Barriers to Antiretroviral Therapy Adherence and Plasma HIV RNA Suppression Among AIDS Clinical Trials Group Study Participants

Abstract

We conducted a secondary data analysis of 11 AIDS Clinical Trials Group (ACTG) studies to examine longitudinal associations between 14 self-reported antiretroviral therapy (ART) adherence barriers (at 12 weeks) and plasma HIV RNA (at 24 weeks) and to discern the relative importance of these barriers in explaining virologic detectability. Studies enrolled from 1997 to 2003 and concluded between 2002 and 2012. We included 1496 (54.2% of the original sample) with complete data. The most commonly selected barriers were “away from home” (21.9%), “simply forgot” (19.6%), “change in daily routine” (19.5%), and “fell asleep/slept through dosing time” (18.9%). In bivariate analyses, “too many pills to take” (OR=0.43, p<0.001), “wanted to avoid side effects” (OR=0.54, p=0.001), “felt drug was toxic/harmful” (OR=0.44, p<0.001), “felt sick or ill” (OR=0.49, p<0.001), “felt depressed/overwhelmed” (OR=0.58, p=0.004), and “problem taking pills at specified time” (OR=0.71, p=0.04) were associated with a lower odds of an undetectable HIV RNA. “Too many pills to take,” “wanted to avoid side effects,” “felt drug was toxic/harmful,” “felt sick/ill,”, and “felt depressed/overwhelmed” had the highest relative importance in explaining virologic detectability. “Simply forgot” was not associated with HIV RNA (OR=0.99, p=0.95) and was ninth in its relative importance. Adherence interventions should prioritize barriers with highest importance in explaining virologic outcomes rather than focusing on more commonly reported barriers.

Introduction

Optimal adherence to antiretroviral therapy (ART) is a consistent predictor of HIV virologic suppression, improved quality of life, reduced health care costs, slower disease progression, and overall survival.^1

–7 Despite the ability to attain virologic suppression with adherence rates as low as 70–80%, depending on the ART regimen,^8
–10 ART adherence in North America has been estimated to be as low as 55%¹¹ and lower mean ART adherence levels of 33–50% have been reported in the US.^12,13 Therefore, it is imperative to understand barriers to adherence and to examine the association between these barriers and HIV treatment goals in order to develop effective ART adherence interventions.

Prior research has heavily focused on commonly reported adherence barriers to determine intervention targets. For example, given that forgetting to take ART has been reported to be one of the most commonly stated adherence barriers,^12

–16 many studies have examined the use of reminder devices in improving adherence.^17

–20 However, most have not revealed clinically significant changes in adherence. Studies have reported other adherence barriers, including feeling depressed or overwhelmed, fear of disclosure, sleeping through a dose, substance use, regimen complexity, not having medication, change in daily routine, and not wanting to be reminded of HIV infection.^12

–16 However, the association between commonly reported adherence barriers and lack of virologic suppression, as well as the relative importance of barriers in predicting plasma HIV RNA, has not been examined. This information is critical in designing future interventions that are more likely to have a positive impact on HIV treatment goals. Therefore, the objectives of our study were: (1) to examine the association between self-reported adherence barriers and virologic detectability, and (2) to establish the relative importance of each adherence barrier in explaining virologic detectability.

Methods

Study design

We conducted a secondary analysis of longitudinal data collected as part of AIDS Clinical Trials Group (ACTG) ART studies that were available for analysis at the time of our request (April 2013). We received approval from the University of California, San Francisco Committee on Human Research and the ACTG's Scientific Agenda Steering Committee of the ACTG Executive Committee.

Setting and study population

We included all ACTG ART studies that were conducted in the US, had used the ACTG adherence barriers questionnaire²¹ at least at 12 weeks (±4 weeks), and had collected participant demographics (i.e., age, sex, race/ethnicity, and HIV risk behaviors) at baseline and plasma HIV RNA at 24 weeks (±4 weeks). These time points were selected to account for the temporal ordering of adherence barriers and HIV RNA and because these were the time points when all studies had conducted measurements for adherence barriers prior to subsequent HIV RNA levels. If studies were conducted at both US and non-US sites, we only included participants from US sites. Studies were excluded if they examined non-oral ART, assessed the impact of treatment interruption before 24 weeks (±4 weeks), or recruited fewer than 10 participants.

