomised controlled trial would have eliminated this selection bias. Based on single centre cohort studies it is estimated that between 16% and 35% of patients intended to receive both resections in a staged approach will not receive the second operation. (Mayo 2013) The majority of these patients, if simultaneous approach was favoured, would have received both parts of the resection. There is possibly a small number of patients for which performing the first resection, without full intention of performing the second resection (i.e. a planned reevaluation of the disease and patient condition) and that a simultaneous approach maybe inappropriate. Whilst this group is not considered by the economic model they are will be included in the retrospective studies. However, as their number will be small it is unlikely that they will alter the results of the economic model.

Overall survival curves for a staged approach were estimated using the 5 year overall survival and assuming an exponential function in the absence of an estimated Kaplan Meier survival curve. Given the biases in the estimates of survival discussed above it was assumed in the base case that 25.5% of patients intended to receive both resections would only receive the initial one, the median of the two proportions discussed above. It was considered by the committee that such patients would instead receive treatment with non-curative intent given their poor prognosis, advanced cancer or frailty. It was considered by the committee that individuals in this group would follow the prognosis of those enrolled in the FOCUS trial (Seymour 2007). This was an RCT involving 2135 previously untreated patients with advanced colorectal cancer receiving treatment with non-curative intent comparing a number of different treatment strategies. All treatment strategies considered by the RCT reported a median overall survival between 13.9 months and 16.7 months. To approximately account for this omitted group in the base-case overall survival for the staged approach was adjusted to assume that 25.5% of patients had been excluded from the analysis and that these would have a mean survival of 16.7 months the largest median value reported in the FOCUS trial. Mean values were not reported in the trial which would be a more appropriate value to adjust by. The median survival is likely to be an underestimate of the true survival. The overall survival of this subgroup did not vary during the PSA although the proportion of the group was varied on a uniform distribution across the range of reported estimates (16%-35%) from the single centre cohort studies.

The population of the FOCUS trial are about 5 years older, more male and have a worse performance scores compared to the population in Mayo 2013 and assumed for this model. As above though it is assumed tha this subsection of the population would be older and frailer than the rest of the population. The focus study is also reasonably old and survival would now likely be greater. This will be somewhat accounted for in the unadjusted population estimates.

Up to 5 years survival curves for the simultaneous approach were calculated from the hazard ratio reported in the accompanying clinical evidence review and the unadjusted overall survival curve for the staged approach following the usual proportional hazard assumptions. After 5 years in the model equal hazards are assumed. This hazard ratio was varied during probabilistic sensitivity analysis following the 95% confidence interval (0.89-1.32) reported in the accompanying clinical evidence review with a log normal distribution.

Disease-free survival was estimated using similar methods. Again an exponential function was assumed for the 52.0% disease-free survival at 2 years estimated in the accompanying clinical evidence review for the staged approach. Both the baseline rate and relative effect were varied for both overall and disease free survival during PSA. To account for the omitted patients who did not receive both resections it was assumed that all patients in this group would not be disease-free after the first year and the disease-free survival curve was adjusted to account for this. It is unlikely that in a patient in which cancer has not progressed significantly but is fit for surgery that the second resection would not be performed. Disease-free survival for patients in the simultaneous approach arm was again estimated using the same assumptions as for overall survival using the estimated hazard ratio for disease-free survival (1.30 95% CI 0.77-2.18). Again the hazard ratio was used against the unadjusted disease-free survival curve for the staged approach. The hazard ratio was varied using a log normal distribution. It was assumed there was no correlation between overall and disease-free survival during the PSA although there is some evidence for a weak to moderate relationship between the two outcomes in metastatic colorectal cancer (Cicero 2018). Where during the PSA disease-free survival was estimated to be greater than overall survival it was adjusted to be equal to that of the overall survival estimate. This was to avoid any logical anomalies in the economic model.

