Longitudinal and cross sectional mortality studies in injecting drug-users.

SIR Dr Frisher et all have recently addressed the issue of mortality in drug injectors. They report a mortality from all causes of 0.54%, which is substantially lower than that described in previous studies. Additional studies, besides those quoted by Frisher et al,26 showed higher mortality in drug injectors. The main difference between these studies and that of Frisher et al is that the former are all longitudinal follow up studies of drug addicts, while the latter estimates numerator and denominator separately. The authors attribute the lower rate shown in their study to an underenumeration of the denominator. We would suggest that a very likely explanation is the estimate of the numerator. The authors used the following criteria to define deaths related to drugs: (1) ICD code on death certificates; (2) reports of drug injection as the mode of transmission in deaths of subjects tested for HIV from the HIV registry; and (3) reports of drug injection in sudden deaths from the Procurator Fiscal's Office. On the basis of these data they found 51 deaths in the age group 15-40 years for the year 1989. Over 90% of these deaths were attributed to overdose and suicide, 2% to AIDS and accidental causes, and only 9.8% to other medical causes (see table IV in the original paper). The results from other studies (4, 6-7) show that the proportion of deaths not directly related to drug use is relevant. In particular, the cohort study we conducted in Rome (6), following up 4200 drug injectors from 1980-88, showed that other medical causes accounted for 28% of the deaths overall. Not only do injectors seem to die from causes directly related to drug use, they have an excess risk for other causes of death, and such an excess risk could have been missed using the sources described by Frisher et al. In conclusion, we think that mortality in drug injectors estimated by the authors cannot be considered a rate for all causes, but only an estimate of the number of deaths in drug injectors attributable to causes directly related to drug abuse. The only way to