Guillermo Prada-Ramallal1, Bahi Takkouche1,2, Adolfo Figueiras3,4. 1. Department of Preventive Medicine and Public Health, Facultad de Medicina, University of Santiago de Compostela, c/San Francisco s/n, 15786, Santiago de Compostela, A Coruña, Spain. 2. Consortium for Biomedical Research in Epidemiology and Public Health (CIBER en Epidemiología y Salud Pública-CIBERESP), Santiago de Compostela, Spain. 3. Department of Preventive Medicine and Public Health, Facultad de Medicina, University of Santiago de Compostela, c/San Francisco s/n, 15786, Santiago de Compostela, A Coruña, Spain. adolfo.figueiras@usc.es. 4. Consortium for Biomedical Research in Epidemiology and Public Health (CIBER en Epidemiología y Salud Pública-CIBERESP), Santiago de Compostela, Spain. adolfo.figueiras@usc.es.
Abstract
INTRODUCTION: Meta-analyses of observational studies represent an important tool for assessing efficacy and safety in the pharmacoepidemiologic field. The data from the individual studies are either primary (i.e., collected through interviews or self-administered questionnaires) or secondary (i.e., collected from databases that were established for other purposes). So far, the origin of the data (primary vs. secondary) has not been systematically assessed as a source of heterogeneity in pharmacoepidemiologic meta-analyses. OBJECTIVE: The aim was to assess the impact of considering the source of exposure data as a criterion in sensitivity and subgroup analysis on the conclusions of drug safety meta-analyses. METHODS: We selected meta-analyses published between 2013 and 2015 in which the intake of frequently used over-the-counter medicines was either the main exposure or a concomitant treatment and the outcome had short latency and induction periods. We stratified the results by origin of data (primary vs. secondary) and compared the new results to those presented originally in the meta-analyses. RESULTS: We used four meta-analyses that fulfilled our criteria of inclusion. The results were selective serotonin reuptake inhibitors and upper gastrointestinal bleeding: original estimate odds ratio (OR) = 1.71 [95% confidence interval (CI) 1.44-2.04], OR primary data = 1.19 (95% CI 0.90-1.58), OR secondary data = 1.81 (95% CI 1.50-2.17); proton pump inhibitors and cardiac events: original estimate hazard ratio (HR) = 1.35 (95% CI 1.18-1.54), HR primary data = 1.05 (95% CI 0.87-1.26), HR secondary data = 1.43 (95% CI 1.23-1.66); non-aspirin non-steroidal anti-inflammatory drugs and myocardial infarction: original estimate risk ratio (RR) = 1.08 (95% CI 0.95-1.22), RR primary data = 0.57 (95% CI 0.34-0.96), RR secondary data = 1.15 (95% CI 1.03-1.28); paracetamol during pregnancy and childhood asthma: original estimate OR = 1.32 (95% CI 1.14-1.52), OR primary data = 1.23 (95% CI 1.06-1.42), OR secondary data = 1.53 (95% CI 1.33-1.75). CONCLUSIONS: The results after stratification are considerably modified. It is crucial to explore the origin of the data, either primary or secondary, as a source of heterogeneity in pharmacoepidemiologic meta-analyses to avoid misleading conclusions.
INTRODUCTION: Meta-analyses of observational studies represent an important tool for assessing efficacy and safety in the pharmacoepidemiologic field. The data from the individual studies are either primary (i.e., collected through interviews or self-administered questionnaires) or secondary (i.e., collected from databases that were established for other purposes). So far, the origin of the data (primary vs. secondary) has not been systematically assessed as a source of heterogeneity in pharmacoepidemiologic meta-analyses. OBJECTIVE: The aim was to assess the impact of considering the source of exposure data as a criterion in sensitivity and subgroup analysis on the conclusions of drug safety meta-analyses. METHODS: We selected meta-analyses published between 2013 and 2015 in which the intake of frequently used over-the-counter medicines was either the main exposure or a concomitant treatment and the outcome had short latency and induction periods. We stratified the results by origin of data (primary vs. secondary) and compared the new results to those presented originally in the meta-analyses. RESULTS: We used four meta-analyses that fulfilled our criteria of inclusion. The results were selective serotonin reuptake inhibitors and upper gastrointestinal bleeding: original estimate odds ratio (OR) = 1.71 [95% confidence interval (CI) 1.44-2.04], OR primary data = 1.19 (95% CI 0.90-1.58), OR secondary data = 1.81 (95% CI 1.50-2.17); proton pump inhibitors and cardiac events: original estimate hazard ratio (HR) = 1.35 (95% CI 1.18-1.54), HR primary data = 1.05 (95% CI 0.87-1.26), HR secondary data = 1.43 (95% CI 1.23-1.66); non-aspirin non-steroidal anti-inflammatory drugs and myocardial infarction: original estimate risk ratio (RR) = 1.08 (95% CI 0.95-1.22), RR primary data = 0.57 (95% CI 0.34-0.96), RR secondary data = 1.15 (95% CI 1.03-1.28); paracetamol during pregnancy and childhood asthma: original estimate OR = 1.32 (95% CI 1.14-1.52), OR primary data = 1.23 (95% CI 1.06-1.42), OR secondary data = 1.53 (95% CI 1.33-1.75). CONCLUSIONS: The results after stratification are considerably modified. It is crucial to explore the origin of the data, either primary or secondary, as a source of heterogeneity in pharmacoepidemiologic meta-analyses to avoid misleading conclusions.
Authors: Chiara Melloni; Jeffrey B Washam; W Schuyler Jones; Sharif A Halim; Victor Hasselblad; Stephanie B Mayer; Brooke L Heidenfelder; Rowena J Dolor Journal: Circ Cardiovasc Qual Outcomes Date: 2015-01-13
Authors: David Madigan; Patrick B Ryan; Martijn Schuemie; Paul E Stang; J Marc Overhage; Abraham G Hartzema; Marc A Suchard; William DuMouchel; Jesse A Berlin Journal: Am J Epidemiol Date: 2013-05-05 Impact factor: 4.897