Stimulant Medication Use and Sudden Death in Youths: Considering Unmeasured Covariates

To the Editor: We would like to commend Dr. Gould et al. (1) on their study examining the association between the use of stimulants among youths and sudden death. We especially appreciated their extensive use of sensitivity analyses and the considerable detail they provided in Table 3, which married the systematic nature of a case-control study with the detail of a case series. This information allows several observations to be made. First, despite the inclusion of more general unexplained deaths, the majority of case subjects were ruled to have died from cardiac causes (dysrhythmia). Second, the activities (e.g., running, sleeping) of those who died of unknown causes were often similar to those of patients who died from dysrhythmias. Third, the authors provided information about concomitant drugs, and most case subjects (eight out of 10) were apparently receiving monotherapy. Fourth, toxicology appeared to have ruled out drug overdose as a likely factor in the majority of deaths.

However, in any case-control study, the comparability of the comparison subjects is crucial. This is especially true for this study, in which the observation of a single additional exposure in the comparison group (corresponding to small change in the stimulant use rate from 0.35% to 0.5%) would have substantially attenuated the observed relationship. It is therefore worth considering whether unappreciated differences in measured or unmeasured covariates plausibly could have caused stimulant use in the comparison group to be biased downward compared with the sudden death group. Among measured covariates, the authors achieved surprisingly good balance in age and year of death, despite using 3-year windows. However, in both these areas, the remaining imbalance, although small, favored lower stimulant use in the comparison group (slightly older age and a slightly earlier year of death). Among unmeasured covariates, could the authors possibly have analyzed the season of death if these data were already recorded, since seasonal differences likely exist in the probability of receiving a stimulant? It also seems to us, as supported by some literature, that juvenile accident victims might be expected to have lower income and live more often in rural settings, both circumstances associated in some (but not all) studies with lower rates of stimulant use. We would like to know the authors' opinion of whether these unmeasured covariates could have influenced their findings. Given the strength of the initial association, our expectation is that such factors might have attenuated, but not likely eliminated, the relationship between stimulant use and sudden death observed in this study. Lastly, we note that the weakest association was observed for the outcome that did not depend on informant reports, suggesting at least the possibility of recall bias favoring the recollection of stimulant use in the sudden death group.

In summary, the study represents a valuable contribution in this area, and we anticipate further studies examining the absolute risks and relative risks and, possibly, the predictors of adverse cardiac events during stimulant use.