Many users of cocaine also use alcohol. A national household survey found that [is greater than]
90% of those respondents who were currently using cocaine also reported current use of alcohol (1). Fischman and Johanson
(2) have reported that from 60% to 90% of persons who abuse cocaine also abuse alcohol. Many studies involving individuals
with cocaine dependency found a high proportion of subjects who also met the diagnostic criteria for alcohol dependence (3-7).
In research by Miller et al. (3), patients diagnosed with cocaine dependence, using DSM-III-R criteria, showed a high prevalence
of alcohol and cannabis dependence as well.
The combination of cocaine and alcohol is physiologically more hazardous than the use of either
substance alone. Two independent studies conducted by Farre et al. (8,9) found that when compared with the effects of cocaine
use alone, the combined effects of cocaine and alcohol produced greater increases in heart rate, which could raise the risk
of cardiovascular toxicity linked with cocaine use. Although excessive use of either cocaine or alcohol can result in severe
liver damage, toxicity is greatly increased when these two substances are used together (10, 11).
Research has described the differences in demographics, substance use, and psychopathology variables
among persons dependent on cocaine alone and persons dependent on cocaine complicated by alcohol use (12). These researchers
found that those subjects who used both alcohol and cocaine had significantly higher depression and global severity scores
and were more likely to experience paranoid psychosis with cocaine use than those who only used cocaine. Users of both cocaine
and alcohol were also more likely to have abused additional substances during the month before the study than users of cocaine
only.
Research has also shown that alcohol may serve as a barrier to recovery from cocaine addiction.
In a study by Brown et al.(13) there was a significantly greater proportion of relapse at post-treatment assessment among
users of both cocaine and alcohol when compared with those who only used alcohol. Higgins and his colleagues (14) also postulated
that alcohol use may serve as both a trigger for cocaine use and a barrier to recovery. They suggested testing the efficacy
of disulfiram therapy in patients who report high levels of simultaneous cocaine and alcohol use.
The present study examined the co-occurrence of cocaine, alcohol, marijuana, and other drug use
among treatment seeking homeless persons to determine whether alcohol use predicted cocaine use differently than marijuana
and other drugs predicted cocaine use. Although the association between alcohol and cocaine is reported in the literature,
it is unclear if alcohol use is related to cocaine any differently than marijuana or other drug use. For this study, alcohol
and other drug use patterns and related retention and abstinence outcomes over 12 months were explored as part of a project
studying behavioral day treatment plus abstinent contingent work and housing among homeless persons with substance use disorders,
particularly crack cocaine. Portions of the findings from this research have been presented at professional conferences (15,16).
METHODS
Participants
Participants were recruited by the Birmingham Health Care for the Homeless Coalition and other
established outreach organizations. Subjects had to meet criteria for (a) homelessness according to the 1985 McKinney Act
(17); (b) self-reported crack cocaine use within the last 2 weeks; and (c) psychological distress defined as any Hopkins Symptoms
Checklist-90-R scale t score of [is greater than] 70. Additionally, participants had to have cognitive ability to provide
informed consent and be free of any severe medical or psychiatric problem requiring inpatient hospitalization.
Context and Treatment
This study was part of a larger treatment outcome project (Homeless II) funded by the National
Institute on Drug Abuse designed to initiate abstinence in dually diagnosed homeless persons in Birmingham, Alabama. This
study consisted of a secondary analysis of the association of cocaine, alcohol, and other drug use using the full sample from
the larger project. Measures of population descriptors, retention, substance use and other mental disorders, and drug and
alcohol use, both from urine testing and self-report, at baseline and follow-up were used.
The larger project was a two-group, randomized, controlled study comparing behavioral day treatment
(DT) to the same day treatment plus an abstinent contingent housing and work therapy intervention (DT+). The behavioral day
treatment program was a 2-month substance abuse treatment program that met 5 days per week for approximately 5 hours per day
(phase I). It consisted of behavioral assessment and treatment planning, psychoeducational groups, goal development and review,
vouchers for goal attainment in social and recreational activities, individual counseling, and a social group called Club
Birmingham. Four months of weekly aftercare followed the first 2 months of behavioral day treatment (phase II). For participants
in the DT+ treatment group, abstinent contingent, rent-free housing was available to those who provided four consecutive drug-free
random urine test results during the first 2 months of treatment. Continued availability of abstinent contingent housing was
offered to the DT+ group for a modest rent during the next 4 months of treatment. Participants eligible for housing were also
eligible for work therapy. Work therapy consisted of supervised and paid work hardening and skills training experiences. Some
work involved refurbishing dilapidated housing stock for use in the study. A positive urine test result for a DT+ client participating
in the abstinent housing or work therapy components would result in immediate eviction and transportation to a shelter and
suspension from work until two consecutive drug-free urine test results were provided. Clients could then return to the housing
and to their work therapy position. Alcohol and drug use (measured by urine testing and selfreport), homelessness, employment,
and psychopathology outcomes were measured at baseline and at 2-, 6-, and 12-month follow-up in the Homeless II study. Treatment
outcome results are reported in another paper (18).