We identified 11 ACTG studies, four of which included ART-naïve participants (ACTG 371²², ACTG 384^23,24, ACTG 746²⁵, and A5073^26,27) and seven that included ART-experienced participants (ACTG 372^28,29, ACTG 398³⁰, ACTG 400, A5025³¹, A5116³², A5126³³, A5143^34,35). The range of enrollment was 1997–2003 and studies were concluded between 2002 and 2012 (median=2009). Each study enrolled an average of 280 participants (range=25–987) and had a mean duration of 93 weeks (range=24–220). We included all participants who had completed the ACTG adherence barriers questionnaire.

Variables

HIV RNA at 24 weeks (±4 weeks) constituted our primary outcome and was dichotomized (detectable/undetectable) based on each study's assay cut-off for the lower limit of quantification to designate viral suppression. Adherence barriers were assessed at 12 weeks (±4 weeks) using the ACTG adherence barriers questionnaire²¹ in which participants are asked: “In the past month, how often have you missed taking your medications because you: (1) were away from home; (2) were busy with other things; (3) simply forgot; (4) had too many pills to take; (5) wanted to avoid side effects; (6) did not want others to notice you taking medication; (7) had a change in daily routine; (8) felt like the drug was toxic/harmful; (9) fell asleep/slept through dose time; (10) felt sick or ill; (11) felt depressed/overwhelmed; (12) had problem taking pills at specified times (with meals, on empty stomach, etc.); (13) ran out of pills; and (14) felt good?” Adherence barriers were dichotomized (yes/no), and summed and categorized based on their distribution (0=no adherence barriers; 1=1–4 barriers; 2=5–14 barriers). These categories were determined post hoc and based on the distribution of the data. Potential confounders included: age, sex, race/ethnicity (white, black, latino, other), HIV risk behaviors [men who have sex with men (MSM), heterosexual, needle sharing, transfusion, other/unknown], and study protocol number. Study protocol number represented the data collection context, reflecting timeframe, geography, population, type of study, and study intervention. By controlling for study protocol number, we thus controlled for a summary measure of the data collection context as it relates to antiretroviral regimens, time and geography.

All ACTG trials used a self-administered questionnaire that was completed by the participant in a quiet secluded place. The questionnaire was written at the sixth-grade level, but subjects may have had the questionnaire read to them, if requested. By not having the study nurses read the questions in a one-on-one interview, social desirability concerns were lessened.^36,37 In addition, these self-report adherence questionnaires have been shown in previous studies to predict both viral suppression and overall HIV progression and death.^37,38

Analysis

After limiting our dataset to participants who had responded to all 14 barriers and had documented plasma HIV RNA results, we described sample characteristics at 12 weeks. Next, we used bivariate logistic regression to examine the association between individual adherence barriers at 12 weeks and virologic detectability at 24 weeks. We also examined the association between total number of reported barriers at 12 weeks (categorized as 0=no adherence barriers; 1=1–4 barriers; 2=5–14 barriers) and virologic detectability at 24 weeks to examine the cumulative effect of barriers on adherence. Inference was performed using robust Huber-White standard errors.³⁹ We used multivariate logistic regression to examine the association between each adherence barrier at 12 weeks and virologic detectability at 24 weeks while adjusting for potential confounders.

Because studies were conducted at different times, geographical locations across the US, and included different ART regimens and study populations, we assessed interaction terms between study protocol number and 14 adherence barriers to ensure that effects of adherence barriers on virologic detectability did not vary by protocol number. These interactions were added to a model containing main effects for study protocol number and the 14 barriers and tested one by one to examine whether the relationships between barriers and virologic detectability varied as a function of study protocol number.

