Combination antiretroviral therapy has demonstrated efficacy in suppressing HIV replication, [1,2] improving immune function  and decreasing HIV-related morbidity and mortality . Despite this potential, studies have shown that traditionally underserved populations, such as minorities and drug users, are less likely to receive such therapy [5,6]. Reasons for suboptimal treatment in these populations include the common perception that such individuals are unable to adhere to complex medication regimens . Concern about non-adherence is warranted, as it may result in both treatment failure and the development of drug resistant virus . Yet the predictive value of individual patient characteristics, including drug use, on adherence remain controversial. Some investigators have demonstrated poorer adherence in minorities [9,10] and drug users, [11,12] while others have not observed these associations [13,14]. However, few studies have examined adherence in HIV-infected women, many of whom are also drug users or members of minority groups.
To date, the HIV adherence literature has focused predominantly on men who have sex with men or male injecting drug users [11,15–20]. However, some studies of drug users and ethnic minorities have found that women are significantly less adherent than men [12,21,22]. Though a growing body of research indicates that psychosocial factors, such as depression and social support, may strongly influence adherence, [15,18,19,21] the reasons for poorer adherence in women remain largely unknown.
Differences in study design and the lack of a standardized measure of adherence have complicated the interpretation of findings across studies. Most studies of antiretroviral adherence have either been cross-sectional, or have based their analyses of longitudinal data on a single summary measure of adherence. Yet more recently, prospective analyses have been performed which demonstrate that adherence is a dynamic process, [23,24] and its predictors vary over time . In addition, while most of the adherence literature has utilized either self-report or pill count, measures which may overestimate adherence [23,26], use of MEMS caps is often considered to be a more accurate method of measuring adherence . However MEMS caps are limited in that they cannot be used by patients who use pill boxes, and cap removal is only a proxy for actual medication ingestion .
This prospective study utilized MEMS caps to describe adherence over time, and to assess its predictors and relationship to viral load in a large cohort of racially diverse women with and without a history of drug use.
Women were recruited from the HIV Epidemiology Research (HER) Study, a multi-center cohort study of HIV infection in women. Participant recruitment and study design have been described in detail elsewhere . Briefly, between 1993 and 1995, 871 HIV-infected women and 439 uninfected women were enrolled in four US cities: New York (Bronx), Baltimore, Providence and Detroit. HER Study participants were followed with semi-annual research visits, at which a standardized interview was administered and blood was obtained for T-lymphocyte studies and HIV viral load measurement.
Between June and September 1999, HER Study participants who were taking at least two antiretroviral medications were co-enrolled in the Adherence Study within 1 month of their semi-annual visit. Research visits for the Adherence Study occurred monthly during the 6-month protocol, and the final visit coincided with a HER Study semi-annual research visit. In keeping with the overall HER Study design, enrollment was monitored so that approximately half of the participants had an injecting drug use history, and half reported only a sexual exposure risk for contracting HIV. The study protocol was approved by the institutional review board for the protection of human subjects of each study site and of the Centers for Disease Control and Prevention. All participants gave written informed consent.
Upon enrollment in the Adherence Study, participants were given a pill bottle fitted with a MEMS cap (Aprex Corporation, Menlo Park, California, USA) for each of their antiretroviral medications. Participants were instructed to store their medications in these pill bottles exclusively, and to open these bottles only when withdrawing a dose for ingestion.
At each monthly research visit, MEMS data were downloaded from the devices using MEMS View 2.61 software (Aprex Corporation) and a MEMS communicator unit. In addition, a brief interview was administered to elicit changes in the antiretroviral regimen, side effects, and dates when medications were taken but the MEMS devices were not used (e.g., during hospitalizations or incarcerations). Whether participants removed more than one dose at a time from the MEMS bottle, however, was not routinely assessed. A new MEMS device was administered when a new medication was prescribed or if a device was lost. Participants received monetary reimbursement for each visit.
Viral load (Quantiplex HIV-1 RNA v3.0 Assay, Chiron Corporation, Emeryville, California, USA) obtained at the semi-annual HER Study visits coinciding with enrollment and completion of the Adherence Study were used as the initial and follow-up viral loads, respectively. The lower limit of detection for this assay is 50 copies/ml. CD4 lymphocyte counts and interview data, including demographic characteristics, medical history, and drug use behaviors, were also obtained at HER Study visits.
