Secondary Logo

Journal Logo


Comparative effectiveness of dual vs. single-action antidepressants on HIV clinical outcomes in HIV-infected people with depression

Mills, Jon C.a; Harman, Jeffrey S.b; Cook, Robert L.c; Marlow, Nicole M.d; Harle, Christopher A.e; Duncan, R. Pauld; Gaynes, Bradley N.f; Pence, Brian W.a

Author Information
doi: 10.1097/QAD.0000000000001618



Depression is the most common psychiatric comorbidity among people living with HIV/AIDS (PLWHA) [1] with prevalence estimates ranging from 20 to 42% [1–7]. Depression has a detrimental impact on antiretroviral therapy (ART) adherence, viral load, and CD4+ T-cell count [1]. Therefore, timely delivery of effective depression treatment is important for PLWHA.

Several antidepressants with various pharmacokinetic properties are used to treat PLWHA for depression. Selective serotonin reuptake inhibitors (SSRIs), serotonin–norepinephrine reuptake inhibitors (SNRIs), bupropion and mirtazapine are commonly used antidepressants for treating depression in this population [8]. SSRIs (e.g. citalopram) are single-action antidepressants given their selective impact on one neurotransmitter, specifically serotonin [9]. SNRIs (e.g. venlafaxine), bupropion, and mirtazapine are considered dual-action antidepressants because they impact two neurotransmitter systems (e.g. serotonin, norepinephrine, or dopamine) in various combinations at the same time [9].

Both single-action and dual-action antidepressants are efficacious for improving depressive symptoms in PLWHA [5,6,10,11]. Researchers have sought to expand this line of work to include outcomes such as viral load and CD4+ T-cell count. The rationale driving these studies is based on the proposition that alleviating depression symptoms should lead to better HIV clinical outcomes through improved ART adherence or direct biological effects on the immune system [12,13]. To date, these studies have produced mixed results. Several randomized controlled trials have not revealed a link between antidepressants and improvements in HIV clinical outcomes, even in the presence of reduced depression symptoms [12,14–16]. However, in a recent pilot study conducted in Sub-Saharan Africa, 55 HIV-positive patients with depression who received an evidence-based antidepressant management intervention experienced improvements in depression symptoms, ART adherence, and HIV clinical outcomes [17]. Supportive evidence is also found in observational studies, which have demonstrated antidepressants have a positive association with ART adherence, viral load, and CD4+ T-cell count [13,18].

Logically, variations in HIV clinical outcomes might be observed between antidepressants if there are differential effects on depression symptoms and subsequent ART adherence. Sparse comparative evidence exists among PLWHA. However, evidence from the general population indicates dual-action antidepressants may have advantages in certain circumstances relevant to PLWHA. Mirtazapine (dual action) has been shown to have a faster onset of action compared with single-action antidepressants [19], which is important to PLWHA given the detrimental effects of depression in this population. Bupropion (dual action) has demonstrated fewer sexual side-effects [19], which is relevant to PLWHA because this population is at risk for sexual dysfunction independent of antidepressant exposure [20]. Finally, SNRIs have demonstrated superior efficacy in more severe cases of depression [21], which is important given PLWHA are prone to worse depression [22].

Unfortunately, the few comparative studies of antidepressants on depression symptoms among PLWHA are inconclusive [23,24]. Moreover, the investigations that were identified did not include HIV clinical outcomes. Furthering the knowledge regarding this relationship through a comparative effectiveness study is important because choosing an antidepressant requires physician consideration of complex factors, including side-effect profile, cost, and past response [8]. A comparative study is well suited to address whether or not differential effectiveness should be included as an additional factor in the choice of an antidepressant.

