4 Evidence and interpretation
4.1.1 One randomised, open-label, active-controlled trial with 1987 participants, the Xeloda – Adjuvant Chemotherapy Trial (X-ACT), investigated the efficacy and safety of capecitabine treatment versus 5-FU/FA treatment (bolus Mayo Clinic regimen) in the postoperative adjuvant setting in patients with stage III (Dukes' C) colon cancer. Apart from the protocol-specified analyses, ad hoc analyses were carried out at the request of the US Food and Drugs Administration (FDA).
4.1.2 For the primary endpoint of disease-free survival, the study was powered to establish non-inferiority, defined so that the upper limit of the 95% confidence interval (CI) around the hazard ratio was no more than 1.25. The median age of participants was 62 years in the capecitabine arm and 63 years in the 5-FU/FA arm.
4.1.3 After a median follow-up of 3.8 years, 35% of patients in the capecitabine arm had experienced disease recurrence (relapse or new occurrence of colon cancer) or died, compared with 39% in the 5-FU/FA arm. The hazard ratio for recurrence was 0.87 (95% CI, 0.75 to 1.00). Updated analyses, not specified in the protocol, showed that with longer follow-up (minimum 3 years and median 4.4 years) capecitabine remained at least as effective as 5-FU/FA.
4.1.4 Overall survival data were not mature at the time of the primary (specified) and secondary (ad hoc) analyses. However, at 3.8 years median follow-up, 80% and 77% of patients were alive in the capecitabine and 5-FU/FA arms, respectively.
4.1.5 Quality of life (QoL) was assessed using the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire (EORTC-QLQ-C30), with global health status being the primary parameter for the QoL evaluation. In both treatment groups, scores for global health status were constant over time (from baseline to 25 weeks of trial treatment) and there were no major (statistically significant) differences between the two groups.
4.1.6 Severe stomatitis and hair loss were significantly more common in the participants treated with 5-FU/FA. In addition, neutropenia, as a clinical adverse event requiring medical intervention, was significantly less common in participants treated with capecitabine. The only treatment-related adverse events occurring statistically significantly more frequently with capecitabine than with 5-FU/FA were hand–foot syndrome (p < 0.001) and hyperbilirubinaemia.
4.1.7 A submission by a professional organisation reports that 'when given a choice, most cancer patients prefer oral instead of intravenous therapy, but only if the treatment is equally effective; patients cite increased convenience, less distress over repeated intravenous access and more control over their own treatment as major factors'.
4.1.8 Two phase III, randomised active-controlled trials that compared oxaliplatin with standard treatment were identified by the Assessment Group. The first was the Multicenter International Study of Oxaliplatin/5-fluorouracil and leucovorin in the Adjuvant Treatment of Colon Cancer (MOSAIC) trial. (This was open-label and had 2246 participants – 60% with stage III and the remainder with stage II colon cancer.) The second trial was the National Surgical Adjuvant Breast and Bowel Project (NSABP C-07) trial. (This had 2492 participants – 71% with stage III and the remainder with stage II colon cancer.) The NSABP C-07 trial was only available in abstract form.
4.1.9 In the MOSAIC trial, oxaliplatin was combined with 5-FU/FA in the de Gramont regimen (an infusional regimen) and compared with 5-FU/FA alone (also given in the de Gramont regimen). In the NSABP C-07 trial, standard treatment consisted of 5-FU/FA administered via a bolus regimen (Roswell Park) and this was subsequently compared with oxaliplatin in combination with the same bolus regimen. In addition to the protocol-specified analyses, ad hoc analyses were carried out at the request of the FDA in the MOSAIC trial. The median age of trial participants was 61 years and 60 years in the oxaliplatin plus 5-FU/FA and the 5-FU/FA alone groups, respectively. NSABP C-07 did not report age at baseline. QoL data were not routinely collected within either trial.
4.1.10 In both trials the addition of oxaliplatin to 5-FU/FA, albeit administered via different regimens, led to a statistically significant reduction in rate of relapse when compared with 5-FU/FA monotherapy. Analysis of disease-free survival at 3 years showed a hazard ratio for recurrence of 0.77 (95% CI, 0.65 to 0.91) in the MOSAIC trial (median follow-up 37.9 months; intention to treat analysis), and 0.79 (95% CI, 0.67 to 0.93) in the NSABP C-07 trial (median follow-up 34 months, according to protocol analysis). Additional analyses on MOSAIC – requested by regulatory authorities – showed a 24% reduction in the rate of relapse (improved disease-free survival) at a median follow-up of 4 years (hazard ratio for recurrence 0.76; 95% CI, 0.65 to 0.90).
