David C. Atkins

Randomized trial of behavioral activation, cognitive therapy, and antidepressant medication in the acute treatment of adults with major depression

Antidepressant medication is considered the current standard for severe depression, and cognitive therapy is the most widely investigated psychosocial treatment for depression. However, not all patients want to take medication, and cognitive therapy has not demonstrated consistent efficacy across trials. Moreover, dismantling designs have suggested that behavioral components may account for the efficacy of cognitive therapy. The present study tested the efficacy of behavioral activation by comparing it with cognitive therapy and antidepressant medication in a randomized placebo-controlled design in adults with major depressive disorder (N = 241). In addition, it examined the importance of initial severity as a moderator of treatment outcome. Among more severely depressed patients, behavioral activation was comparable to antidepressant medication, and both significantly outperformed cognitive therapy. The implications of these findings for the evaluation of current treatment guidelines and dissemination are discussed.

Rethinking how family researchers model infrequent outcomes: a tutorial on count regression and zero-inflated models.

Marital and family researchers often study infrequent behaviors. These powerful psychological variables, such as abuse, criticism, and drug use, have important ramifications for families and society as well as for the statistical models used to study them. Most researchers continue to rely on ordinary least-squares (OLS) regression for these types of data, but estimates and inferences from OLS regression can be seriously biased for count data such as these. This article presents a tutorial on statistical methods for positively skewed event data, including Poisson, negative binomial, zero-inflated Poisson, and zero-inflated negative binomial regression models. These statistical methods are introduced through a marital commitment example, and the data and computer code to run the example analyses in R, SAS, SPSS, and Mplus are included in the online supplemental material. Extensions and practical advice are given to assist researchers in using these tools with their data.

The Risk of Myocardial Infarction Associated With the Combined Use of Estrogens and Progestins in Postmenopausal Women

BACKGROUND While observational studies have suggested that unopposed estrogens reduce the incidence of coronary disease in postmenopausal women, there are few data on the effect of combined therapy with estrogens and progestins--a regimen adopted in recent years to minimize the risk of endometrial hyperplasia and cancer. In clinical trials, the addition of progestins has an adverse effect on serum lipid levels, and these lipid effects have raised the question of whether combined estrogen-progestin therapy increases the risk of coronary disease compared with the use of estrogen alone. METHODS We conducted a population-based, case-control study among enrollees of Group Health Cooperative of Puget Sound. Cases were postmenopausal women who sustained an incident fatal or nonfatal myocardial infarction in 1986 through 1990. Controls were a stratified random sample of female Group Health Cooperative enrollees frequency matched to the cases by age and calendar year. We reviewed the medical records of the 502 cases and 1193 controls and conducted brief telephone interviews with consenting survivors. The health maintenance organizations computerized pharmacy database was used to ascertain the use of postmenopausal hormones. For the primary analysis of current use, we classified women into one of three groups: (1) nonusers of hormones; (2) users of estrogens alone; or (3) users of combined therapy including both estrogens and progestins. Each group of hormone users was compared with nonusers. RESULTS After adjustment for potential confounding factors, the risk ratio of myocardial infarction associated with current use of estrogens alone was 0.69 (95% confidence interval, 0.47 to 1.02); and the risk ratio of myocardial infarction associated with current use of combined therapy was 0.68 (95% confidence interval, 0.38 to 1.22). Duration of combined-therapy use was relatively short, averaging less than 2 years in cases and controls. CONCLUSIONS In this case-control study, the reduced risk of myocardial infarction associated with the use of estrogens alone was consistent with previous observational studies. Although the 95% confidence interval only excluded a risk above 1.22, the current use of combined therapy was not associated with an adverse effect on the incidence of myocardial infarction in postmenopausal women.

