Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Am J Psychiatry. Author manuscript; available in PMC 2010 January 14.
Published in final edited form as:
PMCID: PMC2806486

Do treatment improvements in PTSD severity affect substance use outcomes? A secondary analysis from a randomized clinical trial in NIDA’s Clinical Trials Network



The purpose of the analysis was to examine the temporal course of improvement in Posttraumatic Stress Disorder (PTSD) and substance use disorder symptoms among women in outpatient substance abuse treatment.


Participants were 353 women randomized to 12 sessions of trauma-focused or health education group treatment. PTSD and substance use assessments were conducted during treatment and at 1-week, 3-, 6-, and 12-months post treatment. A continuous Markov model was fit on participants’ four defined responder categories (non-responder, substance use responder, PTSD responder or global responder [improvement in both PTSD and substance use]) to investigate the temporal association between improvement in PTSD and substance use symptom severity during the study’s treatment phase. A generalized linear model was applied to test this relationship over follow-up.


Non-responders, substance use responders and global responders tended to maintain original classification; PTSD responders were significantly more likely to transition to global responders over time, indicating maintained PTSD improvement was associated with subsequent substance use improvement. Trauma-focused treatment was significantly more effective in achieving substance use improvement compared to the health education group, but only among those who were heavy substance users at baseline and had achieved significant PTSD reductions.


PTSD severity reductions were more likely to be associated with substance use improvement, with minimal evidence of substance use symptom reduction improving PTSD symptoms. Results support the self-medication model of coping with PTSD symptoms and an empirical basis for integrated interventions for improved substance use outcomes in patients with severe symptomatology.


Co-occurring Posttraumatic Stress Disorder (PTSD) and substance use disorders are a major public health concern in the U.S. due to negative physical and mental health consequences and poor treatment outcomes. Decades of cross-sectional epidemiologic studies repeatedly show high prevalence rates of PTSD among substance using subjects (1, for review). Advances have been made in behavioral and psychopharmacologic treatments for comorbid PTSD and substance use disorders; however, these studies demonstrate only modest improvements in outcome and have yet to specify mechanisms that influence outcomes. In the absence of clear, empirically-supported treatment guidelines, significant questions remain: 1) should PTSD symptoms be targeted to improve substance use outcomes, 2) should substance use symptoms be targeted to improve PTSD outcomes, or 3) should both sets of symptoms be addressed for clinically meaningful treatment benefits?

A number of conceptual models have hypothesized the causal pathways between PTSD and substance use disorders and the mechanisms responsible for treatment outcome. The predominant models include those that identify PTSD as a risk factor for substance use disorders (i.e., the “self-medication” hypothesis (2, 3)), substance use disorders as a risk factor for PTSD (i.e., substance use increases the likelihood of developing PTSD following exposure), and those who posit a shared neurobiological vulnerability (4). Longitudinal studies demonstrate the most support for the self-medication model (58); however, these studies have not yet accounted for the relationship between PTSD and substance use symptoms in response to interventions (i.e., how do improvements in one disorder affect the other and vice versa?). A recent study by Back and colleagues (9) examined this question by investigating the temporal course of improvement in PTSD and alcohol dependence symptoms among 94 individuals participating in a 12-week outpatient medication treatment trial. Improvements in PTSD led to improvements in alcohol dependence symptoms while the reciprocal relationship — improvements in alcohol dependence reducing PTSD symptoms — was not demonstrated. These preliminary findings suggest that 1) co-occurring PTSD symptoms may have a strong impact on alcohol dependence treatment outcome and 2) concurrent forms of treatment, specifically those that address PTSD symptoms early in treatment, may be important in optimizing outcomes for patients with comorbid PTSD and alcohol dependence.

