Abstract
This was a three group randomized clinical trial of interim methadone and patient navigation involving 225 pre-trial detainees with opioid use disorder in Baltimore. The HIV Risk Assessment Battery (RAB) was administered at baseline (in jail), and at 6 and 12 months post-release. Generalized linear mixed model analyses indicated the condition X time interaction effect failed to reach significance (ps>.05) for both the drug risk and sex risk subscale scores. Therefore, findings suggest that there were no intervention effects on drug or sex risk behaviors. However, increased use of cocaine at baseline was associated with increases in drug- (b = .04, SE = .02) and sex-risk (b = .01, SE = .003) behaviors. These results suggest that interventions targeting cocaine use among pre-trial detainees may serve as a means of reducing HIV risk associated with drug- and sex-risk behaviors.
1. Introduction
The US is confronting intertwined epidemics of HIV [1, 2] and illicit opioid use [3]. Community-based opioid agonist treatment with methadone or buprenorphine plays an important role in addressing both epidemics. Such treatment has been found to reduce illicit opioid use [4–6]. Methadone maintenance treatment has been associated with reduced HIV-risk behaviors, particularly those behaviors associated with increased risk of drug use [7–9]. Therefore, methadone treatment can be effective for primary and secondary HIV prevention by reducing injection drug use behavior [10]. Besides reducing the frequency of injection and the sharing of syringes and works, there is evidence that, with methadone maintenance treatment, the incidence of multiple partners and unprotected sex is also diminished [11].
A study early in the AIDS epidemic found that long-term methadone treatment was associated with HIV negative sero-status [12]. In response to the AIDS epidemic and shortages of methadone treatment in New York City, Yancovitz and colleagues [13] demonstrated that methadone maintenance without counseling (termed “interim methadone”) was effective in reducing heroin injection compared to a waiting list control. Subsequently, in a randomized clinical trial, Schwartz and colleagues [14] found that interim methadone (IM) treatment was associated with reduced heroin use and HIV-risk behavior compared to a waiting list control.
Engaging incarcerated populations in treatment for opioid use disorder (OUD) is a promising approach to address both OUD and HIV because rates of both disorders are considerably higher in incarcerated populations than in community populations [15]. Often, criminal-justice-involved individuals engage in a cluster of risky drug use and sexual behaviors that could lead to acquiring HIV, including injecting, sharing needles, and engaging in unprotected sex [16, 17]. These risk behaviors are elevated during the period following release from incarceration [18–20].
Incarcerated individuals may be detained in jails while awaiting trial or after receiving a short sentence of typically less than one year. Individuals sentenced to greater than one year typically are incarcerated in state or federal prisons. While the daily census of prisons is greater than that of jails, because of the latter’s much shorter sojourns, there were over 10 million admissions to US jails in 2016 [21] and only 626,000 to prisons [22]. For this reason, the jail population is of considerable importance in developing HIV prevention and treatment programs [18, 23]. However, opioid agonist treatment is much less available in jails and prisons in the US than in the community and there has been only limited study of its impact on HIV-risk behaviors following release from incarceration [24].
The present study is a pre-planned secondary analysis of self-reported HIV-risk outcomes over the 12-month post-release period from the Baltimore City Detention Center among 225 newly-arrested adults with OUD. Pre-trial detainees, who were not in treatment for opioid withdrawal at the time of arrest, were enrolled in a randomized clinical trial and assigned to one of three groups: (1) interim methadone with patient navigation (IM+PN); (2) IM only (IM); or (3) enhanced treatment-as-usual (ETAU). The parent study found that both IM groups were more likely than ETAU to be enrolled in OUD treatment at 30 days post-release but there were no significant differences in treatment enrollment by 12-month follow-up and no significant differences in the rates of opioid-positive and cocaine-positive urine screening tests [14]. In this report, we examine the relative effectiveness of the three interventions in reducing drug use and sex-related HIV-risk behaviors.
2. Methods
The study was conducted from 12/16/14 to 10/12/17. It was approved by the Friends Research Institute’s IRB, Western IRB, and the US Office of Human Research Protection.
2.1. Participants
Eligibility criteria included being at least 18 years of age, meeting DSM-5 criteria for OUD, being in detention for at least 48 hours, having a charge that could, if found guilty, result in less than a 1 year sentence, being treated for opioid withdrawal in detention, and having plans to live in the Baltimore area after release. Exclusion criteria included being enrolled in methadone or buprenorphine treatment in the community at the time of arrest, being medically or psychiatrically unstable, being pregnant, and needing treatment for moderate to severe alcohol or sedative hypnotic withdrawal.
2.2. Study Conditions
Study participants were randomized within gender in blocks of 3, 6, or 9 to one of three conditions: (1) IM + PN; (2) IM only; or (3) ETAU. IM consisted of initiating methadone maintenance in detention without counseling. IM+PN included receiving IM along with a patient navigator who helped participants create a post-release plan while in jail and provided case management for up to three months after release while in the community. Both IM groups were able to enter standard methadone treatment with counseling at the time of their release. ETAU participants received a brief medically-supervised withdrawal using methadone, information on drug addiction and overdose prevention, and Baltimore’s assessment/treatment referral helpline number to any type of OUD treatment available in the community.
2.3. Outcomes: Drug and Sex Risk
The outcomes reported in the present paper were drawn from the Risk Assessment Battery (RAB) [25] administered at baseline and 6- and 12-month post-release follow-up. The RAB drug-risk scale score is comprised of 8 items that enquire about activities in the previous 6 months. The first item asks whether the participant injected substances with responses 0 = “No” and 1 = “Yes.” The second item asks whether the participant injected substances or shared needles or works with 0 = “No” and 3 = “Yes.” The number of people with whom the participant shared needles was coded as 0 = “0,” 1 = “1 other person,” 2 = “2 or 3 different people,” and 3 = “4 or more different people.” Frequency of going to a shooting gallery, and having shared rinse water, cookers, cotton and drugs (i.e., backloading) were coded as 0 = “Never,” 1 = “A few times,” 2 = “A few times each month,” and 3 = “Once or more each week”. The RAB scoring system [26] was used to score the drug risk scales by summing the scores on the individual items that comprise the scale, with a range from 0–22.
