Premium
Reporting and variability of constructing medication treatment episodes in pharmacoepidemiology studies: A methodologic systematic review using the case study of DPP ‐4 inhibitors and cardiovascular outcomes
Author(s) -
Weisman Alanna,
King Lauren K.,
Mamdani Muhammad
Publication year - 2020
Publication title -
pharmacoepidemiology and drug safety
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.023
H-Index - 96
eISSN - 1099-1557
pISSN - 1053-8569
DOI - 10.1002/pds.5071
Subject(s) - medicine , pharmacoepidemiology , mace , medical prescription , adverse effect , medline , emergency medicine , pharmacology , political science , law , myocardial infarction , conventional pci
Purpose In pharmacoepidemiologic studies, estimating medication adherence, persistence, and exposure patterns is critical. Constructing medication treatment episodes from prescription claims data involves assumptions related to grace period, carry‐over, and lag effect, but there are no guidelines for these assumptions. We evaluated reporting and variability of these parameters in pharmacoepidemiology studies, using a case study of antihyperglycemic medications and major adverse cardiovascular events (MACE). Methods We conducted a systemic review using MEDLINE and EMBASE for studies published prior to January 2, 2020 comparing the risk of MACE between dipeptidyl peptidase 4 (DPP‐4) inhibitors and active comparators. We extracted study characteristics and results, including grace period, carry‐over, and lag effect. Risk of bias was assessed by the Newcastle‐Ottawa scale, and assessments for prevalent user, immortal time, time lag, and time window biases. Results A total of 14/1850 studies identified were included. Grace period was not reported in 5 (35.7%) studies and ranged from 0 days to 180 days when reported. Carry‐over was not reported in 10 studies (71.4%). Lag effect was not reported in nine (71.4%) studies and ranged from 0 days to 180 days when reported. No studies conducted sensitivity analyses examining the effects of these assumptions on study findings. Predominant biases were inadequate follow‐up time, comparability of cohorts, prevalent use, and lag time bias. Conclusions Use of grace period, carry‐over, and lag effect were poorly reported and highly variable. Future pharmacoepidemiology studies should improve reporting, justify ranges for these parameters, and conduct sensitivity analyses to evaluate effects of these assumptions.