Talwin

INTRODUCTION

Gender differences are known to exist with regard to alcohol and drug use and associated behaviors. For example, it has been observed that women tend to report their first drug use at a later age than men, are often initiated into drug use by their male partners, who in turn also become their main suppliers of addictive substances (1). It has also been reported that compared to men, women drug users have less criminal involvement, especially when it comes to dealing in drugs (2). Hser and colleagues addressed differences in the pattern of treatment utilization and gender-specific "drug treatment careers" (3).

Men and women are known to differ with regard to when, where, and how they access treatment for drug use as well as in their referral sources and social support groups. Women, in general, report a shorter transition from first drug use to addiction, enter treatment sooner than men, and have also been known to report less intense histories of alcohol abuse, yet more impairment (4-11). Though, women tend to be more socially integrated with family and work (12), studies have shown that they receive less spousal or familial support for entering a treatment program (13). Indeed, although positive spousal/partner influence was reported by the men in a study of 50 alcoholics in outpatient treatment (14), a longitudinal study of women opioid addicts (15) found that women seldom cited familial influence as a reason for entering treatment.

Grella and colleagues (16) conducted an analysis of differences in the factors associated with drug treatment history among men and women in the Drug Abuse Treatment Outcome Study (DATOS), a national multisite prospective study sponsored by the National Institute on Drug Abuse. Besides providing a fairly exhaustive yet concise review of the extant literature on the subject, they attempted to put a treatment career perspective to previous studies on traditional gender differences among the participants in DATOS.

Also using the DATOS database, we address whether there are differences between men and women drug addicts even at the clinical level. That is, after we control for the economic, emotional, social support, and similar associated factors, would there still be gender-specific differences in response to treatment?

There is accumulating evidence that the biological effects of drugs of abuse are not always the same for males and females. For example, a study of gender and menstrual cycle difference in response to acute intranasal cocaine reported that mean peak plasma cocaine levels in females were higher in the follicular phase than in the luteal. Males, however, achieved higher mean peak plasma cocaine levels than females, detected cocaine effects significantly faster, and experienced a greater number of episodes of intense good effects. Although men had higher plasma levels than women, men and women experienced the same increased heart rate, suggesting a greater cardiovascular sensitivity to the effects of cocaine for women than for men (17). Behavioral studies with rats have also shown gender differences in response to abused drugs (18). Preliminary data from several studies using varying clinical endpoints are beginning to suggest that drugs of abuse may produce different biologic impairment in males than in females (19-22).

In light of these findings, there has been an increasing realization that the treatment needs of women may be different from those of men. Accordingly, treatment strategies need to be different as well. In our study, we do not have the luxury of looking at the biological and physiological variables that many of the studies just mentioned did. Instead, we shall try to control for various demographic variables that are known to vary across the sexes and examine whether there are still differences in the effectiveness of treatment (measured purely on the basis of drug use frequency) between males and females. This would, in a sense, also address the question of how well the program has been able to tailor itself to the differing needs of the sexes.

Over the years, the National Institute on Drug Abuse has consistently highlighted some of the many research gaps that remain in the study of drug abuse and dependence in its reports (http://www.drugabuse. gov/pdf/NNCollections/NNWomenGender.pdf). The first and foremost of these is the need for more basic research, both human and animal, as well as epidemiological and longitudinal research, directed at identifying gender differences in the etiology and consequences of drug use, abuse, and dependence. By adopting a longitudinal approach (unlike previous studies that mostly focused on descriptive methods comparing responses "before treatment" and "after treatment"), we have been able to assimilate information from different time points during the program and after it. This, in our opinion, should give a more accurate picture of the influence of the explanatory covariates studied. In particular, it would tell us how the sex of the individual and the choice of treatment modality could affect treatment effectiveness for cocaine addiction (as measured by the drop in the frequency of cocaine use). It is usually impossible to come up with a completely general answer in such cases, and for our purpose, we shall restrict our attention to cocaine addiction among the participants of DATOS.