The RAB sex risk scale score summed the responses to questions enquiring about: frequency of crack house visits with responses 0 = “Never,” 1 = “A few times,” 2 = “A few times each month,” and 3 = “Once or more each week;” sexual identity with answer choices 0 = “Heterosexual,” and 3 = “Gay or Bisexual;” numbers of male and female partners were assessed separately with 0 = “0,” 1 = “1”, 2 = “2 or 3” and 3 = “4 or more;” and exchange of sex for drugs or money was assessed with four items with response options of 0 = “Never,” 1 = “A few times,” 2 = “A few times each month,” and 3 = “Once or more each week.” Finally, condom use was measured on a 4-point scale indicating the frequency of condom use with any sexual activity with responses options, 0 = “All of the time or no sexual activity in the last 6 months,” 1 = “Most of the time,” 2 = “Some of the time” and 3 = “None of the time.” Therefore, higher numbers indicated greater risk. As with the drug risk scale, scores from individual items comprising the sex risk scale were summed with a potential range of 0–18.
Additional supplementary analyses were conducted with individual items measuring drug risk items sharing needles, rinse water, cookers, cotton or drugs (i.e., backloading), going to a shooting gallery, and injecting drugs. Because the items had four response options, the analyses that were conducted with the individual items were done with the dichotomized versions of these variables, where 1 = “Yes” versus 0 = “No.” As with drug use items, supplementary analyses were conducted using individual sex risk items that were dichotomized into 1 = “Yes” vs. 0 = “No” with condom use dichotomized as 1 = “Used condoms all of the time or did not engage in sexual activity in the last 6 months” vs. 0 = “Used condoms less than all of the time.”
2.4. Statistical Analyses
Means and frequencies were generated for participant background variables for the total sample. Data from the total sample were examined for the drug and sex risk scale scores. Generalized linear mixed models (GLiMM) were conducted to analyze drug and sex risk outcomes longitudinally at baseline and at 6 and 12 months after release. Of interest were the interaction effects of each of two planned contrasts × time (i.e., combined IM + PN and IM only vs. ETAU × time, and IM + PN vs. IM only × time). Four covariates, which included gender, age, previous methadone maintenance admissions, and baseline number of days of cocaine use in the 30 days before index incarceration were included in the model. Because gender was included as a moderator variable, three additional interaction terms were included in the model (i.e., condition × gender, time × gender and condition × time × gender) along with the main effects for time and condition. The drug and sex risk scale scores were analyzed as count outcomes with a negative binomial distribution and a log link function.
The individual drug and sex risk items were dichotomized and analyzed with GLiMM under the assumption of a binomial distribution and a logit link function. Of primary interest in these analyses were the two planned contrast × time terms (i.e., combined IM + PN and IM only vs. ETAU × time, and IM + PN vs. IM only × time) and, of secondary interest, the four covariates and the three interaction terms discussed above. However, two of the sex risk scale items (paying for sex with drugs or money) were excluded from the analyses due to low frequencies, which made the estimates unstable. All analyses were conducted in SPSS v25.0.
3. Results
3.1. Sample Description
The majority of the sample was male (80.4%) and African American (61.8%) with a mean age of 38.3 years (standard deviation [SD] = 10.4) (Table 1). Approximately half (50.7%) of the sample had injected drugs and the majority (88.0%) had engaged in sex in the previous six months at baseline. In the total sample of 225, one-quarter (25.8%) reported sharing needles or works and 45.1% reported using condoms all of the time. Among the subsample of participants who reported injecting drugs, fully 50.9% reported sharing needles or works. Participants averaged 71.2 months (SD = 75.1) of incarceration during their lifetime and nearly half (48.4%) were on parole or probation at the time of the baseline interview. Baseline background characteristics did not differ significantly among the three conditions in the total sample.
Table 1.
Participant characteristics at baseline (N = 225).
| Total sample | IM+PN | IM | ETAU | Test Statistic | ||
|---|---|---|---|---|---|---|
| (N = 225) | (n = 75) | (n = 74) | (n = 76) | (df) | p | |
| Male, n (%) | 181 (80.4) | 60 (80.0) | 60 (81.1) | 61 (80.3) | χ2(2) = 0.03 | 0.99 |
| Age, mean (SD) | 38.3 (10.4) | 39.6 (10.0) | 37.3 (10.0) | 37.9 (11.1) | F(2, 222) = 0.99 | 0.37 |
| Hispanic, n (%) | 6 (2.7) | 3 (4.0) | 0 (0.0) | 3 (3.9) | χ2(2) = 3.0 | 0.22 |
| Race, n (%) | χ2(2) = 1.0 | 0.60 | ||||
| Black/African American | 139 (61.8) | 51 (68.0) | 45 (60.8) | 43 (56.6) | ||
| White | 56 (24.9) | 17 (22.7) | 17 (23.0) | 22 (28.9) | ||
| Other race/>1 race | 30 (13.3) | 7 (9.3) | 12 (16.2) | 11 (14.5) | ||
| Married, n (%) | 43 (19.1) | 13 (17.3) | 14 (18.9) | 16 (21.1) | χ2(2) = 0.34 | 0.84 |
| Number of days used cocaine 30 days prior to incarceration (baseline), mean (SD) | 15.4 (13.6) | 13.9 (13.4) | 15.5 (13.8) | 16.7 (13.7) | F(2, 222) = 0.83 | 0.44 |
| Injected drugs past 6 months, n (%) | 114 (50.7) | 35 (46.7) | 36 (48.6) | 43 (56.6) | χ2(2) = 1.7 | 0.44 |
| Shared needles in past 6 months, n (%) | 58 (25.8) | 17 (22.7) | 18 (24.3) | 23 (30.3) | χ2(2) = 1.3 | 0.53 |
| Had sex in past 6 months, n (%) | 198 (88.0) | 63 (84.0) | 67 (90.5) | 68 (89.5) | χ2(2) = 1.7 | 0.42 |
| Used a condom all of the time in past 6 months, n (%) | 101 (45.1) | 36 (48.0) | 34 (45.9) | 31 (41.3) | χ2(2) = 0.71 | 0.70 |
| With prior methadone treatment admissions, n (%) | 83 (36.9) | 26 (34.7) | 27 (36.5) | 30 (39.5) | χ2(2) = 0.38 | 0.83 |
| On parole/probation, n (%) | 109 (48.4) | 31 (41.3) | 36 (48.6) | 42 (55.3) | χ2(2) = 2.9 | 0.23 |
| Lifetime months of incarceration, mean (SD) | 71.2 (75.1) | 82.6 (85.3) | 72.3 (72.0) | 58.8 (65.6) | F(2, 222) = 1.9 | 0.15 |
| Arrested during year prior to enrollment, n (%) | 142 (63.1) | 50 (66.7) | 47 (63.5) | 45 (59.2) | χ2(2) = 0.91 | 0.64 |
Notes: IM+PN=Interim Methadone + Patient Navigation; IM=Interim Methadone; ETAU=Enhanced Treatment as Usual. Test statistics are the results of the respective comparison of the three treatment conditions. Test statistic for “Race” was obtained by collapsing data into two categories: White (n = 17, n = 17, and n = 22 for IM+PN, IM, and ETAU, respectively) and African American/other (n = 58, n = 57, and n = 54 for IM+PN, IM, and ETAU, respectively). “Arrested during year prior to enrollment” does not include the index arrest. All variables were self-reported with the exception of “Arrested during year prior to enrollment” which was drawn from official arrest records.
