I am pleased that Jaime Caro and Krista Payne have kept this discussion alive, thus allowing me to explain my position at greater length.1,2 The objective of scientific research is to minimize bias as much as possible, which is why the double-blind randomized trial is the gold standard. In a perfect world, observational studies of adherence to dispensed drugs might be valid. However, the real world is far from perfect.
In my opinion, the studies mentioned by Caro and Payne3,4 are predominantly a measure of the effect of an intensive drug-advertising program. This type of program is able to bias such studies in at least 3 ways.
The first and most obvious way is the effect of free samples. When these studies were carried out, many patients were being given free samples of angiotensin-converting-enzyme (ACE) inhibitors; patients who did well on the sample received a prescription for the drug, and patients who did poorly were tried on something else. In contrast, patients whose initial therapy was thiazides (no free samples) were all given a prescription from the start. Because free samples are not measured as dispensed drugs, the appearance created is that adherence is better for ACE inhibitors than for thiazides.
The second way is the strong physician bias that intensive advertising creates in favour of ACE inhibitors and against thiazides. This bias is conveyed to the patient and affects adherence to treatment.
The third way is that a drug company limits its funding to studies of those populations for which it is relatively certain what the outcome will be, on the basis of its own data. This is called selection bias. Caro and Payne imply that the company funding their study3 was not biased. They neglect to mention that Bristol–Myers Squibb markets fosinopril, an ACE inhibitor.
I reiterate that the best way to measure adherence to different antihypertensive therapies is to use the randomized double-blind design. In this type of study, adherence is affected by the drug's characteristics and not by other influences.