METHODS AND MATERIALS

Study Design

DATOS is a prospective study designed to determine the outcomes of drug abuse treatment delivered in typical, stable, community-based programs and to provide comprehensive information on continuing and new questions about the effectiveness of drug abuse treatment currently available in a variety of publicly funded and private programs. The study examined the role of treatment outcomes and program type, client characteristics (including dependence, treatment history, and physical and mental health comorbidities), treatment received (e.g., length and intensity of services provided), therapeutic approaches, and provision of aftercare. Four types of programs were included: outpatient methadone (OPM), short-term inpatient (STI), long-term residential (LTR), and outpatient drug-free (ODF). Respondents were sampled from among those admitted to treatment in sampled facilities in 1991-1993. Clients entering treatment completed two comprehensive intake interviews (Intake 1 and Intake 2), approximately 1 week apart. These interviews were designed to obtain baseline data on drug use and other behaviors, as well as information on background and demographic characteristics, patterns of dependence, living situation and child custody status, education and training, income and expenditures, and HIV risk behaviors, along with assessments of dependence, mental health, physical health, and social functioning. Data reflecting during-treatment progress, including service delivery and client satisfaction, were collected in the 1- and 3-month in-treatment interviews. The 12-month posttreatment follow-up interview replicated many of the Intake questions and focused on key behaviors in the year after treatment. The substances covered in the study were alcohol, tobacco, marijuana (hashish, THC), hallucinogens or psychedelics such as LSD, mescaline, and PCP, cocaine (including crack), heroin, narcotics or opiates such as morphine, codeine, Demerol, Dilaudid, and Talwin, downers or depressants such as sedatives, barbiturates, and tranquilizers, amphetamines or other stimulants such as speed or diet pills, and other drugs.

A total of 96 treatment programs in 11 mid-size and large United States cities with well-established treatment systems participated in DATOS. Programs were purposively sampled to reflect typical clinical approaches across the four modalities: outpatient methadone, short-term impatient, long-term residential, and outpatient drug-free. Geographic location, type of program, and representativeness of the program and its clients were considered in the three-level process of selecting cities, programs, and clients. Respondents were sampled from among those admitted to treatment in sampled facilities in 1991-1993.

Subjects

A total of 10,010 clients participated in the Intake 1 interview. Of those, 8,755 participated in the Intake 2 interview, 6,148 in the 1-month in-treatment interview, 3,180 in the 3-month in-treatment interview, and 2,966 in the 12-month follow-up. The sample for Intake 1 was 66% male, 47% African American, and 13% Hispanic, with a mean age of 33 years. However, these and other client characteristics varied across modalities, reflecting differing therapeutic and operational characteristics. For the 12-month follow-up sample, 4,299 of the eligible clients who completed the two-stage intake interviews were selected for follow-up using a stratified random design. Of these respondents, 74% (n = 3,147) were located, 70% (n = 2, 966) were successfully interviewed, 1.5% (n = 64) were deceased, and 2.7% (n = 117) refused to participate. Gender, ethnicity, and average age were not significantly different between the intake and follow-up samples. For more information on DATOS, one can point to their excellent website (http://www. datos.org), from where the information in this section has been gleaned.

Variables

The data set we used for analysis contained 10,010 observations and 92 variables, of which a significant number were repeated measurements of the same set of 29 variables. The variables that were included mainly contained information about drug and alcohol use and the psychological functioning of the sampled individuals in each of the four waves. A number of baseline demographic variables were also included. A number of new variables for assessing cigarette, alcohol, and drug use were created by restructuring the categories for the relevant responses. For the sake of consistent comparison, five categories were used for all the variables namely "abstainer," "low user," "moderate user," "heavy user," and "very heavy user."

The variable for assessing comorbid psychology was created by combining a number of variables that measured the extent of anxiety and depression disorders in the four waves. It has four levels namely "neither anxiety nor depression," "only anxiety," "only depression," and "both anxiety and depression."

A measure of problem severity at intake was defined to get a better understanding of the response to treatment vis-a-vis the situation at the time of entry to the program. It is reasonable to think that frequency of drug use or drinking alone cannot be the sole indicator of the breadth and intensity of a patient's addiction problem. Several socioeconomic and emotional factors could confound the addiction problem and thus influence the recovery or rehabilitation process. The index of problem severity at intake was constructed using variables that represent functional domains commonly related to treatment goals and outcomes as assessed in the Addiction Severity Index (23), then (24) seven indicators were scored to reflect a wide variety of problems. When the scores were summed, they defined the Problem Severity Index (PSI) score for each patient. The indicators were multiple drug use (self-reported use of any three or more drug categories in the year before intake), alcohol dependent (either a DSM-III-R or self-reported daily consumption for 1 month or longer during the year before intake), criminally active (being on probation or parole, awaiting trial or case pending at program intake, or a period of weekly involvement in illegal activities during the past year), unemployed (never worked at a full-time job in the year before intake), low social support (having several family members or close friends who used illegal drugs or ones who were incarcerated in the past year), depression or anxiety (having a DSM-III-R diagnosis of depression or anxiety or self-reported suicidal ideation--i.e., having attempted suicide or thought about killing self), and no insurance (having no health insurance). PSI, being the sum of seven binary indicators, would range from 0 (least severe) to 7 (severest). Based on the PSI score, the patients were further categorized into three groups--those with low severity problems (PSI score of 0-3), those with problems of medium severity (PSI score of 4-5), and those with problems of high severity (PSI score of 6-7).