Among the sub-sample of participants who reported injecting drugs (n= 114; 50.7%), 50.9% reported having shared needles in the past six months.
3.2. Outcomes
3.2.1. Drug Risk Scale Score
As shown in Table 2, both the first (combined IM+PN and IM vs. ETAU × time interaction subeffect) and second (IM+PN vs. IM only × time interaction subeffect) planned contrast failed to reach significance (ps = 1.00 and .26, respectively).
Table 2.
Test statistic, degrees of freedom, and p values for Risk Assessment Battery (RAB) drug risk scale score and individual items in the total sample (N=225).
| Drug Risk Scale
Score |
Shared Needles |
Shared Rinse Water |
Shared Cookers |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Source | F | df1, df2 | p | F | df1, df2 | p | F | df1, df2 | p | F | df1, df2 | p |
| Condition | 0.72 | 2, 565 | 0.49 | 1.20 | 2, 569 | 0.30 | 0.44 | 2, 569 | 0.64 | 0.76 | 2, 569 | 0.47 |
| Time | 10.53 | 2, 565 | <0.001 | 8.26 | 2, 569 | <0.001 | 2.39 | 2, 569 | 0.09 | 4.34 | 2, 569 | 0.014 |
| Gender | 0.74 | 1, 565 | 0.39 | 0.002 | 1, 569 | 0.96 | 1.40 | 1, 569 | 0.24 | 0.12 | 1, 569 | 0.73 |
| Prior MMT admissions | 4.38 | 1, 565 | 0.037 | 1.28 | 1, 569 | 0.26 | 5.27 | 1, 569 | 0.022 | 0.88 | 1, 569 | 0.35 |
| Age | 6.49 | 1, 565 | 0.011 | 2.72 | 1, 569 | 0.10 | 6.63 | 1, 569 | 0.010 | 2.81 | 1, 569 | 0.09 |
| Number of days used cocaine past 30 days (baseline) | 14.22 | 1, 565 | <0.001 | 6.35 | 1, 569 | 0.012 | 6.23 | 1, 569 | 0.013 | 7.12 | 1, 569 | 0.008 |
| Condition X Gender | 1.06 | 2, 565 | 0.35 | 1.09 | 2, 569 | 0.34 | 0.02 | 2, 569 | 0.99 | 0.49 | 2, 569 | 0.62 |
| Condition X Time | 0.70 | 4, 565 | 0.60 | 0.69 | 4, 569 | 0.60 | 1.54 | 4, 569 | 0.19 | 0.67 | 4, 569 | 0.61 |
| Planned Contrast: (Combined IM+PN and IM v. ETAU) X Time | 0.003 | 2, 565 | 1.00 | 0.03 | 2, 569 | 0.98 | 0.25 | 2, 569 | 0.78 | 0.14 | 2, 569 | 0.87 |
| Planned Contrast: (IM+PN v. IM) X Time | 1.36 | 2, 565 | 0.26 | 1.26 | 2, 569 | 0.29 | 2.12 | 2, 569 | 0.12 | 1.10 | 2, 569 | 0.34 |
| Time X Gender | 0.38 | 2, 565 | 0.68 | 0.31 | 2, 569 | 0.74 | 0.06 | 2, 569 | 0.94 | 0.51 | 2, 569 | 0.60 |
| Condition X Time X Gender | 1.14 | 4, 565 | 0.34 | -- | -- | -- | -- | -- | -- | -- | -- | -- |
| Shared Cotton |
Shared Drugs
(backloading) |
Shooting Gallery |
Inject Drugs |
|||||||||
| Source | F | df1, df2 | p | F | df1, df2 | p | F | df1, df2 | p | F | df1, df2 | p |
| Condition | 0.18 | 2, 569 | 0.84 | 0.83 | 2, 569 | 0.44 | 0.16 | 2, 565 | 0.85 | 0.34 | 2, 565 | 0.71 |
| Time | 2.22 | 2, 569 | 0.11 | 3.54 | 2, 569 | 0.030 | 2.82 | 2, 565 | 0.06 | 10.68 | 2, 565 | <0.001 |
| Gender | 0.02 | 1, 569 | 0.89 | 0.96 | 1, 569 | 0.33 | 0.40 | 1, 565 | 0.53 | 0.51 | 1, 565 | 0.47 |
| Prior MMT admissions | 3.14 | 1, 569 | 0.08 | 0.07 | 1, 569 | 0.79 | 2.69 | 1, 565 | 0.10 | 6.44 | 1, 565 | 0.011 |
| Age | 2.07 | 1, 569 | 0.15 | 4.35 | 1, 569 | 0.038 | 6.51 | 1, 565 | 0.011 | 7.87 | 1, 565 | 0.005 |
| Days used cocaine past 30 days (baseline) | 6.08 | 1, 569 | 0.014 | 8.15 | 1, 569 | 0.004 | 12.68 | 1, 565 | <0.001 | 11.71 | 1, 565 | 0.001 |
| Condition X Gender | 0.17 | 2, 569 | 0.84 | 0.88 | 2, 569 | 0.41 | 0.91 | 2, 565 | 0.40 | 0.11 | 2, 565 | 0.90 |
| Condition X Time | 0.67 | 4, 569 | 0.61 | 1.87 | 4, 569 | 0.12 | 1.08 | 4, 565 | 0.37 | 0.26 | 4, 565 | 0.90 |
| Planned Contrast: (Combined IM+PN and IM v. ETAU) X Time | 0.18 | 2, 569 | 0.84 | 2.35 | 2, 569 | 0.10 | 1.53 | 2, 565 | 0.22 | 0.03 | 2, 565 | 0.97 |
| Planned Contrast: (IM+PN v. IM) X Time | 1.02 | 2, 569 | 0.36 | 1.73 | 2, 569 | 0.18 | 0.24 | 2, 565 | 0.79 | 0.49 | 2, 565 | 0.61 |
| Time X Gender | 1.17 | 2, 569 | 0.31 | 0.39 | 2, 569 | 0.68 | 0.57 | 2, 565 | 0.57 | 0.88 | 2, 565 | 0.42 |