All data downloaded from the MEMS caps, including the date and time of each cap opening, were exported into the SAS System (SAS Institute Inc., Cary, North Carolina, USA) for analysis. Periods during which adherence could not be measured because the MEMS caps were not used, such as during hospitalizations or incarcerations, were excluded from the MEMS database prior to analysis. A daily adherence rate for each medication was calculated by dividing the number of MEMS openings by the number of doses prescribed in the 24-h period between 0300 h and 0259 h. Next, a composite daily adherence rate was calculated by averaging the daily adherence rate for each medication in a participant's regimen. Monthly adherence rates were then calculated for each participant by averaging the composite daily adherence rates over the number of days in the month that she utilized MEMS caps.
Linear mixed models were used to examine the associations of socio-demographic factors, drug use behaviors, medication-related factors, and CD4 lymphocyte count with the repeated monthly measurements of adherence in participants. The MIXED procedure of the SAS System was used to examine how adherence changed over time within subjects and how between-subjects effects on adherence changed over time. Least-squares means adjusted for time and averaged across the repeated measures were used for all between-subjects comparisons. Generalizing estimating equations (GEE) were applied to determine the factors associated with worsening adherence. The repeated dichotomous outcome variable was set to 1 if the monthly adherence of a given participant decreased by 10 or more percentage points from month i to month i + 1, where i = 1,…,5, and was otherwise set to 0. Factors assessed included those that were significantly associated with adherence in the multivariate linear mixed model, as well as the following dichotomous variables: hospitalization during the study period, change in antiretroviral regimen during the study period, and an undetectable viral load at baseline. The SAS GENMOD procedure was used to assess the effects of the covariates on the multiple binary outcomes. The associations of follow-up CD4 lymphocyte count and viral load (log10 HIV RNA copy number) with each monthly adherence measurement were determined using Spearman's correlation. In addition, logistic regression analysis was performed to determine the odds of virologic failure (HIV viral load > 500 copies/ml) for each quartile of adherence at month 6.
A total of 362 HER Study participants were screened: 215 were found to be eligible, and 175 (81%) of these enrolled in the Adherence Study. Compared to the 40 women who were eligible but did not enroll, the participants who enrolled were more likely to be African–American (P < 0.0005) and unemployed (P = 0.002), but were similar with respect to age, education status, history of injecting drug use, CD4 lymphocyte count and viral load. Adherence data are available for 161 women; the remainder did not return (n = 12) or lost their caps (n = 2) after the first visit. The median length of follow-up for these 161 women was 175 days [interquartile range (IQR), 166–182]; 153 (95%) completed all 6 months of follow-up.
Characteristics of the study population are presented in Table 1. Although over half (56%) of the participants had a history of injecting drug use, recent substance abuse was uncommon: only 9% (n = 14) reported injecting drug use during the 6 months prior to study enrollment and only 13% (n = 21) reported using crack cocaine. The majority (98%) of participants was antiretroviral experienced. Over 4.5 years of follow-up in the HER Study, participants reported that they had taken antiretroviral therapy for a median of 3.0 years (range, 0–4.5 years), and been exposed to a median of six antiretroviral agents (range, 2–14 agents). Most (94%) reported receiving care from an HIV specialist. Nearly half of the participants had advanced disease; the median nadir CD4 lymphocyte count during HER Study follow-up was 203 × 106/l (range, 0–864 × 106/l).
Compared to the remainder of the HIV seropositive HERS cohort, the Adherence Study participants were more likely to have an undetectable viral load (P = 0.001). This is likely because only 60% of the remaining HIV-infected HERS participants were taking antiretroviral drugs (P < 0.0005). The Adherence Study participants were also more likely to be African–American (P = 0.01) and unemployed (P = 0.02), but did not differ from the remainder of the HIV infected cohort with respect to age, education status, injecting drug use history, or CD4 lymphocyte count.
At the time of enrollment in the Adherence Study, 84% of the participants had been taking their current antiretroviral regimen for more than 3 months. Overall, 70% of women were on a triple drug combination, 15% were on fewer than three drugs, and 15% were on more than three antiretroviral agents. Seventy-seven percent (n = 124) were on a twice-daily regimen, and 50% (n = 81) were taking a protease inhibitor. Although almost half (47%) of the participants reported side effects from their antiretroviral medications, the majority (81%) remained on the same regimen throughout the study period.
