Which patient characteristics among cocaine users with HIV relate to drug use and adherence outcomes following a dual-focused intervention?

Abstract

This is a secondary analysis of data from a randomized trial of dually-focused interventions for nonadherent HIV patients with cocaine use disorders (Ingersoll et al., 2011). We examined the relationships among baseline demographic, psychological, psychiatric, and behavioral characteristics and 6-months post-study ART adherence, log Viral Load (VL), ASI Drug Composite Score, and days using cocaine. We used the SAS GLMSELECT procedure to build multivariate models of each post-study outcome. Post-study ART adherence was related to 2 psychological variables; while logVL was related to 2 drug-related behaviors. ASI Drug Composite score was related to 2 psychiatric disorders, 1 demographic, and 1 psychological variable; in contrast, days using cocaine related to 1 behavioral and 3 psychological variables. Analyses show clear, robust relationships among behavioral, psychological and psychiatric diagnosis factors with post-study ART adherence and cocaine use outcomes. Future ART adherence interventions for cocaine users should consider tailoring to these patient characteristics.

Introduction

Adherence to antiretroviral therapy (ART) medications must be nearly perfect to prolong health and to avoid treatment failure, although emerging evidence suggests that some newer ART regimens may allow for slightly more nonadherence (Menendez-Arias, 2010; Nieuwkerk and Oort, 2005; Paredes and Clotet, 2010; Bangsberg et al., 2006; Tozzi et al., 2006). Poor adherence to HAART and frequent drug use each contribute to faster progression of disease, morbidity, and mortality (Baum et al., 2009; Malta et al., 2008a, 2008b).

Crack cocaine use has a particularly negative effect on ART adherence (Arnsten et al., 2002; Ingersoll, 2004; Lucas et al., 2002; Lucas et al., 2007; Sharpe et al., 2004). Crack cocaine use is an independent predictor of running out of HIV medications (Ingersoll, 2004). The binge consumption pattern typical of crack cocaine use can lead to the suspension of medication dosing during seeking, using, and recovery from the drug (Harzke et al., 2009). In addition to its negative effects on adherence, cocaine use itself is independently associated with HIV acceleration and disease progression (Baum et al., 2009; Cook et al., 2008). Cocaine users often enter HIV treatment in more advanced states of disease (Celentano et al., 2001; Wang et al., 2004), making them more vulnerable to poor treatment outcomes and associated morbidity and mortality (Zolopa, 2010).

Efforts to improve adherence to HIV medication regimens began with the advent of ART in 1996. Many studies have tested combination interventions that include counseling, reminders, and strategies to help nonadherent patients achieve and maintain adherence. Unfortunately, behavioral interventions have modest and ephemeral effects and frequently do not result in high adherence, even when theory-based and well-executed (Amico et al., 2006; Fogarty et al., 2002; Ickovics and Meade, 2002; Rueda et al., 2006; Sandelowski et al., 2009; Simoni et al., 2003, 2006). Most adherence studies have excluded active substance users, although, arguably, this is a high risk subpopulation in particular need of intervention. There is an urgent need for more potent and durable ART adherence interventions for street drug users to achieve critical benchmarks on the HIV care cascade (Gardner, McLees, Steiner, del Rio, and Burman, 2011).

To address this gap, we compared two interventions targeting both adherence and drug use, based on the Information–Motivation–Behavioral Skill model of ART adherence (Fisher et al, 2008). The pilot study was conducted from 2004-2009 in two cities in central Virginia. Participants were recruited from infectious disease clinics using flyers, clinician referral, and direct approach. English-speaking HIV positive adults with current crack cocaine use or a crack cocaine use disorder (abuse or dependence) and less than 90% self-reported adherence to a current prescription for ART over the preceding 14 days were eligible to participate. After informed consent, participants completed interviews and self-report measures, and provided urine, blood samples, and vital signs to assess drug use, VL, CD4 count, blood pressure, and pulse rate. They were randomized to a 6-session Motivational Interviewing plus Feedback intervention or to a 6-session informational video plus debriefing intervention. Follow-ups occurred at weeks 12 (post-treatment) and 24 (6 month follow-up). Participants were 54 patients in HIV care. Analyses of outcomes data showed that both interventions improved ART adherence and cocaine-related problems durably. Significant increases in ART adherence and reductions in ASI Drug Composite Scores occurred in both conditions by post-treatment and were maintained at 6 months, representing medium effect sizes. No between-group differences were observed, and no VL changes were observed in either group (Ingersoll et al., 2011). Both interventions improved cocaine use and ART adherence, with good maintenance of changes among previously nonadherent crack cocaine users.

The purpose of this report is to explore which participant characteristics related to post-study outcome variables. We hypothesized that patient characteristics drove behavior change, given that the study found moderate effects, but these did not differ between intervention conditions. Therefore, this study is a secondary, exploratory analysis of the pilot study data to determine which patient characteristics related to 4 important post-study outcomes. This information may help us to better specify the elements that may influence outcome. By attending to them in future interventions, we may tailor and improve dually-focused interventions that target nonadherence and substance use disorders. Refining interventions may increase their potency and durability, and could help to further specify models of ART adherence interventions.

