Drug and Alcohol Dependence 83 (2006) 147–156
Longitudinal predictors of injection cessation and subsequent relapse among
a cohort of injection drug users in Baltimore, MD, 1988–2000
Nina G. Shah a,∗ , Noya Galai b , David D. Celentano b , David Vlahov c , Steffanie A. Strathdee d
a
b
Epidemiology and Response Division, New Mexico Department of Health, 1190 St. Francis Drive, P.O. Box 26110, Santa Fe, NM 87502-6110, USA
Department of Epidemiology, Bloomberg School of Public Health, The Johns Hopkins University, 615 North Wolfe Street, Baltimore, MD 21205, USA
c Center for Urban Epidemiologic Studies, New York Academy of Medicine, 1216 Fifth Avenue, New York, NY 10029, USA
d University of California San Diego School of Medicine, Division of International Health and Cross Cultural Medicine,
9500 Gilman Drive, Mailstop 0622, San Diego, CA 92093, USA
Received 25 May 2005; received in revised form 3 November 2005; accepted 7 November 2005
Abstract
Objective: To determine predictors of injection drug use cessation and subsequent relapse among a cohort of injection drug users (IDUs).
Methods: IDUs in Baltimore, MD were recruited through community outreach in 1988–1989. Among IDUs with at least three follow-up visits,
parametric survival models for time to injection cessation (≥6 months) and subsequent relapse were constructed.
Results: Of 1327 IDUs, 94.8% were African American, 77.2% were male, median age was 34 years, and 37.7% were HIV-infected. Among 936
(70.5%) subjects who ceased injection, median time from baseline to cessation was 4.0 years. Three-quarters subsequently resumed injection drug
use, among whom median time to relapse was 1.0 year. Factors independently associated with a shorter time to cessation were: age <30 years, stable
housing, HIV seropositivity, methadone maintenance treatment, detoxification, abstinence from cigarettes and alcohol, injecting less than daily, not
injecting heroin and cocaine together, and not having an IDU sex partner. Factors independently associated with shorter time to injection relapse
were male gender, homelessness, HIV seropositivity, use of alcohol, cigarettes, non-injection cocaine, sexual abstinence and having a longer time
to the first cessation.
Conclusions: This study provides strong support for targeting cessation efforts among young IDUs and severely dependent, unstably housed, and
HIV-infected individuals.
© 2005 Elsevier Ireland Ltd. All rights reserved.
Keywords: Injection drug use; Cessation; Relapse; Methadone maintenance; HIV infection; Homelessness
1. Introduction
Injection drug use is associated with considerable social,
economic and medical consequences. Drug-related societal
and economic costs are attributable to losses in productivity
due to drug-related morbidity, premature mortality and crime
(Harwood et al., 1999). Injection drug users (IDUs) experience increased risk of developing endocarditis, cellulitis and
abscesses, and acquiring blood borne infections such as human
immunodeficiency virus (HIV) and hepatitis C (HCV) through
needle sharing. In Baltimore, MD, a prospective study of IDUs
found that within 1 year of initiating injection the prevalence
∗
Corresponding author. Tel.: +1 505 476 3607; fax: +1 505 827 0013.
E-mail address: nina.shah@state.nm.us (N.G. Shah).
0376-8716/$ – see front matter © 2005 Elsevier Ireland Ltd. All rights reserved.
doi:10.1016/j.drugalcdep.2005.11.007
of HIV, HBV and HCV was 14%, 50% and 65%, respectively
(Garfein et al., 1996).
Although several studies have shown that IDUs are capable of reducing their frequency of needle sharing, in settings
where HIV infection is prevalent, low levels of risk behaviors can give rise to HIV incidence rates as high as 3–4% per
year (van Ameijden and Coutinho, 1998). These observations
underscore the need to consider strategies to encourage cessation of injection as the primary goal of harm reduction a major
public health priority (Langendam et al., 2000; Stimson et al.,
1996).
For opioid-dependent persons, opiate agonist therapies such
as methadone maintenance (National Consensus Development
Panel on Effective Medical Treatment of Opiate Addiction,
1998; van Ameijden and Coutinho, 2001; Sullivan et al., 2005),
buprenorphine (Raisch et al., 2002) and heroin-assisted treat-
148
N.G. Shah et al. / Drug and Alcohol Dependence 83 (2006) 147–156
ment (Rehm et al., 2001) have been found to significantly reduce
drug use. However, in countries such as the US, drug users experience considerable barriers accessing drug treatment, including
the lack of third party reimbursement and lengthy waiting times
for limited existing drug treatment slots (Shah et al., 2000;
Lewis, 1999). Among populations of IDUs in North America,
only 18–36% are estimated to be receiving some form of drug
treatment at any given time (Joseph et al., 2000).
Relapse is a major characteristic of drug dependence and
remains a primary problem in treating drug abuse. Individuals
need to be engaged in drug treatment for an adequate length
of time in order to reap the greatest benefits (Goldstein et al.,
2000; Hartel and Schoenbaum, 1998). However, half of those
admitted to methadone maintenance treatment (MMT) programs
in the US leave within a year for a variety of reasons (Simpson
et al., 1997). Generally, drug users who are retained in MMT
for only a short time experience higher relapse rates than those
who spend a longer time engaged in treatment (Simpson, 1981;
Toumbourou et al., 1998). Many IDUs require multiple attempts
at drug treatment before becoming drug-free (Raisch et al., 2002;
Bammer et al., 2000).
The relapsing and remitting nature of drug dependence is
a major issue in the study of the natural history of drug use
(Galai et al., 2003). Few empirical investigations have described
predictors of time to injection cessation and relapse to injection
among well-characterized samples of out-of-treatment IDUs. In
the present study, we sought to examine predictors of time to first
injection cessation and, given cessation, to identify predictors
associated with time to subsequent relapse to injection. These
findings may help identify appropriate interventions and support
systems to reduce the negative health and social consequences
of injection drug use.