A deterministic sensitivity analysis was also performed where no adjustments were made to the overall or disease-free survival estimates from the accompanying clinical evidence review.

All patients who had disease recurrence would receive either further treatment or palliative care. The treatments received did not alter the survival in the model as these were already accounted for in the survival estimates from the observational studies survival estimates are based on. These further treatments were added solely as costs and are discussed in detail below.

Mortality

The model split death into disease specific mortality and other cause mortality. Whilst this made no difference to the overall survival in the economic model it was useful for assigning palliative care costs (discussed below). Other cause mortality was assumed to be identical to that of the general population which was estimated using Office of National Statistics Life Tables for England and Wales 2014-2016. Estimated mortality was weighted by the age and percentage male in the model cohort (ONS 2018). All other death was assumed to be as a result of the metastasised colorectal cancer or related complications.

The accompanying clinical evidence review identified 1 study reporting on treatment-related mortality which did not report on any events. The model therefore assumed that there would be no death as a direct result of the treatment approach received. This assumption was held for all sensitivity analyses.

Adverse events

The proportion of adverse events for either approach was taken from the accompanying clinical evidence review. Only grade 3 and grade 4 adverse events, those deemed either severe or potentially life threatening, were included in the economic model as these were considered to be the only adverse events which would add additional costs, relative to the base resection costs, and reduce quality of life over the usual detriment from significant surgery.

The accompanying clinical evidence review only identified 1 study (Moug 2010) which reported grade 3 and grade 4 adverse events. This study reported 1 adverse event, in the simultaneous group, for the 64 patients divided equally across the arms. This difference was not statistically significant. Adverse events for both arms in the economic model were inputted, in absolute terms, identically to the values reported in Moug 2010 (3% staged, 0% simultaneous). The committee highlighted that they thought such adverse events were likely to be very rare in these patient groups and, if there was to be any difference between the two groups, it was most clinically plausible for it to be higher in the staged group where a greater number of operations are needed.

Perioperative period

For all resections there will be some perioperative period given that this is a relatively major operation. Patients will have to spend some period of time in hospital following resection regardless of whether a staged or simultaneous approach is taken. For the purposes of the model the perioperative period was assumed to be equal to the length of stay in hospital although it is acknowledged that recovery after surgery is likely to continue after discharge from hospital. However, the significant health detriments and costs are most likely to occur during this hospitalised period.

Length of stay in the study was estimated from one US costing study a retrospective study at 1 US hospital of 224 patients undergoing resection either as a staged or simultaneous procedure between 1990 and 2012 (Ejaz 2014). The studied reported a median length of stay of 7 days for the simultaneous approach and 13 days (total across both resections) for the staged approach. These perioperative periods were applied to all patients in the model.

Cost of resection and resection related complications

All resection costs were taken from NHS Reference Costs 2016/17 (Department of Health 2018). For the staged approach resection was best matched by the currency description ‘Complex Large Intestine Procedure’ for the colorectal resection and ‘Complex, Hepatobiliary or Pancreatic Procedure’ performed on an inpatient basis. Both procedures reported differing reference costs based on the number of clinical complications of the patient. As the model structure modelled complications from surgery separately the cost of either was based on the no complications (CC score 0-2) reference cost. The total costs of a staged approach was equal to the two weighted mean costs combined.

As the model cohort did not exclude any patient based on comorbidities a weighted mean of all costs reported were used for all currency descriptions with complications reported. This weighted cost then had the no complications cost subtracted to estimate a cost of complications. To avoid double counting of complications between the 2 resections a mean cost of the estimated complications cost for both stages was applied in the model. Costs were weighted based on the total number of full consultant episodes (FCE) reported in the reference costs. ‘Complex Large Intestine Procedure’ reported reference costs for both patients under and over 19 years of age. Given the 60 year average age used for the cohort of the model only the over 19 year of age costs were included (Table 10).