| Condition X Time X Gender | -- | -- | -- | -- | -- | -- | 0.70 | 4, 565 | 0.59 | 1.70 | 4, 565 | 0.15 |
Notes: IM+PN=Interim Methadone + Patient Navigation; IM=Interim Methadone; ETAU=Enhanced Treatment as Usual. df1 and df2 are the degrees of freedom for the respective F test statistic. df1 and df2 are the degrees of freedom for the respective F test statistic.
GLiMM analyses (Table 2) revealed a significant main effect of time (p <.001), such that the drug risk scale scores decreased significantly over time (Estimated Marginal Mean [EMM] (Standard Error [SE]) = 2.59 [.38], 1.81 [.36], and .88 [.24] for baseline, 6 months, and 12 months, respectively). Age (p=.011), prior history of having received methadone maintenance treatment (MMT) (p = .037), and the number of days of cocaine use at baseline (p <.001) were significant covariates. Older participants had, on average, lower scores than did younger participants (b = −.03, SE = .01), participants who had previous MMT admission had higher scores, on average, than participants who had no previous MMT admission (EMM[SE] = 2.07 [.44] vs. EMM[SE] = 1.25 [.26]), respectively, and participants who used cocaine more days at baseline had, on average, higher scores than participants who did not use cocaine at baseline (b = .03, SE = .01). All other effects in the model failed to reach significance.
3.2.2. Sex Risk Scale Score
The first planned contrast (combined IM+PN and IM vs. ETAU × time interaction subeffect) and the second planned contrast (IM+PN vs. IM only × time interaction subeffect) failed to reach significance (p = .66 and p = .33, respectively; see Table 3).
Table 3.
Test statistic, degrees of freedom, and p values for the Risk Assessment Battery (RAB) sex risk scale score and individual sex risk items in the total sample (N = 225).
| Sex Risk Scale
Score |
Condom Use |
Having Sex for
Money |
Having Sex for
Drugs |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Source | F | df1, df2 | p | F | df1, df2 | p | F | df1, df2 | p | F | df1, df2 | p |
| Condition | 0.44 | 2, 563 | 0.65 | 1.33 | 2, 564 | 0.27 | 0.75 | 2, 568 | 0.47 | 0.51 | 2, 567 | 0.60 |
| Time | 6.51 | 2, 563 | 0.002 | 2.46 | 2, 564 | 0.09 | 1.38 | 2, 568 | 0.25 | 0.11 | 2, 567 | 0.89 |
| Gender | 12.95 | 1, 563 | <0.001 | 6.54 | 1, 564 | 0.011 | 33.26 | 1, 568 | <0.001 | 18.37 | 1, 567 | <0.001 |
| Prior MMT admissions | 0.62 | 1, 563 | 0.43 | 0.03 | 1, 564 | 0.86 | 2.33 | 1, 568 | 0.13 | 0.16 | 1, 567 | 0.69 |
| Age | 27.41 | 1, 563 | <0.001 | 1.98 | 1, 564 | 0.16 | 34.76 | 1, 568 | <0.001 | 10.98 | 1, 567 | 0.001 |
| Number of days used cocaine past 30 days (baseline) | 9.06 | 1, 563 | 0.003 | 1.30 | 1, 564 | 0.26 | 3.32 | 1, 568 | 0.07 | 0.09 | 1, 567 | 0.76 |
| Condition X Gender | 1.05 | 2, 563 | 0.35 | 1.09 | 2, 564 | 0.34 | 1.01 | 2, 568 | 0.37 | 0.24 | 2, 567 | 0.79 |
| Condition X Time | 0.77 | 4, 563 | 0.55 | 1.28 | 4, 564 | 0.28 | 1.20 | 4, 568 | 0.31 | 0.47 | 4, 567 | 0.76 |
| Planned Contrast: (Combined IM+PN and IM v. ETAU) X Time | 0.41 | 2, 563 | 0.66 | 1.10 | 2, 564 | 0.33 | 1.51 | 2, 568 | 0.22 | 0.43 | 2, 567 | 0.65 |
| Planned Contrast: (IM+PN v. IM) X Time | 1.12 | 2, 563 | 0.33 | 1.50 | 2, 564 | 0.23 | 1.38 | 2, 568 | 0.25 | 0.61 | 2, 567 | 0.54 |
| Time X Gender | 1.60 | 2, 563 | 0.20 | 1.69 | 2, 564 | 0.19 | 1.13 | 2, 568 | 0.32 | 1.75 | 2, 567 | 0.17 |
| Condition X Time X Gender | 0.21 | 4, 563 | 0.93 | 1.55 | 4, 564 | 0.19 | -- | -- | -- | -- | -- | -- |
Notes: IM+PN=Interim Methadone + Patient Navigation; IM=Interim Methadone; ETAU=Enhanced Treatment as Usual. df1 and df2 are the degrees of freedom for the respective F test statistic.