2. Methods
2.1. Population
The rationale, design and methods of the AIDS link to intravenous experiences (ALIVE) cohort study have been described
previously (Vlahov et al., 1991; Anthony et al., 1991). Briefly,
subjects were eligible to participate if they were at least 18 years
old, were AIDS-free at enrollment and had injected illicit drugs
within the past 10 years. Over 80% of subjects reported having been recruited by another study participant or friend, rather
than referral by a drug treatment program, public health program or trained street outreach worker. A total of 2946 IDUs
were enrolled into the ALIVE Study, though the present study
was restricted to 1327 participants who were active injectors
at baseline and had at least three follow-up visits, to permit
study of factors associated with injection cessation and subsequent relapse to injection drug use. That is, a minimum of three
visits after enrollment was needed for the participant to be “at
risk” for one cessation and one relapse. Recruitment began in
1988–1989 and follow-up for this analysis continued through
December 2000. Study visits were scheduled every 6 months
with ongoing effort to minimize loss to follow-up using mail
and phone reminders and street tracking.
2.2. Data collection
At study entry and semi-annually thereafter, participants
underwent venipuncture for HIV antibody testing and completed an interviewer-administered questionnaire on sociodemographic characteristics, risk behaviors for HIV infection,
health status indicators, and health service utilization. Sociodemographic variables of interest included gender, age, ethnicity,
employment, education, recent homelessness and history of
incarceration. Drug use and sexual risk behaviors referred to
the previous 6 months and included frequency of injection, needle sharing, overdose, shooting gallery attendance, non-injection
drug use, number and type of sexual partners, trading sex for
money or drugs, and unprotected sex. Data on hospital admissions, drug treatment and HIV-related symptoms were also collected at each assessment.
The baseline questionnaire had a number of unique questions
regarding the initiation of injection drug use, including age of
first injection and how long it took to reach the highest frequency
(peak) of use. Based on this information, the variable “time to
peak drug use” was created. The outcome of injection cessation
was defined as the first self-report of not having injected drugs
within the prior 6 months. Relapse was defined as the first selfreport of injection following cessation. Note that the exact time
of relapse within the prior 6-month interval was not known and
thus the event time for analysis was the date of the first report of
relapse to injection. Thus, the variability for injection cessation
was within a window of 6–12 months, and the time window for
relapse was 1 day to less than 6 months.
2.3. Statistical analyses
Characteristics of the study sample were described using
medians and inter-quartile range for continuous variables and
frequencies for categorical variables. The population included
in the present analysis was compared to those excluded by simple
χ2 -tests for categorical variables and t-tests or non-parametric
tests for continuous variables.
Two separate survival analyses were performed: time from
enrollment (where all participants reported injecting) to first
cessation of drug injection was assessed; then, time from cessation to first relapse to injection was assessed among those who
reported cessation and had at least one visit following cessation.
Factors affecting time to outcome (cessation or relapse)
were evaluated using parametric survival models, assuming a
log-normal distribution (accelerated failure time) (Hosmer and
Lemeshow, 1999). In the accelerated failure time model, the
effect of a covariate is expressed as either accelerating (shortening) or decelerating (lengthening) time to the event of interest.
The log-normal survival models assume that the hazard function
has a log-normal shape with two parameters (mean and dispersion) which are estimated from the data. The use of parametric
survival models allows a convenient calculation of the predicted
survival curve for any specific combination of explanatory variables as well as the underlying hazard. Further, covariate effects
are shown as time ratios (TR), rather than hazard ratios, and
hence have an interpretation on the time scale. For example,
N.G. Shah et al. / Drug and Alcohol Dependence 83 (2006) 147–156
a time ratio of 0.5 for a binary factor x, implies that the predicted survival time for the “high-risk” group (x = 1) is shorter
by 50% compared to the survival time for the “low-risk” group
(x = 0). The effects on survival are seen in all the percentiles of
survival time including quartiles and median. Similarly, values
of TR > 1 indicate a protective effect since time to event in the
x = 1 group is longer than the time for the group with x = 0. Note
that factors usually termed “risky” or harmful will be associated
with TR > 1 for the injection cessation analysis but TR < 1 for
the relapse to injection outcome. The overall model fit, assuming the log-normal shape of the hazard function, was assessed
using Cox–Snell residuals and found acceptable (Cox and Snell,
1968).
To ensure that predictors preceded the reported periods of
injection cessation and relapse to injection, the analyses used
data lagged one semi-annual visit. Covariates with statistical significance at levels α ≤ 0.10 in univariate models were entered
into exploratory multivariate models for the respective outcomes. The final multivariate models were developed using a
manual stepwise backward procedure.
3. Results
Of the 2946 participants enrolled in the ALIVE Study, 2232
had at least one follow-up visit. Of these, a total of 905 participants were determined ineligible for this study, where 425 were
not actively injecting at baseline, 401 were actively injecting at
baseline but had less than three follow-up visits, and 79 were
missing baseline data for injection. Ineligible participants differed significantly from eligible participants as they were older
(median age of 35.0 versus 34.0), more likely to be male (83.1%
versus 77.2%), non-African American (14.6% versus 5.2%),
HIV-uninfected (72.3% versus 62.3%), employed (25.1% versus 20.8%), used drugs less often (21.7% injected daily versus
51.4%), and a smaller proportion engaged in risky sexual (54.8%
versus 75.3% reported unprotected sex; 35.4% versus 47.5%
reported sex with an IDU partner) and drug-using behaviors
(29.1% versus 85.2% reported sharing works/needles; 10.8%
versus 23.7% reported shooting gallery attendance), though no
differences were found by history of incarceration and recent
sexual activity. Overall, ineligible participants had a more stable lifestyle and were not as severely addicted as the eligible
participants at baseline.