Methods

Measures

The parent study measured demographics and patient characteristics including age, sex, race, ethnicity, education, employment, sexual orientation, and criminal justice involvement. It measured health variables including lifetime HIV risk behaviors, years since HIV diagnosis, immune health status, and comorbid health conditions. It measured cocaine and alcohol use, abuse and dependence, and comorbid psychiatric disorders with the Mini International Neuropsychiatric Interview (M.I.N.I., (Sheehan et al., 1997, 1998). In the parent study, mean 14-day ART adherence (defined as the percent of prescribed pills taken) was assessed using the timeline follow-back (TLFB) method, (Sobell and Sobell, 1992). logVL was another marker of ART adherence. Drug use outcomes included cocaine-specific Drug Composite Scores calculated from the ASI, a brief semi-structured interview with accepted reliability and validity for assessing the existence, duration, and severity of substance-use-related problems in seven areas (drug, alcohol, medical, employment, legal, family/social, and psychiatric problems) over the previous 30 days (McLellan et al., 1985). ASI Drug Composite Scores can show change in drug use problems over time, and offer an internally consistent estimate of drug use problems, with higher scores reflecting more problems related to drug use (McGahan et al., 1986). A secondary measure of cocaine use was the percent of days using crack cocaine, calculated from self-reports of use on each day on the 14-day TLFB.

Candidate Characteristics

The potential explanatory characteristics used for the current study were chosen because of their relationships with substance abuse and/or ART adherence established in the literature. We selected baseline participant characteristics that could be related to study outcomes at the 6 month follow-up (ASI Cocaine Composite Score, Cocaine Days, ART Adherence, and logVL) based on this literature. In general, these include demographic, health behaviors, psychiatric disorders, and substance use disorders. Additionally, we were interested in psychological variables that could influence ART adherence or cocaine use. Our goal was to identify characteristics that related to improving ART adherence and/or decreasing cocaine use in order to identify either targets for tailoring interventions, or to identify potential content areas for future interventions.

Data Analysis

We used descriptive statistics, t-tests, and chi-square tests to characterize the sample. Our approach was to first examine the univariate relationships between the set of candidate patient characteristics and each outcome variable at the 6 month follow-up (6M). We dropped all variables in each set that were not significantly related to the 6M outcome variable. To further reduce the variable set, we excluded variables that measured overlapping constructs and selected only the most important phenotype. For example, if 3 different markers of depression were related to an outcome, we selected current Major Depressive Disorder (MDD) rather than Dysthymia or lifetime MDD. We generated multivariate models of each outcome using the SAS GLMSELECT procedure. This procedure utilizes stepwise model testing, and allows for the inclusion of both continuous and categorical explanatory variables. We specified that only those explanatory variables with relationships to the outcome whose significance values were 0.1 or less should be retained in the model. Once models were developed, we examined the nature and direction of relationships between each characteristic and the related outcome through correlation analysis.

Results

Participants

Sample characteristics are shown in Tables 1-3. In brief, at baseline the sample of 54 participants was about half female, mostly African-American, unemployed, heterosexual people with at least a high school education. Most participants had detectable viral loads and CD4 counts indicating compromised immune systems. ART adherence over the past 14 days on the TLFB was 58% on average. Participants used cocaine on a third of days on the TLFB, with half positive for cocaine on urine screen at study entry. The most common psychiatric conditions identified by the MINI were Cocaine Dependence (92.3%), current Major Depressive Disorder (48.1%) and current Generalized Anxiety Disorder (30.8%). All participants had one or more psychiatric diagnoses on the MINI.

Table 1.

Demographic characteristics at baseline.

Characteristic	Full sample at baseline			Completers sample at baseline			Test
	n	Mean (SD)	Min-Max	n	Mean (SD)	Min-Max
Continuous Variables
Age	54	44.7 (6.4)	30.00-59.00	36	44.9 (5.9)	33.00-58.00	t44 = 0.64

Characteristic	Full sample at baseline n (%)	Completers sample at baseline n (%)	Test
Categorical Variables
Sex			χ2(2) = 5.00+
Men	25 (46.3%)	18 (42.9%)
Women	28 (51.9%)	24 (57.1%)
Transgender	1 (1.9%)	0
Race			χ2(1) = 0.43
Black	44 (81.5%)	35 (83.3%)
Other	10 (18.5%)	7 (16.7%)
Employment			χ2(1) = 2.24
Not Employed	44 81.5%)	36 (85.7%)
Employed	10 (18.5%)	6 (14.3%)
Education			χ2(1) = 0.96
≥ High School	34 (63.0%)	25 (59.5%)
< High School	20 (37.0%)	17 (40.5%)
Sexuality			χ2(2) = 1.70
Heterosexual	32 (59.3%)	23 (54.8%)
Homosexual	21 (38.9%)	18 (42.9%)
No response	1 (0.02%)	1 (2.4%)
Assignment			χ2(1) = 0.64
MI+	26 (48.2%)	19 (45.2%)
Video+	28 (51.9%)	23 (54.8%)

⁺p < 0.1.

*p < 0.05.

**p < 0.01.

***p < 0.001.

Table 3.

Psychological and psychiatric characteristics of the sample.

