18.4.3  Time-to-event analyses

Collecting IPD that include the time interval between the randomization and the event of interest enables time-to-event analyses to be conducted. These include, for example, time to recovery, time free of seizures, time to conception and time to death. Indeed, one of the main reasons that IPD meta-analyses have been so important in the cancer field is that time-to-event analysis of survival is vital in evaluating therapies. Most interventions are more likely to lead to a prolongation of survival rather than a cure. Therefore, it is important to measure not only whether a death happens, but also the time at which it takes place. To allow this type of analysis one needs to know the time that each individual spends ‘event-free’. This is usually collected as the date of randomization, the event status (i.e. whether the event was observed or not) and the date of last evaluation for the event. Sometimes, it will be collected as the interval in days between randomization and the most recent evaluation for the event. Time-to-event analyses are performed for each trial to calculate hazard ratios, which are then pooled in the meta-analysis (see Section 9.4.9).