Design and Hypothesis
This study analyzed the association of alcohol, cocaine, marijuana, and other drugs including
marijuana at three follow-up by treatment group and for all participants by collapsing treatment groups. Treatment retention
and abstinence outcomes as a function of diagnostic status (cocaine plus alcohol disorder versus cocaine disorder alone) was
conducted. We hypothesized that the association between crack cocaine and alcohol use would be stronger than the association
between crack cocaine and marijuana or other drug use including marijuana at all follow-up for all participants. We also hypothesized
that there would be retention and abstinence outcome differences by diagnostic status, but no directions were predicted.
Variables and Measurement
For this secondary analysis, cocaine was assessed at scheduled baseline and follow-ups (2, 6,
12 months) using urine drug testing. Urine was analyzed using the Ontrak Radioimmunoassay system for abused drug testing manufactured
by Roche Diagnostic Systems (19). This system permitted on-site testing and instant feedback on whether a drug was detected.
Cocaine status were determined by urine testing for the cocaine metabolite, benzoylecgonine (BE). For the purposes of this
study, the cut-off for BE was 300 ng/ml. The approximate duration of detectability for BE is 2-3 days.
Self-reported alcohol, cocaine, marijuana and other drug use (opiates, amphetamines, sedatives
and hypnotics, barbiturates, and hallucinations) including marijuana was measured by the addiction severity index (ASI) (20).
For this study, items that determined day use for each particular drug in the past 30 days were recorded on the ASI. DSM-III-R
Axis I mental disorders were assessed at baseline and at 6 months by the diagnostic checklist (21).
Procedure
Participants were recruited, eligibility was determined, and informed consent was obtained to
participate in the Homeless II research project. Participants were administered a baseline assessment that included the ASI
and a urine sample. A detailed description of the complete assessment battery may be found elsewhere (18). Participants were
then randomly assigned to one of the two treatment intervention groups described earlier. They participated in three phases
of treatment for a total of 12 months. Phase I consisted of 2 months of behavioral day treatment for all participants and
the same day treatment plus rent-free abstinent contingent housing for the DT+ group. Phase II consisted of 4 months of aftercare
for all participants and rented abstinent contingent housing and paid work therapy for the DT+ group. Phase III consisted
of 6 months of aftercare for all participants. Random, twice weekly urine toxicology testing was conducted during phase I,
and weekly testing was conducted during phase II for use in delivering the abstinent contingencies. Assessment of alcohol
and drug use variables for this study were conducted at baseline and scheduled 2-, 6-, and 12-month follow-up.
Statistical Analyses
Statistical analyses were designed to address two primary issues. First, we examined the association
of cocaine use with use of other drugs to address the prior hypothesis that cocaine use would be more strongly associated
with alcohol use that with the use of other drugs. Second, we examined the relationship between treatment retention and alcohol
and cocaine use. Outcomes of cocaine, alcohol, and other drugs were assessed at four times (baseline and at 2, 6, and 12 months
after initiation of treatment). Operationally, cocaine use was treated as a binary outcome with a subject classified as positive
at any time if he or she tested positive for cocaine on a scheduled urine test at that point. Use of alcohol, marijuana, and
other drugs treated as interval level variables, with use defined as number of days use in the last 30 before the assessment
point as reported on the ASI. Alcohol and marijuana were recorded directly on the ASI, and all other drug use was computed
as the maximum number of days use of any drug, including marijuana.
The relationships between the binary outcome of cocaine use and other substance use (measured
as interval variables) were assessed using a series of logistic regression models. This class of generalized linear models
as described in McCullagh and Nelder (22) can be used to model the prevalence of a binary outcome as a nonlinear function
of a continuous predictor variable. One output from this model is an estimate of the area under the ROC curve, which serves
as a measure of the predictive ability of the model. This area has a range of 0.5-1, with a value of 0.5 signifying that the
predictor variable provides no information about the prevalence of the binary outcome and a value of 1 signifying that the
predictor can predict the binary outcome explicitly.
