Cost effectiveness analysis of intensive versus conventional follow up after curative resection for colorectal cancer

Costs

We adopted a health service perspective and calculated patient specific costs related to five year follow up strategies for colorectal cancer (intensive follow up compared with conventional follow up) by estimating costs for surveillance and treatment of recurrences. Costs were based on 2002 UK prices and included direct, indirect, and overhead costs. We used a “bottom-up” approach, rounding costs upward and seeking to be as inclusive as possible in estimates. Sources included the Department of Health reference costs and related cost effectiveness studies (see bmj.com).16^–18 The Department of Health reference costs are aggregate direct costs for trust hospitals nationwide and present costs as mean, minimum, and maximum values.

Because follow up regimens and surveillance tests varied considerably between and within trials, we estimated costs separately for each trial and then calculated the costs for each patient.5 Costs included clinical follow up visits (administration, clinical examination, consumables), laboratory tests (reagents, consumables, staff, service costs, processing and reporting), endoscopic investigations and polypectomy (administration, drugs, consumables, staff, sterilisation, annual equivalent costs for equipment), and radiology (staff, consumables, service costs, processing and reporting, annual equivalent costs for equipment; table 1).19 20 As follow up rates were high for all trials, we modelled full attendance for each prescribed examination and surveillance investigation until five years or death.5 We modelled for additional investigations (namely computed tomography) owing to a 15% false positive rate when testing for serum carcinoembryonic antigen.21

In calculating treatment costs, we incorporated into our model three possible outcomes after the detection of recurrences: inoperable disease requiring symptomatic treatment and palliative care, salvage with cure, and salvage with subsequent failure and palliative care (see figure on bmj.com). For patients followed conventionally, we modelled that the initial assessment of recurrent disease (the decision between palliation and salvage) was based on off-trial investigations, whereas the assessment of recurrent disease was encompassed by protocol investigations in patients followed intensively.14 Treatment costs were then estimated for salvage and palliative treatments for local, hepatic, and other site recurrences. Salvage costs included surgery with average hospital stay, perioperative investigations, high dependency care, blood transfusion, drug usage, stoma care, and care by a district nurse (see bmj.com). A further 35% was added to salvage therapy costs (intensive and conventional follow up) to account for high rates of complications and reoperations associated with these types of treatments.22 23 Palliation costs included palliative surgery and stoma care, palliative radiotherapy, palliative medical care, and cancer care in the community. Chemotherapy regimens were not used in any trial.

Outcome measures

The primary outcome was the extra cost per unit of outcome obtained in comparing one treatment with another (incremental cost effectiveness ratio) for each change in life year for intensive follow up compared with conventional follow up.24 We calculated life years lost and gained for each trial using average life expectancies for the UK population and taking into account the proportions of males and females, the mean ages at initial treatment, and the number of deaths.25 Costs were calculated for each trial and then expressed as costs for each patient and aggregated. The probability that intensive follow up is cost effective depends on how much the NHS is willing to pay for each life year gained—the value of £30 000 (€42 968; $51 888) was used (referred to as NHS cost acceptability), reflecting the currently perceived threshold in the United Kingdom.26

Censoring and discounting

In the absence of individual patient data, we modelled the distribution of censored events (deaths) by constant numbers for each three month period. In practice, surveillance would generally be discontinued (censored) after recurrence, such that this model tends to overestimate surveillance costs. In the base case analysis, we discounted benefit effects at 1.5% and costs at 6.0%.27

Sensitivity analyses

We undertook several sensitivity analyses to illustrate the impact of the principal aspects of uncertainty on the estimate of cost effectiveness. These included, separately and together, changes to discount rates as proposed by the US Panel on Cost Effectiveness in Health and Medicine,24 distributions of deaths (constant number, constant rate, and all deaths at end of five years), false positive test rates, and maximum surveillance and treatment costs.

Using the NHS cost acceptability of £30 000 as reference, we calculated the incremental increase in costs required to increase the incremental cost effectiveness ratio to the limit of NHS cost acceptability (incremental cost threshold) by simultaneous incremental increases in surveillance and treatment costs for both intensive and conventional follow up. We calculated the lowest level of effectiveness—absolute reduction in mortality—required to increase the incremental cost effectiveness ratio to the limit of NHS cost acceptability (threshold of effectiveness). The threshold of effectiveness was calculated by altering differences in benefit between follow up strategies by amounts equivalent to 1% change in mortality (see bmj.com).

Cost effectiveness

The figure shows the Forest plot for all cause mortality with absolute reductions in mortality, together with the calculated life years lost and gained for each patient for each study and associated incremental cost effectiveness ratios. For the five trial model, the lost and gained calculations favoured intensive follow up by 0.73 life years gained for each patient, increasing to 0.82 life years gained for each patient for the four trial model, consistent with the observations from our previous meta-analysis.5 In the five trial model, the adjusted net (extra) cost for each patient was £2479 and for each life year gained was £3402, substantially lower than the current NHS threshold of cost acceptability. The corresponding values for the four trial model were £2529 and £3077, suggesting that targeted surveillance is more cost effective.