Statistical Analysis

Because the data set was mostly categorical, we decided to use the standard contingency table analysis methods such as tests of association/ homogeneity, odds ratios (OR), and log-linear models. Standard SAS procedures such as PROC FREQ and PROC CATMOD were used for most of the initial statistical analysis and OR were obtained. The sampling design for DATOS was such that at any wave (except the last), only a fraction of the cases from the previous wave were followed up. Further, the sampling was not completely random, because care was taken to maintain the racial composition at every wave roughly the same. Such a judicious sampling scheme could lead to disproportionately large or small samples from marginal categories of a variable. The multiplicative invariance property of the OR ensures that it still estimates the same characteristic even when the marginal frequencies are not balanced across the waves. Cumulative logit models were formulated to rigorously test for the significance of the variables including sex and modality on the treatment effectiveness as measured by the drug-use frequency. The cumulative logit analysis was done using the Generalized Linear Models (GLM) routine in S-Plus software. The model formulation and method of analysis are described in greater detail in the next subsection.

A Statistical Model for the Data

Logit models are commonly used for analyzing categorical data. They were originally designed for binary response data but have been generalized to take into account response variables having more than two categories.

In a typical longitudinal data set-up, suppose Y indicates a response variable that can take C ordered categorical values, labeled 0, 1, ..., C - 1. Although the outcomes are ordered, any numerical scale that might be assigned would be arbitrary. Because the response categories for ordinal data are usually arbitrary, we would like a model (a proportional odds model for example) whose coefficients have the same interpretation when we combine or split categories. This is achieved by working with the cumulative probabilities Pr(Y [less than or equal to] a) rather than the cell probabilities, Pr(Y = a).

The proportional odds model for independent observations has the form:

logit Pr(Y [less than or equal to] a) = log[Pr(Y [less than or equal to]/Pr(Y > a)] = [[theta].sub.a] + x'[beta]

It is convenient to introduce the vector of variables [Y.sup.*] = ([Y.sup.*.sub.0], [Y.sup.*.sub.1], ..., [Y.sup.*.sub.c-2]) defined by [Y.sup.*.sub.a] = 1 if Y [less than or equal to] a and 0 otherwise. The proportional odds model is simply a logistic regression for the [Y.sup.*.sub.a], because:

log[Pr(Y [less than or equal to] a)/Pr(Y > a)] = logit Pr([Y.sup.*.sub.a] = 1) = [[theta].sub.a] + x'[beta], a = 0, 1, ..., C - 2.

For our purpose, we shall rewrite the model following the notation of (25).

At occasion g, let [[phi].sub.h](g;x) denote the probability of response h, when a set of covariates takes the value x. When x is categorical, different levels of x usually represent subpopulations whose response distributions we want to compare. The probabilities {[[phi].sub.1] (g;x), ..., [[phi].sub.I] (g; x)} form the gth marginal distribution for the [I.sup.T] contingency table for subpopulation x. At occasion g and covariate value x, the marginal response distribution {[[phi].sub.j](g;x), j = 1, ..., I} has I - 1 cumulative logits, [L.sub.j](g;x) = log{[([[phi].sub.1](g;x) + ... + [[phi].sub.j](g;x)]/[[[phi].sub.j+1](g;x) + ... + [[phi].sub.I](g;x)]}, j = 1, ..., I - 1, one for each response cut point. If we have k categorical covariates {[s.sub.1], ..., [s.sub.k]}, besides the variables for the cut point and the occasion, we could write the model as

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

We could then fit the model in the generalized linear model framework using standard statistical packages. To test for the significance of one or more covariates (i.e., of one or more kinds of categorizations), we would compare the residual deviance of the model to that of an appropriate submodel.

In our analysis, we used the level of cocaine (use in the different waves as our response variable, with sex (1 = males, 2 = females), modality (1 = LTR, 2 = STI, 3 = OPM, 4 = ODF) and problem severity index (1 = low, 2 = medium, 3 = high) were used as covariates. Additionally, the cut-point category (0 = abstainer, 1 = low user or abstainer, 2 = medium user or less, 3 = high user or less) and the occasion when the observation was made (i.e., the wave [0 = intake, 1 = 1-month in-treatment, 2 = 3-month intreatment, 3 = 1-year posttreatment, 4 = 1-year posttreatment (most recent)]) were also thrown into the model as possible explanatory variables. Besides formulating a comprehensive model that incorporated responses from all the five waves, we ran analyses based exclusively on observations in individual waves and also an analysis based on the last four waves while using the values from the first wave as an additional explanatory covariate. The use of the response from the first wave as a covariate in the analysis of the later waves, especially when checking for the significance of the sex factor, is crucial. This would take into account the existing differences between the sexes at the time of entering the program and in a sense make it a level playing field when comparing the effectiveness of treatment in the subsequent waves.