Accordingly, we examined the change in HIV clinical outcomes among PLWHA with depression initiating antidepressants, and compared the effectiveness of dual-action and single-action antidepressants on improving viral load suppression and CD4+ T-cell count. We hypothesized that initiation of antidepressant treatment would be associated with improvements in HIV clinical outcomes for both types of antidepressants. We also hypothesized that dual-action antidepressants would be more effective than single-action antidepressants for improving viral suppression and CD4+ T-cell count. Additionally, we performed a secondary analysis using depression measures as outcomes among a subsample of observations to examine whether improvements in depression symptoms show parallel improvements in HIV clinical outcomes.


Data and participants

The study used data between 2004 and 2014 from the Center for AIDS Research Network of Integrated Clinical Systems (CNICS) [25]. CNICS is a network of clinics located across the United States that provide care to PLWHA. CNICS integrates demographic information, medical records diagnoses, medication utilization, lab results, health service appointment history, and patient-reported outcomes (PROs). Data verification and standardization procedures are described elsewhere [26].

We employed a new user approach [27] to identify the first occurring antidepressant treatment episode for a participant, consisting of a preindex date washout period (baseline), index date, and a 12-month postindex period (follow-up). We required a washout period of at least 90 days where the patient did not receive an antidepressant under investigation. The day immediately following the end of the washout period was considered to be the index date, or date of treatment initiation. Antidepressant treatment episodes were divided into two groups based on the number of neurotransmitters affected by the medication. Dual-action antidepressants included mirtazapine, bupropion, venlafaxine, desvenlafaxine, and duloxetine [9]. Single-action antidepressants included citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, and sertraline [9].

The main approach was intent-to-treat, meaning that episodes were analyzed based on the initial antidepressant prescription regardless of whether the treatment was maintained, discontinued, augmented, or switched during follow-up. As intent-to-treat estimates can be biased because of these therapeutic changes, we also conducted a per protocol sensitivity analysis [28]. We defined per protocol as continuous receipt of the original antidepressant for the entire follow-up period, with no switching or augmentation.

A total of 4985 participants received an antidepressant medication under investigation in this study (Fig. 1). We excluded 950 participants diagnosed with a serious mental illness other than depression (e.g. schizophrenia, bipolar disorder, personality disorders) that could potentially alter the prognosis of antidepressant treatment. Another 50 participants were excluded because of missing date information in their medication utilization records. This left 3985 participants eligible to contribute a treatment episode. Among these participants, 1771 had a diagnosis of depression, with a treatment episode meeting the 90-day washout period that occurred on or after 1 August 2004. Of these 1771 treatment episodes, 361 were excluded because of missing data. This left 1410 intent-to-treat episodes for analysis including 418 (30%) for dual action and 992 (70%) for single-action antidepressants.

Fig. 1
Fig. 1:
Consort diagram.

Main analysis outcomes

The primary outcomes included viral suppression and CD4+ T-cell count which were obtained from lab history files. We defined viral suppression as a binary measure (yes/no) using a threshold of less than 200 HIV-1 RNA copies/ml [29]. CD4+ T-cell count was defined as the mean difference (MD) in the absolute number of CD4+ T lymphocytes/μl. We used observations closest to the index date as baseline measures and the last observed lab in the postindex period for the follow-up measure.

Baseline covariates

Several baseline covariates were identified to control for differences in the treatment groups at baseline. Covariates included exposure to single-action or dual-action antidepressants prior to the current treatment episode; a history of other diagnoses (anxiety, an AIDS-defining illness, history of smoking, at-risk drinking, or drug use) identified with medical records, and whether or not the patient was receiving ART on the index date obtained from medication utilization data. Demographic variables included age at index date, race/ethnicity, sex, and CNICS clinic. We also identified appointments with a psychiatrist using health service appointment data.

Other covariates

For descriptive purposes, several other covariates reflecting events occurring during the postindex period were identified. These covariates included augmentation (the addition of a new psychotropic medication with concurrent receipt of the original antidepressant lasting more than 30 days); switching (the addition of a new psychotropic medication with concurrent receipt of the original antidepressant lasting fewer than 30 days); treatment exposure days (number of days of continuous receipt of original antidepressant beginning on the index date during the follow-up period); and a binary variable (yes/no) indicating whether or not the participant received ART continuously throughout the treatment episode follow-up period. We also identified if the patient had attended an appointment with a psychiatrist.

