Chris Delaney and Michal Pijak and associates argue that observational studies are preferable to the large simple RCTs recommended in my commentary1 because they sometimes yield the same results as RCTs. That is true, but the problem is that often they do not. Therefore, without confirmatory RCT data, we risk making serious mistakes if we advise or prescribe solely on the basis of observational studies. Because of the Women's Health Initiative RCT, we can advise women that the harms of long-term estrogen–progesterone combination therapy outweigh the benefits,2 but on the basis of observational data, physicians were advising the opposite. The details of this debate are well covered in 2 recent articles.3,4
I agree with Delaney that the key to research is to ask the right question, find out if the question has been answered and, if not, use the appropriate study to answer it. Because of the inability to draw conclusions from observational data alone, the appropriate study is almost always an RCT. Unfortunately, this type of study is too infrequently conducted. The ALLHAT trial5 is an exception to this general pattern. As a result of that trial, we can advise patients, with a high degree of certainty, that chlorthalidone, a thiazide- like diuretic, is preferable to amlodipine, a calcium-channel blocker (CCB), as first-line therapy for hypertension; for every 61 patients treated, using a thiazide rather than a CCB prevents one death or hospital admission for heart failure. This finding would not have been discovered from observational data.
Barton,6 in the editorial cited by Pijak and associates, stated that “If high quality randomized trials exist for a clinical question then they trump any number of observational studies.” We need to appreciate that well-designed, large, simple RCTs are not that difficult or expensive to conduct and are highly preferable to widespread empirical prescription of drugs, from which we can learn little or nothing.
James M. Wright Departments of Pharmacology and of Therapeutics and Medicine University of British Columbia Vancouver, BC
Footnotes
-
Competing interests: None declared.