4.1.11 Overall survival results for MOSAIC and NSABP C-07 are to be calculated at follow-up periods of 6 years and 5 years respectively. No mature data are available for MOSAIC at present, and the interim 3-year and 4-year analyses report no statistically significant differences in overall survival between the study groups; 88.2% and 87.0% still alive at 38 months in the oxaliplatin plus 5-FU/FA and the 5-FU/FA arms, respectively, and a hazard ratio for death of 0.89 (95% CI, 0.72 to 1.09) in the 4-year analysis favouring the addition of oxaliplatin. The abstract of the NSABP C-07 trial did not report overall survival.
4.1.12 Only in the MOSAIC study were subgroups prespecified according to stage of the disease, with results reported separately. For participants with stage III colon cancer, the hazard ratio for recurrence was found to be 0.76 (95% CI, 0.62 to 0.92) at 3 years, and 0.75 (95% CI, 0.62 to 0.90) at 4 years. The percentages of people experiencing relapse or death in the oxaliplatin plus 5-FU/FA and the 5-FU/FA arms were 26.9% and 33.5%, respectively. The hazard ratio for death for stage III patients in MOSAIC was 0.86 (95% CI, 0.66 to 1.11) at 3 years. Although the MOSAIC study was adequately powered to demonstrate improved survival outcomes in patients with stage II (40% of total population) or III (60% of total population) disease, the study was not powered to detect a therapeutic effect by subgroup.
4.1.13 In the MOSAIC trial more participants discontinued treatment because of adverse events in the oxaliplatin plus 5-FU/FA group (14.4%) than in the group receiving 5-FU/FA monotherapy (5.6%). Neutropenia and paraesthesia are the toxicities most reported for oxaliplatin plus 5-FU/FA. Grade 3 peripheral neuropathy was observed in 12.4% of patients during treatment (median number of cycles 12; equivalent to about 6 months of chemotherapy), and in 1.1% and 0.5% of patients at 12 months and 18 months follow-up, respectively. Other frequent adverse events in the oxaliplatin plus infusional 5-FU/FA group were diarrhoea, nausea and vomiting.
4.1.14 Grade 3 neurotoxicity was observed in 8% of patients receiving oxaliplatin plus bolus 5-FU/FA in the NSABP C-07 trial compared with 1% of patients receiving 5-FU/FA alone. After 1 year of follow-up, grade 3 neuropathy in the oxaliplatin plus 5-FU/FA group remained in 0.5% of patients. The Assessment Group noted that the incidence of grade 3/4 diarrhoea in the combination arm was much higher than that observed in MOSAIC: approximately 40% and 11%, respectively.
4.1.15 A joint submission by professional organisations reports that oxaliplatin causes a unique cold-related peripheral neuropathy affecting over 90% of patients during treatment, and that symptoms are still present to a greater or lesser degree 18 months after completing treatment in 24% of patients.
4.1.16 Another submission by a professional group refers to the combined incidence of grade 2 and 3 neurosensory symptoms as reported in the MOSAIC trial. It notes that 18 months after completion of treatment, 3.9% of patients had persistent debilitating symptoms.
4.1.17 Three randomised comparisons of bolus versus infusional regimens have been published. Only two studies followed individuals for 5 years – a suitable proxy for long-term survival. The evidence reviewed suggests that infusional intravenous 5-FU-based adjuvant therapy is equivalent to, but has relatively less toxicity than, traditional bolus 5-FU/FA in extending survival and QoL. However, there are concerns about catheter-associated complications, patient inconvenience and the cost of infusional treatment. In the adjuvant setting, the weekly intravenous bolus 5-FU/FA for 30 weeks (QUASAR regimen) is most commonly used in the NHS in England and Wales. However, there remains significant geographical variation in the 5-FU-based regimens currently in use in the UK.
4.2.1 The Assessment Group reviewed three published economic evaluations, two of which were submitted by manufacturers. It also presented its own three-state Markov model to estimate the cost effectiveness of oxaliplatin plus 5-FU/FA versus 5-FU/FA alone, and of capecitabine versus 5-FU/FA alone.