Statistical analysis

All statistical analyses were conducted using SAS version 9.4 [30]. We first ran tests for association on each outcome. For viral suppression, we used SAS PROC FREQ with the McNemar's option and Cochran–Mantel–Haenszel test for paired binary data to test for: a statistical difference in frequencies between the baseline and follow-up period; a statistically significant interaction between time period and treatment group. We used SAS PROC GLM for repeated measures to test for differences in CD4+ T cells/μl between baseline and follow-up and to assess the statistical significance of the interaction between treatment group and time period. We generated average treatment effect estimates with generalized estimating equations (GEE) [31]. GEEs address within-patient correlation for repeated measures and clustering by CNICS site. GEEs were created using the SAS PROC GENMOD procedure with an exchangeable working correlation matrix and the ‘repeated’ option for patient identifier, clustered by CNICS site. In the GEE model for viral load suppression, we used a binomial distribution with a log link to estimate risk ratios. For CD4+ T-cell count, we used a normal distribution with an identity link to estimate MDs.

We generated estimates for three treatment effects. First, we estimated the expected difference in the outcomes at baseline between dual-action and single-action antidepressants. Additionally, we estimated the expected change in the outcome between baseline and follow-up associated with initiating antidepressants independent of the treatment group. Finally, comparative effectiveness was estimated with a difference-in-difference approach by adding an interaction term for treatment group (dual vs. single action) and study period (baseline vs. follow-up). This estimate is the difference in the expected change in the outcome from follow-up between study periods for dual-action antidepressants and the same change for single-action antidepressants.

We used inverse probability of treatment (IPT) weights to address confounding because of absence of randomization. Covariates were selected to create IPT weights based on previous literature and the potential for confounding. IPT weight extreme values were stabilized using a method developed by Harder et al.[32]. We assessed balance on baseline covariates using standardized differences in means and frequencies [33]. We considered a covariate to have good balance if the weighted standardized difference was less than 0.25 [32]. Per protocol results did not significantly deviate from the intent-to-treat evaluation; therefore, we only reported the latter. Per protocol models are contained in the Online Supplemental Appendix,

Secondary analysis

We conducted a secondary analysis to examine the role depression symptoms play in the relationship between antidepressants and HIV clinical outcomes. The impact of depression was assessed using a self-reported Patient Health Questionnaire-9 (PHQ-9), a previously validated instrument for assessing depression symptoms [34]. CNICS began collecting PROs such as the PHQ-9 between 2005 and 2013; however, implementation varied by CNICS clinic and patients only complete PROs at appointments during routine care visits. Lack of control over collection of PROs resulted in the loss of 78% of the main sample (n = 1410) due to missing a PHQ-9 at baseline. This left 306 treatment episodes for the secondary analysis. We added two outcome measures in the secondary analysis including remission from depression and symptom severity. Remission from depression (yes/no), was defined as a PHQ-9 score of less than 5 [34]. Symptom severity was the raw score of the PHQ-9 ranging between 0 and 27 [34].

Owing to the limited sample size, these secondary analyses included only intent-to-treat evaluations. The same statistical methods used in the main analysis were employed for depression outcomes; however, we used inverse probability of observation (IPO) weights [35] in conjunction with IPT weights to address potential bias from missing data, as 22% (n = 72) of the 306 treatment episodes did not have an observed PHQ-9 in the follow-up period. The method for combining IPO and IPT weights is described elsewhere [36]. IPO and IPT weight model details are contained in the Supplemental Appendix,

Ethical reviews

The CNICS Research Review Committee approved this study on 12 December 2014. The University of Florida Institutional Review Board approved this study on 5 March 2015.