The objective of this analysis was to examine the temporal course of improvement in PTSD and related substance use in the context of a clinical trial, addressing the limitations of the extant literature. The current sample of women participated in trauma-focused (Seeking Safety (10)) or health education groups (Women’s Health Education, unpublished treatment manual developed by Miller, Padian, and Tross, 1998) implemented in community-based substance abuse treatment programs within the National Institute on Drug Abuse Clinical Trials Network. Frequent PTSD symptom and substance use measurement were taken longitudinally to facilitate causal modeling. The primary research objectives were: (1) to test temporality in the relationship between PTSD and substance use symptom improvements weekly over the course of active study intervention, and (2) to test the relationships between PTSD reductions and substance use symptom outcomes over the 12 month longitudinal follow-up period, comparing those who received Seeking Safety with those who received Women’s Health Education.



Participants were included on the basis of the following criteria: 1) meeting DSM-IV criteria for full or sub-threshold PTSD (sub-threshold PTSD requires that participants meet either symptom cluster C or D, instead of both), 2) substance use within the past six months and a current diagnosis of drug or alcohol abuse or dependence, 3) aged 18–65, and 4) proficiency in English. Exclusion criteria were: 1) significant risk of suicidal/homicidal intent or behavior, 2) a history of schizophrenia-spectrum diagnosis, or 3) active (past two months) psychosis.

Study Design

Seven community-based treatment programs offering intensive outpatient treatment participated in the study. Sites were situated in urban (n=5) and suburban (n=2) settings and geographically located in the Western (n=1), Midwestern (n=1), Northeastern (n=2), and Southeastern (n=3) United States.

Recruitment occurred over 21-months in 2004 and 2005. Written informed consent was obtained from participants after complete description of the study. After completion of an eligibility screening followed by a baseline assessment, participants were randomly assigned to Seeking Safety or Women’s Health Education. Randomization was stratified by prescription psychotropic medication use and substance use diagnosis (alcohol use disorder only (8.8%) versus drug use disorder only (37.7%) or both drug and alcohol use disorders concurrently (53.5%)). Participants were assessed weekly during treatment for substance use and PTSD symptoms with full assessment repeated at 1-week, 3-, 6-, and 12-months post treatment. Independent assessors, blind to randomized assignment, conducted baseline and post-treatment assessments.

Study Interventions

After randomization, women attended an initial individual session with the counselor to discuss intervention assignment, group format, and rules. Groups had an open, rolling enrollment format, lasted approximately 75–90 minutes, and ran as long as 3 or more women were enrolled. Due to the criterion of needing 2 women present to conduct a group, many women took longer than 6 weeks (2 groups per week) to complete the interventions. All participants continued to attend treatment as usual offered by their treatment programs.

Seeking Safety is a structured cognitive-behavioral treatment with both safety/trauma and substance use components integrated into each session (10). Seeking Safety was abbreviated from 25 to 12 core sessions to better fit standard substance abuse treatment duration. Sessions include basic education on substance use disorders and PTSD, skill-building to prevent drug use and manage PTSD symptoms, cognitive restructuring with attention to maladaptive thoughts linked to substance use and trauma symptoms, and a focus on developing effective communication skills to build healthy support networks. All sessions had the same format: 1) check in, including reports of good coping skills or any “unsafe” behaviors; 2) session quotation, a brief point of inspiration to engage participants and link to session topic; 3) relating the material to patients’ lives, in which hand-outs are used to facilitate discussion and skill practice; and 4) check out, including a commitment to specific between-session skills practice. Each session covered a different topic (e.g., “PTSD: Taking Back Your Power” and “When Substances Control You”).

The Women’s Health Education control condition was adapted from a treatment grant protocol for female partners of injection drug users (unpublished treatment manual, Miller, Padian, and Tross, 1998). It is a psychoeducational, manualized treatment focused on topics such as pregnancy, nutrition, diabetes, hypertension, and HIV/sexually transmitted diseases. Women’s Health Education was designed to provide equivalent therapeutic attention, expectancy of benefit, and an issue-oriented focus, but without the theory-driven techniques of Seeking Safety, nor any explicit focus on substance abuse or trauma. Sessions followed a common format, including reviewing between session assignments, topic presentation with accompanying video or text, group exercises, and goal-setting.

