To the Editor:
The paper by Yun et al1 addresses an important issue-whether treatment of depression improves adherence to antiretroviral therapy (ART) in patients with HIV. There are several aspects of this paper on which we would like to comment.
We believe that the data presented in Table 3 have not been interpreted correctly. The authors seem to have confused column percentages and row percentages. They comment about the top part of Table 3: “The proportion of depressed patients adherent to ART was significantly lower among those not receiving ADT compared with those who received ADT (65% vs. 35%, respectively; P < 0.01).” The 65% and 35% are the percents receiving and not receiving antidepressant treatment (ADT) among those with >95% adherence. The relevant numbers from this part of the table are the row percentages; that is, among those who got ADT (n = 376), 18% (68 of 376 patients) were 95% adherent to ART. Among those who did not get ADT (n = 130), 28% (37 of 130 patients) were 95% adherent to ART. Thus, contrary to what the authors state, the proportion of depressed patients adherent to ART was higher among those not receiving ADT (28%) compared with those who received ADT (18%). This is also shown in the odds ratio (OR) that the authors show in Table 3 but do not comment on in the text. The OR of 0.55 [(68 × 93)/(37 × 308)] is the odds of the patients receiving ADT having >95% adherence to ART compared with those who do not receive ADT. The reference group is those who did not receive ADT. The interpretation is that those receiving ADT are less likely to be >95% adherent to ART than those who are not receiving ADT.
The same row and column issues pertain to the bottom of Table 3, where the authors compare the rates of achieving 95% ART adherence in those adherent and not adherent to ADT. They state: “In addition, when ART adherence was stratified by ADT adherence status, the proportion of depressed patients adherent to ART was significantly higher among ADT-adherent patients compared with ADT-nonadherent patients (69% vs. 31%, respectively; P = 0.001).” Again, the relevant numbers from this part of the table are the row percentages; that is, among those adhering to ADT (n = 194), 24% (47 of 147 patients) were 95% adherent to ART. Among those not adherent to ADT (n = 182), 12% (21 of 182 patients) were 95% adherent to ART. The percents presented in the text therefore greatly overstate the actual percentages and the magnitude of the differences between groups.
Several aspects of the data presented in Table 4 are concerning. One is that Table 3 analyzes 506 cases, and Table 4 analyzes 323; thus, 36% of the sample is somehow lost in the longitudinal analysis. Presumably this is related to people who do not have 6 months of “before” and 6 months of “after” data, but this is not discussed, and no data are presented about how these losses bias the analysis.
Another concerning aspect of Table 4 is that ART adherence before and after receiving ADT (the first 2 rows of the Table) does not seem to be related to ADT adherence. In the 6 months before receiving ADT, those who adhere to ADT and those who do not have the same ART adherence (51% and 53%, respectively), which is what one would expect. In the 6 months after starting ADT, however, those who adhere and those who do not improve to 93% and 90%, respectively. The authors write in their discussion section: “Those individuals adherent to ADT were also more adherent to ART,” but the data in Table 4 suggest, to the contrary, that ART adherence is the same regardless of whether or not patients adhere to ADT.
A final curious aspect of the data presented in Table 4 is that the “after ADT” adherence rates were all extremely high (93%, 90%, and 91%). Why are these depressed patients adhering so well, regardless of depression treatment or depression treatment adherence? This suggests that time is the only important factor here; that is, regardless of depression treatment or ADT adherence, ART adherence quickly improves to 90% or higher. This seems highly unlikely.
Table 5 suggests a strong relation between ADT adherence and ART adherence. The sample used for this regression is not described (eg, there are no analytic sample sizes), however, and the findings for ADT adherence seem to contradict the data presented in Table 4 (ie, rows 1 and 2, which show no relation between ADT adherence and ART adherence).
We share the authors' interest in this study question and agree that there are no longitudinal analyses that directly address the relation between depression treatment and ART adherence. Nevertheless, we do not believe that the analyses they present clarify this important issue.
Ira B. Wilson, MD, MSc*
Denise Jacobson, PhD, MPH†
*Institute for Clinical Research and Health Policy Studies Tufts-New England Medical Center Boston, MA, †Department of Family and Community Medicine Tufts University School of Medicine Boston, MA
1. Yun LW, Maravi M, Kobayashi JS, et al. Antidepressant treatment improves adherence to antiretroviral therapy among depressed HIV-infected patients. J Acquir Immune Defic Syndr