In the unweighted sample (n = 1410), a majority of the participants were men (81%), white (59%), and virally suppressed (67%) with a mean CD4+ T-cell count of 472 in the washout period (Table 1). The frequency of receiving ART on the index date and continuously during following up was 83 and 70%, respectively. Only 31% of the treatment episodes met the per protocol criteria and the mean number of continuous days with the antidepressant initiated on the index date was 217. Compared single-action participants, dual-action participants were more likely to switch or augment treatment during follow-up. Additionally, dual-action participants were more likely to have an appointment with a psychiatrist in either the washout or follow-up period. Also, dual-action participants had a greater frequency of receiving dual-action antidepressants prior to the current treatment episode washout period. The most commonly prescribed single-action antidepressant was citalopram, whereas bupropion represented the majority of dual-action treatment episodes (Supplemental Appendix, The IPT weighted sample was well balanced on baseline confounders.

Table 1
Table 1:
Intent-to-treat treatment episode characteristics (n = 1410).

Weighted intent-to-treat analysis: viral suppression

The frequency of viral suppression at baseline was 67 vs. 78% at follow-up (P ≤ 0.001) (Fig. 2a). Additionally the within treatment group differences from baseline were statistically significant (P ≤ 0.001). In GEE models, initiating antidepressants was associated with a 16% increase in the probability of viral suppression [(risk ratio = 1.16 (1.12, 1.20)] (Table 2). For the difference-in-difference estimate, we did not observe a statistically significant interaction between treatment group and study period.

Fig. 2
Fig. 2:
(a) Results of McNemar's and Cochran–Mantel–Haenszel tests for differences in viral suppression.(b) Results of ANOVA for repeated measures test for differences in CD4+ T cells/μl.
Table 2
Table 2:
Weighted generalized estimating equations intent-to-treat analysis: HIV clinical outcomes.

Weighted intent-to-treat analysis: CD4+ T-cells/μl

There was a statistically significant difference in mean CD4+ T-cell count between study periods (baseline = 472 vs. follow-up = 511; P ≤ 0.001; Fig. 2b). The within treatment group differences from baseline were also statistically significant (P ≤ 0.001). Results for the GEE models show that initiating antidepressants was associated with a mean increase of 39 CD4+ T-cells/μl [MD = 39 (30, 48)] (Table 3). The difference-in-difference estimate was not statistically significant.

Table 3
Table 3:
Secondary analysis: weighted generalized estimating equations intent-to-treat analysis: depression outcomes.a

Secondary analysis


Of the 306 treatment episodes in the secondary analysis, 220 (72%) were for single action and 86 (28%) were for dual-action antidepressants (Supplemental Appendix Table 13, Viral suppression (76%), mean CD4+ T-cell count (526), and receipt of ART (90%) in the preindex period were somewhat higher in this subsample, but otherwise the treatment episode characteristics were similar to the main sample. After applying stabilized IPT weights, preindex period covariates were well balanced between treatment groups except for at-risk drinking, smoking, race (other/unknown), and CNICS site; therefore, we included these unbalance covariates into the regression equations. The HIV clinical outcome analyses, when repeated in this subsample, yielded substantively similar results to those reported from the main sample above (Supplemental Appendix,

Weighted intent-to-treat analysis: depression remission and Patient Health Questionnaire-9 score

The frequency of remission (PHQ-9 <5) at baseline was 26 vs. 35% at follow-up (P = 0.01; Supplemental Appendix Table 13, Within treatment group differences from baseline were statistically significant (P = 0.01). In GEE models (Table 3), there was a 36% increase in the probability of remission associated with initiating antidepressants [risk ratio = 1.36 (1.08, 1.71)]. The interaction between treatment group and study period (difference-in-difference) was not statistically significant.