Finally, we used dominance analysis, a technique which rank-ordered the relative importance or contribution of the 14 adherence barriers at 12 weeks in explaining virologic detectability at 24 weeks. This technique is based on the average pseudo R² explained by each adherence barrier across all possible subsets regression models.^40,41 To ensure replication of our results and to help interpret the dominance analysis, we calculated the population attributable risk (PAR) of having detectable plasma HIV RNA for each barrier.⁴² In a post-hoc analysis, we examined the relative importance of the 14 barriers at 12 weeks using dominance analysis for those who had been ART-naïve at baseline. We used Stata version 13.1 (StatCorp LP, College Station, TX) for our analyses. p Values less than 0.05 were deemed statistically significant.

Results

Of 2759 participants enrolled in the 11 included ACTG trials, 1263 (45.8%) had missing adherence barrier and/or HIV RNA data and were excluded from the analysis. Among those excluded, 806 (63.8%) had no responses to any of the ACTG adherence barriers questions, 353 (27.9%) did not have an HIV RNA reported, and 104 (8.2%) had some missing adherence barrier data. Participants' sex, race/ethnicity, and HIV risk behavior did not differ between those who had a report HIV RNA and those who had missing data. Having a missing HIV RNA was more likely among younger individuals (mean age=36.8 years among those missing HIV RNA data and mean age=39.3 years among those who had a reported HIV RNA, p<0.001), as well as in studies with ART-experienced participants (HIV RNA missing in 9% of studies with ART-naïve participants versus 14% of studies with ART-experienced participants, p<0.001). Participants' age, sex, race/ethnicity, and HIV risk behavior did not differ between those missing or not missing all ACTG adherence barrier responses. However, adherence barrier data was missing in 34.3% of studies with ART-naïve participants versus 24.7% of studies with treatment-experienced participant (p<0.001).

Table 1 summarizes characteristics of 1496 (54.2%) individuals at 12 weeks who met all inclusion/exclusion criteria. On average, participants were 39 years old, 85% male, 51% white, 29% black, 18% Latino, and 61% identifying MSM as their HIV risk behavior. Most participants (53%) were treatment naïve, 46.5% had an undetectable HIV RNA at 12 weeks, and 66% had an undetectable HIV RNA at 24 weeks.

Table 1.

Characteristics of Study Population at Week 12 (N=1496)

Mean age, years (SD)	39.3 (9.2)
Male, N (%)	1275 (85.2)
Race/ethnicity, N (%)
White	756 (50.5)
Black	426 (28.5)
Latino	264 (17.7)
Other	50 (3.3)
HIV risk behavior, N (%)^b
MSM	798 (61.0)
Heterosexual	316 (24.1)
Needle sharing	69 (5.3)
Transfusion	28 (2.1)
Other/do not know	98 (7.5)
Treatment naïve, N (%)	792 (52.9)
Study protocol number, N (%)
ACTG 371^a	102 (6.8)
ACTG 372	42 (2.8)
ACTG 384^a	358 (23.9)
ACTG 398	278 (18.6)
ACTG 400	21 (1.4)
ACTG 746^a	101 (6.8)
A5025	168 (11.2)
A5073^a	231 (15.4)
A5116	132 (8.8)
A5126	21 (1.4)
A5143	42 (2.8)
Mean CD4+ cell count, cells/mL (SD)^c	402.2 (265.2)
Plasma HIV RNA below limit of quantification, N (%)	694 (46.5)
Barriers to adherence, N (%)
Away from home	328 (21.9)
Simply forgot	293 (19.6)
Change in daily routine	292 (19.5)
Fell asleep/slept through dose time	282 (18.9)
Busy with other things	255 (17.1)
Felt sick or ill	186 (12.4)
Problem taking pills at specified time	181 (12.1)
Wanted to avoid side effects	133 (8.9)
Felt depressed/overwhelmed	127 (8.5)
Felt good	111 (7.4)
Not want others to notice you taking medications	105 (7.0)
Felt like the drug was toxic/harmful	82 (5.5)
Too many pills to take	79 (5.3)
Ran out of pills	48 (3.2)
Summed adherence barriers, N (%)
0 (no ACTG barriers reported)	845 (56.5)
1 (1–4 ACTG barriers reported)	446 (29.8)
2 (≥5 ACTG barriers reported)	205 (13.7)

ACTG, AIDS Clinical Trials Group; ART, antiretroviral therapy; MSM, men who have sex with men; SD, standard deviation.