RESULTS

Sex-Specific Comparison of Demographics

The sex-specific distributions of treatment modality, age group, marital status, race, educational level, weeks of full time work, and health insurance status have been outlined in Table 1.

The distribution of men and women in the different treatment modalities are more or less similar. There appears to be a slightly higher percentage of women going in for Methadone maintenance. This ties well with the fact that, typically, Methadone maintenance was used to treat opiate addiction and historically, a greater proportion of women than men have been known to be opiate or narcotic addicts.

Roughly speaking, 51% of the men entering treatment are in the age group 31-44, whereas nearly 56% of the women coming in for treatment are between the ages of 26 and 35. There is also a slightly higher percentage of men over age 44 compared with women in the same age group. This is not surprising, because it is known that women report a shorter transition from first drug use to addiction and tend to enter treatment sooner than men. Historical studies of treatment career pattern of men have shown that men, typically, tend to enter treatment in middle age after a period of social disintegration.

Approximately 31% of the men and 33% of the women are either married or are living as married, whereas about 47% of the men and 42% of the women reported never being married.

The race-wise distribution of men and women in the sample are comparable, with nearly half (46% of men and 48% of women) of the sample being made up of African Americans.

More than 90% (96% of the men and 95% of the women) of the sample attended high school at the least, with the majority of men (39%) having a high school degree and the majority of women (36%) having attended high school (an almost equal percentage actually having a high school degree).

A significantly larger proportion of men (60%) compared with the women (34%) reported having full-time work in at least some part of the year before admission. Although nearly 26% of the men had full-time work during almost the whole year, the corresponding figure for the women was only about 12%.

There are marked differences in the health insurance figures for the two sexes as well. Although more than half the men (54%) did not have any kind of health insurance, less than 40% of the women were without insurance. This is a little surprising because the majority of women (45%) have public insurance plans, but more than 65% of them were unemployed during the year. A possible explanation could be that most of these women were on someone else's insurance plans as dependents. However, the lack of health insurance of any kind among full time working men definitely needs to be looked into.

Sex-Specific Patterns of Drug and Alcohol Use

Sex differences in drug and alcohol use patterns are known to exist according to previous studies. Table 2 compares men and women with regard to their primary drug problem and their frequency of use of the primary drug at the time of admission to the program. Table 3 describes how treatment affects their drug use frequency at different stages of the treatment period. This has been done for some specific substances, such as the ones that were found to be the most commonly used among clients at the time of admission. Figure 1 describes the progression of cocaine use frequency across the waves and for different categories of problem severity. The graphs in Fig. 1 are strikingly similar. The frequency of cocaine use takes a nosedive (correspondingly, the frequency of abstainers shoots up) just a month into treatment and this is true for all levels of problem severity. This is not at all unexpected, because it is known that people most often enter a drug rehabilitation program only when the addiction becomes acute and the immediate drop in drug use that results from suddenly being thrown in a controlled or restricted environment is, understandably, marked. Also, there could be some amount of overreporting of drug use at the time of entry into the program.

[FIGURE 1 OMITTED]

Thereafter, the frequency of use increases slightly in the next wave (3-month in-treatment) and again in the two posttreatment waves. However, the increased use frequency is still far below the pretreatment levels. The odds ratios computed in Tables 4 and 5 further quantify the differential responses of the sexes to treatment.

Cocaine, either in powdered form or as crack, appears to be the drug of choice for both the sexes, with more than half the sample (50% of the men and 56% of the women) reporting it to be their primary drug problem. Heroin, which was the drug of choice in the 1980s, comes in second with about 18% of the men and 20% of the women indicating it as their "problem drug." Although more men (about 4%) than women (about 2%) prefer marijuana, the proportion of addicts in the other drug categories--hallucinogen, narcotics/opiates, sedatives, amphetamines, inhalants, and others--are small and roughly the same for the two sexes. A significantly large proportion of men (more than 14%) compared with women (9%) reported alcohol addiction as their primary problem.

Table 2 tells us that, although men and women have comparable primary drug use frequency in the "abstainer" and "low user" categories, women (64%) tend to outproportion men (56%) as heavy or very heavy users of the primary drug at the time of admission to the program.