The time main effect was significant (EMM [SE] = 4.66 [.21], 4.04 [.21], and 4.17 [.23] for baseline, 6 months, and 12 months, respectively; p = .002). Gender was a significant covariate (p < .001) such that women had higher sex risk scale scores than men (EMM [SE] = 5.01 [.37] vs. 3.66 [.15] for women and men, respectively). Age (p < .001) and number of days of cocaine use in the past 30 days at baseline (p= .003) were significant covariates, such that younger respondents, on average, had higher sex risk scores than older respondents (b = −.02, SE = .003) and more days of cocaine use at baseline was associated with higher sex risk scale scores (b = .01, SE = .003). All other effects in the model failed to reach significance.
3.3. Individual Item Outcomes
3.3.1. Drug Risk Outcomes
Results of the analyses of individual drug risk items are also shown in Table 2. For the sharing needles outcome variable, there was a significant main effect of time (p < .001) for the total sample indicating that needle sharing decreased over time (EMM [SE] = 0.24 [.04], 0.19 [.04], and .06 [.02] for baseline, 6-month, and 12-month assessments, respectively). The baseline number of days of cocaine use was a significant covariate (b = .03, SE = .01, p = .012), such that more days of cocaine use was associated with a greater likelihood of sharing needles.
For sharing rinse water, age (p = .010), prior MMT admissions (p = .022), and baseline number of days of cocaine use (p = .013) were significant covariates. Older participants were less likely to share rinse water (b = −.05, SE = .02) and participants who had previous MMT admissions were significantly more likely than participants who had no previous MMT admissions to have shared rinse water (EMM [SE] = .07 [.03] vs. .03 [.01], respectively). Also, having a greater baseline number of days of cocaine use was associated with a greater likelihood of sharing rinse water (b = .04, SE = .01).
For the sharing cookers outcome, GLiMM analysis revealed a significant main effect of time (EMM [SE] = .17 [.03], .13 [.03], and .06 [.02] for baseline, 6 months, and 12 months, respectively; p = .014). Also, baseline number of days of cocaine use emerged as a significant covariate (b = .03, SE = .01; p = .008) such that more days of cocaine use was associated with a greater likelihood of sharing cookers.
The baseline number of days of cocaine use was a significant covariate for the shared cotton item (p = .014) such that more days of cocaine use at baseline was associated with a greater likelihood of having shared cotton (b = .03, SE = .01).
For the shared drugs (i.e., backloading) outcome, there was a significant main effect for time (p = .030), as well as for the covariates age (p = .038) and baseline cocaine use (p = .004). There was a significant decrease for the time main effect (EMM (SE) = .10 [.03], .07 [.02], and .03 [.02] for baseline, 6 months, and 12 months, respectively). Older participants had, on average, a decreased likelihood of sharing drugs (i.e., backloading) (b = −.04, SE = .02), while having used cocaine on more days at baseline was associated with a greater likelihood of sharing drugs (i.e., backloading) (b = .04, SE = .01).
For the shooting gallery item, age (p = .011) and baseline number of days of cocaine use (p <.001) were significant covariates. Older participants were less likely to use shooting galleries (b = −.04, SE = .02), while participants with more baseline cocaine use were more likely to use shooting galleries (b = .04, SE = .01).
For the injecting drugs outcome, there were significant main effects for time (p < .001), as well as for the covariates age (p = .005), prior MMT admissions (p =.011), and baseline days of cocaine use (p = .001). Over time, the likelihood of injecting drugs declined in the total sample (EMM [SE] = .53 [.05], .36 [.05], and .33 [.05]) for baseline, 6 months, and 12 months, respectively). Younger participants were more likely to have injected than older participants (b = −.04, SE = .01), participants with one or more previous MMT admission had a significantly greater likelihood than participants who had no previous MMT admissions of having injected (EMM [SE] = .50 (.06) vs. .32 [.05]), and participants who had used cocaine more days at baseline were more likely to have injected (b = .04, SE = .01).
3.3.2. Sex Risk Outcomes
The analyses for the individual sex risk items are also shown in Table 3. Gender was associated with the condom use variable (p = .011), such that females had a greater likelihood of reporting the use of condoms all the time compared to males (EMM [SE] = .59 [.07] vs. .39 [.03]).
Gender was associated with the having sex for money and the having sex for drugs items (both ps < .001). Age was a significant covariate for both having sex for money and having sex for drugs (p<.001 and p=.001, respectively). Women were more likely than men to have sex for money (EMMs[SEs] = .38 [.10] vs. .01 [.004]) or drugs (.30 [.11] vs. .02 [.01]), while younger respondents were more likely to have sex for money (b = −.17, SE = .03) or drugs (b = −.11, SE = .03) than older respondents.
Discussion
This paper reports on the 12-month post-release drug- and sex-risk findings from a randomized clinical trial that compared starting interim methadone during pre-trial detention with three months of community-based patient navigation (IM+PN) versus interim methadone without patient navigation (IM) versus an enhanced treatment-as-usual condition (ETAU) that provided a medically supervised opioid withdrawal using methadone with a referral to treatment in the community. Notwithstanding the possibility to transfer from interim methadone treatment in the jail to standard methadone treatment with counseling in the community, there were no significant differences in post-release HIV drug and sex risk scores among the three study conditions. We also found no significant differences in drug and sex risk scale scores and individual variables for the planned contrasts comparing the two IM groups versus the ETAU condition as well as between the IM+PN and IM alone conditions.