ACTG ART naïve studies; ^bN=1309; ^c N=1483.

Over 56% reported no adherence barriers. The most commonly selected barriers were “away from home” (21.9%), “simply forgot” (19.6%), “change in daily routine” (19.5%), and “fell asleep/slept through dosing time” (18.9%). “Ran out of pills” (3.2%), “too many pills to take” (5.3%), and “felt drug was toxic/harmful” (5.5%) were the least common barriers.

In bivariate analyses, “too many pills to take,” “wanted to avoid side effects,” “felt drug was toxic/harmful,” “felt sick or ill,” “felt depressed/overwhelmed,” and “problem taking pills at specified time” were the only barriers that were associated with a lower odds of having an undetectable HIV RNA (Table 2). In multivariate analyses, “felt sick or ill” (OR=0.53, 95% CI=0.37–0.76, p<0.001) was significantly associated with lower odds of undetectable HIV RNA while adjusting for all confounders. “Too many pills to take” (OR=0.61, 95% CI=0.37–1.01, p=0.06) and “felt like the drug was toxic/harmful” (OR=0.62, 95% CI=0.37–1.04, p=0.07) were marginally associated with a lower odds of an undetectable HIV RNA when controlling for all confounders. In comparison to not reporting any adherence barriers, those reporting 1–4 barriers had a 0.85 odds of an undetectable HIV RNA (95% CI=0.67–1.09, p=0.2) and those reporting ≥5 barriers had a 0.64 odds of an undetectable HIV RNA (95% CI=0.46–0.87, p=0.005). Study protocols ACTG 398, ACTG 746, A5116, and A5126 were also significantly associated with lower odds of undetectable HIV RNA. Among these studies, ACTG 746 was the only study conducted in treatment-naïve individuals. Interactions between study site and adherence barriers were not statistically significant and therefore dropped from analyses.

Table 2.

Odds Ratios (ORs) and 95% Confidence Intervals (CIs) for the Bivariate Association Between Adherence Barriers at 12 Weeks and Undetectable Plasma HIV RNA at 24 Weeks (N=1496)

	OR (95% CI)	p Value
Adherence barriers
Away from home	0.88 (0.68–1.13)	0.32
Busy with other things	0.95 (0.72–1.26)	0.73
Simply forgot	0.99 (0.76–1.30)	0.95
Too many pills to take	0.43 (0.27–0.68)	<0.001
Wanted to avoid side effects	0.54 (0.38–0.77)	0.001
Not want others to notice you taking medications	0.94 (0.62–1.43)	0.77
Change in daily routine	0.82 (0.63–1.07)	0.14
Felt like the drug was toxic/harmful	0.44 (0.28–0.70)	<0.001
Fell asleep/slept through dose time	0.84 (0.64–1.10)	0.20
Felt sick or ill	0.49 (0.36–0.66)	<0.001
Felt depressed/overwhelmed	0.58 (0.40–0.84)	0.004
Problem taking pills at specified time	0.71 (0.52–0.98)	0.04
Ran out of pills	1.26 (0.67–2.37)	0.48
Felt good	0.80 (0.54–1.19)	0.27
Summed adherence barriers^a	-	0.02^b
1 (1–4 ACTG barriers reported)	0.85 (0.67–1.09)	0.20
2 (≥5 ACTG barriers reported)	0.64 (0.46–0.87)	0.005

“No adherence barriers” as reference category; ^bOmnibus Wald test.