Of the 1327 IDUs eligible for analysis at baseline, 94.8%
were African American, 77.2% were male, 42.2% were high
school graduates, 79.2% were unemployed and 37.7% were
HIV-infected. The median age was 34 (1st, 3rd inter-quartile
range (IQR): 30, 38 years). The types of drugs injected during the
6 months prior to enrollment were heroin alone (71.6%), cocaine
alone (81.4%), and heroin and cocaine together, referred to as
speedball (69.2%). Nearly one-fifth (18.3%) gave a recent history (prior 6 months) of drug detoxification at baseline, and 9.7%
had recently been receiving methadone treatment. The median
number of visits per year for participants over the follow-up
period was 1.8 (mean = 1.7 visits).
Fig. 1 shows the crude time-to-event graphs for first injection
cessation and subsequent relapse to injection drug use. Of the
149
Fig. 1. Kaplan–Meier of observed time from baseline to first cessation of injection and given cessation, time to first relapse to injection, in the ALIVE Study,
Baltimore, MD, 1988–2000.
1327 IDUs contributing 5553 person-years of observation, 936
(70.5%) reported at least one episode of cessation of injection
drug use corresponding to a cumulative incidence rate of 16.8 per
100 person years. Note that although the total follow-up period
was 12 years, the person-time used in the analysis was censored
at the first occurrence of the event. The crude estimated median
time from baseline to first cessation was 4.0 years (1st, 3rd IQR:
1.7, 8.5 years). Of the initial group, 40 (3.0%) IDUs died before
the median time to cessation (mid-1992) and did not report a single cessation episode, and 106 (8%) were lost to follow-up (did
not return after 1 January 1999 but were not deceased according to the National Death Index records). Among the 936 IDUs
reporting cessation of injection drug use, 898 (95.9%) had at
least one follow-up visit following cessation and were thus eligible for the analysis of predictors of relapse to injection. Of the
898 persons contributing 1727 person-years over the follow-up
period, 678 (75.5%) reported at least one subsequent episode of
relapse to injection, corresponding to a cumulative relapse incidence rate of 39.2 per 100 person years. The crude estimated
median time from injection cessation to first relapse to injection
was 1.0 year (1st, 3rd IQR: 0.5, 3.2 years). Of the IDUs reporting
cessation, 83 (9.2%) died thereafter with no report of subsequent
relapse.
3.1. Univariate associations
3.1.1. Cessation of injection. Factors found to be associated
with shorter time to cessation included age younger than 30
years (time-ratio (TR) = 0.69), being employed (TR = 0.84),
being HIV-infected (TR = 0.78), being admitted to a hospital
(TR = 0.78), reporting MMT (TR = 0.58) and detoxification
(TR = 0.70) (Tables 1 and 2). In contrast, median time to
cessation was longer for African Americans compared to
non-African Americans by 34% (TR = 1.34), by 44% for those
reporting recent homelessness (TR = 1.44), by 21% for those
with a lengthy duration of injection drug use (>10 years,
TR = 1.21), by 50% for smoking cigarettes (TR = 1.50), 59% for
alcohol use (TR = 1.59), 31% for sharing needles (TR = 1.31),
48% for drug overdose (TR = 1.48), 73% for injecting speedball
150
N.G. Shah et al. / Drug and Alcohol Dependence 83 (2006) 147–156
Table 1
Time from baseline to first cessation of injection among 1327 active injection drug users in the ALIVE Study, Baltimore, MD, 1988–2000 (univariate associations
for baseline characteristics and time-dependent data based on log-normal survival models: time ratios and median years to first cessation)
Person-years
Time to cessation of injection (years)
Time ratioa (95% confidence interval)
Median
5301
251
1.34 (1.00, 1.78)
1.0
3.8
2.9
Age groupings
<30 years
30–34 years
35–39 years
≥40 years
515
1115
1519
2404
0.69 (0.57, 0.85)
0.87 (0.73, 1.03)
0.98 (0.83, 1.16)
1.0
2.9
3.6
4.1
4.1
Employed at baseline
Yes
No
1094
4452
0.84 (0.72, 0.99)
1.0
3.3
3.9
Duration of injection at baseline
>10 years
≤10 years
3539
2014
1.21 (1.06, 1.39)
1.0
4.1
3.4
2300
3253
0.78 (0.68, 0.89)
1.0
3.3
4.2
4170
807
1.44 (1.18, 1.74)
1.0
5.4
3.8
Baseline characteristics
Race
African American
Non-African American
Time-dependent datab
HIV status, past 6 months
HIV-infected
HIV-uninfected
Homelessness, past 6 months
Yes
No
a
b
Note that TR > 1 represents longer time to cessation of injection and hence “higher risk”.
Time-dependent variables concurrent with the outcome.
compared to heroin or cocaine alone (TR = 0.1.73) and 79% for
daily injection (TR = 1.79).
For sexual behaviors, the estimated median time to injection
cessation was prolonged for those reporting anonymous sex partners by 58% (compared to sexual abstinence, TR = 1.58), IDU
sex partners by 53% (compared to sexual abstinence, TR = 1.53),
trading sex by 34% (TR = 1.34) and for those reporting unprotected sex by 18% (TR = 1.18). However, men having sex with
men had a shorter time to cessation (TR = 0.65). There was no
effect on time to cessation of injection drug use for gender, education, history of incarceration, age of first injection, time to peak
drug use, use of a needle exchange program, shooting gallery
attendance, illicit non-injection drug use or HIV-related symptoms (data not shown).
