Publishing data from substudies

An increasing number of trials collect not just the clinical data necessary to evaluate a treatment, but also additional data or information in all or a subset ofpatients. Examples include quality of life or health economics data as well as data arising from sub-studies such as tissue collection for evaluation of prognostic and predictive markers. The question then arises as to how much of these data should be included in the main study report without making it unwieldy. The key question to ask is, could these additional data potentially influence the overall assessment of the relative merits of the treatments? If the answer is yes, then at least the key results should be included in the main study report. For example, if some specific quality of life (QL) hypotheses were specified in advance, data pertaining to these should be included alongside the clinical data. As QL can generate a vast amount of data, it may well be appropriate to publish a more detailed analysis separately. In such a report, data on all the QL items and subscales can be presented, and exploratory analyses conducted, with appropriate cautionary notes relating to multiplicity of outcome measures and absence of pre-defined hypotheses (see Chapter 9). One such example is the MRC/EORTC factorial trial of three versus four cycles of BEP chemotherapy over three versus five days for good prognosis metastatic germ cell tumours. This was an equivalence trial and as such the QL results could potentially have had a major influence on the interpretation of the results. Therefore, the main study report [10] focused on the clinical data but included a summary of the findings of the QL data. A further report will present detailed analyses of the QL data, which have not previously been reported in a randomized trial in this disease.

If the additional data are not essential in order to interpret the main clinical results, then in the interests of brevity it may well be more appropriate to include all the data in a separate report and cross reference it in the main study report. Studies of potential new predictive factors - molecular markers for example - may be difficult to place. They will certainly be relevant to the main study results since the aim is to determine whether they can identify true interactions with treatment effect, isolating subgroups who may benefit to a substantially greater or lesser extent than patients overall. However, it will generally be the case that these are exploratory analyses, requiring independent confirmation, particularly if the trial was not specifically powered to look at treatment interactions (this will nearly always be the case, especially when tissue collection is planned in only a subset of trial patients). In this situation, it would be appropriate to publish the results separately as the overall trial result remains the most important and can be interpreted and acted upon without reference to substudy results. As the sub-study analyses would need to refer to the overall trial treatment effect, it would of course be important that analyses of sub-studies are only published after, or in parallel with, the main trial results.

It is important not to clutter reports with unnecessary data. A prime example is the common practice of including results of an analysis of clinical prognostic factors at the end of a trial report. This is only of interest if it provides new insight - for example if it includes an evaluation of new potential prognostic factors rarely if ever studied before, or if the trial is in a rare disease for which data on prognostic factors in substantial numbers of patients maybe lacking. There is really no point in providing an analysis of common prognostic factors in trials in common diseases. The temptation to include an exciting p-value, however irrelevant to the trial results, somewhere in the report should be resisted.

0 0

Post a comment