Traditional Versus Integrative Behavioral Couple Therapy for Significantly and Chronically Distressed Married Couples

A randomized clinical trial compared the effects of traditional behavioral couple therapy (TBCT) and integrative behavioral couple therapy (IBCT) on 134 seriously and chronically distressed married couples, stratified into moderately and severely distressed groups. Couples in IBCT made steady improvements in satisfaction throughout the course of treatment, whereas TBCT couples improved more quickly than IBCT couples early in treatment but then, in contrast to the IBCT group, plateaued later in treatment. Both treatments produced similar levels of clinically significant improvement by the end of treatment (71% of IBCT couples and 59% of TBCT couples were reliably improved or recovered on the Dyadic Adjustment Scale; G. B. Spanier, 1976). Measures of communication also showed improvement for both groups. Measures of individual functioning improved as marital satisfaction improved.

Using multilevel models to analyze couple and family treatment data : Basic and advanced issues

Couple and family treatment data present particular challenges to statistical analyses. Partners and family members tend to be more similar to one another than to other individuals, which raises interesting possibilities in the data analysis but also causes significant problems with classical, statistical methods. The present article presents multilevel models (also called hierarchical linear models, mixed-effects models, or random coefficient models) as a flexible analytic approach to couple and family longitudinal data. The article reviews basic properties of multilevel models but focuses primarily on 3 important extensions: missing data, power and sample size, and alternative representations of couple data. Information is presented as a tutorial, with a Web appendix providing datasets with SPSS and R code to reproduce the examples.

Cholesterol Reduction and the Risk for Stroke in Men: A Meta-Analysis of Randomized, Controlled Trials