Characteristic	Full sample at baseline			Completers sample at baseline			Test
	n	Mean (SD)	Min-Max	n	Mean (SD)	Min-Max
Continuous Variables
TOFHLA Score Totala	53	29.45 (11.10)	0 - 36	42	28.95 (11.53)	0 - 36	t51 = 0.64
NEO-FFI T-scorea
Neuroticism	54	59.52 (11.88)	14 - 75	42	61.31 (9.73)	37 - 75	t13 = 1.61
Extroversion	54	47.44 (9.23)	26 - 65	42	47.12 (8.84)	27 - 65	t52 = 0.48
Openness	54	47.78 (10.68)	28 - 75	42	47.38 (9.81)	33 - 75	t52 = 0.51
Agreeableness	54	39.89 (12.41)	25 - 68	42	39.52 (12.51)	25 - 68	t52 = 0.40
Conscientiousness	54	42.57 (11.47)	25 - 68	42	42.64 (11.66)	25 - 68	t52 = 0.08
MOS Social Supporta
Tangible	54	3.48 (1.20)	1.25 - 5	42	3.43 (1.22)	1.25 - 5	t52 = 0.48
Emotional Informational	54	3.40 (1.01)	1.29 - 5	42	3.41 (0.97)	1.29 - 5	t52 = 0.09
Affectionate	54	3.57 (1.29)	1 - 5	42	3.66 (1.30)	1 - 5	t52 = 0.90
Social	54	3.40 (1.12)	1.33 - 5	42	3.48 (1.11)	1.33 - 5	t52 = 0.92
Functional	54	3.24 (1.26)	1 - 5	42	3.29 (1.31)	1 - 5	t52 = 0.49
Total Support	54	3.42 (1.01)	1.33 - 5	42	3.45 (1.00)	1.33 - 5	t52 = 0.46
Total Friends	46	3.63 (7.26)	0 - 50	35	4.14 (8.25)	0 - 50	t40 = 1.46
Treatment Motivation Scalea
Problem Recognition	54	0.69 (0.13)	0.37-0.98	42	0.69 (0.13)	0.37-0.98	t52 = 0.23
Desire for Help	54	0.73 (0.14)	0.09-0.94	42	0.71 (0.14)	0.09-0.91	t52 = 1.46
Treatment Readiness	54	0.66 (0.12)	0.43-0.97	42	0.64 (0.10)	0.48-0.92	t13 = 1.10
External Pressure	54	0.51 (0.17)	0.20-0.96	42	0.51 (0.16)	0.20-0.96	t52 = 0.41
Self-Efficacy
Confidence Cocaine Use	54	52.46 (23.24)	13 - 109	42	50.57 (24.01)	13 - 109	t52 = 1.12
Temptation Cocaine Use	54	62.70 (23.75)	2 - 113	42	62.19 (24.83)	2 - 113	t52 = 0.29
Confidence Medication Adherence	54	30.20 (10.51)	0 - 55	42	32.02 (10.15)	14 - 55	t52 = 2.50*
Temptation Medication Adherence	54	25.54 (11.94)	0 - 51	42	24.64 (12.75)	0 - 51	t52 = 1.03
Processes of Change for Cocaine Use
Consciousness Raising	54	10.30 (2.79)	3 - 15	42	10.12 (2.83)	3 - 15	t52 = 0.87
Dramatic Relief	54	10.54 (3.06)	3 - 15	42	10.24 (3.32)	3 - 15	t38 = 1.94+
Environmental Reevaluation	54	10.37 (3.37)	3 - 15	42	10.10 (3.45)	3 - 15	t52 = 1.13
Self-Reevaluation	54	11.28 (2.88)	4 - 15	42	10.86 (3.00)	4 - 15	t52 = 2.07*
Social Liberation	54	8.93 (3.20)	3 - 15	42	8.93 (3.49)	3 - 15	t32 = 0.02
Counter-Conditioning	54	10.57 (2.54)	3 - 15	42	10.29 (2.57)	3 - 15	t52 = 1.58
Helping Relationships	54	7.85 (3.28)	3 - 15	42	7.98 (3.34)	3 - 15	t52 = 0.52
Reinforcement Management	54	8.44 (3.75)	3 - 15	42	8.33 (3.54)	3 - 15	t52 = 0.40
Self-Liberation	54	10.72 (2.56)	5 - 15	42	10.38 (2.61)	5 - 15	t52 = 1.88+
Stimulus Control	54	8.22 (3.39)	3 - 15	42	8.33 (3.48)	3 - 15	t52 = 0.45
Processes of Change for Medication Adherence
Consciousness Raising	54	6.52 (3.01)	3 - 13	42	6.31 (3.09)	3 - 13	t52 = 0.95
Dramatic Relief	54	9.43 (3.16)	3 - 15	42	9.38 (3.33)	3 - 15	t52 = 0.19
Environmental Reevaluation	54	10.57 (3.31)	3 - 16	42	10.38 (3.55)	3 - 16	t52 = 0.80
Self-Reevaluation	54	11.85 (2.50)	4 - 15	42	11.71 (2.64)	4 - 15	t52 = 0.75
Social Liberation	54	10.94 (2.62)	5 - 15	42	10.88 (2.77)	5 - 15	t52 = 0.33
Counter-Conditioning	54	10.20 (3.17)	3 - 15	42	9.93 (2.97)	3 - 15	t52 = 1.20
Helping Relationships	54	11.61 (3.59)	3 - 15	42	11.64 (3.48)	3 - 15	t52 = 0.12
Reinforcement Management	54	11.70 (2.80)	5 - 15	42	11.74 (2.93)	5 - 15	t52 = 0.17
Self-Liberation	54	10.65 (3.00)	3 - 15	42	10.52 (3.09)	3 - 15	t52 = 0.57
Stimulus Control	54	6.28 (3.23)	3 - 15	42	6.12 (3.05)	3 - 14	t52 = 0.67
Visual Analog Scaling Rulers (0-100)
Importance of Changing Cocaine Use	53	88.32 (15.15)	46 - 100	41	87.76 (15.46)	46 - 100	t51 = 0.50
Confidence to Change Cocaine Use	53	74.98 (24.20)	16 - 100	41	72.54 (25.83)	16 - 100	t51 = 1.37
Readiness to Change Cocaine Use	53	77.34 (27.22)	2 - 100	41	75.51 (28.58)	2 - 100	t51 = 0.90
Importance of Changing Medication Adherence	49	90.92 (11.97)	47 - 100	42	91.36 (10.57)	61 - 100	t7 = 0.41
Confidence to Change Medication Adherence	49	79.7 6 (23.30)	4 - 100	42	78.55 (23.37)	4 - 100	t47 = 0.89
Readiness to Change Medication Adherence	49	81.69 (22.91)	0 - 100	42	82.69 (20.49)	3 - 100	t7 = 0.50
Self-Regulation Questionnaire Scorea
Informational Input	52	27.31 (5.53)	13 - 38	40	27.53 (5.28)	16 - 38	t50 = 0.51
Self-Evaluation	52	27.25 (4.12)	19 - 37	40	27.45 (3.99)	19 - 37	t50 = 0.64
Instigation to Change	52	29.94 (3.28)	23 - 39	40	29.90 (3.30)	25 - 39	t50 = 0.17
Search	52	34.81 (3.87)	26 - 44	40	34.78 (3.79)	26 - 44	t50 = 0.11
Planning	52	26.25 (5.58)	19 - 45	40	26.45 (5.70)	19 - 45	t50 = 0.46
Implement	52	27.69 (6.22)	13 - 45	40	27.60 (6.11)	18 - 45	t50 = 0.19
Plan Evaluation	52	28.81 (4.37)	14 - 38	40	29.05 (4.06)	22 - 38	t50 = 0.73
Total Score	52	202.06 (23.00)	129 - 263	40	202.80 (21.62)	168 - 263	t50 = 0.39
Discrepancy Scale Medication Total Score	51	39.18 (13.08)	13 - 67	41	38.85 (13.46)	13 - 67	t49 = 0.35
Inventory of Interpersonal Problemsa
Dominance	53	−1.91 (3.25)	−11.55 - 5.30	41	2.07 (3.51)	−11.55 - 5.3	t51 = 0.63
Love	53	1.93 (3.85)	−5.14 - 10.33	41	2.19 (3.91)	−4.78 - 10.33	t51 = 0.90