In dominance analysis, we examined the relative importance of each barrier in its association with virologic detectability. The top five barriers with the most impact on virologic detectability were: 1. “felt sick or ill,” 2. “had too many pills to take,” 3. “felt like the drug was toxic/harmful,” 4. “wanted to avoid side effects,” and 5. “felt depressed/overwhelmed.” The PAR resulted in the same top five barriers but ranked them slightly differently from the dominance analysis. The PAR associated with “felt sick or ill,” “wanted to avoid side effects,” “felt depressed/overwhelmed,” “felt like the drug was toxic/harmful,” and “had too many pills to take” were 6.2%, 3.8%, 3.2%, 3.2%, and 3.2%, respectively. The dominance analysis ranking of the remaining barriers was as follows with the PAR reported in parentheses: 6. “ran out of pills” (−0.5), 7. “were busy with other things” (0.6), 8. “had problem taking pills at specified times” (2.8), 9. “simply forgot” (0.1), 10. “did not want others to notice you taking medication” (0.3), 11. “felt good” (1.1), 12. “had a change in daily routine” (2.6), 13. “fell asleep/slept through dose time” (2.2), and 14. “were away from home” (1.9). In a post-hoc analysis, the relative importance of adherence barriers at 12 weeks in their association with virologic detectability for individuals who were ART-naïve at study entry was generally similar to the overall population with “felt sick or ill” as the most important barrier.

Discussion

In our analysis of 11 ACTG studies, individuals reporting a higher number of adherence barriers had lower odds of attaining virologic suppression, but the majority of participants did not endorse any of the listed adherence barriers. Individual barriers were reported at a low frequency, with the highest prevalence at approximately 21.9% (“away from home”). Given the association between adherence barriers and virologic detectability and that 53% of participants were not virologically suppressed at 12 weeks (time point when adherence barriers were assessed), the identification and understanding of adherence barriers are critical.

Specific adherence barriers related to ART regimens, such as pill burden or perceived/actual medication adverse effects, and feeling depressed or overwhelmed were significantly associated with viral detectability. Despite being some of the least frequently reported barriers, dominance analysis ranked these barriers as the highest in their relative importance in predicting virologic detectability. The PARs associated with these barriers were generally small but were higher than other barriers. Unexpectedly, other frequently reported barriers, such as forgetfulness, were not associated with viral suppression and were ranked relatively low in explaining virologic outcome in both dominance analysis and PAR. It may be that forgetfulness is multi-faceted and may include other barriers such as stigma, depression, drug and alcohol use, and lack of social support. We believe these findings challenge the notion that the justification for an intervention to overcome an adherence barrier be based solely on the frequency of reporting that barrier. This type of reasoning may potentially lead investigators in the wrong direction and may result in ineffective interventions and inefficient allocation of time and resources.

The association between depression and non-adherence has been well-documented.⁴³ Across 95 independent samples, depression was strongly associated with non-adherence, and this relationship was not limited to individuals with clinical depression. Many studies have suggested a 20–50% prevalence of depression and depressive symptoms in HIV-infected patients.^44
–46 Our results add to this body of evidence that feeling depressed is highly correlated with viral detectability and ranks high in its relative importance in explaining this outcome. These findings underscore the need for behavioral interventions aimed at reducing clinical and subclinical depression.

Our findings should be viewed in the context of the following limitations. We conducted a secondary analysis of data collected for other purposes and which contained a substantial number of research participants with missing data. We relied on self-reported adherence barriers, which may be prone to recall and social desirability bias and may not have fully captured all adherence barriers (e.g., substance use,⁴⁷ trust in HIV care provider, pregnancy,^36,48 etc.). We examined adherence barriers at only one time point early in the ART course. However, these barriers will likely vary during an individual's life; therefore, their assessment over time is critical to address in order to overcome non-adherence. Due to a great deal of heterogeneity within and across studies with regard to ART regimen type and time on ART, we were unable to fully control for these variables and were unable to estimate their unique effects accurately. Because the effects of these variables are reflected in the different protocols, inclusion of the protocol number in the multivariate analysis partly accounted for their influence. Lastly, our results capture information from a time period preceding newer and better tolerated ART regimens; therefore, our results may be less generalizable to the present ART era.

Based on our findings, we believe that the assessment of adherence barriers in clinical encounters is critical. ART-related barriers such as high pill burden and adverse effects, as well as feeling depressed or overwhelmed are particularly important barriers to inquire about and address. Interventions should focus on barriers that have been associated with poor virologic outcomes rather than focusing on the most commonly reported barriers.

Currently, ART regimens are more potent, require fewer daily pills, and have improved tolerability profiles. However, lifelong adherence to ART is still a prevailing predicament and many studies continue to examine clinical tools, programs, and interventions to achieve and maintain a high level of adherence.^49
–51 Therefore, assessing adherence barriers and examining the association between barriers and HIV RNA detectability for newer regimens can be used to design and develop effective interventions to improve HIV treatment outcomes.