3.1.2. Relapse to injection. The estimated median time from
cessation of injection to first relapse to injection was significantly shorter for persons who were HIV-infected (TR = 0.80),
who reported recent homelessness (TR = 0.54) and a history of
incarceration (TR = 0.70) (Table 3). Considering substance use,
smoking cigarettes shortened median time to relapse by 36%
(TR = 0.64), 29% for alcohol use (TR = 0.71), 31% and 38% for
non-injection cocaine and heroin use, respectively (TR = 0.69
and 0.62). Regarding initiation of drug injection, older age
at first injection was associated with longer time to relapse
(TR = 1.07 per 5 years). On the other hand, the estimated median
time from cessation to relapse was extended for females by
22% (TR = 1.22), 40% for African Americans (TR = 1.40) and
31% for those who experienced less than 1.5 years to first cessation (TR = 1.31). For sexual behaviors, those reporting sex,
but not with an IDU or anonymous partner, was estimated to
lengthen time to relapse by 34% and 30%, respectively (compared to sexual abstinence, TR = 1.34 and 1.30). There were no
statistical differences in estimated time to relapse to injection
for age, education, health care utilization, duration of injection drug use at baseline, engaging in sex trade, unprotected
sex, using marijuana, crack or HIV-related symptoms (data not
shown).
As an example, Fig. 2 panel shows the crude observed timeto-event graphs for homelessness and time to injection cessation and subsequent relapse to injection drug use. The time
to cessation for IDUs reporting homelessness was 44% longer
(TR = 1.44) than for those who reported stable housing. The
time to relapse to injection was nearly 50% faster (TR = 0.54)
for those reporting homelessness relative to those who did not.
This example shows the “harmful” effects of being homeless
for both outcomes, where homelessness lengthened the time to
injection cessation (TR > 1) and accelerated the time to relapse
to injection (TR < 1).
3.2. Multivariate models
Table 4 presents results from the multivariate models.
Independent predictors of longer time to first cessation of
151
N.G. Shah et al. / Drug and Alcohol Dependence 83 (2006) 147–156
Table 2
Time from baseline to first cessation of injection among 1327 active injection drug users in the ALIVE Study, Baltimore, MD, 1988–2000 (univariate associations
for lagged time-dependent data based on log-normal survival models: time ratios and median years to first cessation)
Person-years
Time to cessation of injection (years)
Time ratioa (95% confidence interval)
Median
1195
4353
0.78 (0.67, 0.90)
1.0
3.1
4.0
Methadone maintenance treatment
No methadone maintenance treatment
601
4946
0.58 (0.47, 0.71)
1.0
2.3
4.1
Detoxification program
No detoxification program
759
4783
0.70 (0.59, 0.83)
1.0
2.8
4.0
Injection behavior
Needle sharing
No needle sharing
3834
1319
1.31 (1.11, 1.55)
1.0
4.4
3.4
Drug overdose
No drug overdose
312
3434
1.48 (1.12, 1.96)
1.0
5.6
3.8
Daily injection frequency
Less often than daily injection
2928
2625
1.79 (1.57, 2.04)
1.0
5.2
2.9
Injected speedball
Injected cocaine or heroin alone
4187
1363
1.73 (1.52, 1.99)
1.0
4.5
2.6
5151
400
1.50 (1.19, 1.88)
1.0
3.9
2.6
4541
1007
1.59 (1.35, 1.87)
1.0
4.2
2.6
4347
1202
1.27 (1.08, 1.51)
1.0
4.0
3.1
No sex partners
One sex partner
Two or more sex partners
1478
2226
1848
1.0
1.15 (0.96, 1.37)
1.32 (1.10, 1.59)
3.2
3.7
4.3
Sex with anonymous partner
Sex but not with anonymous partner
No sex
662
3686
1202
1.58 (1.24, 2.00)
1.23 (1.03, 1.45)
1.0
4.9
3.8
3.1
Sex with IDU partner
Sex with non-IDU partner
No sex
2264
2083
1202
1.53 (1.28, 1.84)
1.07 (0.89, 1.28)
1.0
4.8
3.3
3.1
Trading sex
No trading sex
1124
4429
1.34 (1.14, 1.59)
1.0
4.8
3.6
Any unprotected sex
Always protected sex
3061
2491
1.18 (1.02, 1.35)
1.0
4.0
3.4
MSM sex
No MSM sex
283
5270
0.65 (0.47, 0.91)
1.0
2.5
3.8
Lagged time-dependent datab
Health care utilization
Hospital admission
No hospital admission
Non-injection drug use
Cigarette use
No cigarette use
Alcohol use
No alcohol use
Sexual behavior
Any sex
No sex
a
b
Note that TR > 1 represents longer time to cessation of injection and hence “higher risk”.
To ensure that predictors preceded the reported period of injection cessation, the analyses used data lagged one semi-annual visit.
injection drug use included recent homelessness (adjusted
(adj) TR = 1.36), daily injection (adj TR = 1.55), cigarette
smoking (adj TR = 1.49), alcohol use (adj TR = 1.33), having an IDU sexual partner (compared to sexual abstinence,
adj TR = 1.37) and injecting speedball (adj TR = 1.39). Conversely, factors predictive of a shorter time to cessation of
injection included age less than 30 years (adj TR = 0.79),
being HIV-infected (adj TR = 0.83), and enrollment in
MMT (adj TR = 0.70) and detoxification programs (adj TR =
0.61).
