Reducing uncertainties about the effects of chemoradiotherapy for cervical cancer: individual patient data meta-analysis

For the main analysis, individual patient data were provided for 15 of the 25 trials identified, with two of the included trials having additional, adjuvant chemotherapy for the women in the chemoradiotherapy group. The data could not be located for six trials (814 patients) and the reviewers were unable to make contact with the investigators for four trials (299 patients). The available data represented a total of 3452 women, who had been recruited to the trials between 1987 and 2006. It included 118 patients who had been excluded from the analyses done by the original researchers. In total, 85% of the women in trials of platinum-based chemoradiotherapy (using cisplatin) were available for analyses, along with nearly 80% from studies with 5-Fluorouracil, mitomycin-C or both as the chemotherapy.

The data collected for each woman included prognostic variables recorded at randomisation, details of the type of treatment used and follow-up information. The baseline variables included characteristics of the woman (such as her age and performance status) and of her cancer (such as histology, stage and grade). In addition to her random allocation, the treatment information included the type and dose of chemotherapy and radiotherapy dose and duration. Dates of randomisation, local and distant recurrence, and survival or death were also collected, along with details on the cause of death and any early or late adverse effects.

The data were used to analyse the time-to-event for the different outcomes in each trial, so that these could produce hazard ratios for the meta-analyses as a whole, employing a fixed-effect model with the logrank test stratified by trial. The results are displayed with Kaplan-Meier curves and estimates of the absolute effects of treatment were calculated by applying the hazard ratios to the observed survival pattern for women in the control groups of the trials.

The main analyses were based on 3104 women in the 13 trials that did not use adjuvant chemotherapy following the chemoradiotherapy. The meta-analysis of these trials produced a hazard ratio of 0.81 (95% CI 0.71 to 0.91, p<0.001) which, when applied to the control group survival of 60% at five years, would give an absolute improvement in survival of 6% at that time. The meta-analysis for disease-free survival gave a hazard ratio of 0.78 (95% CI 0.70 to 0.87, p<0.001), which translates to an absolute benefit of 8% at 5 years (58% compared to 50% for the radiotherapy group). It was possible to estimate the hazard ratios for survival in three of the ten trials for which individual patient data were not available. Two of these trials would contribute to the main group of 13 trials without adjuvant chemotherapy and one to the group of trials that used additional chemotherapy after chemoradiotherapy. The reviewers found that incorporating the results from these trials into the individual patient data meta-analysis did not materially change the results for either group but do not include the actual results in the Cochrane review.

When the individual patient data meta-analyses were subdivided to investigate the effects of treatment in various types of trial and for different groups of women, the results tended to be consistent across the subgroups. There was no evidence of a difference in the size of the effect between the trials that used platinum-based chemoradiotherapy and those that used non-platinum-based regimens. The hazard ratios for survival were 0.83 (95% CI 0.71 to 0.97) and 0.77 (95% CI 0.63 to 0.94) respectively, with the test for subgroup interaction being non-significant (p=0.57). Similarly, there was no evidence of different effects for different radiotherapy doses, total planned duration of radiation or cycle length or dose intensity of cisplatin regimens. In the patient subgroup analyses, there was a suggestion of a trend in the relative effect of chemoradiotherapy on overall survival by tumour stage, with the benefit of chemoradiotherapy decreasing with increasing stage (p=0.017). When the hazard ratios for each stage were applied to the survival pattern of women with the relevant stage in the control groups of the trials, the 5-year survival benefits for chemoradiotherapy were 10% for stages 1b to 2a, 7% for stage 2b, and 3% for women with stage 3 to 4a cancer. However, this trend was not repeated for the outcome of disease free survival (p=0.073). None of the other subgroup analyses suggested that the effect of chemoradiotherapy differed in groups of women defined by age, histology, tumour grade or whether they had pelvic lymph node involvement.