In the staged approach arm of the model for patients who receive the first stage (colorectal cancer resection) but do not proceed to the second stage their total cost of procedure will be equal just the primary colorectal cancer resection cost.

Table 10. NHS reference costs used to estimate the cost of staged resection

To estimate the cost of the resections in the simultaneous approach it was not appropriate to combine the 2 costs as undertaken for the staged approach. This is because both sets of costs would include costs for pre-assessment, surgical preparation, anaesthesia, time in surgical theatre and many other items that can be ‘shared’ between the 2 procedures by combining them. This will lead to both reduced overall operating time and resource use.

No currency description was identified in the NHS Reference Costs which was applicable to the simultaneous approach. To estimate the cost for the simultaneous approach the cost of the staged approach was adjusted using data on total operative time and length of stay in hospital. Abbott 2012, identified in the accompanying clinical evidence review, estimated that total operating time in the simultaneous approach was 144 minutes shorter compared to the staged approach. This was converted into a cost using estimates from Ramsay 2012. Ramsey 2012 was a systematic review and economic model of laparoscopic and robotic surgery for the removal of the prostate in individuals with prostate cancer. The economic model estimated a cost per hour of operating time of £1,266. This was converted into a cost for the 144 minutes difference and subtracted from the staged approach cost. Whilst removal of a prostate is fundamentally a different type of surgical operation, the pay grade and number of personnel in attendance and size of operating theatre would be almost identical between the 2 and it was the committee’s opinion that this was a reasonable estimate of operating costs.

Length of stay in hospital also differed between the 2 arms of the model with simultaneous approach resulting in 6 less days in hospital as discussed above. This reduced time in hospital was again converted into a cost. This was done by using the excess bed day costs for ‘Complex Large Intestine Procedures, 19 years and over’ again weighted against the FCEs in the NHS Reference Costs. This mean cost was then multiplied by the reduction in days and subtracted from the procedure cost for the staged approach. This was in addition to the adjustment for operative time. The cost of complications was assumed to be identical to that of the staged approach. Full details of how costs for the simultaneous approach were calculated are presented in Table 11.

Table 11. Estimation of the cost of a simultaneous approach

Resource use and cost of further treatment

All patients who have disease recurrence will go on to receive further treatment or if not appropriate palliative care. Of those patients with recurrent disease who go on to receive further treatment three broad types of treatment were identified by the committee-hepatic resection, extrahepatic resection and chemotherapy. Hepatic resection is an identical surgical operation to the liver resection stage of the staged approach. It was assumed that given the post-surgical surveillance these patients would be under that any recurrence would be identified reasonably quickly and that complications from any operation in this group would be minimal. Extrahepatic resection is defined as where disease has recurred near to but outside of the liver most commonly in the bile ducts. Chemotherapy was the treatment choice for patients with recurrence in either the hepatic or extra-hepatic regions where this is not amenable to treatment or to other parts of the body where surgery would not be considered appropriate.

The proportion of patients going on to receive further treatment was taken from a UK cost utility study comparing operable to non-operable treatments for liver metastases (Roberts 2015). The study was excluded from the economic evidence review as comparisons of operable to non-operable treatments were outside of the scope of the protocol. The proportion for each type of treatment was taken from one prospectively maintained database of individuals undergoing surgery or chemotherapy for liver metastases at 1 UK hospital between 1992 and 2001. Estimates of the proportion of patients undergoing each type of further treatment are presented in Table 12.

Table 12. Estimate of proportions undergoing each type of treatment from Roberts 2015

Costs for both hepatic and extrahepatic resection were again taken from NHS Reference Costs. Hepatic resection was costed identically to the resection part of the staged approach although the assumption was made, given the follow-up of these patients, that the recurrence would be identified rapidly and the ‘Complex, Hepatobiliary or Pancreatic Procedure - without complications-with CC score 0-2’ description and costs were assumed. Extrahepatic resection was costed as ‘Complex Thoracic Procedures, 19 years and over’ and as for other procedures in the model a weighted average of all CC scores was calculated based on FCEs in an inpatient setting. The estimated costs for both approaches are presented in Table 13.

