And enough with the notion that medication compliance is a good proxy for overall efficacy.

All of these horrible psychiatry studies-- CATIE, Lamictal and Depakote maintenance trials, etc-- keep telling us how long patients stay on medications, because they say this means the drugs are working.  The authors think that if a drug is working, they patient will stay on it.  But you would think this only if you didn't actually treat many patients.  I can make a similar argument that staying on a medication is inversely related to efficacy-- because when a patient feels better, they simply stop taking their meds.  

Think about antibiotics.  People don't finish the full 14 day course, precisely because they feel well.  If they felt sick, they would probably take them longer than 14 days.  In fact, people overuse these antibiotics even when its a virus, despite the antibiotic having no efficacy at all.    They will demand an antibiotic even though know that it shouldn't be doing anything.

Same with pain meds.  Oh, that's an acute problem?  How about the chronic problems of diabetes and hypertension.  People will skip/miss/forget doses when they feel asymptomatic, and will be more compliant when they have symptoms associated with these illnesses (e.g. headache, dizziness, etc.)

Look, I'm not telling you that compliance and efficacy aren't related.  I am saying that if you want to measure efficacy, don't use compliance as a proxy-- go measure actual efficacy.   And don't tell me it's too hard.  You got $67 million for this study.  Find a way.