Baseline mean PHQ-9 was 10.2 compared with 7.8 at follow-up (P < 0.001; Supplemental Appendix Table 13, Within treatment group differences were statistically significant for both single and dual-action antidepressants (P < 0.001). In GEE models (Table 3) there was a 2.5 point decrease in the mean PHQ-9 score associated with initiating antidepressants [MD = −2.5(−3.5,−1.6)]. The difference-in-difference estimate was not statistically significant.


Our results demonstrated that initiating antidepressant treatment in the course of routine HIV care is associated with improvements in viral suppression and CD4+ T-cell count. Additionally, we found that initiation of antidepressant treatment was associated with reductions in depression symptoms. Such findings suggest improvements in depression correspond to improvements in HIV clinical outcomes following the initiation of antidepressant treatment. It is, therefore, possible the relationship between initiating antidepressant treatment and HIV clinical outcomes is mediated by reductions in depression symptoms and subsequent improvements ART adherence. However, we cannot speak definitively to this mediation pathway because data restrictions (lack of treatment episodes with ART adherence PROs) prevented us from conducting a mediation analysis. Despite the inability to conduct a mediation analysis, our results strengthen evidence supporting the mediation pathway generated from past studies [13,17,18].

Contrary to our second hypothesis, single-action and dual-action antidepressants appear to have comparable effectiveness on the outcomes in this study. Out results for depression are consistent to a related study we previously reported using a different sample from the same cohort which showed comparable effectiveness between single and dual-action antidepressants [23]. However, this previous analysis was not able to control for baseline depression severity as done here. Our depression results are also consistent with a prior randomized controlled trial comparing mirtazapine (dual action) to escitalopram (single action) [24]. Although consistency with past study results may explain our findings, another potential explanation exists. Specifically, postindex period therapeutic changes in the present study may be driving the depression results. Specifically, we observed a greater frequency of switching and augmentation in the dual-action group. These occurrences may be because of inadequate treatment response; however, data restrictions prevented identifying what drove these changes.

Given that we observed similar changes in depression symptoms, it is plausible that participants experienced comparable improvements in ART adherence. As such, the observed comparable improvements in HIV clinical outcomes are not surprising. Nevertheless, we cannot definitely speak to the impact of ART adherence because of data restrictions as noted above.

It is important to note that participants experienced a relatively small therapeutic response in depression symptoms. We observed a 2.5 point decrease PHQ-9 scores (Supplemental Appendix Table 13,, which is below the defined level of a clinically meaningful change (≥5 points) [35]. Additionally, the mean PHQ-9 at follow-up was 7.8 which is still indicative of unresolved depression [34]. Moreover, only 35% of participants met the criteria for remission at follow-up which represents a relatively small increase (11% points) in remission rates. These findings indicate more research is needed to improve the overall effectiveness of depression treatment for PLWHA.

Our results should be interpreted with caution given the limitations encountered in this study. We were unable to control for baseline depression severity in the main analysis. However, we were able to control for baseline depression severity in the secondary analysis which produced similar results for HIV clinical outcomes as those observed in the main analysis. Consistency in the findings for HIV clinical outcomes between the main and secondary analyses indicates depression severity was most likely not a significant confounder. We were also unable to account for ART adherence because of data restrictions. As such we cannot determine if differences in baseline ART adherence biased our estimates for HIV clinical outcomes. However, ART adherence is theoretically a mediator rather than a confounder; therefore, this omission does not raise significant concerns about uncontrolled confounding. Another limitation comes from combining multiple medications into our treatment groups. This approach assumes comparable effectiveness across individual medications within each group; however, our grouping is theoretically justified based on the ‘dual-action’ hypothesis which suggests that medications with dual action may have systematic differences in effectiveness compared to SSRIs [9,36]. Data restrictions also prevented us from fully accounting for important aspects of treatment beyond antidepressants. Specifically, we could not determine the impact of variations in nonpharmacological interventions, prescribing physician type and medication dosage. However, we were able to balance treatment groups on a rich set of covariates, which likely mitigated some bias from these confounders (e.g. past treatment experiences, psychiatric appointments).