Counselors and supervisors from each treatment program were centrally trained in their respective study interventions and later certified upon successful completion of a training group of at least 4 sessions. All intervention sessions were videotaped and 50% rated for adherence by supervisors. The coordinating site randomly selected and rated 25% of tapes reviewed by local supervisors to assure fidelity and inter-rater reliability.


Post Traumatic Stress Disorder

PTSD was measured with the Clinician Administered PTSD Scale (11), a structured interview that measures DSM-IV PTSD diagnosis and the frequency and intensity of symptoms over the past 30 days. The total scale severity score has a range of 0–136. The scale was administered at baseline and all follow up timepoints. The lead team conducted diagnostic reliability checks by listening to a subset (20%) of audiotaped scale assessments and held weekly conference calls with independent assessors to maintain competency and discuss challenging clinical issues. The Post Traumatic Stress Disorder Symptom Scale-Self Report was administered at all assessment timepoints, including the treatment phase, to measure the frequency and intensity of PTSD symptoms (12).

Substance Use

The Addiction Severity Index-Lite (13) was used to assess prior 30-day substance use at baseline and follow-up timepoints. The maximum number of days of use across 10 substance use categories (alcohol, cocaine, marijuana, opiates, sedatives, stimulants, heroin, barbiturates, inhalants, and hallucinogens) was categorized into three levels: abstinence (no use), light use (used 1–12 days), and heavy use (used 13 or more days [i.e., more than three days per week]), (14). The Addiction Severity Index Alcohol and Drug Composite Scores ranged from 0–1 (15). Each Alcohol composite score was recoded into four levels: abstinence, light (0.01–0.15), median (0.16–0.4), and heavy (0.41 or higher); for the Drug composite score, the levels were abstinence, light (0.01–0.1), median (0.11–0.2), and heavy (0.21 or higher). The cut points were determined statistically, chosen to equalize the sample in each level among users at baseline. The Substance Use Inventory is a series of self-report questions about quantity and frequency of substance use in the past 7 days adapted from the Time Line Follow-Back measure (16). The Substance Use Inventory was administered at all assessment time points, including the treatment phase.

Defining treatment responders during the treatment phase

Consistent with Brady and colleagues’ (17) scoring practices for the measurement of clinically significant changes in PTSD symptoms, improvement was defined as a 30% or greater reduction from baseline to each intervention visit. Substance use improvement was defined by the scoring conventions of Nunes and colleagues (18) as a 75% or greater reduction of drug/alcohol using days per week, measured by the Substance Use Inventory, compared with baseline levels. At each intervention phase visit, participant improvement was classified into one of four categorical variables: (1) non-responder: no improvement in either PTSD or substance use severity, (2) substance use responder: improvement in substance use only, (3) PTSD responder: improvement in PTSD only, and (4) global responder: improvement in both PTSD and substance use. A fifth category, drop-outs, was used for those who were no longer in treatment.

Statistical analyses

The first of two analytic methods was applied to investigate the temporality of the association between improvement in PTSD severity measured by the Posttraumatic Stress Disorder Symptom-Self Report and improvement in drug and alcohol use during the 6-week treatment phase of the study. A continuous Markov model was fit on the participants’ four defined responder categories described above.

The second strategy was applied longitudinally to test the relationship between PTSD and substance use symptom changes over the course of the one year follow-up and involved the application of generalized linear models. The generalized linear model was applied for repeated outcome measures, observed at 1-week, 3-months, 6-months, and 12-months after treatment. The three main outcomes were ordinal measures of 1) the maximum number of days used across 10 substances in the past 30 days, 2) Addiction Severity Index Alcohol composite, and 3) Addiction Severity Index Drug composite. Each of the three outcomes was modeled as a function of PTSD changes from baseline to each assessment point; intervention type (Seeking Safety or Women’s Health Education); time of assessment; baseline level of the relevant outcome variable; and pre-selected baseline covariates (race, age, education, and marital status). The possible interactions between PTSD changes, intervention type, and the baseline level of the outcome measure were tested and included in the final model if significant. In order to examine the difference in treatment effect among programs, site was included as an additional fixed effect, while the participant was a random variable. Generalized estimating equations (20) were used to estimate and test the models. The generalized estimating equations methodology is able to handle within-subject correlation arising from repeated measurements, requires no parametric distribution assumption for the outcomes, provides robust inference with respect to misspecification of the within-subject correlation, and considers missing at random. PROC GENMOD in SAS (SAS 9.1.3) was used to conduct the analysis.