Footnotes

Acknowledgments

The authors would like to thank Mr. Justin Ritz and Ms. Ann Walawander for their help with data assembly and management, and for the many sites and patients that participated in the 11 ACTG trials included in this analysis.

Funding: The project described was supported by Award Number U01AI068636 from the National Institute of Allergy and Infectious Diseases and supported by National Institute of Mental Health (NIMH), National Institute of Dental and Craniofacial Research (NIDCR). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Allergy and Infectious Diseases or the National Institutes of Health. In addition, the project described was supported by NIH/NIMH Grant Numbers K23MH097649, K24MH087220, and U01 AI69471.

Author Disclosure Statement

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Bangsberg

, Perry

, Charlebois

, et al. Non-adherence to highly active antiretroviral therapy predicts progression to AIDS. AIDS, 2001; 15:1181–1183.

Garcia de Olalla

, Knobel

, Carmona

, Guelar

, Lopez-Colomes

, Cayla

. Impact of adherence and highly active antiretroviral therapy on survival in HIV-infected patients. J Acq Immun Def Synd, 2002; 30:105–110.

Hogg

, Heath

, Bangsberg

, et al. Intermittent use of triple-combination therapy is predictive of mortality at baseline and after 1 year of follow-up. AIDS, 2002; 16:1051–1058.

World Health Organization. Adherence to Long-Term Therapies: Evidence for Action. In: World Health Organization, ed. Geneva, 2003.

Chesney

. The elusive gold standard. Future perspectives for HIV adherence assessment and intervention. J Acq Immun Def Synd, 2006; 43:S149–S155.

Nachega

, Hislop

, Dowdy

, Chaisson

, Regensberg

, Maartens

. Adherence to nonnucleoside reverse transcriptase inhibitor-based HIV therapy and virologic outcomes. Ann Int Med, 2007; 146:564–573.

Aitken

, Valkova

. Avoidable Costs in U.S. Healthcare: The 200 Billion Opportunity from Using Medicines More Responsibly. 2013. Parsippany, NJ: IMS Institute for Healthcare Informatics. Available at: http://www.imshealth.com/deployedfiles/imshealth/Global/Content/Corporate/IMS%20Institute/RUOM-2013/IHII_Responsible_Use_Medicines_2013.pdf

Shuter

, Sarlo

, Kanmaz

, Rode

, Zingman

. HIV-infected patients receiving lopinavir/ritonavir-based antiretroviral therapy achieve high rates of virologic suppression despite adherence rates less than 95%. J Acq Immun Def, 2007; 45:4–8.

Bangsberg

. Less than 95% adherence to nonnucleoside reverse-transcriptase inhibitor therapy can lead to viral suppression. Clin Infect Dis, 2006; 43:939–941.

10.

Maggiolo

, Airoldi

, Kleinloog

, et al. Effect of adherence to HAART on viroloaic outcome and on the selection of resistance-conferring mutations in NNRTI- or PI-Treated patients. HIV Clin Trials, 2007; 8:282–292.

11.

Mills

, Nachega

, Buchan

, et al. Adherence to antiretroviral therapy in sub-Saharan Africa and North America: A meta-analysis. JAMA, 2006; 296:679–690.

12.

MacDonell

, Naar-King

, Huszti

, Belzer

. Barriers to medication adherence in behaviorally and perinatally infected youth living with HIV. AIDS Behav, 2013; 17:86–93.

13.

Murphy

, Sarr

, Durako

, et al. Barriers to HAART adherence among human immunodeficiency virus-infected adolescents. Arch Pediatr Adolesc Med, 2003; 157:249–255.

14.

Buchanan

, Montepiedra

, Sirois

, et al. Barriers to medication adherence in HIV-infected children and youth based on self- and caregiver report. Pediatrics, 2012; 129:e1244–e1251.

15.

Mills

, Nachega

, Bangsberg

, et al. Adherence to HAART: A systematic review of developed and developing nation patient-reported barriers and facilitators. PLoS Med, 2006; 3:e438.