A separate series of logistic models was run at each time, with the same general modeling structure
used at all three postintervention time points. For all models, cocaine use was the outcome (response) variable. First, separate
models were fit for alcohol, marijuana, and all other drugs, including marijuana, with baseline cocaine use as the controlling
variable. Then two models, one with alcohol and marijuana use and one with alcohol and all other drug use as explanatory variables
with baseline cocaine use as the control variable, were fit to assess the cocaine/alcohol relation when other drugs were included
in the model. Finally, relationships among alcohol use, treatment, and cocaine use were examined with a model that included
alcohol use and treatment controlling for baseline cocaine use.
For the comparison of retention and abstinence outcomes between groups defined by diagnostic
status, results from the diagnostic checklist were used to identify those participants who had a dual diagnosis of alcohol
and cocaine abuse/ dependence versus those who had a diagnosis of cocaine abuse/dependence only. For these analyses, retention
was defined as a dichotomous measure for phase I and phase II, with a subject defined as retained in each phase if they attended
at activities on average of 0.5 days per week. Abstinence was defined as continuous weeks abstinent and either positive or
negative for cocaine based on assessment urine toxicology results at each assessment. For these dichotomous outcomes, chi-square
tests based on contingency tables were used to assess differences in diagnosis groups. For consecutive weeks abstinent, the
Wilcoxon Rank Sum test was used.
RESULTS
Participant Characteristics
Participants were 141 homeless persons with substance use and other nonpsychotic mental disorders
seeking drug treatment at a metropolitan health care for the homeless agency. Of the participants, 72.3% were male and 27.7%
were female; 82.7% were African American and 17.3% Caucasian. The average age of participants was 37.7 (SD 7.1) years; with
13.1 (SD = 2.4) average years of education. Participants were randomly assigned to DT (n = 69) or DT+ (n = 72).
For this study, 141 persons were administered baseline assessments. Follow-up rates were 78.0%
at 2 months (DT+ = 86.1%, DT = 69.6%), 77.3% at 6 months (DT+ = 84.7%, DT = 69.6%), and 70.9% at 12 months (DT+ = 79.2%, DT
= 62.3%). In phase I, 83.7% were exposed to treatment (DT+ = 93.1%, DT = 73.9%) and 52.5% (DT+ = 63.9%, DT = 40.6%) were exposed
to phase II treatment. A two-way contingency table analysis of numbers of participants in each group exposed to some treatment
(attendance of at least four morning meetings in phase I and at least two aftercare meetings in phase II) at 2 and 6 months
showed that DT+ had significantly fewer "drop outs" from treatment at 2 months (chi-square = 9.46, p = 0.002) and 6 months
(chi-square 7.68, p = 0.006). To determine if participants remaining in treatment were representative of those randomly assigned,
treatment exposed participants were compared with drop-outs on baseline demographic variables. No differences between treatment
groups were found.
Substance Use and Other Mental Disorders
Among this sample, psychoactive substance use disorders were diagnosed in the following proportions:
57.8% alcohol, 96.9% cocaine, 18.0% marijuana, and 10.9% other drug disorders. Additionally, 60.2% had two or more psychoactive
substance use disorders, and 39.8% had only one. At baseline, 63.8% of participants met criteria for one or more DSM-III-R
Axis I mental disorder other than a psychoactive substance use disorder (23); 46.9% mood, 35.9% anxiety, 7.0% adjustment,
and 4.7% other disorders. Effect of comorbid substance use and other mental disorders on substance use and other treatment
outcome variables are beyond the scope of this article and will be presented elsewhere.
Association of Cocaine Use and Alcohol, Marijuana, and Other Drug Use
Figure 1 shows the days of cocaine, alcohol, marijuana, and other drug use at baseline and follow-up.
Cocaine and alcohol use accounted for most of the total days of any substance use in the last 30 days. The proportion of days
of cocaine and alcohol use in the last 30 days was similar. These patterns remained constant across all time despite the significant
decrease in days of any substance use from baseline. Figure 2 shows the days of alcohol use by cocaine toxicology status at
baseline and follow-up. These data revealed that alcohol use was significantly greater among persons who were cocaine positive
than those who were cocaine negative at all times (p [is less than] 0.01).