Demographic and diagnostic characteristics for the randomized sample (N=353) are displayed in Table 1. There were no significant differences between treatment groups on sociodemographic factors, PTSD symptoms, or drug use at baseline. The average age of the sample was 39.2 years. Forty-six percent were Caucasian and one-third were African American. About half were divorced or separated and 37% were never married.

Table 1
Sociodemographic and Diagnostic Characteristics at Baseline for the Randomized Sample (N=353)

As specified by study eligibility criteria, all participants met current DSM-IV criteria for either full (80.4%) or subthreshold PTSD (19.6%). The average Clinician Administered PTSD Scale total score was in the severe range with a mean of 62.9 (SD = 19.4, range 19–119). The most frequent substances used in the past 30 days were cocaine and alcohol; given that the sample was recruited within substance abuse treatment programs, 40% were abstinent at baseline.

Eighty-two percent of participants attended at least one treatment session, with 56% attending 6 or more (the a priori definition of treatment completion). Retention rates were similar at each follow-up point (61%–63%) and did not differ significantly by study intervention type or frequency of drug use at baseline. Eighty-two percent of the sample had at least one post-treatment assessment.

Temporal association between PTSD and substance use improvement during treatment

In order to test our first hypothesis that PTSD reductions lead to changes in substance use, we examined data collected weekly during active study treatment. Six hundred and thirty-nine transition events were modeled: 144 non-responders, 289 substance use responders, 49 PTSD responders, and 157 global responders. Eighty transition events were to drop-out (27 non-responders, 33 substance use responders, 7 PTSD responders, and 13 global responders). Table 2 displays the proportion of transitions to each of the five classifications in successive weeks following baseline. Non-responders, substance use responders and global responders tended to maintain their original classification (e.g., if they improved in substance use only, they were most likely to remain a substance use responder). Participants who were initially classified as PTSD responders, however, were significantly more likely to transition to a global responder over time, indicating maintained PTSD improvement associated with subsequent substance use improvement.

Table 2
Transition Matrix showing the number of patients transitioning weekly from one to another of four response classes (Non-responders, Substance Use Responders, Post Traumatic Stress Responders, and Substance Use/Post Traumatic Stress Responders (Global ...

Inferential tests of treatment effects in the Markov model yielded no overall significant treatment effect (χ2(7) = 9.72, p = 0.21) and no significant treatment effects on individual transition intensity. That is, the association between PTSD and drug use did not differ between the two interventions. Table 3 displays the estimated transition probability matrix in successive weeks based on model 1. PTSD responders (3 → 4) were approximately 2.80 (= 0.37/0.13) times (95% CI: 1.52–4.58, based on 1,000 bootstrapping repetitions) more likely (in probability) than substance use responders (2 → 4) to change to global responders within one week.

Table 3
Probability Matrix of Averaged Weekly Transition Frequencies for Non-responders, Substance Use Responders, Post Traumatic Stress Responders, and Substance Use/Post Traumatic Stress Responders (Global Responders) over the Active Treatment Period

Association between PTSD change and substance use outcomes assessed longitudinally