16.

Murphy

, Roberts

, Hoffman

, Molina

, Lu

. Barriers and successful strategies to antiretroviral adherence among HIV-infected monolingual Spanish-speaking patients. AIDS Care, 2003; 15:217–230.

17.

Andrade

, McGruder

, Wu

, et al. A programmable prompting device improves adherence to highly active antiretroviral therapy in HIV-infected subjects with memory impairment. Clin Infect Dis, 2005; 41:875–882.

18.

Mannheimer

, Morse

, Matts

, et al. Sustained benefit from a long-term antiretroviral adherence intervention. Results of a large randomized clinical trial. J Acq Immun Def Synd, 2006; 43:S41–S47.

19.

Powell-Cope

, White

, Henkelman

, Turner

. Qualitative and quantitative assessments of HAART adherence of substance-abusing women. AIDS Care, 2003; 15:239–249.

20.

Safren

, Hendriksen

, Desousa

, Boswell

, Mayer

. Use of an on-line pager system to increase adherence to antiretroviral medications. AIDS Care, 2003; 15:787–793.

21.

Chesney

, Ickovics

, Chambers

, et al. Self-reported adherence to antiretroviral medications among participants in HIV clinical trials: The AACTG Adherence Instruments. AIDS Care, 2000; 12:255–266.

22.

Volberding

, Demeter

, Bosch

, et al. Antiretroviral therapy in acute and recent HIV infection: A prospective multicenter stratified trial of intentionally interrupted treatment. AIDS, 2009; 23:1987–1995.

23.

Smeaton

, DeGruttola

, Robbins

, Shafer

. ACTG (AIDS Clinical Trials Group) 384: A strategy trial comparing consecutive treatments for HIV-1. Controlled Clin Trials, 2001; 22:142–159.

24.

Shafer

, Smeaton

, Robbins

, et al. Comparison of four-drug regimens and pairs of sequential three-drug regimens as initial therapy for HIV-1 infection. New Engl J Med, 2003; 349:2304–2315.

25.

Fischl

, Ribaudo

, Collier

, et al. A randomized trial of 2 different 4-drug antiretroviral regimens versus a 3-drug regimen, in advanced human immunodeficiency virus disease. J Infect Dis, 2003; 188:625–634.

26.

Gross

, Tierney

, Andrade

, et al. Modified directly observed antiretroviral therapy compared with self-administered therapy in treatment-naive HIV-1-infected patients. A randomized trial. Arch Intern Med, 2009; 169:1224–1232.

27.

Flexner

, Tierney

, Gross

, et al. Comparison of once-daily versus twice-daily combination antiretroviral therapy in treatment-naive patients: Results of AIDS Clinical Trials Group (ACTG) A5073, a 48-week randomized controlled trial. Clin Infect Dis, 2010; 50:1041–1052.

28.

Hammer

, Bassett

, Squires

, et al. A randomized trial of nelfinavir and abacavir in combination with efavirenz and adefovir dipivoxil in HIV-1-infected persons with virological failure receiving indinavir. Antivir Ther, 2003; 8:507–518.

29.

Hammer

, Ribaudo

, Bassett

, et al. A randomized, placebo-controlled trial of abacavir intensification in HIV-1-infected adults with virologic suppression on a protease inhibitor-containing regimen. HIV Clin Trials, 2010; 11:312–324.

30.

Hammer

, Vaida

, Bennett

, et al. Dual vs single protease inhibitor therapy following antiretroviral treatment failure - A randomized trial. JAMA, 2002; 288:169–180.

31.

Havlir

, Gilbert

, Bennett

, et al. Effects of treatment intensification with hydroxyurea in HIV-infected patients with virologic suppression. AIDS, 2001; 15:1379–1388.

32.

Fischl

, Collier

, Mukherjee

, et al. Randomized open-label trial of two simplified, class-sparing regimens following a first suppressive three or four-drug regimen. AIDS, 2007; 21:325–333.

33.

