Capecitabine for the treatment of advanced gastric cancer

The manufacturer's submission considered the use of capecitabine with epirubicin plus cisplatin or oxaliplatin compared with fluorouracil with epirubicin plus cisplatin or oxaliplatin, and capecitabine plus cisplatin compared with fluorouracil plus cisplatin. The population was people with locally advanced (that is, the disease had spread to regional lymph nodes) or metastatic (that is, the disease had spread beyond the regional lymph nodes to other parts of the body) inoperable gastric cancer. This was in line with the scope, which restricted the population to people with inoperable advanced gastric cancer. The manufacturer reported details of two phase 3 multicentre randomised controlled trials (REAL-2 and ML17032). These trials assessed the non-inferiority of capecitabine compared with fluorouracil for the treatment of advanced gastric cancer. REAL-2 compared ECX and EOX combinations with epirubicin/cisplatin/fluorouracil (ECF) and epirubicin/oxaliplatin/fluorouracil (EOF) combinations. ML17032 compared CX with cisplatin/fluorouracil (CF).

The REAL-2 trial was an open-label UK multicentre study that enrolled adults with advanced carcinoma of the oesophagus, oesophageal-gastric junction or stomach. People were included in the trial if they had locally advanced or metastatic adenocarcinoma, squamous cell carcinoma or undifferentiated carcinoma. In addition the primary tumour had to be classified as inoperable. People were randomised to receive capecitabine plus platinum-based chemotherapy regimens: ECX regimen (n=241), ECF regimen (n=249), EOX regimen (n=239) or EOF regimen (n=235). The doses were as specified in the individual SPCs of each drug. In all cases treatment was repeated every 3 weeks for 8 cycles in the absence of progressive disease or unacceptable toxicity. The primary endpoint was to determine non-inferiority of overall survival in people receiving capecitabine compared with those receiving fluorouracil, and non-inferiority of overall survival in people receiving oxaliplatin compared with those receiving cisplatin. The null hypothesis of non-inferiority of the capecitabine regimen was rejected if the upper limit of the 95% confidence interval (CI) around the hazard ratio (HR) for overall survival was more than 1.23.

The manufacturer reported that REAL-2 met the two primary non-inferiority endpoints, and that there was a trend towards improved survival in favour of both capecitabine over fluorouracil and oxaliplatin over cisplatin. The manufacturer also reported that the trial showed non-inferiority in terms of overall survival for capecitabine; the HR adjusted for performance status, extent of disease and age was 0.89 (95% CI 0.77 to 1.02) in the per-protocol population. This was based on the comparison of ECF and EOF versus ECX and EOX. The manufacturer also reported that for the secondary endpoints, there was no significant difference in progression-free survival between the capecitabine and the fluorouracil arms or between the cisplatin and the oxaliplatin arms. There was minimal quality-of-life data reported in REAL-2 and there were no differences between the mean scores at baseline and 12 weeks between any of the groups on the General Health Status subscale of the European Organization for Research and Treatment of Cancer (EORTC)-30 questionnaire.

ML17032 was an open-label study that enrolled people with advanced gastric adenocarcinoma. Adults were included in the trial if they had histologically confirmed gastric adenocarcinoma with advanced and/or metastatic disease and at least one measurable lesion according to Response Evaluation Criteria in Solid Tumors (RECIST) that had not been irradiated and a Karnofsky performance status score of 70% or higher. People were randomised to receive cisplatin (80 mg per m² intravenously, day 1) plus fluorouracil (800 mg per m² intravenously, days 1 to 5 as a continuous infusion; (n=156) or cisplatin (80 mg per m² intravenously, day 1) plus capecitabine (1,000 mg per m² orally, twice daily, days 1 to 14; n=160). The primary endpoint was non-inferiority of progression-free survival. The null hypothesis of non-inferiority of the capecitabine regimen was rejected if the upper limit of the 95% CI around the HR for progression-free survival was more than 1.25.

In the per-protocol population, there was non-inferiority of progression-free survival in people receiving CX compared with those receiving CF (adjusted HR 0.85, 95% CI 0.65 to 1.11). The manufacturer reported that the results showed a trend towards improved progression-free survival with CX compared with CF in the unadjusted analysis. The median overall survival was 10.5 months for CX compared with 9.3 months for CF (HR 0.85, 95% CI 0.64 to1.13). It was also reported by the manufacturer that the trial demonstrated non-inferiority of capecitabine compared with fluorouracil for the secondary end-points of overall survival, response rate, mean time of response and duration of response. No quality of life data were collected in ML17032.

The manufacturer also reported a published meta-analysis that combined the individual data from 1,318 people taking part in the REAL-2 and ML17032 trials. The aim of the meta-analysis was to test whether capecitabine was superior to fluorouracil within the double and triple combination chemotherapy regimens for people with advanced oesophago-gastric cancer. The primary endpoint was overall survival and the secondary endpoints were progression-free survival and response rate. The median overall survival for the intention-to-treat population was 285 days (95% CI 265 to 305 days) for people treated with fluorouracil (n=664) and 322 days (95% CI 300 to 343 days) for people treated with capecitabine (n=654). This resulted in an unadjusted HR of 0.87 (95% CI 0.77 to 0.98, p=0.027) in favour of capecitabine. There was no evidence of any statistically significant heterogeneity of treatment effect according to baseline patient characteristics (such as age, disease site and histology). The meta-analysis reported that superiority of capecitabine over fluorouracil was maintained with multivariate analyses (adjusted HR 0.87, 95% CI 0.77 to 0.98, p=0.02). The meta-analysis also reported a non-statistically significant trend towards improved progression-free survival in people receiving capecitabine (unadjusted HR 0.91, 95% CI 0.81 to 1.02, p=0.093).