The generalized multinomial logistic models of substance use over the follow-up period (1-week, 3- 6- and 12-month post treatment) indicated a significant effect of changes in Clinician Administered PTSD Scale total score from baseline to each follow-up point (Table 4). During the 12-month follow-up, maximum days of use, Addiction Severity Index Alcohol and Drug composite scores were all significantly related to the baseline level of substance use and PTSD improvements from baseline to each follow-up point. Further, for both maximum days of use and Drug composite scores, a three-way interaction between treatment group, baseline level of substance use, and PTSD improvements from baseline to each follow-up point was found such that the impact of PTSD improvement on substance use at follow-up significantly differed by treatment group and baseline level of drug use (χ2(2)=8.07, p = 0.02 for maximum days of use, χ2(2)=9.19, p = 0.03 for Drug composite). In the experimental group (Seeking Safety), one unit of improvement on Clinician Administered PTSD Scale scores for those who were heavy substance users at baseline decreased the odds of being in the heavy users group at follow-up by 4.6% (z score=4.35, p < 0.001), 1.3% (z score=1.49, p = 0.13) for light substance users, and no impact for those abstinent at baseline. In the control group (Women’s Health Education), one unit of improvement on Clinician Administered PTSD Scale scores for those who were heavy substance users at baseline decreased the odds of being in the heavy users group at follow-up by 0.6% (z score=0.75, p = 0.45), 2.3% (z score=2.60, p = 0.009) for light substance users, and 0.6% (z score=0.66, p = 0.51) for those who were abstinent. The effect of the improvement of Scale scores was significantly different between Seeking Safety heavy substance users and Women’s Health Education heavy substance users at baseline (z score=2.95, p = 0.003), but not statistically different between light substance users at baseline (z score=0.79, p = 0.43).

Table 4
The effects of Total Clinician Administered PTSD Scale (CAPS) Severity Score Change from Baseline to Each Follow-up Timepoint on the Maximum Number of Days of Use Across 10 Drug Categories by Treatment Group (N=353)

The effects on the Alcohol composite differed from the two drug use outcomes revealing a 2-way interaction effect between Clinician Administered PTSD Scale improvements and baseline alcohol use. The effect of one unit of improvement of Clinician Administered PTSD Scale severity on the probability of being a heavy alcohol user was stronger for baseline heavy alcohol users than light users (χ2(3)=15.85, p = 0.001).


When the comorbidity between PTSD and substance use disorders during active study intervention was examined, PTSD changes were found to impact substance use outcomes. Specifically, PTSD severity reductions were associated with substance use disorder improvement, with minimal evidence of substance use reduction improving PTSD symptoms. The findings, derived from two different sets of analyses spanning week-to-week probability data and longitudinal follow-up data, support the self-medication model as applied to populations with comorbid PTSD and addictive disorders.

Moreover, as predicted, PTSD-targeted treatment (Seeking Safety) was significantly more effective in achieving substance use improvement than the comparison group — but, only among those with heavy baseline substance use who had achieved significant PTSD reductions. Indeed, Seeking Safety is an integrated cognitive behavioral approach that actively links PTSD symptoms with “unsafe” substance use behaviors and whose efficacy has been largely demonstrated with active substance users (21). Accordingly, those in the sample who were abstinent at baseline may have benefited less from a focus on PTSD symptoms as related to substance use behaviors compared to those with active substance use. We further speculate that Seeking Safety was superior to the control condition for this group because those with more substance use also had more severe PTSD. In fact, baseline PTSD scores were statistically different among three levels of substance use defined by the maximum number of days of use (p < 0.05), such that those with heavy substance use at baseline had more severe PTSD. If substances are used to self-treat PTSD symptoms, then daily substance use may be considered a proxy for greater PTSD severity. This finding suggests that while for the average dually-diagnosed patient the additional benefits of PTSD-targeted interventions may be limited, PTSD-focused treatments like Seeking Safety offer an advantage to patients with more severe PTSD and substance use symptomatology.

Although only the subset of study participants with more severe baseline substance use appeared to benefit from the specific elements of Seeking Safety, this finding is in line with other treatment studies that show the largest treatment effects among those with the most severe problems (22). Because the detection of mediation in clinical trials is inherently a low power endeavor, it would be expected that differences would be most clearly observed in the subset of the sample where the intervention effect is most powerful, namely those with a high level of the baseline treatment targets (i.e., higher levels of substance use).