Forest plot of randomised trials. Pooled analysis with summary estimates (Mantel-Haenszel fixed effects method) are for all cause mortality (adapted from Renehan et al 20025). Data are categorised into extramural and intramural detection groups. Positive difference indicates time gained (improved survival). See bmj.com for calculation of estimates. Incremental cost effectiveness ratio=[costs for each patient for intensive follow up minus control follow up]/[discounted life years either gained or lost]

• Download figure
• Open in new tab
• Download powerpoint

Changes in benefit and discount rates and in false positive test rates made little impact on the incremental cost effectiveness ratios in both models (data for four trial model not shown; table 2). Changing the distributions of deaths yielded similar incremental cost effectiveness ratios, suggesting that the assumptions within the primary models were robust. Worst case scenario analyses showed that substituting with maximum surveillance costs produced greater increases in incremental cost effectiveness ratios compared with substituting with maximum treatment costs, indicating that surveillance cost is the most important determinant of cost effectiveness. All incremental cost effectiveness ratios for maximum cost estimates fell within the NHS cost acceptability threshold. The incremental cost threshold was near ninefold for both the five trial (£30 620) and four trial (£27 695) models. For both models the threshold of cost effectiveness was an absolute reduction in mortality of 3%.

Strengths compared with other studies

Previous studies have examined economic issues related to follow up after treatment for colorectal cancer,28^–33 but predated the effectiveness data provided by the two recently published meta-analyses.5 6 The present analysis is strengthened by the inclusion of a comprehensive search to include many aspects of cost estimates (a “bottom-up” approach) and sensitivity analyses at several levels. Further strengths were the robustness of the results: adjustments to maximise costs yielded incremental cost effectiveness ratios well within the threshold for NHS cost acceptability, the incremental cost threshold was high (ninefold increases), and the thresholds of effectiveness were 3% for both models.

Our study emphasises a further aspect, that around equal proportions (one third) of patients with colorectal cancer develop recurrences, irrespective of follow up strategy. After detection of recurrence, palliative therapy (more common in control patients) is not without expense and partially offsets the costs of salvage therapy (more common in patients intensively followed).

Study limitations

The first limitation of our study was the clinical heterogeneity of follow up regimens among the five trials.5 No study directly compared specific tests, and therefore it was not possible to evaluate the cost effectiveness of specific surveillance tools. A second limitation was the lack of quality of life data for follow up of patients with colorectal cancer, although one study has assigned a quality adjusted life year of 0.83 based on data from a chemotherapy trial.29 As intensive follow up does not significantly raise patient anxieties, this may prove a reasonable estimate, but further studies are required before these values can be incorporated into health economic calculations.34 The third limitation was the absence of individual patient data, such that we had to make assumptions about survival and recurrence patterns. These assumptions, however, held firm, as shown by the sensitivity analysis. The fourth limitation was that the included trials were conducted over a decade ago, and thus did not include aspects of contemporary practice (for example, multidisciplinary management) that could impact on costs or effectiveness. The final limitation is that we were unable to compare intensive follow up with absolute no follow up (zero surveillance costs), although in modern practice the acceptability of a no follow up strategy is questionable.35

Implications for health policy

The strategy of intensive follow up after curative resection for colorectal cancer is economically justified and should be normal practice. As a screening type health strategy, intensive follow up of patients with colorectal cancer compares favourably with screening for breast and colorectal cancers and for abdominal aortic aneurysms.18 36 37 Moreover, compared with population screening, these patients are self defined, well motivated, and compliant.38

Factors that could affect future results

Large randomised trials are needed to evaluate the efficacy of specific surveillance tools.5 35 These studies must include cost effectiveness analyses and quality of life assessments, in addition to other issues that were beyond the scope of our analysis. Firstly, costs beyond five years after initial treatment need to be considered as a proportion of patients with recurrences undergoing salvage therapy will have delayed second recurrences. Secondly, the included trials predated the widespread use of chemotherapy for advanced colorectal cancer or as an adjunct to salvage reoperation, approaches that may improve survival.39 This potential improvement in effectiveness will be coupled with increased costs. However, as maximising treatments costs (without changes in surveillance costs) had only minor effects on incremental cost effectiveness ratios, it is tempting to speculate that chemotherapy, expensive when considered in selected groups, may be cost effective in the wider context.39 Thirdly, our analysis treated local, hepatic, and other site recurrences as mutually exclusive events. In practice there is overlap, which will tend to attenuate costs. Fourthly, we did not address societal perspectives, such as travel and time off work, which are important determinants of compliance in screening for colorectal cancer and may thus be relevant in cancer surveillance.40 Fifthly, it was not possible to estimate capital costs for additional laboratory and clinical capacity. Finally, we did not consider the benefits of intensive follow up from factors other than salvage for recurrent disease—for example, psychosocial or improved treatment of coincidental disease. Evidence from our meta-analysis shows that some reductions in all cause mortality associated with intensive follow up are attributable to factors other than salvage therapy (unpublished data).5