Characteristic	Full sample at baseline n (%)	Completers sample at baseline n (%)	Test
Categorical Variables
Cocaine Stage of Change			χ2(4) = 1.08
Precontemplation	1 (1.9%)	1 (2.4%)
Contemplation	3 (5.6%)	2 (4.8%)
Preparation	48 (88.9%)	37 (88.1%)
Action	1 (1.9%)	1 (2.4%)
Maintenance	1 (1.9%)	1 (2.4%)
Medication Stage of Change			χ2(4) = 2.02
Precontemplation	1 (1.9%)	1 (2.4%)
Contemplation	8 (14.8%)	5 (11.9%)
Preparation	22 (40.7%)	17 (40.5%)
Action	14 (25.9%)	11 (26.2%)
Maintenance	9 (16.7%)	8 (19.1%)
DSM-IV substance use disorders
Cocaine Dependence	48 (92.3%)	37 (90.2%)	χ2(1) = 1.16
Cocaine Abuse	2 (3.9%)	2 (4.9%)	χ2(1) = 0.56
Alcohol Dependence	13 (25.0%)	10 (24.4%)	χ2(1) = 0.04
Alcohol Abuse	7 (13.5%)	5 (12.2%)	χ2(1) = 0.27
DSM-IV psychiatric disorders
MDD Current	25 (48.1%)	21 (51.2%)	χ2(1) = 0.77
Panic D/O Current	3 (5.8%)	3 (7.3%)	χ2(1) = 0.85
Agoraphobia Current	12 (23.1%)	7 (17.1%)	χ2(1) = 3.94*
Social Phobia	8 (15.4%)	6 (14.6%)	χ2(1) = 0.08
OCD	10 (19.2%)	8 (19.5%)	χ2(1) = 0.01
PTSD	6 (11.5%)	5 (12.2%)	χ2(1) = 0.08
Generalized Anxiety D/O	16 (30.8%)	13 (31.7%)	χ2(1) = 0.08
Suicidality			χ2(3) = 2.12
High	4 (7.84%)	4 (10.0%)
Low	8 (15.7%)	7 (17.5%)
Medium	6 (11.8%)	4 (10.0%)
None	33(64.7%)	25 (62.5%)
Homicidality	7 (13.5%)	7 (17.1%)	χ2(1) = 2.17

Note: Full sample is the sample at baseline when 2 mistaken enrollments were removed.

Completers sample at baseline is the sample completing the 6 month follow-up (6M). Some n's are lower than the full sample due to missing data for that variable. Some percentages do not total to 100% due to rounding, or due to responses such as “refused” or “don't know,” (data not shown). D/O is Disorder. MDD is Major Depressive Disorder. OCD is Obsessive Compulsive Disorder. PTSD is Post Traumatic Stress Disorder.

⁺p < 0.1.

^*p < 0.05.

