To the Editor The incidence of drug-induced liver injury (DILI) due to specific drugs in the general population is an important but challenging topic with few reliable figures. In a prospective population-based study from Iceland, with careful exclusion of other potential causes of liver injury, the risk of DILI was calculated with the drugs implicated, with a denominator based on a national prescription database.¹

In a recent issue of JAMA Internal Medicine, Torgersen et al² provided estimates of the incidence of acute liver injuries associated with potentially hepatoxic drugs based on the US Department of Veterans Affairs (VA) health care system’s electronic health records.² Although the data on the total number of prescriptions should be accurate, the study is limited because it relies on medical record data without carefully assessing causality. The problem with such an approach was shown in a recent study from Iceland on the risk of liver injury associated with oral anticoagulants.³ Among 15 819 patients prescribed medication, 2300 subsequently had documented elevated liver biomarker levels, matching the criteria in the VA study.³ A manual medical record review identified only 3 instances caused by DILI (<0.1%). Most patients’ elevations were caused by hepatobiliary diseases, ischemic injury, alcoholic hepatitis, viral hepatitis, and several other causes of liver injury.³ Thus, in the current study, many diagnoses may have been overlooked and misattributed to DILI based on the study methods used. Unfortunately, the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes are frequently just as unreliable. In the study by Torgersen et al,² hepatologists assessed a random sample of 75 patients (out of 1750 patients) for whether a medication possibly caused the liver injury, but the criteria used for assigning causality were unclear, and causality methods, such as the Roussel Uclaf Causality Assessment Method (RUCAM) or revised electronic version of RUCAM (RECAM), were not used.⁴