A total of 501 MEMS caps were administered, and of these, 489 (98%) were analyzed. Reasons for lack of analysis included: cap lost (n = 8), cap malfunctioned (n = 2), and medication discontinued (n = 2). Adherence was measured with a mean of 3.0 MEMS caps per participant for a median of 170 days (IQR, 142–179 days). Adherence to medications in liquid or powder form (n = 13) was not measured. In 21 women, periods during which the MEMS caps were not used (median 21 days; IQR, 10–32 days) were excluded from the database prior to analysis. Seventy-three percent of women contributed six monthly adherence measurements, 18% contributed five measurements, and the remaining 9% contributed between two and four measurements.
The mean monthly adherence rate was 64%, 55%, 52%, 49%, 47%, and 45% during the first, second, third, fourth, fifth, and sixth month of observation, respectively (Fig. 1). Of the 117 women with six monthly adherence measurements, only 2% (n = 2) had greater than 95% adherence at each time point. Adherence varied significantly over time (P < 0.001). Mean adherence was significantly different between months 1 and 2 (P < 0.0001 for paired t test), months 2 and 3 (P = 0.001), and months 3 and 4 (P = 0.02).
To investigate the variability in adherence further, the change in monthly adherence was calculated by subtracting the adherence rate in month i + 1 from the adherence rate in month i. As shown in Fig. 2, the median (IQR) change in adherence between months 1 and 2 was −6% (−14 to 0); between months 2 and 3, −2% (−10 to 3); months 3 and 4, −2% (−9 to 3), months 4 and 5, −2% (−10 to 4), and months 5 and 6, −1% (−9 to 4). Thirty-seven percent of participants (n = 60) had a decrease in adherence of 10% or more between months 1 and 2; 24% (n = 38) between months 2 and 3; 24% (n = 36) between months 3 and 4; 24% (n = 35) between months 4 and 5; and 21% (n = 25) between months 5 and 6. An increase in adherence of 10% or more between consecutive months was found in 7%, 5%, 12%, 11%, and 13% of participants at each successive comparison.
Factors associated with adherence: univariate analysis
Race, education, and living with a child were not significantly associated with adherence (Table 2). When age was treated as a continuous variable, older age was associated with greater adherence (P = 0.05). Employed women were more adherent than those who were not employed (67% versus 48%; P = 0.001).
Although a history of ever injecting drugs was not associated with poor adherence (P = 0.28), active drug use during the Adherence Study was. The mean adherence among those reporting use of cocaine or heroin by any route during the study period was 34%, compared with 57% among those not using drugs (P = 0.0002). Similarly, women who reported alcohol intake on at least one day per week during the study period were less adherent than those who did not (43% versus 56%; P = 0.02).
Medication-related factors and CD4 lymphocyte count
Greater duration of antiretroviral use was associated with greater adherence. The mean adherence among participants on antiretroviral therapy for more than 2 years was 56%, compared with 42% among those on therapy for 2 years or less (P = 0.005). Protease inhibitor use, in contrast, was associated with lower adherence (47% versus 57%; P = 0.04). This may have been due to the fact that women on twice-daily regimens were more adherent than those on thrice-daily regimens (55% versus 42%; P = 0.01). Medication side effects (P = 0.79), number of antiretroviral agents (P = 0.74), and daily number of pills (P = 0.62) were not associated with adherence. Women with a CD4 lymphocyte count > 500 × 106/l at enrollment were more adherent (65%) that those with a CD4 lymphocyte count between 201 × 106 and 500 × 106/l (51%; P = 0.005), and those with a CD4 lymphocyte count of ≤ 200 × 106/l (40%; P < 0.0001).
Predictors of adherence: multivariate analysis
The factors that remained significantly associated with lower adherence upon multivariate analysis included active drug use (P = 0.002), alcohol use at least one day per week during the study period (P = 0.04), more frequent antiretroviral dosing (P = 0.001), shorter duration of antiretroviral drug use (P = 0.005), younger age (P = 0.047), lower initial CD4 lymphocyte count (P < 0.0001), time (P < 0.0001), and an interaction term representing active drug use and time (P = 0.002).