Given cessation of injection, independent predictors of
a longer duration from cessation to subsequent relapse to
152
N.G. Shah et al. / Drug and Alcohol Dependence 83 (2006) 147–156
Table 3
Time from first cessation of injection to relapse to injection among 898 injection drug users who reported a cessation in the ALIVE Study, Baltimore, MD, 1988–2000
(univariate associations based on log-normal survival models: time ratios and median years from first cessation to relapse)
Person-years
Time ratioa (95% confidence interval)
Median
455
1272
1.22 (1.01, 1.46)
1.0
1.60
1.31
1660
67
1.40 (1.00, 1.94)
1.0
1.40
1.00
990
736
0.70 (0.60, 0.83)
1.0
1.21
1.73
–
1.07 (1.00, 1.15)
392
833
502
1.17 (0.94, 1.44)
1.19 (1.00, 1.43)
1.0
1.43
1.46
1.23
712
1015
1.31 (1.11, 1.54)
1.0
1.64
1.26
736
991
0.80 (0.68, 0.93)
1.0
1.22
1.53
104
1255
0.54 (0.43, 0.68)
1.0
0.76
1.42
1413
313
0.64 (0.50, 0.82)
1.0
1.30
2.04
Alcohol use
No alcohol use
780
947
0.71 (0.61, 0.83)
1.0
1.17
1.64
Heroin use
No heroin use
97
1629
0.69 (0.51, 0.93)
1.0
0.98
1.41
Cocaine use
No cocaine use
54
1671
0.62 (0.46, 0.85)
1.0
0.88
1.41
1256
471
1.30 (1.10, 1.53)
1.0
1.49
1.15
Sex with anonymous partner
Sex but not with anonymous partner
No sex
111
1145
471
1.31 (0.95, 1.82)
1.30 (1.09, 1.54)
1.0
1.51
1.49
1.15
Sex with IDU partner
Sex with non-IDU partner
No sex
248
1008
471
1.17 (0.92, 1.49)
1.34 (1.12, 1.52)
1.0
1.35
1.53
1.15
Baseline characteristics
Gender
Female
Male
Race
African American
Non-African American
History of incarceration at baseline
Yes
No
Injection initiation
Age of first injection (per 5 year increase)
Time to peak use
Quick peak, within 1 year
1–5 years
≥6 years
<1.5 years to first cessation
≥1.5 years to first cessation
Time-dependent datab
HIV status, past 6 months
HIV-infected
HIV-uninfected
Homelessness, past 6 months
Yes
No
Lagged time-dependent datac
Non-injection drug use
Cigarette use
No cigarette use
Sexual behavior
Any sex
No sex
a
b
c
Time to relapse to injection (years)
Note that TR > 1 represents longer time to relapse to injection and hence “lower risk”.
Time-dependent variables concurrent with the outcome.
To ensure that predictors preceded the reported period of relapse to injection, the analyses used data lagged one semi-annual visit.
injection were female gender (adj TR = 1.22), non-IDU sexual partner (compared to sexual abstinence, adj TR = 1.26)
and less than 1.5 years from study enrollment to cessation
(adj TR = 1.43). Independent predictors of a shorter median time
from injection cessation to relapse to injection were being HIVinfected (adj TR = 0.83), recently homeless (adj TR = 0.59), use
of cigarettes (adj TR = 0.73), alcohol (adj TR = 0.76) and noninjection cocaine (adj TR = 0.46).
N.G. Shah et al. / Drug and Alcohol Dependence 83 (2006) 147–156
153
Table 4
Multivariate associations based on log-normal survival models for time from baseline to first cessation of injection and given cessation, time to first relapse to
injection, in the ALIVE Study, Baltimore, MD, 1988–2000: adjusted time ratios with 95% confidence intervals (95% C.I.)
Female
Age <30 years
HIV-infected
Homelessness
Methadone maintenance treatment
Detoxification program
Daily injection
Injected speedball
Cigarette use
Alcohol use
Non-injection cocaine use
Sex with IDU partnerb
Sex with non-IDU partnerb
<1.5 years to first cessation
a
b
Cessation of injection adjusted time ratio (95% C.I.)
Relapse to injection adjusted time ratio (95% C.I.)
NSa
0.79 (0.65, 0.94)
0.83 (0.73, 0.96)
1.36 (1.12, 1.65)
0.70 (0.56, 0.86)
0.61 (0.51, 0.72)
1.55 (1.35, 1.79)
1.39 (1.20, 1.62)
1.49 (1.17, 1.88)
1.33 (1.12, 1.57)
NSa
1.37 (1.13, 1.65)
1.07 (0.89, 1.29)
N/A
1.22 (1.02, 1.46)
NSa
0.83 (0.72, 0.97)
0.59 (0.47, 0.73)
N/A
N/A
N/A
N/A
0.73 (0.58, 0.92)
0.76 (0.66, 0.89)
0.62 (0.46, 0.84)
1.18 (0.93, 1.49)
1.26 (1.06, 1.49)
1.43 (1.23, 1.68)
NS indicates not statistically significant.
Reference group are those reporting sexual abstinence in past 6 months.
Fig. 2. Effect of being homeless on time to cessation and subsequent time to
relapse: crude observed time-to-event from baseline to first cessation of injection and given cessation, time to first relapse to injection, in the ALIVE Study,
Baltimore, MD, 1988–2000.
4. Discussion
An encouraging finding of the study was that 71% of active
IDUs in the cohort were successful in abstaining from drug
injection for at least 6 months at some point during follow-up;
however, of these, 76% relapsed, half within just 1 year. IDUs
participating in this study routinely received drug and HIV risk
reduction counseling at the end of each visit (i.e., after questionnaire was completed), assisted referrals for treatment and learned
about harm reduction strategies at each study visit. Despite these
efforts, only half of the cohort reported a first 6-month interval
of non-injection drug use within 4 years of enrollment. This
is a considerably long time frame during which participants
continued to engage in risky behaviors with major health and
social consequences. Similar difficulties have been observed in
Amsterdam where long-standing harm reduction strategies exist,
but risky injection behaviors such as borrowing/sharing needles
have proven to be difficult to eliminate (van Ameijden et al.,
1998).
An important finding is that IDUs younger than 30 years had
a shorter time to cessation compared to older IDUs. At the study
baseline, older IDUs had been injecting for an average of 17
years, compared to 7 years for IDUs younger than 30 years old.
A hypothesis is that younger IDUs were less heavily entrenched
in the drug-using lifestyle, and less likely to have established stable injector partnerships and consistent drug sources, compared
to older IDUs. Despite challenges inherent in working with more
recent initiates (Doherty et al., 2000), younger drug users are an
important group to reach. Not surprisingly, IDUs with a more
severe dependence, indicated by injecting at least daily or using
speedball, had an almost 50% longer time to injection cessation than those with less severe dependence. These results are
consistent with other studies (Colón et al., 2001; Hemby et al.,
1999).
