Citation:
Jackson JW, Gagne JJ. SMART designs in observational studies of opioid therapy duration. J Gen Intern Med. 2014;29 (3) :429-31.
Joshua Gagne
Associate Professor
Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women's Hospital and Harvard Medical School
Department of Epidemiology, Harvard T.H. Chan School of Public Health
Contact
-
3⃣The problem occurs only with truly persistent users -- i.e., those who will continue to be exposed until they die -- and not with long-term exposures as long as a steady-state of starting and stopping is reached. See box plot in 1⃣ for unbiased fixed 180-day exposure estimate.
-
2⃣When its assumptions are met (which is easy to do in a simulation study, not so easy in real data), the case-time-control approach can completely eliminate the bias. t.co/3KRLfqh8Ro
-
1⃣Bias is small but apparent when 10% of patients are persistent users (see figure: biased odds ratio = 1.11; true OR = 1.00). Bias is sizable (OR = 1.43) when 30% of patients are persistent users and it grows quickly from there. t.co/r0QXhXLvk0