Overall, our results have important clinical implications for treating women with comorbid PTSD and substance use disorders. The findings further contradict conventional wisdom that addressing trauma-related symptoms will negatively impact substance use recovery. Instead, we demonstrate that trauma-focused treatment can lead to improvements in substance use outcomes in the context of PTSD symptom reductions, without decreasing participant attendance. Thus, we contend that the most effective treatment models are those that address PTSD before substance use or simultaneously. We propose this course of treatment, in contrast to treatment commonly offered in substance abuse treatment settings that lack a trauma-focus, especially because of the high prevalence rates of trauma histories and PTSD among such patients.

Our study has several limitations. First, 40% of the sample was abstinent at baseline which restricted the variability in alcohol and drug outcomes and, thereby, could have diluted the overall treatment effect. This is particularly true with respect to alcohol outcomes, as the vast majority of the sample met drug abuse or dependence criteria, with or without concurrent alcohol abuse or dependence (91.2%). Therefore, the findings may not generalize to a primarily alcohol dependent sample. Second, the sample was entirely comprised of women, which precludes extrapolation of results to men. A third consideration is that the participants received study interventions while enrolled in substance abuse treatment; receiving additional treatment focused on managing addictive behavior may have influenced outcomes.

The present study is only the second attempt of which we are aware to examine and test the temporal course of PTSD and substance use disorder symptom change. These data afforded a unique opportunity to discriminate between different mechanisms of comorbidity due to repeated and longitudinal measurement of symptoms. Our results offer support for the self-medication model and an empirical basis for PTSD-focused and integrated interventions for improved substance use outcomes in patients with severe symptomatology. Future studies should investigate these issues with men and examine the efficacy of PTSD-focused treatments for patients with varied substance use patterns to determine if such treatments are superior for them, as well.


Disclosures and Acknowledgements. Dr. Nunes has served on the advisory panel and speakers’ bureau of Alkermes/Cephalon, Inc. (resigned October 2007). All other authors report no competing interests. The authors would like to acknowledge the Clinical Coordinating Center at the Clinical Trials Network who collaborated in the design and conduct of the study and assisted with data management, quality assurance during the trial, and comments for consideration on the final draft of the manuscript. The authors also acknowledge the work of the research and clinical staff at the seven participating community substance abuse treatment programs and the women who participated in this clinical trial.

Grant Support. This study was supported by the following National Institute on Drug Abuse Clinical Trials Network grants: U10DA13035 (Edward Nunes, PI), U10DA13714 (Dennis Donovan, PI), U10DA13038 (Kathleen Carroll, PI), U10DA13732 (Eugene Somoza, PI), U10DA13727 (Kathleen Brady, PI), U10DA013720 (Jose Szapocznik, PI), U10DA013046 (John Rotrosen, PI). The following grant support is also acknowledged: National Institute on Drug Abuse, K24DA022412 (Edward Nunes).


Portions of these data were presented at the American Society for Addictions Medicine 40th Annual Medical-Scientific Conference, New Orleans, LA, April 30-May 3, 2009 and the 162nd American Psychiatric Association Annual Meeting, San Francisco, CA, May 16–21, 2009.

Clinical Trial Identification Number: National Institute on Drug Abuse, NCT00078156,