Table 13. Estimation of hepatic and extrahepatic resection costs

For disease recurrence, which was considered not amenable to resection, patients received systemic chemotherapy treatment. Two chemotherapy regimens were used by the economic model which were considered to cover the majority of chemotherapy received in the NHS for this patient group following inoperable disease recurrence-FOLFOX and FOLFIRI. An annual cost for each regimen was estimated based on committee estimates of the quantity of each component and the number of cycles. Patients who had recurred disease and received chemotherapy were assumed to continue receiving it annually until death and the cost was applied for every 1 year model cycle. The resource use and quantity for the chemotherapy regimens used in the study are presented in Table 14. Prices were taken from the ‘Drug and Pharmaceutical Electronic Market Information Tool’ (eMit) [Accessed March 2019] for all drug components. Administration costs were taken from NHS Reference Costs 2016/17 using the currency description ‘Deliver Complex Chemotherapy, including Prolonged Infusional Treatment, at First Attendance’ on a day case and regular day/night basis. Given the costs for both chemotherapy regimens were almost identical (less than £1 difference) no differentiation was made between them in the model and a mean cost of the two regimens was applied to all patients receiving chemotherapy following recurrence.

Table 14. Cost of chemotherapy treatment following recurrence

The committee also considered that cetuximab or panitumumab may be used for patients whose disease recurs but is not amenable to further surgery. Cetuximab or panitumumab is used in the NHS in conjunction with Cetuximab and panitumumab for previously untreated metastatic colorectal cancer (TA439) alongside either the FOLFOX or FOLFIRI chemotherapy regimens. This may either be given to shrink cancer tumours to allow surgery or to palliate. Resource impact tools for TA439 estimated that 65% of patients receiving either FOLFIRI or FOLFOX for metastatic cancer would receive this alongside cetuximab or panitumumab.

The list prices for cetuximab is £890.50 for a 100ml vial and panitumumab £1,517.16 per 20ml vial Both drugs are currently provided to the NHS, by the manufacturers, under the terms of a confidential patient access scheme (for panitumumab) and commercial access agreement (for cetuximab) and the cost to the NHS is likely to be significantly less than the list price. It has not been possible to transparently include these costs in the economic model. Therefore, costs for cetuximab and panitumumab were not included in the base-case analysis. This does not imply that these costs will not be significant but that the access agreements are priced so that the net benefit, assuming NICE’s threshold of £20,000 per QALY gained is identical to society’s willingness to pay per QALY gained, then the addition of either drug to FOLFIRI or FOLFOX will have no impact upon the incremental cost effectiveness ratios or ultimately the decision around the preferred approach to resection. A deterministic sensitivity analysis was also undertaken which used the list price for both drugs to investigate if this would alter the preferred approach to resection. These 2 values were considered to cover all plausible costs.

Cost of palliative care

Given the relatively short life expectancy of the model cohort and that the majority of patients would die as a result of their disease a one off cost of palliative care was applied to the entirety of the cohort during their final year of life. This is to represent the increase in resource use experienced during the final months of a patient’s life. This one off cost was taken from Georghiou 2014. The study used medical records of over 1,836 patients with cancer at multiple UK hospitals and hospices to estimate resource use and publically available UK costs to estimate a total cost for the final 90 days of life. An average cost for patients with cancer was used from the report. These costs are presented in Table 15.

Table 15. Costs of palliative care for patients with cancer from Georghiou 2014

The above costs includes ‘local authority-funded care’. The methods of calculation from the original report may include costs, such as personal contributions to care, which are not strictly covered by the NHS & PSS perspective used for this economic model. A deterministic sensitivity analysis was therefore undertaken which removed this cost from the total palliative care cost estimate.