4.2.2 In the Assessment Group model, hypothetical individuals were assumed to move between three states: alive without relapse, alive with relapse, and dead. Transition probabilities in the Assessment Group model and one of the manufacturer models were estimated from the disease-free survival curve and the partitioned overall survival curves for patients with and without relapse. This joint modelling of disease-free and overall survival differs from the approach adopted in the model submitted by the manufacturer of capecitabine, where there was independent modelling of relapse-free survival and overall survival with inconsistent results.
4.2.3 Key assumptions used in the Assessment Group model were as follows.
Overall survival of people who relapse is assumed to be independent of the time of relapse.
Overall survival of people who relapse is equivalent to that of patients who are initially diagnosed with advanced (stage IV – Dukes' D) colorectal cancer.
All relapses occur within the 5 years following resection of the primary tumour.
Overall survival of people alive and disease free at 5 years is similar to the survival in the general population, adjusting for age and sex.
People who relapse are assumed to receive first-line 5-FU/FA followed upon progression by single-agent irinotecan.
People receiving 5-FU/FA via the de Gramont regimen are assumed to receive their treatment on an outpatient basis.
All of these assumptions, except for the last two, are also used in the model submitted by the manufacturer of oxaliplatin. Instead of using the cost of a specific chemotherapy regimen to estimate cost of relapse, the manufacturer's model uses an average cost of relapse that is calculated from a distribution using costs of treatment for four different types of relapse.
4.2.4 Evidence for estimating preference-based utilities for the different health states is scarce. The submissions from the manufacturers of both drugs based their utility estimates on a study of 173 patients with colorectal cancer (40 of whom had stage III disease). In this study, generic and cancer-specific QoL tools were administered at regular intervals following diagnosis, starting at 13 months post diagnosis. Although the study did not differentiate between patients who relapsed and those who did not, both submissions used a disutility of approximately 0.2 for people who experienced relapse. In the manufacturer submission for oxaliplatin, utilities while on treatment were also corrected for adverse events.
4.2.5 The Assessment Group noted that because the study used in the manufacturers' submissions started long after diagnosis, and a relatively small proportion of patients had stage III disease, they could only use data from this study to estimate the utility for people in remission (0.92).
4.2.6 From a second study that elicited utilities from 81 patients with colorectal cancer with all stages of the disease (including those with stage III undergoing resection and chemotherapy), utilities were taken for those people undergoing treatment without adverse events (0.7) and with adverse events (0.63), as well as for those who relapse (0.24).
4.2.7 The key cost driver of the economic analysis submitted by the manufacturer was the difference in the drug acquisition and administration costs between the capecitabine and 5-FU/FA (Mayo Clinic regimen) arms. Acquisition costs were approximately £1400 higher for the capecitabine arm, whereas administration costs and costs associated with adverse events were lower for the capecitabine arm – approximately £4750 and £300 per patient for 5-FU/FA and capecitabine, respectively.
4.2.8 Primarily because of reduced drug administration costs associated with capecitabine (long-term costs were assumed to be approximately equal), the manufacturer's submission concluded that capecitabine is cost saving compared with 5-FU/FA, costing on average £3653 less per patient. Combined with lifetime extrapolated relapse-free and overall survival benefits, treatment with capecitabine also leads to a gain of 0.75 quality-adjusted life years (QALYs) in the manufacturer's model. The one-way sensitivity analyses and extreme analysis showed that the only significant uncertain driver for varying cost effectiveness is the cost per administration visit. Scenario analyses on the regimen used for 5-FU/FA indicate that capecitabine remains cost saving whichever regimen is used.
4.2.9 In the Assessment Group model, total cost savings from the use of capecitabine compared with the Mayo Clinic 5-FU/FA regimen are slightly less than those reported in the manufacturer's submission (£3320). This is primarily due to the differences between the two models in the costs associated with relapse and a difference in pharmacy costs between capecitabine and 5-FU/FA that was included in the Assessment Group model but not in the manufacturer's submission. The higher QALY gain associated with capecitabine in the Assessment Group model (0.98 QALYs) appears to be attributable to the different methods used to estimate survival. In all the one-way sensitivity analyses, capecitabine treatment results in a cost saving when compared with 5-FU/FA in the Mayo Clinic regimen.