We are not aware of other randomized trials comparing starting methadone treatment during pre-trial detention with medically managed withdrawal. The study by Wilson and colleagues [27] measuring HIV-risk behavior using the Texas Christian University AIDS Risk Assessments (ARA) at 1, 3, 6, and 12 months following release from prison offers an interesting contrast to the present study, albeit with some limitations. In that three group-random assignment study, 211 male pre-release sentenced prisoners with histories of opioid dependence were assigned to begin methadone either during or following release from prison compared with assignment to counseling without medication in prison along with a passive treatment referral. The two conditions in which participants were assigned to begin methadone reported fewer occasions of injecting drugs and of sharing needles post-release. These findings are in contrast to the present study in which there were no significant differences between conditions assigned to initiating methadone in jail versus receiving a medically supervised withdrawal and treatment referral in terms of drug injection or of sharing needles. There were a number of differences between these two studies, which could account for the divergent findings. First, the studies were conducted more than 10 years apart. In contrast to the Wilson et al. study, the present study included women as well as men and used dichotomous responses to the RAB rather than continuous responses on the ARA. Moreover, Wilson et al. was conducted with pre-release prisoners who had relatively long periods of opioid abstinence prior to initiating methadone treatment in contrast to the present study’s participants who initiated methadone soon after detention and had relatively brief stays in jail.
The only other randomized trial that compared starting methadone versus buprenorphine treatment in jail lacked a no medication control group, and did not report HIV-risk behavior [28]. Aside from the report by Wilson and colleagues [27] of starting methadone in prison compared to starting after release, the only other randomized trial of starting methadone in a US prison did not find significant differences between conditions in drug injection [29].
The present study also found no significant differences among conditions in terms of either RAB sex risk scores or each of the individual sex risk items (i.e., self-reported use of condoms, or having sex for drugs or money). We also found no significant differences in the planned contrasts comparing the two IM conditions combined versus ETAU and between the IM+PN versus IM alone. These non-significant findings are consistent with the report by Wilson and colleagues [27].
Women represent an important and growing population in correctional settings. Incarcerated women who have engaged in sex trade and used illicit substances have 12 times the prevalence of HIV infection compared to women not in the sex trade [30]. There were significant gender effects in the present study in terms of sex risk scores, condom use, and having sex for money or for drugs. Women had higher sex risk scores and greater likelihood of having sex for money or for drugs than men. There were no significant differences among conditions in HIV sex risk behaviors among the men in the study by Wilson and colleagues and such risk behaviors were not reported by Magura et al. [28] or McKenzie et al. [29]. In contrast to other research conducted in a large, multi-site sample of criminal justice-involved men and women recruited from jails, prisons, and community corrections that found women were less likely than men to report using condoms [31], our study found that significantly more women reported using condoms than men. This may be attributable to differences in the study sample populations, study methods, or local practices. In our latent class analysis of the baseline characteristics of the present study sample, we found that the majority of women were engaged in sex trade and that these women were more likely than men to use condoms consistently [32]. It is possible that the women who were involved in transactional sex were more likely to use condoms because they were acutely aware of their high risk behavior. It is also possible that they misrepresented their condom use during the interview.
More days of cocaine use in the 30 days prior to the baseline interview was associated with both increased drug and sex risk. In addition to higher drug and sex risk scale scores, the number of days of cocaine use was associated with injecting drugs, sharing needles, cookers, cotton, rinse water or drugs (i.e., backloading), and visiting shooting galleries.. The parent study did not find an effect of study interventions on cocaine use at follow-up. This null finding is in keeping with the randomized clinical trial of extended-release naltrexone vs. treatment-as-usual on cocaine use among criminal justice involved outpatients [33]. Evidence of positive associations between cocaine use and HIV-risk behaviors, including both drug and sex risk behaviors, has been shown over decades of research [34, 35]. However, reducing cocaine use continues to be a challenge.
Criminal justice supervision, including drug courts and parole and probation, have been utilized in the US for decades to address co-occurring substance use disorders and criminal justice involvement. However, medication treatments, especially methadone or buprenorphine, have been underutilized in the treatment of OUD among individuals under criminal justice supervision [36]. The extent to which supervision combined with medication reduces cocaine use among individuals with opioid and cocaine use disorder is unclear because there are limited data with contradictory findings [37–39]. Conducting random assignment studies addressing this question would be difficult. In any case, the present study was conducted in a pre-trial population with heterogeneous involvement in supervision but was not focused on the impact of supervision.
A potential advantage of initiating pharmacotherapy prior to release from jail or prison would be to reduce or eliminate opioid use and drug injection during incarceration. Drug injection in prison has been reported throughout the world [40–42]. In the only randomized trial of which we are aware that compared starting methadone in prison compared to a wait list control, Dolan, Wodak, and Hall [43]in New South Wales, Australia, found that the methadone treatment group had significantly lower rates of drug injection in prison. The present study did not collect data on drug injection in prison.
The present study had a number of limitations. HIV-risk behaviors were self-reported during a face-to-face interview. Although confidentiality of responses was stressed prior to and during the research interviews, it is possible that responses were influenced by social desirability [44, 45]. The study was conducted in only one city in the US. In addition, there was a relatively small sample of women. The Risk Assessment Battery’s use of categorical variables for calculating risk scores and our use of dichotomous responses to individual items may have been less sensitive to change compared with continuous responses. Moreover, there are limitations in the RAB itself, such as scoring individuals who identify as gay or bisexual at higher HIV risk compared to individuals who identify as heterosexual. Additionally, due to the different timeframes for measuring baseline cocaine use in the past 30 days and HIV-risk behaviors in the past six months, it is possible that baseline HIV-risk behaviors could have occurred before cocaine use at baseline. Finally, we did not examine the effect of ongoing cocaine use on these behaviors.
This is the first randomized trial of initiating methadone treatment for a pre-trial population that had a no methadone maintenance control group and measured post-release HIV-risk behavior. No significant differences in HIV drug use and sex risk behaviors were found between groups starting methadone maintenance during detention compared with the group that received a medically supervised withdrawal. Cocaine use plays a significant role in these risk behaviors and should be addressed with specific interventions to reduce both drug and sex risk. Specific HIV risk interventions should be added to future research in this at-risk population in order to determine optimal ways of reducing the spread of HIV in this population, especially during the potentially high-risk time immediately post-release [18].
Acknowledgments
Funding
This work was supported by National Institute on Drug Abuse (NIDA) Grant No. 2 U01 DA01363 and the Laura and John Arnold Foundation. NIDA, the National Institutes of Health, and the Arnold Foundation had no role in the design and conduct of the study; data acquisition, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funders.