1. Chilcoat HD, Menard C. In: Epidemiological investigations: comorbidity of posttraumatic stress disorder and substance use disorder, in Trauma and substance abuse: causes, consequences and treatment of comorbid disorders. Ouimette PC, Brown PJ, editors. Washington, DC: American Psychological Association; 2003. pp. 9–28.
2. Khantzian EJ. Treating Addiction as a Human Process. London: Aronson; 1999.
3. Khantzian EJ. The ego, the self, and opiate addiction: theoretical and treatment considerations. Int J Psychoanal. 1977;5:189–199. [PubMed]
4. Chilcoat HD, Breslau N. Investigations of casual pathways between PTSD and drug use disorders. Addict Behav. 1998;23:827–840. [PubMed]
5. Dixon L, Haas G, Weiden PJ. Drug abuse in schizophrenia patients: clinical correlates and reasons for use. Am J Psychiatry. 1991;148:224–230. [PubMed]
6. Noordsy DL, Drake RE, Teague GB. Subjective experiences related to alcohol use among schizophrenics. J Nerv Ment Dis. 1991;179:410–414. (1991) [PubMed]
7. Addington J, Duckak V. Reasons for substance use in schizophrenia. Acta Psychiatr Scand. 1997;96:329–333. [PubMed]
8. Goswami S, Mattoo SK, Basu D, Singh G. Substance-abusing schizophrenics: do they self-medicate? Am J Addict. 2004;13:139–150. [PubMed]
9. Back SE, Brady KT, Sonne SC, Verduin M. Symptom improvement in co-occurring PTSD and alcohol dependence. J Nerv Ment Dis. 2006;194(9):690–696. [PubMed]
10. Najavits LM. Seeking safety: a treatment manual for PTSD and substance abuse. New York: Guilford Press; 2002.
11. Blake DD, Weathers FW, Nagy LM, Kaloupek DG, Gusman FD, Charney DS, Keane TM. The development of a Clinician-Administered PTSD Scale. J Trauma Stress. 1995;8(1):75–90. [PubMed]
12. Foa E, Riggs DS, Dancu CV, Constance V, Rothbaum BO. Reliability and validity of a brief instrument for assessing posttraumatic stress disorder. J Trauma Stress. 1993;6(4):459–473.
13. McLellan AT, Alterman AI, Cacciola J, Metzger D, O'Brien CP. A new measure of substance abuse treatment: initial studies of the treatment services review. J Nerv Ment Dis. 1992;180:101–110. [PubMed]
14. McLellan AT, Luborsky L, Cacciola J. New data from the Addiction Severity Index: reliability and validity in three centers. J Nerv Ment Dis. 1985;173:412–423. [PubMed]
15. McGahan PL, Griffith JA, Parente R, McLellan AT. Addiction Severity Index: composite scores manual. Philadelphia (PA): The University of Pennsylvania, Veterans Administration Centre for Studies of Addiction; 1986.
16. Sobell LC, Sobell MB. In: Timeline Followback: a technique for assessing self-reported alcohol consumption, in Measuring alcohol consumption: psychosocial and biological methods. Litten RZ, Allen JT, editors. Towota, NJ: Humana Press; 1992. pp. 41–72. (1992)
17. Brady K, Pearlstein T, Asnis GM, Baker D, Rothbaum B, Sikes CR, Farfel GM. Efficacy and safety of Sertraline treatment of posttraumatic stress disorder: a randomized controlled trial. JAMA. 2000;283:1837–1844. [PubMed]
18. Nunes EV, Quitkin FM, Donovan SJ, Deliyannides D, Ocepek-Welikson K, Koenig T, Brady R, McGrath PJ, Woody G. Imipramine treatment of opiate-dependent patients with depressive disorders: a placebo-controlled trial. Arch Gen Psychiatry. 1998;55(2):153–160. [PubMed]
19. Marshall G, Jones RH. Multi-state models and diabetic retinopathy. Stat Med. 1995;14:1975–1983. [PubMed]
20. Diggle PJ, Liang KY, Zeger SL. Analysis of longitudinal data. Oxford, UK: Oxford University Press; 2004.
21. Hien DA, Cohen LR, Miele GM, Litt LC, Capstick C. Promising treatments for women with comorbid PTSD and substance use disorders. Am J Psychiatry. 2004;161:1426–1432. [PubMed]
22. Klein DF, Ross DC. Reanalysis of the National Institute of Mental Health Treatment of Depression Collaborative Research Program General Effectiveness Report. Neuropsychopharmocology. 1993;8:241–251. [PubMed]