Potential benefits of the IPD approach

As implied in the conclusions of the above study, a major benefit of IPD is that it allows time-to-event analysis (see Section 11.5.7). Although there are methods that allow us to estimate and combine time-to-event summary statistics from published reports [26], in practice this can prove extremely difficult as the required information is seldom presented in a usable format. Thus, the ability to conduct time-to-event analyses across all trials might be reason in itself for collecting IPD in illnesses, including many cancers, where prolongation of survival, or time to recurrence of disease or onset of symptoms is important. The IPD approach is the only practical way to carry out analyses to investigate whether any observed treatment effect is consistent across well-defined groups of patients, indeed this might be the only context in which it is reasonable to explore subgroups (see Section 11.5.7).

Where trials have used different classifications or scales of measurement e.g. different tumour staging systems, or different units for recording haemoglobin, IPD may provide the opportunity to translate between scales and combine data, a process that might not otherwise be feasible. It also allows detailed data checking to ensure consistency and quality of randomization and follow-up. Importantly, it also means that up-to-date follow-up information can be collected and thereby provides a valuable opportunity to look at longer-term outcomes. Some additional advantages of direct collaboration with trialists, when collecting either IPD or summary data from source, is that this contact can result in more complete identification of relevant trials and better compliance in providing missing data. Discussions with the trialists during the project can also aid interpretation and wider endorsement and dissemination of the results and to better clarification of, and collaboration in, further research. For these reasons IPD metaanalyses are considered by many to be the 'yardstick' against which other forms of systematic review should be measured [27].

0 0

Post a comment