Despite these limitations, this study makes several contributions to the field of HIV research. First, to our knowledge this is the only study that has compared the impact of dual-action and single-action antidepressants on depression symptoms and HIV clinical outcomes. Second, the results from this study strengthen findings from prior work that have demonstrated support for the connection between antidepressants and HIV clinical outcomes. Such evidence highlights the potential for circumventing the deleterious impact depression has on HIV disease progression with antidepressants. Third, our evidence adds support for current guidelines that suggest available antidepressants have roughly comparable effectiveness, and antidepressant selection can, therefore, be guided by patient preferences, side-effect profiles, previous responses, and potential drug interactions [8]. Finally, we used observational data collected in the course routine HIV care; therefore, our results are generalizable to the complex circumstances physicians face when treating depression among PLWHA.

In conclusion, we found that dual-action and single-action antidepressants have a comparable positive impact on depression symptoms and HIV clinical outcomes for PLWHA diagnosed with depression. Additional studies should build upon this investigation by comparing specific medications, therapeutic classes, and existing collaborative care interventions. Future studies should also model ART adherence and depression symptom severity as mediators. Uncovering the nature of this complex relationship should prove useful in advancing the treatment of psychiatric comorbidities in PLWHA.


All authors for this manuscript meet each of the three authorship requirements as stated in the Uniform Requirements for Manuscripts Submitted to Biomedical Journals.

All authors have reviewed and approved the manuscript for submission to AIDS.

J.C.M.: developed the study concept, design, and analytic methods; conducted the data analyses, interpreted the results, primary role of writing the manuscript, and implementing suggested revisions from coauthors and journal reviewers; J.S.H.: aided in development of the study concept, design, and analytic methods; provided guidance with data analysis and interpreting the results – critical review and editing of the manuscript; provided guidance on addressing reviewer comments; R.L.C.: aided in development of the study concept and design; provided clinical consultation regarding treatment in the course of routine clinical care; aided in the interpretation of the results; critical review and editing of the manuscript. Provided guidance on addressing reviewer comments; N.M.M.: aided in development of the study concept, design and analytic methods; provided guidance with data analysis and interpreting the results; critical review and editing of the manuscript; C.A.H.: aided in development of the study concept and design; provided guidance with interpreting the results and communication of implication of results; critical review and editing of the manuscript; provided guidance on addressing reviewer comments; R.P.D.: aided in development of the study concept; provided guidance with interpreting the results and communication of implication of results; critical review and editing of the manuscript; B.N.G.: aided in development of the study concept and design; provided clinical consultation regarding treatment in the course of routine clinical care; aided in the interpretation of the results; critical review and editing of the manuscript; provided guidance on addressing reviewer comments; B.W.P.: aided in development of the study concept, design, and analytic methods; provided guidance with data analysis and interpreting the results – critical review and editing of the manuscript.

The work was supported by National Institute of Allergy and Infectious Diseases of the National Institutes of Health (T32AI007001). Additional support was provided by CFAR-Network of Integrated Clinical Systems (CNICS), a NIH funded program (R24 AI067039) made possible by the National Institute of Allergy and Infectious Diseases (NIAID) and the National Heart, Lung, and Blood Institute (NHLBI). The CFAR sites involved in CNICS include Univ of Alabama at Birmingham (P30 AI027767), Univ of Washington (P30 AI027757), Univ of California San Diego (P30 AI036214), Univ of California San Francisco (P30 AI027763), Case Western Reserve Univ (P30 AI036219), John Hopkins Univ (P30 AI094189, U01 DA036935), Fenway Health/Harvard (P30 AI060354), and Univ of North Carolina Chapel Hill (P30 AI50410).

We thank the patients, providers, and research staff from the CNICS for providing access to the required data to complete this study.