Footnotes
Clinical Trials Registration: Clinicaltrials.gov NCT 02334215
Conflicts
Dr. Schwartz reports providing consultation to Verily Life Sciences. No financial disclosures were reported by the other authors.
Publisher's Disclaimer: This Author Accepted Manuscript is a PDF file of an unedited peer-reviewed manuscript that has been accepted for publication but has not been copyedited or corrected. The official version of record that is published in the journal is kept up to date and so may therefore differ from this version.
References
- 1.Harris NS, Johnson AS, Huang YA, Kern D, Fulton P, Smith DK, et al. Vital Signs: Status of Human Immunodeficiency Virus Testing, Viral Suppression, and HIV Preexposure Prophylaxis - United States, 2013–2018. MMWR Morb Mortal Wkly Rep. 2019;68(48):1117–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Maruschak L, Bronson J. HIV in Prisons, 2015 - Statistical Tables. In: U.S. Department of Justice, ed. Washington: Bureau of Justice Statistics. 2017. [Google Scholar]
- 3.Rudd RA, Aleshire N, Zibbell JE, Gladden RM. Increases in Drug and Opioid Overdose Deaths--United States, 2000–2014. MMWR Morb Mortal Wkly Rep. 2016;64(50–51):1378–82. [DOI] [PubMed] [Google Scholar]
- 4.Mattick RP, Breen C, Kimber J, Davoli M. Methadone maintenance therapy versus no opioid replacement therapy for opioid dependence. Cochrane Database Syst Rev. 2009(3):CD002209. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Mattick RP, Breen C, Kimber J, Davoli M. Buprenorphine maintenance versus placebo or methadone maintenance for opioid dependence. Cochrane Database Syst Rev. 2014(2):CD002207. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Nielsen S, Larance B, Degenhardt L, Gowing L, Kehler C, Lintzeris N. Opioid agonist treatment for pharmaceutical opioid dependent people. Cochrane Database Syst Rev. 2016(5):CD011117. [DOI] [PubMed] [Google Scholar]
- 7.Edelman EJ, Chantarat T, Caffrey S, Chaudhry A, O’Connor PG, Weiss L, et al. The impact of buprenorphine/naloxone treatment on HIV risk behaviors among HIV-infected, opioid-dependent patients. Drug Alcohol Depend. 2014;139:79–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Gowing LR, Farrell M, Bornemann R, Sullivan LE, Ali RL. Brief report: Methadone treatment of injecting opioid users for prevention of HIV infection. J Gen Intern Med. 2006;21(2):193–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Lott DC, Strain EC, Brooner RK, Bigelow GE, Johnson RE. HIV risk behaviors during pharmacologic treatment for opioid dependence: a comparison of levomethadyl acetate [corrected] buprenorphine, and methadone. J Subst Abuse Treat. 2006;31(2):187–94. [DOI] [PubMed] [Google Scholar]
- 10.Bruce RD. Methadone as HIV prevention: high volume methadone sites to decrease HIV incidence rates in resource limited settings. Int J Drug Policy. 2010;21(2):122–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Karki P, Shrestha R, Huedo-Medina TB, Copenhaver M. The impact of methadone maintenance treatment on HIV risk behaviors among high-risk injection drug users: a systematic review. Evid Based Med Public Health. 2016;2. [PMC free article] [PubMed] [Google Scholar]
- 12.Novick DM, Joseph H, Croxson TS, Salsitz EA, Wang G, Richman BL, et al. Absence of antibody to human immunodeficiency virus in long-term, socially rehabilitated methadone maintenance patients. Arch Intern Med. 1990;150(1):97–9. [PubMed] [Google Scholar]
- 13.Yancovitz SR, Des Jarlais DC, Peyser NP, Drew E, Friedmann P, Trigg HL, et al. A randomized trial of an interim methadone maintenance clinic. Am J Public Health. 1991;81(9):1185–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Schwartz RP, Highfield DA, Jaffe JH, Brady JV, Butler CB, Rouse CO, et al. A randomized controlled trial of interim methadone maintenance. Arch Gen Psychiatry. 2006;63(1):102–9. [DOI] [PubMed] [Google Scholar]
- 15.Beckwith CG, Zaller ND, Fu JJ, Montague BT, Rich JD. Opportunities to diagnose, treat, and prevent HIV in the criminal justice system. J Acquir Immune Defic Syndr (1999), 55 Suppl 1(Suppl 1), S49–S55. 10.1097/QAI.0b013e3181f9c0f7. 2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Magura S, Rosenblum A, Joseph H. AIDS-risk among intravenous drug-using offenders. Crime Delinq. 37(1); 86–100. 1991. [Google Scholar]
- 17.Underhill K, Dumont D, Operario D. HIV prevention for adults with criminal justice involvement: a systematic review of HIV risk-reduction interventions in incarceration and community settings. Am J Public Health, 104(11), e27–e53. 2014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Adams LM, Kendall S, Smith A, Quigley E, Stuewig JB, Tangney JP. HIV risk behaviors of male and female jail inmates prior to incarceration and one year post-release. AIDS Behav. 2013;17(8):2685–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.MacGowan RJ, Margolis A, Gaiter J, Morrow K, Zack B, Askew J, et al. Predictors of risky sex of young men after release from prison. Int J STD AIDS. 14(8):519–23. PubMed PMID: 12935380.523. 10.1258/095646203767869110. 2003. [DOI] [PubMed] [Google Scholar]
- 20.Milloy MJ, Buxton J, Wood E, Li K, Montaner JS, Kerr T. Elevated HIV risk behaviour among recently incarcerated injection drug users in a Canadian setting: a longitudinal analysis. BMC Public Health. 2009;9:156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Zeng Zhen. Jail Inmates in 2016 U.S. Department of Justice Office of Justice Programs Bureau of Justice Statistics. NCJ 251210.12. 2018. [Google Scholar]
- 22.Carson EA. Bureau of Justice Statistics | Office of Justice Programs. U.S. Department of Justice. Prisoners in 2016. Summary| NCJ 251149 2018. [Google Scholar]