Table. SI Units and Drug Stroke and ischemic heart disease share important risk factors such as age, hypertension, diabetes, and cigarette smoking [1]. Although serum cholesterol is a strong and consistent risk factor for coronary heart disease, its importance in stroke remains controversial [1-5]. Although the incidence of thromboembolic stroke appears to increase with elevated serum cholesterol levels, a J-shaped association between cholesterol and risk for total stroke has been observed in several studies, due to an increased number of hemorrhagic strokes at the lowest cholesterol levels [6, 7]. Although interventions to lower serum cholesterol reduce mortality from heart disease, treatments have not affected total mortality rates in trials of high-risk persons without heart disease [8, 9]. This observation has raised concerns that treatments adversely affect noncardiac mortality and has fueled a debate over the risks and benefits of treating elevated serum cholesterol in asymptomatic patients [10, 11]. Although stroke remains the third leading cause of death in the United States, it is not known whether cholesterol-lowering treatments have favorable or unfavorable effects on stroke. None of the completed clinical trials of cholesterol reduction reported a sufficient number of strokes to estimate precisely the effect of treatment on stroke. Because the number of deaths from stroke in middle-aged U.S. men is roughly one tenth of that resulting from heart disease [12], much larger trials are needed to confirm an effect on stroke mortality. Using meta-analysis to combine results from independent studies [13-15], however, we evaluated completed primary and secondary prevention trials to examine the association between cholesterol reduction and risk for fatal or nonfatal stroke. We conducted a meta-analysis of randomized trials of cholesterol reduction to address three questions: 1) Does reducing serum cholesterol, through diet or medication, affect mortality risk for stroke in men? 2) Does reducing cholesterol affect the incidence of fatal and nonfatal stroke in men? and 3) Do the various interventions to lower cholesterol have the same effect on risk for stroke? Because nearly all trials of cholesterol reduction enrolled men exclusively, we could not explore similar questions for women. Methods Our study used a meta-analysis of all published randomized trials that examined the cardiovascular effects of diet or medications used to lower cholesterol. The end points of interest were fatal and nonfatal strokes occurring within the treatment period of the trial. We did not include transient ischemic attacks because diagnostic criteria for such events are more variable than are those for stroke. Before reviewing studies, we established eligibility criteria requiring that individual participants were randomized to active treatment or to a control group; that end points were assessed without knowledge of treatment group; and that stroke deaths or morbidity were reported separately. We included trials enrolling participants free of heart disease at baseline (primary prevention), trials selecting participants with a history of heart disease (secondary prevention), and trials that included interventions in addition to cholesterol reduction (multiple intervention). Controls received placebo or standard treatment in various trials. We excluded crossover trials [16, 17], short-term efficacy trials using serum cholesterol level as the primary end point, trials using surrogate end points such as coronary stenosis [18-20], and trials that randomized groups rather than individuals [21-23]. To ensure comparability of included studies, we excluded one study that was restricted to patients with previous strokes [24] and one study that used surgical interventions (ileal bypass) to lower cholesterol level [25]. Eligible trials were located using three independent approaches: 1) a computed MEDLINE search of all English-language reports published between 1966 and 1992 that cited cholesterol and clinical trials as major descriptors; 2) a review of references from five recent meta-analyses that examined the effects of cholesterol reduction on ischemic heart disease and total mortality rate [8, 26-29]; and 3) a review of bibliographies of each eligible trial identified using the first two methods. When multiple reports were published from a single trial, we relied on the report that contained the most complete data on cause of death. Where possible, data on women was excluded from trials that enrolled both men and women. Some women were included, however, in the Stockholm Ischemic Heart Disease study (20% of 555 participants) because the investigators did not report sex-specific outcomes [30]. All trials reported results based on an intention-to-treat analysis. Events that occurred after the conclusion of any trial [31-35] were not included in the quantitative analysis but were examined for evidence of a delayed effect of treatment. Numbers of participants and deaths in excluded trials were recorded to estimate the potential influence of excluded trials on the observed results. We used Petos modification of the Mantel-Haenszel method for combining data from 2 2 tables to examine the relation between cholesterol reduction and the risks for fatal stroke, nonfatal stroke, and fatal coronary heart disease in our primary analysis [14, 36]. Study size was the major determinant of the statistical weight given to individual trial results using this model. Although the study population (middle-aged, white men with elevated serum cholesterol levels) and the interventions (cholesterol reductions from 6% to 14%) were similar among included trials, we tested the assumption of a uniform effect of treatment by comparing treatment effect in subgroups of studies. Homogeneity of the treatment effect among the eligible trials was tested as described by Peto. A finding of significant heterogeneity indicated that the variation in treatment effect among studies exceeded that expected from random variation, possibly due to fundamental differences in the interventions, study samples, or designs. Summary estimates of effect may be inappropriate in cases of significant heterogeneity [37, 38]. A pooled estimate of the mean effect of treatment among all studies was calculated and expressed as an odds ratio, using the Peto method. We also used the maximum likelihood estimates described by Rothman [39] to calculate a pooled effect of treatment in terms of strokes per person-year in treated patients compared with controls. Finally, we explored whether the treatment effect was constant for different treatments or different study samples. Estimates of treatment effects pooled across all trials may not be appropriate if important differences exist between subgroups, and these differences may not be evident in an overall test of homogeneity [37, 38]. Using the methods described by Greenland [37], the effect of treatment was compared among the following subgroups: single-intervention compared with multiple-intervention trials; primary-prevention compared with secondary-prevention trials; drug compared with dietary treatments; clofibrate compared with nonclofibrate therapies; long ( 5 years) compared with shorter interventions; and greater ( 10%) compared with lesser reductions in cholesterol level. The effect of cholesterol reduction on the end points of fatal and nonfatal stroke, fatal coronary heart disease, and cerebral infarction was also compared. Because the smoking and blood pressure interventions included in multiple-intervention trials may have independent effects on stroke, all analyses were repeated after excluding these three trials. To test for a dose-response relation between magnitude of cholesterol reduction and risk for fatal stroke, we used a weighted linear regression of individual trial results, with percentage change in cholesterol as the independent variable and log relative risk as the dependent variable [37]. Results Our search located 18 independent trials of cholesterol-lowering therapy that met the first two criteria for trial design outlined previously. Five of the 18 (all secondary prevention trials begun before 1965) were excluded because they did not report separate data on strokes for experimental and control participants [40-44]. Study Characteristics The 13 remaining eligible trials are described in Table 1. Four primary-prevention trials used different drugs to lower cholesterol [45-48]. Two trials used modified diets among institutionalized patients with and without established heart disease [49, 50]. Two secondary-prevention trials used drugs [30, 51] and two used dietary supplements [52, 53]; the Coronary Drug Project, which included separate treatment arms using clofibrate and niacin, accounted for most participants in secondary trials. We did not consider data from three other treatment arms in this trial (dextrothyroxine; estrogen, 5 mg/d; or estrogen, 2.5 mg/d), which were terminated early [54]. The clofibrate and niacin groups from the Coronary Drug Project were combined in our initial analyses but were treated as separate trials when we compared the effects of different therapies. Three multiple-intervention trials enrolled men who were identified as high-risk patients by a multifactorial risk score [55-58]. The interventions in one trial [57] varied among treated patients, using either diet or drugs to lower cholesterol. Table 1. Design and Patient Population of Included Trials* The 13 trials together randomized 46 538 men, roughly two thirds in the 10 single-intervention trials. The age of the enrolled participants and the effects of treatment are summarized in Table 2. Due to variation in sample sizes, overall averages are presented by study and by participant. The average baseline cholesterol level for randomized participants was 6.53 mmol/L (252 mg/dL). Although the mean cholesterol reduction from baseline in treated patie