Table 3 provides information about the progressive decline in the use of drugs of some specific categories for the two sexes at different stages of the program. For both sexes, the frequency of drug use for all kinds of drugs and alcohol takes a virtual nosedive at the end of a month's stay in the program. The frequency picks up a little bit at the end of the third month and a lot more at the time of the 1-year posttreatment survey. However, the decline in drug use frequency even a year after treatment is still remarkable. The proportion of heavy and very heavy cocaine users decreased from 35% to 10% among the men and from 46% to 12% among the women, whereas the number of abstainers or complete nonusers increased from 18% to 63% among men and from 15% to 67% among women. The figures for heroin, marijuana, amphetamine, and opiate/narcotic are even better (e.g., the proportion of heavy heroin users declined from more than 50% [more than 60% for women] to below 7% for both the sexes).

Although this is very encouraging, we do need to exercise some amount of caution when considering this as a measure of the program's success. It is known that people usually go in for treatment when the problem becomes severe and there is always a tendency of overreporting at the time of admission to the drug treatment program.

When we compare the figures for men and women, a couple of interesting points reveal themselves. It appears as though improvement from treatment is more pronounced for men than for women 1 month into the program. However, the women seem to catch up as they spend more time in the program and the differences are much less pronounced at the time of the 1-year posttreatment survey. To further investigate this, we decided to restrict our attention to those in the sample who were heavy or very heavy drug users at the time of admission. The idea was to follow this group (or rather a subset of this group, as allowed by the sampling design) through the four waves, record what proportion of them came down from a heavy to a non--heavy user or a complete nonuser. Tables 4 and 5 list our findings. As far as progression from being a heavy to a non-heavy user is concerned, the general picture that emerges from Table 4 is that men are more likely than women to change from being a heavy user to a non-heavy user after a month in treatment. In particular, the odds of men, who were heavy cocaine users, to change to less than heavy users were more than twice that for women, and the odds for the two sexes also proved to be significantly different at the 5% level. However, the odds ratio gets closer to one after 3 months in treatment, and by the time of the 1-year posttreatment interview, the odds for men and women are almost the same. If we want to compare how the heavy addicts among the men and the women fared in terms of completely kicking a specific addiction, we need to look at Table 5. Once again, it appears as though the odds for a man to kick an addiction within a month of being in treatment are higher than that for women. We stress, however, that these differences, although sometimes quite large, were not statistically significant in most cases. The only significant differences were in the use of opiates/narcotics assessed at the point of 3 months into treatment, where it turns out that the odds of a man kicking the habit are about four times the corresponding odds for a woman, and, in the use of marijuana at the 1-month point into treatment where the odds for a man becoming substance-free were more than one and a half times that for a woman.

Problem Severity and Demographics

PSI categories were found to be significantly associated with the sex of the patient. The percentages of women with medium and high severity problems were greater (slight but significant) than the corresponding figures for the men. Among men, as well as among women, the PSI category was also found to be significantly associated with treatment modality, age group, race/ethnicity, and educational level. Here, we need to note that none of these variables were used in the construction of PSI and hence the association is not artificial. We shall briefly discuss some of the specific findings:

The choice of treatment modality is heavily influenced by the severity of the problem. Among men, a little more than half of the "high severity" cases and about a third of the medium severity cases opt for the residential modality, whereas a majority (about 40%) of the low severity cases prefer short-term inpatient treatment. The corresponding figures for the women are, respectively, 50%, 30%, and 37%.

The association between PSI and a patient's age group seems to follow a very consistent pattern. As we move from the lower age groups to the higher ones, the percentages for medium and high severity cases steadily decrease, whereas that of low severity cases shows a corresponding increase. This trend exhibits dramatic consistency among the men and is also unmistakable among the women, though in the latter case there is little more "noise" in the middle age groups.

When it comes to PSI and race/ethnicity, we find that Caucasian and "other race" men are more prone to having medium and high severity problems compared with African American or Hispanic men. Approximately 57% of the male patients who are Caucasians or belong to the "other races" have medium to high severity problems at intake, whereas the same proportion among African-Americans or Hispanics is about one half. The differences are even more pronounced among the women, whereas more than 60% of Caucasians and "other races" could be classified as having medium to high severity problems compared with a little more than 50% among African-Americans and Hispanics.

The pattern of association between PSI and educational level is almost a mirror image of that between PSI and age group. As the level of education increases, the percentages of medium and high severity cases decrease very consistently among the men and to a slightly lesser extent among the women as well.

