To the Editor: The study by Michael D. Stein, M.D., et al. (1) expanded on previous work by assessing the association between depression severity and HIV risk behavior among the users of injected drugs. The findings implied that the reduction in depression severity might result in a decrease in the rate of high-risk injection behavior.

However, the sampling and methods used in the study may undermine its findings. Although the authors used bootstrap resampling to increase the power, 63.3% of 109 subjects reported contact with drug treatment agencies, and 10% were HIV positive. Treatment for drug use and HIV may moderate rather than mediate the relationship between depression severity and risk behavior in patients who inject drugs. Further efforts are needed to explore these complicated relationships.

As the authors pointed out, the continuous diagnosis approach would better capture and measure a variable and its relation to others. While the authors used a continuous measure in the case of instances of injection risk, ranging from 0 to 750, they reconstructed it into a four-category response: no injection risk, one to five instances, six to 89 instances, and 90 instances or more. The reasons for such cutoffs were not articulated, and the question arises of whether there was any difference in use of the continuous measures to assess its relationship with depression severity through ordinary least-square regression analysis. What would have occurred had they used different cutoff points and categorized the data into more or less than four groups?

Problems may arise in measuring drug injection frequency and risk behavior and in using them as independent and dependent variables, respectively. It can be clearly seen that risk behavior in drug injection is highly correlated with injection frequency. That is, as the risk behaviors increase, the injection frequency rises because the risk behavior and the nonrisk behavior consist of the total frequency of drug injections. Such an overlap in measurement may give rise to the high correlation between these two variables, and it may not be appropriate to use one of them as the predictor of the other; however, the article told us, "Injection frequency was not correlated with depression severity (r=0.08, N=109, p=0.40)" (p. 1660). Although the authors discussed the possible mechanisms by which depression severity is significantly associated with risk behavior but not with injection frequency, it will require more effort to examine the association between depression severity and the frequency of drug injection.

The limitations of the study group and the potential measurement errors affected the generalizability of the findings. These questions notwithstanding, the article offers new evidence into an increasingly important area of risk-reduction research among drug users by using diagnostic assessment of depression severity and identifying its independent relationship with HIV risk behavior in those who inject drugs.