In REAL-2, grade 3 and 4 neutropenia was more common in the ECX arm compared with the ECF arm and there was an increased level of fatigue in the EOF arm compared with the EOX arm. Stomatitis occurred more frequently and with greater severity in the CF arm than in the CX arm in ML17032, while hand–foot syndrome was more common in the CX arm. The ML17032 trial also reported that adverse events leading to dose modification were more common in the CX arm (62%) compared with the CF arm (48%).

On the basis of equivalent clinical effectiveness, similar safety and improved patient convenience, a cost-minimisation model was developed to evaluate the costs for each regimen. The manufacturer reported that this captured all significant incremental direct costs relating to switching from fluorouracil-based therapies to capecitabine-based therapies. The model considered both the drug acquisition and drug administration costs for all the regimens evaluated. People entered the model at the start of treatment when they received either capecitabine or fluorouracil and left the model after 5.5 cycles (21 days per cycle), which was the time horizon of the model. The costs of treatment-related adverse events were not included. The overall tolerability profile of capecitabine was considered by the manufacturer to be similar and at least as good as fluorouracil. The manufacturer also stated that the adverse events associated with the method by which fluorouracil is administered, such as central-line complications, can be expensive to manage. This meant that the costs associated with the management of adverse events with capecitabine were unlikely to be higher than those associated with fluorouracil. Therefore, the non-inclusion of adverse events costs in the model would be expected to favour fluorouracil.

The manufacturer stated that the economic evaluation of capecitabine was undertaken within its licensed indication for the first-line treatment of advanced gastric cancer in combination with a platinum-based regimen. There were three sets of analyses: a comparison of ECX with ECF, EOX with EOF, and CX with CF. A total of six regimens were therefore analysed in the cost-minimisation model. The dosages in each regimen were analysed according to the SPC for each treatment and no drug wastage was taken into account. The manufacturer also conducted some additional sensitivity analyses that included one-way sensitivity analyses, scenario analyses, a worst case scenario analysis and threshold analyses.

The base-case results included all the drug acquisition and administration costs for all the regimens considered in the submission. All capecitabine-based regimens were shown to be cost saving compared with equivalent fluorouracil-based regimens. The overall NHS cost saving for switching from ECF to ECX regimens was £1,620. The overall NHS cost saving for switching from EOF to EOX regimens was £1,572. For the double combination chemotherapy regimens the overall NHS saving of switching from CF to CX was £4,210. All the results of the sensitivity and scenario analyses conducted by the manufacturer supported the base case and suggested that capecitabine-based regimens were cost saving compared with fluorouracil-based regimens.

The ERG considered that the clinical-effectiveness evidence presented by the manufacturer reflected the available relevant evidence. It noted that the ML17032 trial was underpowered since the trial had only 50% power to detect statistically significant non-inferiority. The REAL-2 and the ML17032 trials used appropriate outcomes, but there were limited data on health-related quality of life. The ERG also noted that in clinical practice, fluorouracil and capecitabine would be administered in lower doses in the double combination chemotherapy regimen compared with the doses used in ML17032.

Overall, the ERG considered the manufacturer's approach to the economic evaluation reasonable and that the cost-minimisation analysis used in the submission was acceptable. The ERG stated that there was minimal change to the savings when many of the assumptions used were explored. The ERG pointed out that treatments cannot be considered to be exactly equivalent due to uncertainty in estimating their effectiveness. By conducting a cost-minimisation analysis, the manufacturer did not address the uncertainty around the estimates of efficacy. However, they performed a threshold analysis but the ERG noted that modelling a scenario in which fluorouracil was more effective was unlikely to be relevant to the determination of capecitabine's cost effectiveness. The ERG noted that adverse events had not been included in the cost-minimisation model because the manufacturer assumed that the costs associated with adverse event management were unlikely to be higher for capecitabine than fluorouracil. The ERG highlighted that rare adverse events may not have been identified because of the large sample sizes needed to detect these. Therefore, the ERG felt that some uncertainty remained about treatment-related adverse events associated with capecitabine and fluorouracil regimens. The ERG also noted that number of cycles used in the model did not represent the number used in UK clinical practice: the maximum number of cycles in clinical practice is usually six, but in REAL-2 the median (rather than the maximum) number of cycles received was six.

The ERG noted some additional areas of uncertainty. In the model, the manufacturer calculated capecitabine costs based on milligrams used. The ERG considered that in clinical practice, this would be rounded to match the available tablets. It also considered that in calculating fluorouracil costs, wastage had not been taken into account. The ERG noted that when calculating epirubicin, cisplatin and oxaliplatin costs, the manufacturer used an average of the NHS list prices. In practice, the NHS is likely to prefer the cheapest product. The ERG also considered that, in practice, the dose of capecitabine may be reduced by 25% to 50% because of toxicity. The ERG undertook exploratory analyses that took into account many of the above areas of uncertainty. In all analyses capecitabine was cost saving compared with fluorouracil.

Full details of all the evidence are in the manufacturer's submission and the ERG report.