Participation in drug treatment was associated with shorter
time to injection cessation, consistent with numerous other studies. The use of opioid agonist therapies reduce heroin withdrawal
symptoms and craving, thus reducing injection frequency (Booth
et al., 1996; Kwiatkowski and Booth, 2001; Shore et al., 1996;
Gossop et al., 2002; Metzger and Navaline, 2003) or result in
complete abstinence (Langendam et al., 2000). Unfortunately,
information on drug treatment was not consistently collected
from this community sample during periods of non-injection, so
we were not able to assess the overall effect of drug treatment
154
N.G. Shah et al. / Drug and Alcohol Dependence 83 (2006) 147–156
on time to relapse to injection. Although it is generally recognized that detoxification alone is inadequate as a treatment for
opiate addiction, IDUs attending a detoxification program had
a faster time to cessation than those who did not, after adjusting
for enrollment in MMT. This could be a reflection of these IDUs’
stronger motivation for quitting drug use, known to be an important predictor of cessation (Simpson and Joe, 1993; Simpson et
al., 1997).
It is important to note that although IDUs may stop drug use
by injection, they may actively continue non-injection drug use.
In fact, 34% (n = 302) of IDUs who reported injection cessation subsequently used non-injection heroin and/or cocaine, and
13% (n = 117) did so within the first year following cessation.
Notably, non-injection cocaine use predicted a more rapid time
to relapse to injection. It is possible that among out-of-treatment
IDUs, efforts to sustain abstinence from injection are weakened
by non-injection cocaine use.
IDUs reporting unstable housing, alcohol and cigarette use
had a longer time to cessation and a shorter time to subsequent relapse. Unstable housing can impede opportunities for
achieving injection cessation and create an unfavorable setting
to remain drug-free (Song et al., 2000; Celentano et al., 1991).
Similarly, the effects of alcohol and cigarette use showed that
addiction is a multi-faceted disease and clinicians should address
nicotine and alcohol use in the opiate-dependence recovery process.
Interestingly, HIV-infected persons had a shorter time to first
injection cessation but also a shorter time to subsequent relapse.
This suggests that IDUs may be particularly motivated to stop
injecting after learning of their HIV-positive status (Celentano
et al., 1994), indicating the need to target referrals to drug treatment among this vulnerable population. A previous study in
this cohort and other research has found that HIV-infected IDUs
transition between periods of non-injection and injection more
often than HIV-uninfected IDUs (Galai et al., 2003; Rehm et
al., 2001). Studies show that HIV-infected IDUs lack adequate
social support and coping abilities (Avants et al., 2001; Belding
et al., 1996), indicating the need for wide-spectrum interventions, especially considering the risk of HIV transmission.
A noteworthy finding in this study was that while sociodemographic and drug-using behaviors were predictive of cessation
and relapse, sexual partnerships were also important. Compared
to IDUs reporting no sex in the previous 6 months, IDUs who
reported having an IDU sexual partner had a longer time to cessation, while those who reported non-IDU sex partners during
periods of injection abstinence had a longer time to subsequent
relapse. Social networks have shown to influence injection cessation. IDUs who report fewer personal network members with
whom they used drugs were more likely to stop injecting than
those with larger, drug-related networks (Latkin et al., 1999;
Schroeder et al., 2001). IDUs tend to select sexual partners and
form emotional bonds with persons from within their network
(Johnson et al., 2002) and who may interact in similar social
scenes (Friedman et al., 1999). Sexual activity with a non-IDU
partner lengthened time to relapse; intimate relationships outside
the drug-using network may have tacitly supported abstinence
(Sibthorpe and Lear, 1994). Companionship outside the drug-
using network also provides a supportive environment to prevent
relapsing to injection.
A longer time to relapse was associated with being female,
possibly reflecting stronger social support, presence of familial
obligations (Lundy et al., 1995) and sexual partnership with nondrug users (Warner et al., 2004). Social psychological research
finds that social networks among women are particularly salient
and interventions such as participation in peer support groups
during periods of abstinence may be more beneficial to women
than men (DiNitto et al., 2002). The gender difference for
relapse should be further investigated and replicated in other
contexts.
Several study limitations need to be addressed. First, injection frequency was self-reported; report of any injection versus
none is less likely to be subject to recall bias compared to
injection frequency. However, we and others have found selfreported drug use to be both accurate and reliable (Latkin et
al., 1993; Safaeian et al., 2001; McElrath et al., 1994). As mentioned, joining this cohort study also reflected entry into risk
reduction. The median duration of injection for IDUs at study
baseline was 14 years and considering the long history of drug
use prior to study enrollment, the first reported cessation during
the follow-up period might not have been the first time that participants were able to stop using drugs by injection. Third, results
from this study may be most applicable to inner-city, predominantly African American IDUs. Lastly, participants who were
excluded initially or later lost to follow-up were more likely to be
HIV-uninfected and less severely addicted at baseline than those
included and not lost to follow-up. It is possible that IDUs who
were HIV-uninfected and not as severely drug-dependent were
somewhat underrepresented in this sample; the proportion lost
to follow-up was small (8%) and likely did not unduly affect our
findings.
In conclusion, this analysis showed that there are a number
of behavioral and social factors affecting the time it takes IDUs
to stop injection and relapse. Even among severely dependent
IDUs, the majority was successful in achieving injection abstinence for at least 6 months. Based on the results, we can better
understand the circumstances affecting the time to cessation and
resuming injection: (1) Severely dependent IDUs, users of other
substances (alcohol, cigarettes, cocaine) and homeless persons
should be targeted since they take longer, on average, to stop
injection and are faster to relapse; (2) Younger IDUs and being
HIV-infected may warrant special attention in cessation efforts;
(3) Drug network peers and sexual partners of IDUs should be
encouraged to jointly attend treatment in an effort to expand
programs to meet the needs of participants, promote injection
cessation and improve the sustainability of recovery.