GEE were applied to determine the best predictive model for worsening adherence, defined as a decrease in adherence of 10% or more between two consecutive months. The factors associated with worsening adherence were: hospitalization during the period of observation [odds ratio (OR) 1.6; 95% confidence interval (CI), 1.0–2.6], a change in antiretroviral regimen during the study period (OR, 1.9; 95% CI, 1.2–3.0), and an interaction term representing duration of antiretroviral use and baseline viral load (OR, 0.2; 95% CI, 0.1–0.4). This interaction term indicated that among women with a detectable viral load, those who were on antiretroviral therapy for 2 years or less were less likely to have worsening adherence than those who were more antiretroviral experienced; while among women with an undetectable viral load, there was no association between duration of therapy and worsening adherence.
Adherence and immunologic and virologic outcomes
To investigate the construct validity of the adherence measure, the correlations between each monthly adherence measure and both CD4 lymphocyte count and HIV viral load at follow-up were determined. There was a significant direct correlation between CD4 lymphocyte count and adherence in months 1 through 5 (R, 0.29 to 0.31; P < 0.0005). There was also a significant inverse correlation between viral load and adherence at each time point (R, −0.37 to −0.48; P < 0.0001), with the strongest correlation found at month 6. In addition, the mean adherence during month 6 among participants who achieved an undetectable viral load was significantly higher than that among women with a detectable viral load (66% versus 38%; P < 0.0005, t test).
The relationship between adherence at month 6 and viral load was examined further by calculating the percent of participants with virologic failure (> 500 RNA copies/ml) at follow-up in each quartile of adherence. Only 17% of participants in the highest quartile of adherence (≥ 88%) had virologic failure (Fig. 3). The proportion of women with virologic failure was higher in the third quartile (28%), however this difference was not statistically significant. Compared to women in the fourth quartile, the odds of virologic failure in the lowest two quartiles were: 3.6 (95% CI, 1.1–12.2; P = 0.04) in women with adherence of 13–44%, and 11.7 (95% CI, 3.4–40.4; P < 0.0001) in women with adherence of ≤ 12%.
In this cohort of racially diverse, heavily-pretreated HIV-infected women, mean adherence measured by MEMS caps ranged from 64% in the first month of observation, to 45% in month 6. Although the study population was biased towards participants who were more adherent and in better health than the overall cohort, antiretroviral adherence in these women was still poor. Adherence in this population was lower than that found in a cohort of protease inhibitor-naive patients (mean 80%),  and in a study of predominantly white men who have sex with men (mean 90%) . Bangsberg reported a similar rate (median 67%) in a cohort of homeless individuals,  and we previously found a comparable rate (mean 53%) among opiate-addicted drug users .
The longitudinal nature of this study allowed us to demonstrate that adherence changes over time. Although on a population level, mean adherence stabilized after the first few months of observation, on an individual level, adherence continued to vary throughout the study period. We found that adherence increased between consecutive measurements in a considerable minority of patients. However even after the initial drop in adherence between months 1 and 2, nearly one fourth of participants had a 10% or greater decrease in adherence between consecutive monthly measurements. The risk of worsening adherence was greater among women who were hospitalized, and among those who started a new antiretroviral regimen during the study period. Although on average, women who were more antiretroviral experienced were more adherent, if they had taken antiretroviral therapy for more than 2 years and had a detectable viral load, they were also at a greater risk of worsening adherence over time. Based on these findings, we recommend that in clinical practice, the assessment of adherence should be an ongoing process, and may be especially important during periods of illness or when a patient starts a new regimen. Furthermore, future studies of adherence-enhancing interventions should include longitudinal strategies to account for the dynamic nature of adherence behavior.
Similar to other studies, [16,26] we found that the risk of virologic failure rose with decreasing levels of adherence. However, we also found that a significant proportion of subjects achieved virologic suppression with adherence levels that were slightly lower than others have reported,  suggesting that 95% adherence (measured by MEMS) may not be necessary to achieve virologic suppression. Our findings are consistent with recent findings by Liu et al. who demonstrated that adherence of 74% (measured by MEMS) was sufficient to achieve virologic suppression . Although MEMS adherence correlates strongly with viral load in this study and in others, [17,26] electronic monitors may underestimate adherence if patients take their medications from somewhere other than the MEMS bottles (`pocket doses') . As we did not adjust for pocket doses in this analysis, the true mean adherence among women with an undetectable viral load may be somewhat higher than the 66% we report. Self-report, which was not assessed in this study, has been shown to produce higher levels of adherence than MEMS . Thus in clinical practice, where self-report is most often used to assess adherence, we believe that patients and their providers should continue to strive for greater than 95% adherence.