**p < 0.01.

***p < 0.001.

In terms of psychological characteristics, participants showed a normal range of personality traits on the NEO-FFI. They were interpersonally submissive, with strong warm/friendly characteristics on the Inventory of Interpersonal problems. They reported low to average social support on the MOS Social Support Scale and moderate treatment motivation on the Treatment Motivation Scale. Self-efficacy was measured in terms of temptation for nonadherence and confidence for adherence, and temptation for cocaine use and confidence to refrain from cocain use. Self-efficacy for ART adherence showed higher confidence to adherence than temptation for nonadherence. In contrast, self-efficacy for cocaine use was lower, with temptation to use cocaine exceeding confidence to refrain from cocaine use. A majority of participants was classified in the preparation stage of change for cocaine use (88.9%) and medication adherence (40.7%). Participants reported using a range of processes of change for cocaine use, but used action-oriented strategies like reinforcement management and stimulus control less than strategies consistent with contemplation like consciousness raising, self-liberation, and self-reevaluation. In general, participants reported higher use of more strategies for ART adherence. On visual analog scales (0-100) for their perceived importance, confidence, and readiness to change cocaine use or ART adherence, participants rated all three higher for ART adherence than for cocaine use. On the Self-Regulation Questionnaire, their scores were consistent with average levels of self-regulation, with highest scores on the Search and Instigation to Change subscales.

Relationships of participant characteristics and outcomes

Table 4 presents the relationships between participant characteristics and each 6M outcome. Characteristics related to ASI drug composite score at 6M included demographic variables, psychological variables, psychiatric disorders and drug use variables. Characteristics related to days using cocaine at 6M included one demographic variable, psychological variables, a psychiatric disorder, drug use variables, and a health indicator, CD4 count. Characteristics related to ART adherence at 6M included demographics, psychological variables, and psychiatric disorders. Characteristics related to Log Viral Load at 6M included health variables, psychological variables, psychiatric variables, and drug variables. Table 5 presents the final multivariate models.

Table 4.

Univariate relationships of selected baseline characteristics to outcome variables at 6 month follow-up.

Baseline characteristics with Univariate Relationships to ASI Drug Composite Score at 6M	n	Mean (SD)	Test
Continuous Variables
Age	36	44.97 (5.91)	F(1,32) = 3.68+
NEO-FFI Openness T-Score	42	49.17 (9.81)	F(1,38) = 5.26*
ASI Cocaine Composite Score	42	0.45 (0.28)	F(1,38) = 5.25*
ASI Family Composite Score	41	0.11 (0.14)	F(1,37) = 4.49*
TMS Desire for Help	42	0.71 (0.14)	F(1,38) = 5.52*
TMS External Pressure	42	0.51 (0.16)	F(1,38) = 4.51*
Self-Efficacy Temptation Medication Adherence	42	24.64 (12.75)	F(1,38) = 4.44*
Inventory of Interpersonal Problems Dominance	41	−2.07 (3.51)	F(1,37) = 4.40*
Categorical Variables
DSM-IV Major Depressive Disorder Current			t37 = 2.67*
Negative	19	0.06 (0.07)
Positive	20	0.13 (0.09)
Homicidality			t37 = 1.69+
Negative	32	0.08 (0.08)
Positive	7	0.14 (0.10)

Baseline Characteristics with Univariate Relationships to Cocaine Days at 6M	n	Mean (SD)	Test
Continuous Variables
Proportion of days in 90 using Cocaine	42	0.36 (0.30)	F(1,40) = 5.17*
NEO-FFI Conscientiousness T-score	42	42.64 (11.66)	F(1,40) = 3.03+
CD4 Count	42	493.52 (327.41)	F(1,40) = 4.60*
Visual Analog Scaling Rulers (0-100)
Readiness to Change Cocaine Use	41	75.51 (28.58)	F(1,39) = 4.14*
Confidence of Changing Nonadherence	42	78.55 (23.37)	F(1,40) = 3.00+
Categorical Variables
Race			t40 = 3.30**
Black	35	0.13 (0.19)
Other	7	0.02 (0.03)
DSM-IV Major Depressive Disorder Current			t27.4 = 2.20*
Negative	20	0.05 (0.09)
Positive	21	0.16 (0.21)
Cocaine Use			t24 =3.71**
Negative Urine Toxicology	17	0.03 (0.04)
Positive Urine Toxicology	23	0.19 (0.21)
Marijuana Use			t28 =2.13*
Negative Urine Toxicology	21	0.06 (0.12)
Positive Urine Toxicology	19	0.18 (0.22)

Baseline Characteristics with Univariate Relationships to Medication Adherence at 6M	n	Mean (SD)	Test
Continuous Variables
TMS Desire for Help	42	0.71 (0.14)	F(1,40) = 8.90**
TMS External Pressure	42	0.51 (0.16)	F(1,40) = 5.61*
Medication Dramatic Relief	12	9.58 (3.33)	F(1,40) = 4.92*
Medication Helping Relationships	12	11.50 (3.48)	F(1,40) = 4.35*
Medication Reinforcement Practices	12	11.58 (2.93)	F(1,40) = 5.10*
SRQ Total Score	12	199.75 (28.07)	F(1,38) = 3.68+
Categorical Variables
Race			t39 = 2.23*
Black	35	0.88 (0.25)
Other	7	0.98 (0.03)
Employment			t40 = 2.10*
Not Employed	36	0.88 (0.25)
Employed	6	0.98 (0.04)
DSM-IV Cocaine Dependence			t36 = 2.85**
Negative	4	1.00 (0)
Positive	37	0.88 (0.25)
DSM-IV Panic Disorder Current			t38 = 2.41*
Negative	38	0.89 (0.25)
Positive	3	0.99 (0.02)
DSM-IV Post-Traumatic Stress Disorder			t39 = 2.33*
Negative	36	0.88 (0.25)
Positive	5	0.99 (0.03)
Homicidality			t36 = 2.52*
Negative	34	0.88 (0.26)
Positive	7	0.99 (0.03)

