Data in longitudinal randomised controlled trials in cancer pain: is there any loss of the information available in the data? Results of a systematic literature review and guideline for reporting.

Background: Given the prevalence of untreated pain among cancer patients, there have been calls for more and better research in the domain. Increasingly, calls for less waste and more optimal use of trial data collected are being made. Waste of data includes non-optimal statistical analysis and non-presentation of interpretable effect size as a measure of effectiveness of an intervention which also enable comparisons across studies.

Methods: We reviewed the recent literature on randomised trials on longitudinal cancer pain to identify sources of loss of data information by collecting material on the nature of outcomes collected, analysed, the method of analysis and what was presented as a result of the trial. Illustrated with real data, we propose some guidelines on how to adequately analyse longitudinal data and report the results using mixed models.

Results: We identified some major source of data information loss, one of which is the transformation of a continuous pain outcome. Not adjusting for the collected outcome baseline value is moreover a source of bias. Multiple testing by analysing the data cross-sectionnally at each time-point leads to loss of information and power. Finally, effect sizes reflecting the effectiveness of the intervention were never reported.

Conclusions: We identified several sources of information loss in the way longitudinal trials on pain were analysed and reported. However these problems could be easily solved by using regression methods like mixed models and presenting regression parameters to provide a concrete quantitative effect of the intervention.