Acknowledgments
The authors are indebted to the ALIVE participants for
imparting their experiences. We also thank the ALIVE Study
staff for their generous time and support in conducting the study.
This research was funded by grants DA04334 and DA12568
from the National Institute on Drug Abuse.
N.G. Shah et al. / Drug and Alcohol Dependence 83 (2006) 147–156
References
Anthony, J.C., Vlahov, D., Nelson, K.E., 1991. New evidence on intravenous
cocaine use and the risk of infection with human immunodeficiency virus
type 1. Am. J. Epidemiol. 134, 1175–1189.
Avants, S.K., Warburton, L.A., Margolin, A., 2001. How injection drug users
coped with testing HIV-seropositive: implications for subsequent healthrelated behaviors. AIDS Educ. Prev. 13, 207–218.
Bammer, G., Battison, L., Ward, J., Wilson, S., 2000. The impact of retention
of expansion of an Australian public methadone program. Drug Alcohol
Depend. 58, 173–180.
Belding, M.A., Iguchi, M.Y., Lamb, R.J., Lakin, M., Terry, R., 1996. Coping
strategies and continued drug sue among methadone maintenance patients.
Addict. Behav. 21, 389–401.
Booth, R.E., Crowley, T.J., Zhang, Y., 1996. Substance abuse treatment entry,
retention and effectiveness: out-of-treatment opiate injection drug users.
Drug Alcohol Depend. 2, 11–20.
Celentano, D.D., Vlahov, D., Cohn, S., Anthony, J.C., Solomon, L., Nelson,
K.E., 1991. Risk factors for shooting gallery use and cessation among
intravenous drug users. Am. J. Publ. Health 81, 1291–1295.
Celentano, D.D., Munox, A., Cohn, S., Nelson, K.E., Vlahov, D., 1994. Drugrelated behavior change for HIV transmission among American injection
drug users. Addiction 89, 1309–1317.
Colón, H.M., Robles, R.R., Deren, S., Sahai, H., Finlinson, H.A., Andı́a, J.,
Cruz, M.A., Kang, S.Y., Oliver-Vélez, D., 2001. Between-city variation
in frequency of injection among Puerto Rican injection drug users: East
Harlem, NY and Bayamon, PR. J. Acquir. Immune Defic. Syndr. 27,
405–413.
Cox, D.R., Snell, E.J., 1968. Analysis of Survival Data. Chapman & Hall,
London.
DiNitto, D.M., Webb, D.K., Rubin, A., 2002. Gender differences in duallydiagnosed clients receiving chemical dependency treatment. J. Psychoactive Drugs 34, 105–117.
Doherty, M.C., Garfein, R.S., Monterroso, E., Brown, D., Vlahov, D., 2000.
Correlates of HIV infection among young adult short-term injection drug
users. AIDS 14, 717–726.
Friedman, S., Curtis, R., Neaigus, A., Jose, B., Des Jarlais, D., 1999. Social
Networks, Drug Injectors’ Lives and HIV/AIDS. Kluwer Academic, New
York.
Galai, N., Safaeian, M., Vlahov, D., Bolotin, A., Celentano, D.D., ALIVE
Study, 2003. Longitudinal patterns of drug injection behavior in the
ALIVE Study cohort, 1988–2000: description and determinants. Am. J.
Epidemiol. 158, 695–704.
Garfein, R.S., Vlahov, D., Galai, N., Doherty, M.C., Nelson, K.E., 1996. Viral
infections in short-term injection drug users: the prevalence of the hepatitis C, hepatitis B, human immunodeficiency, and human T-lymphotropic
viruses. Am. J. Publ. Health 86, 655–661.
Goldstein, M.F., Deren, S., Magura, S., Kayman, D.J., Beardsley, M., Tortu,
S., 2000. Cessation of drug use: impact of time in treatment. J. Psychoactive Drugs 32, 305–310.
Gossop, M., Marsden, J., Stewart, D., Treacy, S., 2002. Change and stability
of change after treatment of drug misuse: 2-year outcomes from the
National Treatment Outcome Research Study (UK). Addict. Behav. 27,
155–166.
Hartel, D.M., Schoenbaum, E.E., 1998. Methadone treatment protects against
HIV infection: two decades of experience in the Bronx, New York City.
Publ. Health Rep. 113, 107–113.
Harwood, H.J., Fountain, D., Fountain, G., 1992. Economic cost of alcohol
and drug abuse in the United States, 1992. Addiction 94, 631–635.
Hemby, S.E., Co, C., Dworkin, S.I., Smith, J.E., 1999. Synergistic elevations in nucleus accumbens extracellular dopamine concentrations during
self-administration of heroin/cocaine combinations (speedballs) in rats. J.
Pharmacol. Exp. Ther. 288, 274–280.
Hosmer Jr., D.W., Lemeshow, S., 1999. Applied Survival Analysis:
Regression Modeling of Time to Event Data. John Wiley and Sons,
New York.
Johnson, R.A., Gerstein, D.R., Pach III, A., Cerbone, F.G., Brown, J., 2002.
HIV risk behaviors in African-American drug injector networks: implica-
155
tions of injection-partner mixing and partnership characteristics. Addiction
97, 1011–1024.
Joseph, H., Stancliff, S., Landgrod, J., 2000. Methadone maintenance treatment (MTT): a review of historical and clinical issues. Mt. Sinai J. Med.
67, 347–364.
Kwiatkowski, C.F., Booth, R.E., 2001. Methadone maintenance as HIV risk
reduction with street-recruited injecting drug users. J. Acquir. Immune
Defic. Syndr. 26, 483–489.