Baseline Variables with Univariate Relationships to Log Viral Load at 6M	n	Mean (SD)	Test
Continuous Variables
Log Viral Load	42	2.71 (1.03)	F(1,37) = 13.90***
ASI Alcohol Composite Score	42	0.15 (0.19)	F(1,37) = 2.95+
Self-Efficacy Temptation for Cocaine Use	42	62.19 (24.83)	F(1,37) = 4.18*
Visual Analog Scaling Rulers (0-100) Confidence to change Nonadherence	41	72.54 (23.37)	F(1,37) = 9.59**
Self-Efficacy Temptation Medication Adherence	42	24.64 (12.75)	F(1,37) = 4.14*
Discrepancy Scale Medication Total Score	41	38.85 (13.46)	F(1,37) = 14.54***
Inventory of Interpersonal Problems Dominance	41	−2.07 (3.51)	F(1,36) = 4.29*
Categorical Variables
DSM-IV Cocaine Abuse			t35 = 6.31***
Negative	36	2.93 (1.20)
Positive	2	1.67 (0)
Marijuana Use			t24 =2.28*
Negative Urine Toxicology	20	2.35 (0.75)
Positive Urine Toxicology	17	3.18 (1.34)

Note: In some instances degrees of freedom does not match the full completers sample at baseline due to missing responses at the 6-month follow-up (6M).

⁺p < 0.1.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

Table 5.

Final models of outcome.

Summary of GLMSELECT stepwise regression analysis predicting ASI Drug Composite Score at 6M
Analysis of Variance	DF	Sum of Squares	Mean Square	F Value	R2
Model	5	0.12	0.02	5.38**	0.51
Error	26	0.11	0.004
Corrected Total	31	0.23

Parameter Values	DF	Estimate	Standard Error	t Value
DSM-IV MDD Current
Negative	1	−0.06	0.03	2.40*
Positive	0	0	.	.
Self-Efficacy: Temptation Medication Adherence	1	0.002	0.001	2.40*
Homicidality
Negative	1	−0.09	0.04	2.36*
Positive	0	0	.	.
Age	1	0.004	0.002	1.84+
NEO-FFI Openness T-Score	1	0.002	0.001	1.57

Summary of GLMSELECT stepwise regression analysis predicting Cocaine Days at 6M
Analysis of Variance	DF	Sum of Squares	Mean Square	F Value	R2
Model	4	0.41	0.1	8.01***	0.49
Error	33	0.43	0.01
Corrected Total	37	0.84

Parameter Values	DF	Estimate	Standard Error	t Value
NEO-FFI Conscientiousness T-Score	1	−0.01	0.002	3.31**
Visual Analog Scaling Rulers (0-100)
Readiness to Change Cocaine Use	1	−0.002	0.001	2.39*
Confidence to change Medication Adherence	1	0.002	0.001	2.13*
Cocaine Use
Negative Urine Toxicology	1	−0.13	0.04	3.39**
Positive Urine Toxicology	0	0	.	.

Summary of GLMSELECT stepwise regression analysis predicting Medication Adherence at 6M
Analysis of Variance	DF	Sum of Squares	Mean Square	F Value	R2
Model	2	0.76	0.38	9.29***	0.34
Error	36	1.48	0.04
Corrected Total	38	2.24

Parameter Values	DF	Estimate	Standard Error	t Value
TMS Desire for Help	1	0.84	0.23	3.68***
SRQ Total Score	1	0.004	0.002	2.56*

Summary of GLMSELECT stepwise regression analysis predicting Log Viral Load at 6M
Analysis of Variance	DF	Sum of Squares	Mean Square	F Value	R2
Model	2	9.43 × 109	4.72 × 109	3.51*	0.24
Error	22	2.96 × 1010	1.34 × 109
Corrected Total	24	3.90 × 1010

Parameter Values	DF	Estimate	Standard Error	t Value
ASI Alcohol Composite Score	1	−7.23 × 104	4.18 × 104	1.73*
Marijuana Use
Negative Urine Toxicology	1	−3.19 × 104	1.47 × 104	2.16*
Positive Urine Toxicology	0	0	.	.

Note: In some instances degrees of freedom does not match the full completers sample at baseline due to missing responses at the 6-month follow-up (6M).

⁺p < 0.1.

^*p < 0.05.

^**p < 0.01.

^***p < 0.001.

Characteristics related to cocaine use outcomes

The model of ASI Cocaine Composite score was significant (F_{(5, 26 df)} = 5.38; p,.01) and explained 50.87% of the variance. Significant characteristics and their correlations with the outcome (shown in Supplementary Table 6) were Major Depressive Disorder current (r=.40, p<.05), temptation for ART nonadherence (r=.32, p<.05), Homicidality (no correlation calculated), and age (r=.32, p < 10). NEO Openness was retained in the model but was not independently related to outcome.

The model of days using cocaine was significant (F_{(4, 33 df)} = 8.01; p<.001) and explained 49.25% of the variance. Participant characteristics that related to this outcome included NEOFFI Conscientiousness (r=−.27, p<.10), readiness to change cocaine use (r=−.31, p<.05), confidence to change ART adherence (r=.26, p<.10), and negative urine toxicology for cocaine at study entry (correlation not calculated).