4.2.10 Two published economic analyses that considered oxaliplatin plus 5-FU/FA in the adjuvant setting were included in the Assessment Report. One of these analyses was conducted from a non-UK perspective and used survival estimates from trials of oxaliplatin plus 5-FU/FA in advanced colorectal cancer that are unlikely to be representative of survival outcomes for patients receiving adjuvant chemotherapy. Further analysis by the Assessment Group of the marginal cost and survival results given in the paper suggested that the cost per life year gained of the addition of oxaliplatin to 5-FU/FA is £24,952. An abstract of another economic analysis was presented at the 2005 meeting of the American Society of Clinical Oncology (ASCO) and updated to form the basis of the manufacturer's submission to the appraisal (see 4.2.11 below). The cost per life year gained associated with oxaliplatin plus infusional 5-FU/FA in this study was estimated to be US$27,300.
4.2.11 The economic model submitted by the manufacturer was based on patient-level data from the MOSAIC trial. It used observed mortality and disease-free survival, as well as the relationship between disease-free survival and overall survival, to estimate the difference in overall survival between the two treatment arms. Data from MOSAIC that relate to the cohort of patients with stage III colon cancer were used to report a base-case cost per QALY gained (CQG) of £4805 for oxaliplatin plus infusional 5-FU/FA versus infusional 5-FU/FA alone, calculated over a 50-year time horizon. This CQG estimate consists of a difference in costs of £3267 and a difference in benefits of 0.68 QALYs. When a one-way sensitivity analysis was performed, and benefits and costs were limited to those within trial data, the CQG increased to £56,780. There was no other one-way sensitivity analysis that resulted in a very different estimate from that of the base case – not even a doubling of the disutility for relapse (to 0.4). The manufacturer suggests that the difference between its base-case results (for stage II and III combined – CQG of £7210) and those of the published economic analyses is probably due to the lower drug acquisition costs of oxaliplatin in the UK compared with the US.
4.2.12 In an addendum to its submission the manufacturer presented a second cost-effectiveness analysis, now based on the NSABP C-07 trial. Equivalent efficacy (0.68 QALYs gained) was assumed for oxaliplatin plus bolus 5-FU/FA (Mayo Clinic regimen) and oxaliplatin plus 5-FU/FA (de Gramont). When combined with a cost difference of £4246 between oxaliplatin plus bolus 5-FU/FA (Mayo Clinic regimen) and 5-FU/FA alone (Mayo Clinic regimen), this analysis resulted in a CQG estimate of £6244 for oxaliplatin plus 5-FU/FA relative to 5-FU/FA alone.
4.2.13 Incremental benefits in the Assessment Group model were greater than those in the manufacturer's submission (1.33 versus 0.68 QALYs) when oxaliplatin plus 5-FU/FA (de Gramont) was compared with 5-FU/FA alone (de Gramont). Reasons for this lie in the differences in methods used for long-term extrapolation of survival curves and utility estimates used for those people that relapse in the economic model. When incremental benefits were combined with a cost difference that was also greater than that in the manufacturer's submission (£3940), the Assessment Group model resulted in an estimated CQG of £2970. This cost difference arises because the manufacturer's model uses differential costs of relapse for the 5-FU/FA and combination arms, whereas the Assessment Group model uses costs of relapse unrelated to the intervention received in the adjuvant setting. Furthermore, unlike the manufacturer's submission, the Assessment Group model included differences in pharmacy costs between oxaliplatin plus 5-FU/FA and 5-FU/FA alone. Finally, by setting the model parameters to the worst-case scenario, the estimated CQG in the Assessment Group model was increased to £7587.
4.2.14 In the absence of studies directly comparing oxaliplatin plus 5-FU/FA (de Gramont) with capecitabine, the Assessment Group modelled indirect comparisons of oxaliplatin plus 5-FU/FA (de Gramont) versus capecitabine, and oxaliplatin plus 5-FU/FA versus bolus 5-FU/FA, in the adjuvant treatment of stage III colon cancer.
For the first comparison, two approaches were taken. The first used the absolute predicted long-term survival and cost data of the Assessment Group model, and the second used the marginal cost effectiveness of oxaliplatin plus 5-FU/FA (de Gramont) and of capecitabine against the comparator arms of MOSAIC and X-ACT, respectively. The estimated CQGs for oxaliplatin plus 5-FU/FA (de Gramont) compared with capecitabine were £12,874 (£16,283 additional costs and 1.26 QALYs) and £46,814 (£16,283 additional costs and 0.35 QALYs) for the first and second approach, respectively.
