Article type
Year
Abstract
Objective: We aim to evaluate the performance of meta-analyses using median survival times by comparing its results with the results of IPD meta-analyses of the same randomized trials. We will focus on meta-analyses involving cancers with high mortality rates, e.g. lung cancer.
Methods: We will calculate a pooled ratio of median survival times (Simes, Statist. Med. 6, 11-29). The variance of the ratio of the median survival times in each trial will be approximated using the total number of patients. The median survival times will be calculated from the IPD database rather than taken from publications. In this way we will compare like with like. The same trials, patients and extended follow-up will be used in each analysis. The IDP meta-analyses on which the comparisons are made, consist of meta-analyses carried out by the Institut Gustave Roussy, the Meta-Analysis Group in Cancer (MAGIC) and the Meta-Analysis Group from the Medical Research Council Clinical Trials Unit.
Results: The three lung cancer meta-analyses, for which the results of homogenous trial sets are tabulated below, are the prophylactic cranial irradiation overview on small cell lung cancer (lung 1), the locally advanced (lung 2) and advanced settings of the non-small cell lung cancer meta-analysis on chemotherapy (lung 3). More comparisons with other meta-analyses will be presented at the meeting. The same patterns can be discovered in the analyses based on the MST method as in the real IPD meta-analyses. The differences in the estimation of the overall hazard ratio (HR) for the IDP method and the estimation of the pooled median ratio (MR) for the proposed MST method stem from differences of the estimations of the hazard ratio at the trial level, in particular when the differences between the survival curves are not constant over time.
[Table omitted]
Conclusions: The results of a simple summary data method based on the median survival time are compared with the results from IDP meta-analyses with high mortality rates. The only difference in estimation relates to bias of the statistical method. Under- or overestimations of the overall hazard ratio may occur. However, based on the comparisons with IDP meta-analyses, the simple MST method seems to have practical and exploratory value. Other differences will come into play if used as part of a meta-analysis of the published literature (unpublished trials, patient exclusion, duration of follow-up).
Methods: We will calculate a pooled ratio of median survival times (Simes, Statist. Med. 6, 11-29). The variance of the ratio of the median survival times in each trial will be approximated using the total number of patients. The median survival times will be calculated from the IPD database rather than taken from publications. In this way we will compare like with like. The same trials, patients and extended follow-up will be used in each analysis. The IDP meta-analyses on which the comparisons are made, consist of meta-analyses carried out by the Institut Gustave Roussy, the Meta-Analysis Group in Cancer (MAGIC) and the Meta-Analysis Group from the Medical Research Council Clinical Trials Unit.
Results: The three lung cancer meta-analyses, for which the results of homogenous trial sets are tabulated below, are the prophylactic cranial irradiation overview on small cell lung cancer (lung 1), the locally advanced (lung 2) and advanced settings of the non-small cell lung cancer meta-analysis on chemotherapy (lung 3). More comparisons with other meta-analyses will be presented at the meeting. The same patterns can be discovered in the analyses based on the MST method as in the real IPD meta-analyses. The differences in the estimation of the overall hazard ratio (HR) for the IDP method and the estimation of the pooled median ratio (MR) for the proposed MST method stem from differences of the estimations of the hazard ratio at the trial level, in particular when the differences between the survival curves are not constant over time.
[Table omitted]
Conclusions: The results of a simple summary data method based on the median survival time are compared with the results from IDP meta-analyses with high mortality rates. The only difference in estimation relates to bias of the statistical method. Under- or overestimations of the overall hazard ratio may occur. However, based on the comparisons with IDP meta-analyses, the simple MST method seems to have practical and exploratory value. Other differences will come into play if used as part of a meta-analysis of the published literature (unpublished trials, patient exclusion, duration of follow-up).