Sex-Specific Comparison of Cocaine Use Controlling for Problem Severity

A natural question of interest was to investigate whether there were significant differences between the sexes in their response to treatment, when the severity of problem at intake was controlled. We decided to focus on cocaine addiction, which was the primary problem drug for more than half of the population, among both men and women. A binary variable indicating use or non-use of cocaine was used for the purpose. The association of this variable with sex of the patient was assessed through the odds ratio in each of the four waves ("1-month in treatment," "3 months in treatment," "1-year posttreatment," and "most recent") subsequent to intake. All in all, there would be 12 odds ratios (3 PSI categories x 4 waves) to look at and of these just two were significant (marginally). Among patients with medium severity problems and 1 month into treatment, the odds of a male patient becoming cocaine-free appear to be one and a half times that of a female patient. On the other hand, among low severity cases and at the time of the most recent posttreatment interview, the OR is almost the same, but in favor of the women. The overall (i.e., collapsing over the PSI categories) OR turns out to be 1.34 in favor of the men. In the face of these contradictory patterns, it is reasonable to think that we need to further control for other factors. An obvious factor that could confound the effect of sex is the treatment modality (i.e., to take into account whether the patient was availing the "residential," "short-term inpatient," "outpatient drug free," or "methadone maintenance" modality). Controlling for both problem severity and treatment modality, we then had 48 (4 x 3 x 4) OR to consider, of which only two turned out to be significant. For patients with low severity problems who were using the "outpatient drugfree" modality, at the time of the 1-year posttreatment as well as the most recent interviews, the odds for a woman to be cocaine-free was well over three times that for a man. Thus, on the whole, men and women do not seem to differ significantly, with regard to their "clinical" response to treatment, when the severity of problem (at Intake) and treatment modality are controlled.

Modality-Wise Comparison of Cocaine Use

Having observed that sex is not much of a determining factor for response to treatment, our next objective would be to investigate to what extent choice of modality influences the same. We constructed, for that purpose, four binary variables, each indicating whether or not a patient belonged to a particular modality, the idea being to compare the odds for a patient belonging to a certain modality to that of someone who doesn't.

We shall briefly discuss the salient features of these ORs as listed in Table 6. First, if success is defined by the ability of the program to make the patient completely cocaine-free, then, the residential mode of treatment seems to be the best and the methadone maintenance program by far the worst. The odds of remaining cocaine free a year after treatment for a nonresidential patient is about three quarters that for a residential patient, in case of medium severity problems, and only about half, in case of low severity problems. If a patient is not in the methadone maintenance program, the odds are about one and a half times that of someone in the program, for low and medium severity problems, and more than twice for high severity cases.

Second, the difference in the success rates of the different modalities is more pronounced in the two in-treatment waves (wave 1 and wave 2) than in the two posttreatment waves (wave 3 and wave 4). This is particularly prominent in the figures for methadone maintenance, where the ORs are well above 10 during the first two at the 1-month and 3-month points but fall to more reasonable levels (less than 3), a year from treatment. This is understandable because the retentivity (retaining the treatment effect after the treatment period has ended) of a program will tend to decrease with the passage of time.

Third, choice of modality has a greater impact on patients with problems of low and medium severity at intake than on the high severity cases. With the exception of the methadone maintenance program, the majority of the ORs for the high severity cases turn out to be nonsignificant.

Results of Cumulative Logit Analysis

The results for some of the significance tests for covariates of interest are given in Table 7. We shall note some of the salient features of the analysis. When all five waves are considered simultaneously, all covariates--sex, modality, PSI, wave, and cut-point--are highly significant. However, when the last four waves are considered together, with the response from the first wave being taken as a covariate, sex is no longer significant. Sex does not turn out to be a significant factor even when the last four waves are considered separately, and the response from the first wave is taken as a covariate. It is negative and marginally insignificant in case of the second wave (the 1-month in-treatment wave) and is highly insignificant for the later waves. This corroborates what we saw in the course of our descriptive analysis using percentages and odds ratios. Men seem to make more improvement than women after a month in treatment; however, with time, the women catch up, and there is hardly any difference in the treatment effectiveness between the sexes at the 1-year posttreatment point. The fact that the estimate for sex was negative and highly significant in wave 0 indicates that women entering the program were more likely than men to be taking cocaine in greater frequency. After this baseline difference is taken into account through a baseline cocaine-intake covariate, there is not much difference, sex-wise, in the subsequent waves (Model 3-Model 7).