In this study, neither race nor a history of ever injecting drugs predicted a woman's adherence. Although this was a study of a select group of women who chose to participate in a long-term study, we believe that based on these findings, minority race and past injecting drug use should not be regarded as contraindications to the prescription of potent antiretroviral therapy. However, like others, [11,21] we found that active use of either illicit drugs or alcohol was associated with worse adherence. These data underscore the need for drug and alcohol treatment programs in HIV-infected women. Although antiretroviral regimens were not randomly assigned in this study, the finding that less frequent antiretroviral dosing was associated with greater adherence supports the development of simpler antiretroviral regimens to aid patients with their adherence.
In summary, this study demonstrates that antiretroviral adherence among HIV-infected women is poor, and that long-term interventions aimed at improving their adherence are urgently needed. As women constitute a steadily increasing proportion of US AIDS cases,  finding ways to maximize their adherence to combination antiretroviral therapy should be a major public health priority.
Sponsorship: Supported by the Centers for Disease Control and Prevention (U64/CCU206798-01) and the National Institute on Drug Abuse.
1.Collier AC, Coombs RW, Schoenfeld DA, Bassett RL, Timpone J, Baruch A, et al.Treatment of human immunodeficiency virus infection with saquinavir, zidovudine, and zalcitabine.N Engl J Med
2.Gulick RM, Mellors JW, Havlir D, Eron JJ, Gonzalez C, McMahon D, et al.Treatment with indinavir, zidovudine, and lamivudine in adults with human immunodeficiency virus infection and prior antiretroviral therapy.N Engl J Med
3.Autran B, Carcelain G, Li TS, Blanc C, Mafher D, Tubiana R, et al.Positive effect of combined antiretroviral therapy on CD4+ T cell homeostasis and function in advanced HIV disease.Science
4.Palella FJ, Delaney KM, Moorman AC, Loveless MO, Fuher J, Satten GA, et al
.and the HIV Outpatient Study Investigators. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection.N Engl J Med
5.Shapiro MF, Morton SC, McCaffrey DF, Andersen RM, Berk ML, Bing EG, et al.Variations in the care of HIV-infected adults in the United States: results from the HIV Cost and Services Utilization Study.JAMA
6.Celentano DD, Vlahov D, Cohn S, Shadle VM, Obasanjo O, Moore RD. Self-reported antiretroviral therapy in injection drug users.JAMA
7.Bassetti S, Battegay M, Furrer H, Rickenbach M, Flepp M, Kaiser L, et al.Why is highly active antiretroviral therapy (HAART) not prescribed or discontinued?J Acquir Immune Defic Syndr
8.Wainberg MA, Friedland G. Public health implications of antiretroviral therapy and HIV drug resistance.JAMA
9.Singh N, Squier C, Sivek C, Wagener M, Hong Nguyen M, Yu VL. Determinants of compliance with antiretroviral therapy in patients with human immunodeficiency virus: prospective assessment with implications for enhancing compliance.AIDS Care
10.Wenger N, Gifford A, Liu H, Chesney M, Golin C, Crystal S, et al.Patient characteristics and attitudes associated with antiretroviral adherence.Sixth Conference on Retroviruses and Opportunistic Infections.
Chicago, February 1999 [abstract 98].
11.Haubrich RH, Little S, Currier JS, Forthal D, Kemper C. The value of patient-reported adherence to antiretroviral therapy in predicting virologic and immunologic response.AIDS
12.Montessori V, Heath KV, Yip B, Hogg RS, O'Shaughnessy MV, Montaner JSG. Predictors of adherence with triple combination therapy.Seventh Conference on Retroviruses and Opportunistic Infections.
San Francisco, 2000 [abstract 72].