The second comparison, using data from the MOSAIC and X-ACT trials, resulted in an estimated CQG of £5777 for oxaliplatin plus 5-FU/FA (de Gramont) versus bolus 5-FU/FA (Mayo Clinic regimen), consisting of £12,963 in additional costs and 2.24 QALYs.
4.3.1 The Committee reviewed the data available on the clinical and cost effectiveness of capecitabine as monotherapy, and oxaliplatin in combination with 5-FU/FA, in the adjuvant treatment of stage III (Dukes' C) colon cancer. It considered evidence on the nature of the condition and on the value that people with colon cancer, those who represent them, and clinical experts placed on the benefits of the two drugs in the adjuvant treatment of the condition. It was also mindful of the need to take account of the effective use of NHS resources.
4.3.2 In reviewing the evidence of clinical effectiveness for both capecitabine and oxaliplatin the Committee noted that to date, no statistically significant benefit on overall survival of these interventions over their comparators has been reported from the trial populations of X-ACT, MOSAIC and NSABP C-07. However, the Committee considered it reasonable to assume that for the purpose of the economic model, the 3-year disease-free survival benefits reported in these clinical trials would predict 5-year overall survival benefits for the adjuvant treatment of people with stage III (Dukes' C) colon cancer.
4.3.3 The Committee considered the fact that participants in trials for adjuvant treatment of stage III colon cancer are often younger than those who would be treated in clinical practice. It heard testimony from clinical experts that it is reasonable to extrapolate these results to older patients, that appropriately selected older people show a relatively good tolerability profile to the drugs, and that the effect on overall survival in those older people in clinical practice is comparable with that seen in the younger trial participants. Additionally, the Committee heard evidence from the Assessment Group that, if an older cohort of people was used in the model that was more representative of the population under consideration, and survival benefits for this group were assumed to be equivalent to those for the group of trial participants, the cost-effectiveness estimates would not materially change.
4.3.4 The Committee carefully considered the rates of adverse events reported for capecitabine and oxaliplatin in the three pivotal clinical trials. It particularly noted sensory neuropathy following treatment with oxaliplatin. It heard testimony from clinical experts that not only grade 3 but also grade 2 neuropathies can be severely debilitating, and continue long term in a significant percentage of patients. This is particularly problematic in adjuvant treatment. It further heard that the appearance of sensory neuropathy was not predictable, but the degree to which individuals are affected by such adverse events depends to some extent on their fitness.
4.3.5 The Committee considered the Assessment Group's assumptions and sensitivity analyses used in its economic model and noted the following points.
They expressed some concerns regarding the utility values adopted, but accepted that these were the best available from the literature and gave a plausible set of results.
They were concerned that the adverse effects of the drugs, particularly oxaliplatin-induced sensory neuropathy, could have been undervalued by the Assessment Group.
They noted that recurrence of a tumour more than 5 years after first receiving adjuvant treatment is possible; however, the relevant sensitivity analysis from the economic model did not affect the cost effectiveness materially.
They were aware that when findings from the FOCUS and GERCOR trials (which looked at advanced colorectal cancer) are used in estimating costs of relapse, these costs are likely to be an underestimate of the real costs; however, it was accepted that imputing higher costs of relapse would lead to more favourable cost effectiveness for capecitabine and oxaliplatin plus 5-FU/FA.
They considered that the use of oxaliplatin in the adjuvant treatment of colon cancer could restrict its use in advanced colorectal cancer, but that this would depend on when the relapse was experienced after first use of the drug in the adjuvant setting. They were persuaded that it was most important to achieve the benefits of adjuvant treatment early in order to avoid or delay relapse.
4.3.6 Overall, the Committee accepted that capecitabine as monotherapy, and the combination of oxaliplatin plus 5-FU/FA, should be considered as cost-effective options for the adjuvant treatment of people with stage III (Dukes' stage C) colon cancer.
4.3.7 The Committee was mindful of the substantial uncertainty within the indirect comparisons reported in the economic analyses by the Assessment Group. It therefore did not consider the comparison between oxaliplatin plus 5-FU/FA and capecitabine to be informative for guidance.
4.3.8 The Committee was clear that, given the different toxicities of the drugs, particularly the risk of longer-term sensory neuropathy with oxaliplatin, the choice of therapy should be made in clear consultation with the patient, and in careful consideration of the patient's performance status.