[ILLUSTRATIONS OMITTED]
The results of the logistic regression analysis at the 6-month follow-up are summarized in Table
1. Because findings at 2 and 12 months were similar to the 6-month point and the 6-month point represents completion of the
most intensive phase, only 6-month data are presented. In general, results supported the assertion that cocaine use was strongly
associated with extent of alcohol use and that the association between cocaine and alcohol was stronger than the association
between cocaine and marijuana or other drug use. This assertion was supported by both the univariate models (controlling for
baseline cocaine use) and the multi variable models that included alcohol and other drugs. At all three times, the area under
the ROC curve was larger for alcohol than for any of the other drugs, and, in general, adding other drug use to the alcohol
model added little predictive value. The only exception to the last statement is that, for marijuana at 6-months, some additional
predictive value was obtained by adding marijuana to the alcohol model. However, the improvement was slight.
Table 1. Comparison of Alternative Logistic Regression Models
for Predicting Cocaine Use at 6-Month Follow-Up
Model Parameter Estimate SE p Value
1 Alcohol 0.11 0.036 0.002
Baseline coke 0.52 0.49 0.29
2 Marijuana 0.39 0.20 0.051
Baseline coke 0.91 0.47 0.052
3 Other drugs 0.099 0.047 0.035
Baseline coke 0.97 0.46 0.035
4 Alcohol 0.097 0.035 0.006
Marijuana 0.23 0.17 0.18
Baseline coke 0.59 0.51 0.25
5 Alcohol 0.10 0.036 0.006
Other drugs 0.068 0.045 0.13
Baseline coke 0.62 0.50 0.22
6 Alcohol 0.11 0.057 0.048
Treatment -0.87 0.54 0.11
Odds Ratio
ROC
Model Parameter Estimate CI Area
1 Alcohol 1.75 1.28,2.69 .756
Baseline coke 1.68 0.63,4.40
2 Marijuana 6.95 1.59,70.6 .698
Baseline coke 2.47 1.00,6.30
3 Other drugs 1.65 1.11,3.00 .696
Baseline coke 2.63 1.08,6.60
4 Alcohol 1.63 1.19,2.48 .780
Marijuana 3.21 1.15,28.2
Baseline coke 1.80 0.66,4.90
5 Alcohol 1.65 1.21,2.54 .771
Other drugs 1.40 0.91,2.43
Baseline coke 1.86 0.69,5.02
6 Alcohol 1.75 1.10,3.60 .764
Treatment 0.42 0.14,1.19
CI, confidence interval.
The results in Table 1 also contain the "best fitting" models for the relationship of alcohol
use and treatment (DT or DT+) to cocaine use. In developing these models, we considered the main effects of alcohol use and
treatment and the alcohol by treatment interaction, as well as other drug use and baseline cocaine use. Model fit statistics
were examined in a series of reduced models, and any variables that did not contribute significantly to the model at the 0.05
level or affect the magnitude of the coefficients of primary interest (alcohol, treatment, and their interaction) were removed
from the model. No potential confounding variables were found to contribute substantively to the model, and the alcohol by
treatment interaction was found to be negligible. The model showed that when alcohol was included in the model, the treatment
effect was nonsignificant. We interpreted this finding to mean that, at these later times, the treatment procedure acted on
alcohol use and cocaine use in a similar fashion. Consequently when alcohol use was included in the model as a potential confounder,
the treatment effect was diminished.
Retention and Abstinence Outcomes
Of the 141 participants, thirteen had no diagnostic checklist results and four had alcohol disorders
only, leaving 124 participants available for these analyses. Of these 124 participants, 70 (56.5%) were positive for alcohol
and cocaine, whereas 54 (43.5%) were positive for cocaine only. Participants were distributed relatively evenly across treatment
groups, with 40 of 68 DT+ participants (59%) dually diagnosed and 30 of 56 DT participants (54%) dually diagnosed and with
no evidence of a difference in proportions dually diagnosed in the two treatment groups (chi-square 0.35, p = 0.56). During
phase I, the two diagnosis groups showed a treatment retention difference with 70 of the dually diagnosed participants (100%)
retained in treatment and 43 of the cocaine only participants (80%) retained in treatment (chi-square 15.6, p [is less than]
0.001). However, the two groups showed no retention difference in phase II with 20 of the dually diagnosed (29%) and 15 of
the cocaine only participants (28%) retained in treatment (chi-square = 0.009, p = 0.92). The two groups also showed comparable
levels of average consecutive weeks abstinent at 2 months (dually diagnosed 4.2, cocaine only 3.5, p = 0.16) and at 6 months
(dually diagnosed 7.5, cocaine only 6.7, p = 0.18). These results were also reflected in the assessment urine toxicology results
with 51 of 60 dually diagnosed participants (85%) and 33 of 43 cocaine only participants (77%) abstinent at 2 months (chi-square
1.14, p = 0.29), 38 of 52 dually diagnosed participants (73%) and 21 of 37 cocaine only participants (57%) abstinent at 6
months (chi-square 2.58, p = 0.11), and 32 of 49 dually diagnosed participants (65%) and 23 of 37 cocaine only participants
(62%) abstinent at 12 months (chi-square 0.09, p = 0.76). Consequently, other than retention at 2 months, the two diagnostic
status groups as classified at baseline showed no difference in primary outcomes.