Modality turns out to be a highly significant factor no matter how many of the waves we choose to consider simultaneously. By virtue of the identifiability conditions imposed, the first level of all the categorical variables are constrained to be necessarily zero, and have not been listed in the table. A glance at the nontrivial estimates for the modality levels in models 3 through 7 tells us that the third level (OMT) has a consistently large negative value compared to the other levels and hence is clearly the worst performing modality. Also, we notice that the estimates for the three free levels are mostly negative (and even if positive, never large), in the in-treatment and posttreatment waves, implying that the first level, namely, LTR (constrained to be zero), by and large, performs the best. Interestingly, the estimates for all three free levels for wave 0 are positive. This might appear to be inconsistent with the largely negative estimates of the same in the subsequent waves, but, actually, makes perfect sense when we recall that the choice of modality at intake was strongly motivated by the seriousness of the problem, with the heavier addicts mostly opting for LTR. Our initial exploratory analysis indicated as much.

Though not the main focus of our analysis, the signs and magnitudes of the estimates for the different levels of the variables PSI, wave, and cut-point also follow common-sense logic and are consistent with our previous findings (wherever available) from exploratory analysis. For example, in Models 1 through 7, we observe that for higher levels of PSI, performance declines, whereas for higher levels of cut-point (that amounts to choosing less conservative criteria for defining "success" while assessing effectiveness of treatment), performance is improved. From Model 1, it is clear that the greatest improvement seems to occur at the time of wave 1 (1-month in-treatment), followed by waves 2, 3, and 4, in that order. When the estimate for wave 1 is constrained to be zero (as in Model 7), and the subsequent waves are compared against this baseline, quite naturally the estimates for waves 2, 3, and 4 are found to be increasingly negative. The index of problem severity (PSI) is marginally insignificant in wave 1 but becomes significant in the following waves. Thus PSI seems to have a greater impact on treatment effectiveness in the longer term.

DISCUSSION

As is usual with any field-based longitudinal study, the DATOS study failed to locate and interview all the patients targeted for follow-up. However, the conventional threshold of 70% for an acceptable relocate rate was reached and there was no evidence of sampling bias when follow-up respondents and nonrespondents in DATOS were compared (26). The other common concern regarding large-scale treatment evaluations is the reliability of self-report data. Previous articles on the DATOS study (24) have addressed this issue. The most recent article reports that urine specimens collected from a random subsample of the DATOS study indicated that self-reported drug use was reasonably accurate and not biased systematically across the subgroups compared, which are, in fact very similar to the ones we have used here.

The major findings of this paper are the following. There are differences between men and women entering treatment on basic demographics such as age and marital status with women generally tending to enter the program at a younger age and being more likely to be living with a partner. Men are more likely to have full-time employment but (strangely) less likely to have health insurance. Although it is known that men and women differ widely with regard to their treatment history and drug-use pattern, our analyses reveal that when they do not differ significantly in their response to treatment, especially if the severity of the problem (at admission) and the modality of treatment are controlled. On the other hand, if one assumes (reasonably so) that counseling and other treatment methods were indeed fine-tuned to take into account the different demographic and psychosocial backgrounds of the sexes, then our analysis shows that the treatment program was equally effective for men and women.

The different treatment modalities, however, do differ in their success rate, even when treating patients with similar severity of problem with LTR seemingly the best and Methadone maintenance by far the worst. One important feature of our analysis was the use of an index of problem severity, created by combining several "problem" indicators. Women tended to report a higher problem severity than men, which ties well with the historical evidence of the "gentler" sex suffering greater impairment for the same level of abuse. The problem severity index was found to be significantly associated with a number of common demographic variables in both sexes, with high severity cases being more likely to be younger, lesser educated, Caucasian (or having some other non-Hispanic and non-African American ethnicity), and opting for the long-term residential mode of treatment.

ACKNOWLEDGMENTS

This work is supported in part by NIH grants AA-12044 and DA-12468. The authors thank Dr. Jody Sindelar for her assistance with the DATOS database.

REFERENCES

(1.) Amaro H, Hardy-Fanta C. Gender relations in addiction and recovery. J Psychoact Drugs 1995; 27:325-337.

(2.) Anglin MD, Hser Y. Addicted women and crime. Criminology 1987; 25:359-396.

(3.) Hser YI, Anglin MD, Grella CE, Longshore D, Prendergast ML. Drug treatment careers: a conceptual framework and existing researching findings. J Subst Abuse Treat 1997; 14:1-16.

(4.) Brennan PL, Moos RH, Kim JY. Gender differences in the individual characteristics and life contexts of late-middle-aged and older problem drinkers. Addiction 1993; 88:781-790.

(5.) Weisner C, Greenfield T, Room R. Trends in the treatment of alcohol problems in the U.S. general population 1979 through 1990. Am J Public Health 1995; 85:55-60.

