PURPOSE: Inconsistencies in boxed warnings between drug information resources and the manufacturer's prescribing information (PI) were evaluated. METHODS: This study was a cross-sectional evaluation of boxed warnings in Black-BoxRx, DrugDex, Facts and Comparisons, Epocrates, Lexicomp, and PDR.net conducted in June 2010. New molecular entities with boxed warnings and PI that conformed to current Food and Drug Administration labeling requirements were included. Each resource was reviewed for warnings that appeared verbatim with the full boxed warning in the PI. Two drug information pharmacists independently reviewed the remaining nonverbatim boxed warnings to determine concordance with the boxed warning summary in the "highlights" section of the PI. Tests of proportions were used to examine differences among resources in the proportion of warnings concordant with the PI. Interrater reliability was assessed with the kappa statistic. RESULTS: A total of 71 drugs with unique boxed warnings were included in the evaluation. Resources revealed varying degrees of discordances with the boxed warning in the PI. The resource with the lowest number of verbatim warnings contained a significantly higher percentage of warnings with discordant information when compared with all other resources (p < 0.0001 for all paired comparisons). Interrater reliability was excellent (kappa = 0.86). CONCLUSION: Boxed warning information presented in major drug information resources may be missing key elements of the official boxed warning in the current PI. The current PI may be the most reliable approach to accessing the complete, up-to-date boxed warning for a given drug.
PURPOSE: Inconsistencies in boxed warnings between drug information resources and the manufacturer's prescribing information (PI) were evaluated. METHODS: This study was a cross-sectional evaluation of boxed warnings in Black-BoxRx, DrugDex, Facts and Comparisons, Epocrates, Lexicomp, and PDR.net conducted in June 2010. New molecular entities with boxed warnings and PI that conformed to current Food and Drug Administration labeling requirements were included. Each resource was reviewed for warnings that appeared verbatim with the full boxed warning in the PI. Two drug information pharmacists independently reviewed the remaining nonverbatim boxed warnings to determine concordance with the boxed warning summary in the "highlights" section of the PI. Tests of proportions were used to examine differences among resources in the proportion of warnings concordant with the PI. Interrater reliability was assessed with the kappa statistic. RESULTS: A total of 71 drugs with unique boxed warnings were included in the evaluation. Resources revealed varying degrees of discordances with the boxed warning in the PI. The resource with the lowest number of verbatim warnings contained a significantly higher percentage of warnings with discordant information when compared with all other resources (p < 0.0001 for all paired comparisons). Interrater reliability was excellent (kappa = 0.86). CONCLUSION: Boxed warning information presented in major drug information resources may be missing key elements of the official boxed warning in the current PI. The current PI may be the most reliable approach to accessing the complete, up-to-date boxed warning for a given drug.