Eron

Jr. , Park

, Haubrich

, et al. Predictive value of pharmacokinetics-adjusted phenotypic susceptibility on response to ritonavir-enhanced protease inhibitors (PIs) in human immunodeficiency virus-infected subjects failing prior PI therapy. Antimicrob Agents Chemother, 2009; 53:2335–2341.

34.

Kashuba

, Tierney

, Downey

, et al. Combining fosamprenavir with lopinavir/ritonavir substantially reduces amprenavir and lopinavir exposure: ACTG protocol A5143 results. AIDS, 2005; 19:145–152.

35.

Collier

, Tierney

, Downey

, et al. Randomized study of dual versus single ritonavir-enhanced protease inhibitors for protease inhibitor-experienced patients with HIV. HIV Clin Trials, 2008; 9:91–102.

36.

Cohn

, Umbleja

, Mrus

, Bardeguez

, Andersen

, Chesney

. Prior illicit drug use and missed prenatal vitamins predict nonadherence to antiretroviral therapy in pregnancy: Adherence analysis A5084. AIDS Patient Care STDs, 2008; 22:29–40.

37.

Cohn

, Kammann

, Williams

, Currier

, Chesney

. Association of adherence to Mycobacterium avium complex prophylaxis and antiretroviral therapy with clinical outcomes in Acquired Immunodeficiency Syndrome. Clin Infect Dis, 2002; 34:1129–1136.

38.

Cohn

, Jiang

, McCutchan

, et al. Association of ongoing drug and alcohol use with non-adherence to antiretroviral therapy and higher risk of AIDS and death: Results from ACTG 362. AIDS Care, 2011; 23:775–785.

39.

White

. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrics, 1980; 48:817–838.

40.

Azen

, Traxel

. Using dominance analysis to determine predictor importance in logistic regression. J Edu Behav Statistics, 2009; 34:319–347.

41.

Budescu

. Dominance analysis: A new approach to the problem of relative importance of predictors in multiple regression. Psychol Bull, 1993; 114:542–551.

42.

Laaksonen

, Harkanen

, Knekt

, Virtala

, Oja

. Estimation of population attributable fraction (PAF) for disease occurrence in a cohort study design. Statistics Med, 2010; 29:860–874.

43.

Gonzalez

, Batchelder

, Psaros

, Safren

. Depression and HIV/AIDS treatment nonadherence: A review and meta-analysis. J Acq Immun Def Synd, 2011; 58:181–187.

44.

Burack

, Barrett

, Stall

, Chesney

, Ekstrand

, Coates

. Depressive symptoms and Cd4 lymphocyte decline among HIV-infected men. JAMA, 1993; 270:2568–2573.

45.

Ickovics

, Hamburger

, Vlahov

, et al. Mortality, CD4 cell count decline, and depressive symptoms among HIV-seropositive women. Longitudinal analysis from the HIV epidemiology research study. JAMA, 2001; 285:1466–1474.

46.

Lyketsos

, Hutton

, Fishman

, Schwartz

, Treisman

. Psychiatric morbidity on entry to an HIV primary care clinic. AIDS, 1996; 10:1033–1039.

47.

Langebeek

, Gisolf

, Reiss

, et al. Predictors and correlates of adherence to combination antiretroviral therapy (ART) for chronic HIV infection: A meta-analysis. BMC Med, 2014; 12:142.

48.

Gertsch

, Michel

, Locatelli

, et al. Adherence to antiretroviral treatment decreases during postpartum compared to pregnancy: A longitudinal electronic monitoring study. AIDS Patient Care STDs, 2013; 27:208–210.

49.

Genberg

, Lee

, Rogers

, Willey

, Wilson

. Stages of change for adherence to antiretroviral medications. AIDS Patient Care STDs, 2013; 27:567–572.

50.

Navarro

, Perez

, Curran

, et al. Impact of an adherence program to antiretroviral treatment on virologic response in a cohort of multitreated and poorly adherent HIV-infected patients in Spain. AIDS Patient Care STDs, 2014; 28:537–542.

51.

Williams

, Wang

, Li

, Chen

, Li

, Fennie

. Efficacy of an evidence-based ARV adherence intervention in China. AIDS Patient Care STDs, 2014; 28:411–417.