Article type
Year
Abstract
Introduction/Objective: One of the advantages of systematic review using individual patient data (IPD), where 'raw' data is centrally collected, checked and re-analysed is that it allows more sensitive and flexible analyses. In particular, full time to event (TTE) analyses can be done. While not all systematic reviews of 'survival' endpoints have time and resource available to use IPD, they can and should make the most efficient and complete use of the summary statistical data available.
Methods: Many systematic reviews of TTE endpoints are done by collecting or extracting the total number of patients and events at a fixed point in time, commonly by reading from a survival curve, and using these to calculate an odds ratio (OR). ORs calculated in this way relate only to the point in time at which they are calculated and may not make the most efficient use of the information reported. In contrast, the TTE of individual patients can be used to calculate a hazard ratio (HR) representing the entire survival experience and also taking account of censoring. However, even when IPD are not available, more appropriate use of summary statistics can also provide a HR or a reasonable estimation of it.
Results: Individual RCTs may report various statistical summaries of their TTE analysis e.g. log HRs, HRs or log rank 0-E's plus variances, chi-square and p-values. HRs can be calculated from these, using different combinations of information on a per trial basis, depending upon what is reported or available. Methods for making such calculations will be presented. If no summary statistics are presented it may still be possible to estimate HRs using data from multiple points on a survival curve.
Discussion: Although IPD systematic reviews may offer a 'gold standard' for the analysis of 'survival' endpoints and offer the additional advantages of extended follow-up and ability to undertake subgroup analyses, analyses of summary data can be improved by making better use of data presented in publications. TTE analyses can be done, if the appropriate statistical methods are adopted.
Methods: Many systematic reviews of TTE endpoints are done by collecting or extracting the total number of patients and events at a fixed point in time, commonly by reading from a survival curve, and using these to calculate an odds ratio (OR). ORs calculated in this way relate only to the point in time at which they are calculated and may not make the most efficient use of the information reported. In contrast, the TTE of individual patients can be used to calculate a hazard ratio (HR) representing the entire survival experience and also taking account of censoring. However, even when IPD are not available, more appropriate use of summary statistics can also provide a HR or a reasonable estimation of it.
Results: Individual RCTs may report various statistical summaries of their TTE analysis e.g. log HRs, HRs or log rank 0-E's plus variances, chi-square and p-values. HRs can be calculated from these, using different combinations of information on a per trial basis, depending upon what is reported or available. Methods for making such calculations will be presented. If no summary statistics are presented it may still be possible to estimate HRs using data from multiple points on a survival curve.
Discussion: Although IPD systematic reviews may offer a 'gold standard' for the analysis of 'survival' endpoints and offer the additional advantages of extended follow-up and ability to undertake subgroup analyses, analyses of summary data can be improved by making better use of data presented in publications. TTE analyses can be done, if the appropriate statistical methods are adopted.