Cesari raises concerns regarding the categorization of the variable “number of drugs.” This variable was categorized not on the basis of arbitrary cutoffs but on tertiles calculated in the GIFA (Gruppo Italiano di Farmacoepidemiologia nell’Anziano [Italian Group of Pharmacoepidemiology in the Elderly]) population. In addition, Cesari suggests that the simple count of the overall number of drugs taken may be as good as the full score in predicting the risk of ADRs. The number of drugs was the strongest predictor of ADR in our model, but other variables assessing relevant processes involved in the onset of ADR also entered in the risk score. In this context, comorbidity may increase the risk of ADR by increasing the likelihood of the drug-disease interactions, heart failure, hepatic disease, and renal failure by impairing drug distribution and metabolism, and history of ADR may indicate a susceptibility to negative effects of drugs due to ethnic, genetic, or cultural factors. Noticeably, these risk factors were also identified by other studies in the medical literature.1-5 In addition, in the GIFA population, the area under the receiver operating characteristic curve that measures the ability of the risk score to predict ADR was significantly greater (0.71; 95% confidence interval, 0.68-0.73) than that for number of drugs alone (0.67; 95% confidence interval, 0.64-0.70) (P < .01), indicating that the risk score is more powerful than the single variable for number of drugs in predicting the risk of ADR.
Onder G, Petrovic M, Rajkumar C, van der Cammen TJM. The Easiest Way to Predict Adverse Drug Reactions in Older Persons—Reply. Arch Intern Med. 2011;171(1):91–92. doi:10.1001/archinternmed.2010.490
Artificial Intelligence Resource Center
Customize your JAMA Network experience by selecting one or more topics from the list below.