Participant characteristics related to ART adherence outcomes

The model of ART adherence was significant (F_{(2, 36 df)} = 9.29; p<.001) and explained 34.04% of the variance. Participant characteristics that related to this outcome included desire for help on the Treatment Motivation Scale (r=.43, p<.01), and self-regulation total score (r=.30, p<.10).

The model of logVL was significant (F_{(2, 22 df)} = 3.51*) and explained 24.19% of the variance. Participant characteristics related to this outcome included ASI alcohol composite score (r=−.10, ns) and a negative toxicology screen for marijuana (correlation not calculated).

Discussion

This secondary analysis of data from a trial that compared two behavioral interventions targeting both ART adherence and cocaine use found that different sets of participant characteristics were related to post-study ART adherence and cocaine use outcomes. Behavioral, psychological, and psychiatric diagnosis variables were independently related to outcomes, and explained a large proportion of the variance. In general, the models of cocaine use outcomes accounted for more variance than the models of ART adherence outcomes.

Understanding drug use outcomes

Age, current MDD, and Temptation not to adhere to ART regimens were significantly related to ASI Cocaine Composite scores at the six month follow-up. These relationships may occur for several reasons. Age could be related to both treatment-seeking for drug use problems, and recognition of the problems the person has related to cocaine use, both potentially increasing response to the intervention. Struggling with depression may have been related to cocaine “crashes” (experiencing a range of negative mood and psychological symptoms related to cocaine withdrawal) and could have also served as a motivator for treatment. It is less clear why being tempted NOT to adhere to HIV medications predicts cocaine use outcomes, but it is common for HIV patients who use street drugs to view them as dangerous to mix with medications. Thus, the temptation not to adhere to ART may represent their tendency to skip medications when using cocaine and other street drugs due to this belief. Alternatively, ART medications may have been viewed as lower in importance than obtaining substance abuse treatment.

Different characteristics related to days using cocaine at the 6 month follow-up. Both readiness to change cocaine use and conscientiousness were significantly inversely related to days using cocaine, meaning that more conscientious people and those with higher levels of readiness to change their cocaine use had fewer days using cocaine at the six month follow-up. Higher conscientiousness and readiness may have meant that people were willing to work harder to eliminate cocaine use habits. Participants with negative urine toxicology screens for cocaine at baseline also had fewer days using cocaine at follow-up. These people may have already begun changing their cocaine use behavior, compared to those who provided cocaine-positive urine samples at baseline. In contrast, higher confidence about ART adherence related to more days using cocaine at follow-up. It is possible that this subset of the sample was more focused on HIV treatment than substance abuse treatment.

Understanding ART Adherence Outcomes

Two psychological characteristics composed the multivariate model of ART adherence. Patients’ baseline desire for help with cocaine-related problems was significantly directly related to ART adherence at follow-up, while self-regulation total score was correlated to ART adherence at a trend level. These variables could have increased the person's engagement in the intervention, and their efficacy to enact recommended changes.

The model of logVL at 6 month follow-up included only two patient characteristics, both related to cocaine use behaviors. Alcohol composite scores and Marijuana-negative urine screens were both inversely related to logVL at 6M, meaning that higher baseline problems from drinking and the absence of marijuana in urine samples were related to better logVL at follow-up. These results are difficult to explain because they imply that drinking problems but the absence of current marijuana use are both related to improved (lower) levels of HIV in the blood.

The relationship between ART adherence and viral load is not necessarily direct or linear. Perhaps it is not surprising that psychological factors influence taking ART medication, while cocaine use behaviors are related to a direct clinical marker of immune health.

Limitations and Strengths

This study is exploratory in nature, and intended to identify promising targets for development of future interventions focused on cocaine use and ART adherence. This study used data from a small sample from a pilot dual-focused study, and the results from a stepwise model building method could over-estimate the relationships between explanatory variables and outcomes. The findings are specific to crack cocaine users, and may not apply to HIV patients who use other types of street drugs. Further investigations of the importance and patterns of the relationships between the participant characteristics that related to cocaine use and ART adherence outcomes in this study are needed. We modelled characteristics that related to 4 different markers of outcome, including self-reports of Art adherence and cocaine use on the TLFB. It could be argued that only biological or objective markers of outcomes or benchmarks on the HIV care cascade are important. However, to move the ART adherence intervention field forward, a better understanding of the characteristics that relate to the most common types of outcome measures is needed.

This analysis also has strengths. It identified an intriguing set of participant characteristics that influenced ART adherence and cocaine use outcomes for patients with HIV who use cocaine. While the sample size was small in the parent study, the follow-up rate was excellent, yielding a representative sample at the six month follow-up point.

Clinical Implications

How can these findings help to develop tailored, more effective ART adherence interventions for people who use cocaine and are living with HIV? We were unable to find a previous study that tailored interventions to specific patient characteristics to improve their potency or lengthen their duration of impact. Therefore, we stick closely to these exploratory findings and their potential implications for intervention development.

The tested interventions targeting both cocaine use and ART adherence reduced drug-related problems among older, depressed patients with beliefs that they should skip HIV medications when using drugs. The idea that ART should be skipped during periods of drug use is common, and more education about this within interventions is indicated. When dealing with younger patients, with less awareness of cocaine-related problems, or those who are not depressed, interventions that raise consciousness or even increase their distress about cocaine use may be needed. Additionally, the interventions worked well to decrease cocaine-using days for those ready to change, and those who have already started to change. More motivational components in interventions may be required for cocaine-using patients with lower readiness, lower levels of conscientiousness, and who present for treatment with cocaine-positive urines. Additionally, these findings imply that interventions that build skills in self-regulation or increase the desire for help as a motivation for treatment could enhance ART adherence outcomes.