Langendam, M.W., Van Brussel, G.H.A., Coutinho, R.A., van Ameijden,
E.J.C., 2000. Methadone maintenance and cessation of injection drug
use: results from the Amsterdam cohort study. Addiction 95, 591–600.
Latkin, C.A., Vlahov, D., Anthony, J.C., 1993. Socially desirable responding
and self-reported HIV infection risk behaviors among intravenous drug
users. Addiction 88, 517–526.
Latkin, C.A., Knowlton, A.R., Hoover, D., Mandell, W., 1999. Drug network characteristics as a predictor of cessation of drug use among adult
injection drug users: a prospective study. Am. J. Drug Alcohol Abuse 25,
463–473.
Lewis, D.C., 1999. Access to narcotic addiction treatment and medical care:
prospects for the expansion of methadone maintenance treatment. J.
Addict. Dis. 18, 5–21.
Lundy, A., Gottheil, E., Serota, R.D., Weinstein, S.P., Sterling, R.C., 1995.
Gender differences and similarities in African American crack cocaine
abusers. J. Nerv. Ment. Dis. 183, 260–266.
McElrath, K., Chitwood, D.D., Griffin, D.K., Comerford, M., 1994. The consistency of self-reported HIV risk behavior among injection drug users.
Am. J. Publ. Health 84, 1965–1970.
Metzger, D.S., Navaline, H., 2003. HIV prevention among injection drug
users: the need for integrated models. J. Urban Health 80 (4 Suppl. 3),
iii.59–iii.66.
National Consensus Development Panel on Effective Medical Treatment of
Opiate Addiction, 1998. Effective medical treatment of opiate addiction.
JAMA 280, 1936–1943.
Raisch, D.W., Fye, C.L., Boardman, K.D., Sather, M.R., 2002. Opioid dependence treatment, including buprenorphine/naloxone. Ann. Pharmacother.
36, 312–321.
Rehm, J., Gschwend, P., Steffen, T., Gutzwiller, F., Dobler-Mikola, A., Uchtenhagen, A., 2001. Feasibility, safety, and efficacy of injectable heroin
prescription for refractory opioid addicts: a follow-up study. Lancet 358,
1417–1420.
Safaeian, M., Brookmeyer, R., Vlahov, D., Latkin, C., Marx, M., Strathdee,
S.A., 2001. Validity of self-reported needle exchange attendance among
injection drug users: implications for program evaluation. Am. J. Epidemiol. 155, 169–175.
Schroeder, J.R., Latkin, C.A., Hoover, D.R., Curry, A.D., Knowlton, A.R.,
Celentano, D.D., 2001. Illicit drug use in one’s social network and in one’s
neighborhood predicts individual heroin and cocaine use. Ann. Epidemiol.
11, 389–394.
Shah, N.G., Celentano, D.D., Vlahov, D., Stambolis, V., Johnson, L., Nelson,
K.E., Strathdee, S.A., 2000. Correlates of enrolment in methadone maintenance treatment programs differ by HIV-serostatus. AIDS 14, 2035–
2043.
Shore, R.E., Marmor, M., Titus, S., Des Jarlais, D.C., 1996. Methadone
maintenance and other factors associated with intraindividual temporal
trends in injection-drug use. J. Subst. Abuse Treat. 13, 241–248.
Sibthorpe, B., Lear, B., 1994. Circumstances surrounding needle use transitions among injection drug users: implications for HIV intervention. Int.
J. Addict. 29, 1245–1257.
Simpson, D.D., 1981. Treatment for drug abuse. Follow-up outcome and
length of time spent. Arch. Gen. Psychiatr. 38, 875–880.
Simpson, D.D., Joe, G.W., 1993. Motivation as a predictor of early dropout
from drug abuse treatment. Psychotherapy 30, 357–367.
Simpson, D.D., Joe, G.W., Rowan-Szal, G.A., 1997. Drug abuse treatment
and process effects on follow-up outcomes. Drug Alcohol Depend. 47,
227–235.
Song, J.Y., Safaeian, M., Strathdee, S.A., Vlahov, D., Celentano, D.D., 2000.
The prevalence of homelessness among injection drug users with and
without HIV infection. J. Urban Health 77, 678–687.
156
N.G. Shah et al. / Drug and Alcohol Dependence 83 (2006) 147–156
Stimson, G.V., Adelekan, M., Rhodes, T., 1996. The diffusion of drug injecting in developing countries. Int. J. Drug Policy 7, 245–255.
Sullivan, L.E., Metzger, D.S., Fudala, P.J., Fiellin, D.A., 2005. Decreasing
international HIV transmission: the role of expanding access to opioid
agonist therapies for injection drug users. Addiction 100, 150–158.
Toumbourou, J.W., Hamilton, B., Fallon, B., 1998. Treatment level progress
and time spent in treatment in the prediction of outcomes following drugfree therapeutic community treatment. Addiction 93, 1054–1064.
van Ameijden, E.J.C., Coutinho, R.A., 1998. Maximum impact of HIV prevention measures targeted at injection drug user. AIDS 12, 625–633.
van Ameijden, E.J.C., Coutinho, R.A., 2001. Large decline in injecting drug
use in Amsterdam, 1986–1998: explanatory mechanisms and determinants
of injecting transitions. J. Epidemiol. Community Health 55, 356–363.
Vlahov, D., Anthony, J.C., Munoz, A., Margolick, J., Nelson, K.E., Celentano,
D.D., Solomon, L., Polk, B.F., 1991. The ALIVE Study, a longitudinal
study of HIV-1 infection in intravenous drug users: description of methods
and characteristics of participants. NIDA Res. Monogr. 109, 75–100.
Warner, L.A., Alegria, M., Canino, G., 2004. Remission from drug dependence symptoms and drug use cessation among women drug users in
Puerto Rico. Arch. Gen. Psychiatr. 61, 1034–1041.