- 23.Chandler R, Gordon MS, Kruszka B, Strand LN, Altice FL, Beckwith CG, et al. Cohort profile: seek, test, treat and retain United States criminal justice cohort. Subst Abuse Treat Prev Policy, 12(1), 24. 2017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Sharma A, O’Grady KE, Kelly SM, Gryczynski J, Mitchell SG, Schwartz RP. Pharmacotherapy for opioid dependence in jails and prisons: research review update and future directions. Subst Abuse Rehabil. 2016;7:27–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Metzger D, Woody GE, Navaline H, McLellan AT, Meyers K, Boney T, et al. The risk assessment battery (RAB): Validity and reliability. In Sixth Annual Meeting of National Cooperative Vaccine Development Group for AIDS. University of Pennsylvania, Center for Studies on Addition, Philadelphia: PA. 1993. [Google Scholar]
- 26.UPENN. Risk Assessment Battery. HIV/AIDS Research Prevention Division. Perelman School of Medicine. University of Pennsylvania. Available at: https://www.med.upenn.edu/hiv/rab_download.html. Accessed on April 20, 2020. [Google Scholar]
- 27.Wilson ME, Kinlock TW, Gordon MS, O’Grady KE, Schwartz RP. Postprison release HIV-risk behaviors in a randomized trial of methadone treatment for prisoners. Am J Addict. 2012;21(5):476–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Magura S, Lee JD, Hershberger J, Joseph H, Marsch L, Shropshire C, et al. Buprenorphine and methadone maintenance in jail and post-release: a randomized clinical trial. Drug Alcohol Depend. 2009;99(1–3):222–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.McKenzie M, Zaller N, Dickman SL, Green TC, Parihk A, Friedmann PD, et al. A randomized trial of methadone initiation prior to release from incarceration. Substance abuse : official publication of the Association for Medical Education and Research in Substance Abuse. 2012;33(1):19–29. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Strathdee SA, West BS, Reed E, Moazen B, Azim T, Dolan K. Substance Use and HIV Among Female Sex Workers and Female Prisoners: Risk Environments and Implications for Prevention, Treatment, and Policies. J Acquir Immune Defic Syndr. 2015;69 Suppl 2:S110–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Loeliger KB, Biggs ML, Young R, Seal DW, Beckwith CG, Kuo I, et al. Gender Differences in HIV Risk Behaviors Among Persons Involved in the U.S. Criminal Justice System and Living with HIV or at Risk for HIV: A “Seek, Test, Treat, and Retain” Harmonization Consortium. AIDS Behav. 2017;21(10):2945–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Mitchell MM, Gryczynski J, Mitchell SG, Kelly SM, O’Grady KE, Monico LB, et al. A Latent Class Analysis of HIV Risk Factors Among Men and Women with Opioid Use Disorder in Pre-trial Detention. AIDS Behav. 2019. [DOI] [PubMed] [Google Scholar]
- 33.Lee JD, Nunes EV, Bailey GL, Brigham GS, Cohen AJ, Fishman M, et al. NIDA Clinical Trials Network CTN-0051, extended-release naltrexone vs. buprenorphine for opioid treatment (X: BOT): study design and rationale. Contemp Clin Trials. 2016;50:253–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Booth RE, Watters JK, Chitwood DD. HIV risk-related sex behaviors among injection drug users, crack smokers, and injection drug users who smoke crack. Am J Public Health. 1993;83(8):1144–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Bux DA, Lamb R, Iguchi MY. Cocaine use and HIV risk behavior in methadone maintenance patients. Drug Alcohol Depend. 1995;37(1):29–35. [DOI] [PubMed] [Google Scholar]
- 36.Morris NP, Bentzley BS. Supporting the use of medications for addiction treatment in US drug courts: opportunities for health professionals. J Addict Med. 2020;14(4):277–9. [DOI] [PubMed] [Google Scholar]
- 37.Baughman M, Tossone K, Singer MI, Flannery DJ. Evaluation of treatment and other factors that Lead to drug court success, substance use reduction, and mental health symptomatology reduction over time. Int J Offender Ther Comp Criminol. 2019;63(2):257–75. [DOI] [PubMed] [Google Scholar]
- 38.Kelly SM, O’Grady KE, Jaffe JH, Gandhi D, Schwartz RP. Improvements in outcomes in methadone patients on probation/parole regardless of counseling early in treatment. J Addict Med. 2013;7(2):133. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Mitchell SG, Gryczynski J, Kelly SM, O’Grady KE, Jaffe JH, Olsen YK, et al. Treatment outcomes of African American buprenorphine patients by parole and probation status. J Drug Issues. 2014;44(1):69–82. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Azbel L, Wegman MP, Polonsky M, Bachireddy C, Meyer J, Shumskaya N, et al. Drug injection within prison in Kyrgyzstan: elevated HIV risk and implications for scaling up opioid agonist treatments. Int J Prison Health. 2018;14(3):175–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Culbert GJ, Waluyo A, Iriyanti M, Muchransyah AP, Kamarulzaman A, Altice FL. Within-prison drug injection among HIV-infected male prisoners in Indonesia: a highly constrained choice. Drug Alcohol Depend. 2015;149:71–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Dolan KA, Wodak A. HIV transmission in a prison system in an Australian State. Med J Aust. 1999;171(1):14–7. [DOI] [PubMed] [Google Scholar]
- 43.Dolan KA, Wodak AD, Hall WD. Methadone maintenance treatment reduces heroin injection in New South Wales prisons. Drug Alcohol Rev, 17(2), 153–158. 1998. [DOI] [PubMed] [Google Scholar]
- 44.Rao A, Tobin K, Davey-Rothwell M, Latkin CA. Social Desirability Bias and Prevalence of Sexual HIV Risk Behaviors Among People Who Use Drugs in Baltimore, Maryland: Implications for Identifying Individuals Prone to Underreporting Sexual Risk Behaviors. AIDS Behav. 2017;21(7):2207–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Latkin CA, Vlahov D, Anthony JC. Socially desirable responding and self-reported HIV infection risk behaviors among intravenous drug users. Addiction. 1993;88(4):517–26. [DOI] [PubMed] [Google Scholar]