Conflicts of interest

There are no conflicts of interest.


1. Nanni MG, Caruso R, Mitchell AJ, Meggiolaro E, Grassi L. Depression in HIV infected patients: a review. Curr Psychiatry Rep 2015; 17:530.
2. Asch SM, Kilbourne AM, Gifford AL, Burnam MA, Turner B, Shapiro MF, et al. Underdiagnosis of depression in HIV: who are we missing?. J Gen Intern Med 2003; 18:450–460.
3. Bess KD, Adams J, Watt MH, O’Donnell JK, Gaynes BN, Thielman NM, et al. Providers’ attitudes towards treating depression and self-reported depression treatment practices in HIV outpatient care. AIDS Patient Care STDS 2013; 27:171–180.
4. Olatunji BO, Mimiaga MJ, O’Cleirigh C, Safren SA. Review of treatment studies of depression in HIV. Top HIV Med 2006; 14:112–124.
5. Pence BW, O’Donnell JK, Gaynes BN. Falling through the cracks: the gaps between depression prevalence, diagnosis, treatment, and response in HIV care. AIDS 2012; 26:656–658.
6. Primeau MM, Avellaneda V, Musselman D, Jean GS, Illa L. Treatment of depression in individuals living with HIV/AIDS. Psychosomatics 2013; 54:336–344.
7. Blank MB, Himelhoch S, Walkup J, Eisenberg MM. Treatment considerations for HIV-infected individuals with severe mental illness. Curr HIV/AIDS Rep 2013; 10:371–379.
8. Adams JL, Gaynes BN, McGuinness T, Modi R, Willig J, Pence BW. Treating depression within the HIV ‘medical home’: a guided algorithm for antidepressant management by HIV clinicians. AIDS Patient Care STDS 2012; 26:647–654.
9. Rakesh J. Single-action versus dual-action antidepressants. Prim Care Companion J Clin Psychiatry 2004; 6 (supp 1):7–11.
10. Hill L, Lee KC. Pharmacotherapy considerations in patients with HIV and psychiatric disorders focus on antidepressants and antipsychotics. Ann Pharmacother 2013; 47:75–89.
11. Yanofski J, Croarkin P. Choosing antidepressants for HIV and AIDS patients: insights on safety and side effects. Psychiatry (Edgmont) 2008; 5:61–66.
12. Tsai AC, Karasic DH, Hammer GP, Charlebois ED, Ragland K, Moss AR, et al. Directly observed antidepressant medication treatment and HIV outcomes among homeless and marginally housed HIV-positive adults: a randomized controlled trial. Am J Public Health 2013; 103:308–315.
13. Tsai AC, Weiser SD, Petersen ML, Ragland K, Kushel MB, Bangsberg DR. A marginal structural model to estimate the causal effect of antidepressant medication treatment on viral suppression among homeless and marginally housed persons with HIV. Arch Gen Psychiatry 2010; 67:1282–1290.
14. Rabkin JG, Wagner G, Rabkin R. Effects of sertraline on mood and immune status in patients with major depression and HIV illness: an open trial. J Clin Psychiatry 1994; 55:433–439.
15. Rabkin JG, Wagner GJ, Rabkin R. Fluoxetine treatment for depression in patients with HIV and AIDS: a randomized, placebo-controlled trial. Am J Psychiatry 1999; 156:101–107.
16. Pence BW, Gaynes BN, Adams JL, Thielman NM, Heine AD, Mugavero MJ, et al. The effect of antidepressant treatment on HIV and depression outcomes: results from a randomized trial. AIDS 2015; 29:1975–1986.
17. Gaynes BN, Pence BW, Atashili J, O’Donnell JK, Njamnshi AK, Tabenyang ME, et al. Changes in HIV outcomes following depression care in a resource-limited setting: results from a pilot study in Bamenda, Cameroon. PLoS One 2015; 10:e0140001.