(6.) Weisner C, Schmidt L. Gender disparities in treatment for alcohol problems. J Am Med Assoc 1992; 268:1872-1876.

(7.) Anglin MD, Hser Y, Booth MW. Sex differences in addict careers. 4. Treatment. Am J Drug Alcohol Abuse 1987; 13:253-280.

(8.) Anglin MD, Hser Y, McGlothlin WH. Sex differences in addict careers. 2. Becoming addicted. Am J Drug Alcohol Abuse 1987; 13:59-71.

(9.) Hser Y, Anglin MD, McGlothlin W. Sex differences in addict careers. 1. Initiation of use. Am J Drug Alcohol Abuse 1987; 13:33-57.

(10.) Hser Y, Anglin MD, Booth MW. Sex differences in addict careers. 3. Addiction. Am J Drug Alcohol Abuse 1987; 13:231-251.

(11.) Griffin ML, Weiss RD, Mirin SM, Lange U. A comparison of male and female cocaine abusers. Arch Gen Psychiatry 1989; 46:122-126.

(12.) John U. Alcohol-dependent men and women in detoxification: some comparisons. Alcohol Clin Exp Res 1987; 11:152-157.

(13.) Beckman LJ, Amaro H. Personal and social difficulties faced by women and men entering alcoholism treatment. J Stud Alcohol 1986; 47:135-145.

(14.) Thom B. Sex differences in help-seeking for alcohol problems: 2. Entry into treatment. Br J Addict 1987; 82(9):989-997.

(15.) Marsh KL, Simpson DD. Sex differences in opioid addiction careers. Am J Drug Alcohol Abuse 1986; 12:309-329.

(16.) Grella CE, Joshi V. Gender differences in drug treatment careers among clients in the national drug abuse treatment outcome study. Am J Drug Alcohol Abuse 1999; 25(3):385-406.

(17.) Lukas SE, Sholar M, Lundahl LH, Lamas X, Kouri E, Wines JD, Kragie L, Mendelson JD. Sex differences in plasma cocaine levels and subjective effects after acute cocaine administration in human volunteers. Psychopharmacology 1996; 125:346-354.

(18.) Roberts DCS, Bennett SAL, Vickers GJ. The estrous cycle effects cocaine self-administration on a progressive ration schedule in rats. Psychopharmacology 1989; 98:408-411.

(19.) Levin JM, Holman BL, Mendelson JH, Teoh SK, Garada B, Johnson KA, Springer S. Gender differences in cerebral perfusion in cocaine abuse: Technetium-99m-HMPAO SPECT study of drug-abusing women. J Nucl Med 1994; 35:1902-1909.

(20.) Chang L, Ernst T, Strickland TL. Neurochemical abnormalities and gender effects in abstinent asymptomatic cocaine users, Abstract from the Society of Magnetic Resonance Fourth Scientific Meeting, New York, NY, 1996.

(21.) Stein RA, Strickland TL, Khalsa-Dennison E, Andre K. Gender differences in neuropsychological test performance among cocaine abusers. Arch Clin Neuropsychol 12:410-411.

(22.) Levin JM, Ross MH, Mendelson JH, Mello NK, Cohen BM, Renshaw PF. Sex differences in blood-oxygenation-level-dependent functional MRI with primary visual stimulation. Am J Psychiatry 155:434-436.

(23.) McLellan AT, Kushner H, Metzger D, Peters R, Smith I, Grissom G, Pettinati H, Argeriou M. The fifth edition of the addiction severity index: cautions, additions and normative data. J Subst Abuse Treat 1992; 9:261-275.

(24.) Simpson DD, Joe GW, Fletcher BW, Hubbard RL, Anglin MD. A national evaluation of treatment outcomes for cocaine dependence. Arch Gen Psychiatry 1999; 56(6):507-514.

(25.) Agresti A. Categorical Data Analysis. New York: Wiley-Interscience, 1990.

(26.) Hubbard RL, Craddock SG, Flynn PM, Anderson J, Etheridge RM. Overview of 1-year follow-up outcomes in the Drug Abuse Treatment Outcomes Study (DATOS). Psychol Addict Behav 1997; 11:261-278.

Suddhasatta Acharyya, Ph.D., and Heping Zhang, Ph.D. *

Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, Connecticut, USA

* Correspondence: Heping Zhang, Department of Epidemiology and Public Health, Yale University School of Medicine, 60 College Street, New Haven, CT 06520-8034, USA; Fax: (203) 785-6912; E-mail: heping.zhang@yale.edu.

COPYRIGHT 2003 Marcel Dekker, Inc.
COPYRIGHT 2003 Gale Group