13.Kalichman SC, Ramachandran B, Catz S. Adherence to combination antiretroviral therapies in HIV patients of low health literacy
. J Gen Intern Med
14.Mostashari F, Riley E, Selwyn PA, Altice FL. Acceptance and adherence with antiretroviral therapy among HIV-infected women in a correctional facility.J Acquir Immune Def Syndr Hum Retrovirol
15.Singh N, Berman SM, Swindells S, Justice JC, Mohr JA, Squier, C et al.Adherence of human immunodeficiency virus-infected patients to antiretroviral therapy.Clin Infect Dis
16.Paterson DL, Swindells S, Mohr J, Brester M, Squier C, Wagener MM, et al.Adherence to protease inhibitor therapy and outcomes in patients with HIV infection.Ann Intern Med
17.Bangsberg DR, Hecht FM, Charlebois ED, Zolopa AR, Holodniy M, Sheiner L, et al.Adherence to protease inhibitors, HIV-1 viral load, and development of drug resistance in an indigent population.AIDS
18.Gordillo V, del Amo J, Soriano V, Gonzalez-Lahoz H. Sociodemographic and psychological variables influencing adherence to antiretroviral therapy.AIDS
19.Holzemer WL, Corless IB, Nokes KM, Turner JG, Brown MA, Powell-Cope GM, et al.Predictors of self-reported adherence in persons living with HIV disease.AIDS Patient Care and STDs
20.Gao X, Nau DP, Rosenbluth SA, Scott V, Woodward C. The relationship of disease severity, health beliefs and medication adherence among HIV patients.AIDS Care
21.Arnsten JH, Demas PA, Grant RW, Gourevitch MN, Farzadegan H, Howard AA, et al.Impact of active drug use on antiretroviral therapy adherence and viral suppression in HIV-infected drug users.J Gen Intern Med
22.McNabb J, Ross JW, Abriola K, Turley C, Nightingale CH, Nicolau DP. Adherence to highly active antiretroviral therapy predicts virologic outcome at an inner-city human immunodeficiency virus clinic.Clin Infect Dis
23.Liu H, Golin CE, Miller LG, Hays RD, Beck CK, Sanandaji S, et al.A comparison of multiple measures of adherence to HIV protease inhibitors.Ann Intern Med
24.Carrieri P, Cailleton V, Le Moing V, Spire B, Dellamonica P, Bouvet E, et al.The dynamic of adherence to highly active antiretroviral therapy: results from the French National APROCO cohort.J Acquir Immune Def Syndr
25.Mannheimer S, Friedland G, Matts J, Child C, Chesney M,for the Terry Beirn Community Programs for Clinical Research on AIDS. The consistency of adherence to antiretroviral therapy predicts biologic outcomes for human immunodeficiency virus-infected persons in clinical trials.Clin Infect Dis
26.Arnsten JH, Demas PA, Farzadegan H, Grant RW, Gourevitch MN, Chang CJ, et al.Antiretroviral therapy adherence and viral suppression in HIV-infected drug users: Comparison of self-report and electronic monitoring.Clin Infect Dis
27.Feinstein A. On white-coat effects and the electronic monitoring of compliance.Arch Intern Med
28.Stone VE. Strategies for optimizing adherence to highly active antiretroviral therapy: lessons from research and clinical practice.Clin Infect Dis
29.Smith DK, Warren DL, Vlahov D, Schuman P, Stein MD, Greenberg BL, et al.Design and baseline participant characteristics of the Human immunodeficiency virus Epidemiology Research (HER) Study: A prospective cohort study of human immunodeficiency virus infection in US women.Am J Epidemiol
30.Golin C, Liu H, Hays R, Ickovics J, Beck K, Miller L, et al
. Self-reported adherence to protease inhibitors substantially overestimates an objective measure.Sixth Conference on Retroviruses and Opportunistic Infections.
Chicago, 1999 [abstract 95].
31.Melbourne KM, Geletko SM, Brown SL, Willey-Lessne C, Chase S, Fisher A. Medication adherence in patients with HIV infection: A comparison of two measurement methods.AIDS Reader
32.Centers for Disease Control and Prevention. HIV/AIDS among U.S. women: minority and young women at continuing risk.CDC Fact Sheet
The HER Study group
The HER Study group consists of: R. S. Klein, E. E. Schoenbaum, J. H. Arnsten, R. D. Burk, C. J. Chang, P. Demas, A. A. Howard, Montefiore Medical Center and the Albert Einstein College of Medicine; P. Schuman, J. Sobel, the Wayne State University School of Medicine; A. Rompalo, D. Vlahov, D. Celentano, the Johns Hopkins University School of Medicine; C. Carpenter, K. Mayer, the Brown University School of Medicine; A. Duerr, L. I. Gardner, S. Holmberg, D. Jamieson, J. Moore, R. Phelps, D. Smith, D. Warren, the Centers for Disease Control and Prevention; K. Davenny, National Institute on Drug Abuse.