18. Horberg MA, Silverberg MJ, Hurley LB, Towner WJ, Klein DB, Bersoff-Matcha S, et al. Effects of depression and selective serotonin reuptake inhibitor use on adherence to highly active antiretroviral therapy and on clinical outcomes in HIV-infected patients. J Acquir Immune Defic Syndr 2008; 47:384–390.
19. Gartlehner G, Hansen RA, Morgan LC, Thaler K, Lux L, Van Noord M, et al. Comparative benefits and harms of second-generation antidepressants for treating major depressive disorder: an updated meta-analysis. Ann Intern Med 2011; 155:772–785.
20. Bouhnik AD, Preau M, Schiltz MA, Obadia Y, Spire B. VESPA study group. group vs. Sexual difficulties in people living with HIV in France: results from a large representative sample of outpatients attending French hospitals (ANRS-EN12-VESPA). AIDS Behav 2008; 12:670–676.
21. Bradley AJ, Lenox-Smith AJ. Does adding noradrenaline reuptake inhibition to selective serotonin reuptake inhibition improve efficacy in patients with depression? A systematic review of meta-analyses and large randomised pragmatic trials. J Psychopharmacol 2013; 27:740–758.
22. Gaynes BN, O’Donnell J, Nelson E, Heine A, Zinski A, Edwards M, et al. Psychiatric comorbidity in depressed HIV-infected individuals: common and clinically consequential. Gen Hosp Psychiatry 2015; 37:277–282.
23. Mills JC, Harman JS, Cook RL, Marlow NM, Harle CA, Duncan RP, et al. Comparative effectiveness of dual-action versus single-action antidepressants for the treatment of depression in people living with HIV/AIDS. J Affect Disord 2017; 215:179–186.
24. Patel S, Kukreja S, Atram U, Sousa A, Shah N, Yadav S, Sonavane, Sushma, et al. Escitalopram and Mirtazapine for the treatment of depression in HIV patients: A randomized controlled open label trial. ASEAN J Psychiatry 2013; 14:1.
25. CNICS. Center for AIDS Research Network of Inegrated Clinical Systems. Birmingham, Alabama: University of Alabama; 2014.
26. CNICS. CNICS Data Elements. Birmingham, Alabama: University of Alabama; 2014.
27. Ray WA. Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol 2003; 158:915–920.
28. Brookhart SB, Roy JA, Sauer B, VanderWeele TJ. Developing a protocol for observational comparative effectiveness research: a user's guide. Rockville: Agency for Healthcare Research and Quality; 2013.
29. Bradley H, Hall HI, Wolitski RJ, Van Handel MM, Stone AE, LaFlam M, et al. Vital signs: HIV diagnosis, care, and treatment among persons living with HIV: United States, 2011. MMWR Morb Mortal Wkly Rep 2014; 63:1113–1117.
30. SAS®. SAS®. In. 9.4 ed. Cary, NC: SAS® Institute, Inc.; 2012
31. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986; 42:121–130.
32. Harder VS, Stuart EA, Anthony JC. Propensity score techniques and the assessment of measured covariate balance to test causal associations in psychological research. Psychol Methods 2010; 15:234–249.
33. Austin PC, Stuart EA. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat Med 2015; 34:3661–3679.
34. Kroenke K, Spitzer RL, Williams JB. The PHQ-9: validity of a brief depression severity measure. J Gen Intern Med 2001; 16:606–613.
35. McMillan D, Gilbody S, Richards D. Defining successful treatment outcome in depression using the PHQ-9: a comparison of methods. J Affect Disord 2010; 127:122–129.
36. Delgado PL. Neurobiology of serotonin norepinephrine reuptake inhibitors. Prim Psychiatry 2009; 16:8–15.

CD4+; comparative effectiveness research; depression; HIV/AIDS; second-generation antidepressive agents; viral load

Supplemental Digital Content

Copyright © 2017 Wolters Kluwer Health, Inc.