Conclusions

In this exploratory study of a small sample of cocaine users who were nonadherent to ART treatment, we identified characteristics that related to drug use and ART adherence outcomes at the 6 month follow-up. The relationship found among psychiatric, psychological, and behavioral variables and cocaine-related problems (and days using cocaine) at 6M fits with the literature. In general, those who were older, depressed, and tempted to skip HIV medications had fewer problems related to cocaine use at 6M. There was also a logical relationship between psychological traits like conscientious, constructs like readiness to change cocaine use, behaviors like beginning to reduce cocaine use (as evidenced by a cocaine-negative screen), and days using cocaine at 6M.

Similarly, there were logical relationships between psychological variables such as being motivated for treatment by a desire for help and higher use of self-regulation skills and the behavior of taking ART medications. In contrast, the reason for the relationships between drug use behaviors and logVL is less clear and should be studied further.

We recommend that future dual-focused interventions contain content and intervention processes that are tailored to the behavioral, psychological, and psychiatric characteristics that related to cocaine use and ART adherence outcomes. Overall, these results suggest a potential benefit of adding information on myths such as avoiding ART during cocaine use, along with self-regulation and motivational strategies, to future interventions targeting both ART adherence and cocaine use. These findings may help to extend and further specify the Information-Motivation-Behavioral Skills model of ART adherence.

Table 2.

HIV adherence and drug use characteristics of the sample.

Characteristic	Full sample at baseline			Completers sample at baseline			Test	Completers sample at 6M
	n	Mean (SD)	Min-Max	n	Mean (SD)	Min-Max		n	Mean (SD)	Min-Max
Continuous Variables
Log Viral Load	53	2.99 (1.20)	1.67-5.79	42	2.71 (1.03)	1.67-5.28	t51=3.74***	39	2.83 (1.20)	1.67-5.72
CD4 Count	54	441.44 (319.87)	46-1516	42	493.50 (327.40)	46-1516	t52=2.33*	40	467.23 (258.66)	16-1126
Proportion of ART Medications Taken/ Prescribed across 14 days	54	0.58 (0.27)	0-0.89	42	0.63 (0.24)	0-0.89	t52=2.25*	42	0.90 (0.24)	0-1
Proportion of Days in 90 Using Cocaine	54	0.33 (0.30)	0-1	42	0.36 (0.30)	0-1	t52=1.59	42	0.12 (0.18)	0-0.64
ASI Composite Score
Alcohol	54	0.16 (0.20)	0-0.77	42	0.15 (0.19)	0-0.77	t52=0.74	40	0.11 (0.14)	0-0.45
Drug	54	0.17 (0.10)	0-0.46	42	0.17 (0.10)	0-0.46	t52=0.25	40	0.09 (0.09)	0-0.27
Family	53	0.12 (0.16)	0-0.59	41	0.11 (0.14)	0-0.52	t14=0.49	40	0.06 (0.13)	0-0.51
Legal	54	0.08 (0.15)	0-0.60	42	0.06 (0.13)	0-0.60	t14=1.27	40	0.05 (0.13)	0-0.60
Education	54	0.77 (0.26)	0.04-1	42	0.80 (0.25)	0.09-1	t52=1.55	40	0.70 (0.29)	0-1
Medication	54	0.38 (0.37)	0-1	42	0.36 (0.36)	0-1	t52=0.96	40	2.39 (12.92)	0-82
Psychiatric	54	0.37 (0.27)	0-0.91	42	0.40 (0.27)	0-0.91	t52=1.19	40	0.31 (0.26)	0-0.80
Cocaine	54	0.44 (0.27)	0-1	42	0.45 (0.28)	0-1	t52=0.66	40	0.29 (0.26)	0-0.83

Characteristic	Full sample at baseline n (%)	Completers sample at baseline n (%)	Test
Categorical Variables
Health Literacy Score			χ2(1) = 0.12
Inadequate	6 (11.1%)	5 (11.9%)
≥ Marginal	48 (88.9%)	37 (88.1%)
Primary Drugs			χ2(1) = 0.55
Cocaine with other drug(s)	22 (40.7%)	16 (38.1%)
Cocaine only	32 (59.3%)	26 (61.9%)
Positive Urine Toxicologya
Cocaine	26 (50.0%)	23 (57.5%)	χ2(1) = 3.90*
Methamphetamines	1 (1.9%)	1 (2.5%)	χ2(1) = 0.31
Marijuana	20 (38.5%)	19 (47.5%)	χ2(1) = 5.98*
Opioids	4 (7.69%)	3 (7.5%)	χ2(1) = 0.01
Benzodiazepines	3 (5.9%)	3 (7.7%)	χ2(1) = 0.98
Barbiturates	2 (3.9%)	1 (2.5%)	χ2(1) = 0.85

Note: Full sample is the sample at baseline when 2 mistaken enrollments were removed.

Completers sample at baseline is the sample completing the 6 month follow-up (6M).

Completers sample at 6M is the values re-measured at 6 months after baseline. Some n's are lower than the full sample due to missing data for that variable. Some percentages do not total to 100% due to rounding, or due to responses such as “refused” or “don't know,” (data not shown).

^aWhile a full drug panel was collected, no participants were positive for drugs other than those shown.

⁺p < 0.1.

^*p < 0.05.

**p < 0.01.

^***p < 0.001.