# How to best respond to investigator who wants to study secondary outcome

I'm co-investigator on a clinical trial where the we are studying the effect of an intervention. The study is powered to detect a clinically meaningful change for some primary endpoint. My questions are around the planned secondary endpoints.

There are two secondary endpoints: 1) whether or not individuals randomized in the trial were prescribed medications (in this case statins) and 2) among those who were prescribed medications, whether or not individuals adhered to those medications. The first secondary endpoint is the main one and we have decided to use a regression model to estimate the intervention effect. The latter secondary endpoint hinges on the results of the first one. The PI wants to look at whether the intervention had an effect on statin adherence among those who were prescribed. Specifically, he wants to use a regression model to do this.

My concerns are 1) by subsetting to this group, we no longer have the randomization with respect to the intervention, so that the "effect of intervention on adherence" is meaningless, and 2) in line with the first point, the underlying mechanisms may vary between the main secondary outcome and the adherence outcome. In other words, we may have many unmeasured factors. My suggestion was to keep the analysis descriptive: tabulate the binary adherence outcome by the binary intervention, and restricting the analysis to those who were prescribed medications. However, a reviewer (non-statistician) insists that we can still use regression.

Are there things that I can add in order to make a convincing case for keeping it descriptive?

• I find this not so easy to follow. Could you maybe draw a diagram of the situation? – Sextus Empiricus Nov 11 '20 at 19:01
• @SextusEmpiricus - I've simplified the language to clarify the main points that I would like addressed. Please let me know if it's still unclear. – stats134711 Nov 11 '20 at 19:22
• Can you give a sense of what the intervention is? It sounds like these are mediation analyses, which would be perfectly reasonable. The main analysis for the primary outcome would be intervention -> improved biomarker. The 1st secondary analysis would be intervention -> increased Rx rate. The 2nd secondary analysis would be intervention -> adherence rate (among patients w/ Rx). – gung - Reinstate Monica Nov 11 '20 at 19:36
• It sounds to me like the implied full causal theory is that the intervention increases the rate of prescriptions (to patients who would not otherwise have received them), and adherence does not decrease amongst patients who normally don't get the Rx, so the intervention ultimately leads to improvements in the biomarker. For example, is the intervention something like helping patients w/o insurance afford the Rx? If so, it all seems perfectly sensible to me. – gung - Reinstate Monica Nov 11 '20 at 19:37
• The main target is primary care patients who are at high risk for CVD and eligible for statin therapy based on current guidelines, but have not received one. The intervention is essentially additional pharmacogenetic testing, the results of which the care-provider and patient can discuss together to deliver appropriate statin prescribing. The control is the usual care based on currently established guidelines. – stats134711 Nov 11 '20 at 19:46