DISCUSSION
The co-occurring use of cocaine and alcohol is widely known among researchers and clinicians
alike. However, the sole emphasis on either one has blurred the focus on their relatedness and synergistic effects. Researchers
tend to ignore either alcohol or cocaine when the other drug defines the target population of inquiry. Fragmentation is evident
in the names of federal research institutions (e.g., National Institute on Alcohol Abuse and Alcoholism and National Institute
on Drug Abuse) and community self-help groups (e.g., Alcoholics Anonymous and Narcotics Anonymous). The isolation of "drug
of choice" by clinicians and epidemiologists disguises the fact that it is more common to use alcohol and other drugs together
than to use these substances alone.
These findings revealed that among a sample of homeless persons participating in behavioral day
treatment, use of crack cocaine was strongly associated with the extent of alcohol consumption and that the association between
cocaine and alcohol was stronger than the association between cocaine and other drug use, including marijuana, These results
are consistent with previous findings that cocaine use and alcohol use frequently co-exist and the new notion that cocaine
use occurs differently with alcohol than with other drugs.
The present study not only supports the notion that cocaine and alcohol strongly coexist among
homeless persons seeking treatment, it also implies that cocaine is used with alcohol more than with other drugs, including
marijuana. We also find that persons with dual diagnoses of cocaine plus alcohol tended to be better retained in this day
treatment than persons with a cocaine disorder alone. Persons with dual diagnoses may present with more complicated clinical
pictures and have more issues to address in treatment than persons with cocaine disorders alone. No differences between diagnostic
status groups in treatment outcome as measured by cocaine abstinence rates were found. Both groups appeared to respond similarly
despite the cocaine only disorder group spending less time in treatment. Adjunct treatments like the use of disulfrum (Antabuse)
may be used to improve the treatment outcome for persons with additional alcohol diagnoses.
Note the generalizability limitation of these findings to homeless persons who used primarily
crack cocaine with coexisting mental disorders not requiring hospitalization. These findings are further limited to the nature
of the drug abuse treatment. The context of this study was a controlled clinical trial utilizing a behavioral day treatment
and contingency management program with services lasting over 6 months. Owing to the experimental nature of the services provided,
further generalization to participants in other or more typical drug abuse treatment programs should be tested. Future research
should broaden the scope of these findings by targeting samples representative of other outpatient and residential drug abuse
treatment centers.
This study investigated the use of crack cocaine only in relation to alcohol and other drug use.
One might predict differences in use patterns with different modes of cocaine administration (e.g., intranasal powder and
intravenous use) because differences in addictive potential and sociodemographic variations. Finally, this study did not assess
the exact co-occurrence of cocaine and alcohol use, only the approximate combined use within the past month. Additionally,
this study did not test for the presence of cocaethylene deleterious toxin resulting from the combined use of cocaine and
alcohol.
When used in combination with each other, cocaine and alcohol can be especially dangerous through
the formation of a toxic substance called cocaethylene. Fischman and Johanson (2) identified cocaethylene as a chemical product
developed in the liver as a reaction to the presence of both cocaine and alcohol. Randall (24) reported that cocaethylene
was present in both humans and animals when alcohol and cocaine were used in combination. The formation of cocaethylene may
enhance the toxic effects of either cocaine or alcohol when these substances are used alone. Future research should attempt
to assess simultaneous use of alcohol and cocaine, the order in which they are used, and the delay between use when both alcohol
and cocaine are used to ascertain the precise cooccurrence of these drugs.
This study and previous research documenting the coexistence of cocaine and alcohol use, dangers
unique to the combined use of cocaine and alcohol, the discovery of the toxic substance cocaethylene, and differential demographic
and psychopathological variables associated with cocaine and alcohol demonstrate the need to pay more attention to their combined
use and interactive effects.