Brief Intervention for Problem Drug Use in Safety-Net Primary Care Settings: A Randomized Clinical Trial

IMPORTANCE Although brief intervention is effective for reducing problem alcohol use, few data exist on its effectiveness for reducing problem drug use, a common issue in disadvantaged populations seeking care in safety-net medical settings (hospitals and community health clinics serving low-income patients with limited or no insurance). OBJECTIVE To determine whether brief intervention improves drug use outcomes compared with enhanced care as usual. DESIGN, SETTING, AND PARTICIPANTS A randomized clinical trial with blinded assessments at baseline and at 3, 6, 9, and 12 months conducted in 7 safety-net primary care clinics in Washington State. Of 1621 eligible patients reporting any problem drug use in the past 90 days, 868 consented and were randomized between April 2009 and September 2012. Follow-up participation was more than 87% at all points. INTERVENTIONS Participants received a single brief intervention using motivational interviewing, a handout and list of substance abuse resources, and an attempted 10-minute telephone booster within 2 weeks (n = 435) or enhanced care as usual, which included a handout and list of substance abuse resources (n = 433). MAIN OUTCOMES AND MEASURES The primary outcomes were self-reported days of problem drug use in the past 30 days and Addiction Severity Index-Lite (ASI) Drug Use composite score. Secondary outcomes were admission to substance abuse treatment; ASI composite scores for medical, psychiatric, social, and legal domains; emergency department and inpatient hospital admissions, arrests, mortality, and human immunodeficiency virus risk behavior. RESULTS Mean days used of the most common problem drug at baseline were 14.40 (SD, 11.29) (brief intervention) and 13.25 (SD, 10.69) (enhanced care as usual); at 3 months postintervention, means were 11.87 (SD, 12.13) (brief intervention) and 9.84 (SD, 10.64) (enhanced care as usual) and not significantly different (difference in differences, β = 0.89 [95% CI, -0.49 to 2.26]). Mean ASI Drug Use composite score at baseline was 0.11 (SD, 0.10) (brief intervention) and 0.11 (SD, 0.10) (enhanced care as usual) and at 3 months was 0.10 (SD, 0.09) (brief intervention) and 0.09 (SD, 0.09) (enhanced care as usual) and not significantly different (difference in differences, β = 0.008 [95% CI, -0.006 to 0.021]). During the 12 months following intervention, no significant treatment differences were found for either variable. No significant differences were found for secondary outcomes. CONCLUSIONS AND RELEVANCE A one-time brief intervention with attempted telephone booster had no effect on drug use in patients seen in safety-net primary care settings. This finding suggests a need for caution in promoting widespread adoption of this intervention for drug use in primary care. TRIAL REGISTRATION clinicaltrials.gov Identifier: NCT00877331.

A tutorial on count regression and zero-altered count models for longitudinal substance use data

Critical research questions in the study of addictive behaviors concern how these behaviors change over time: either as the result of intervention or in naturalistic settings. The combination of count outcomes that are often strongly skewed with many zeroes (e.g., days using, number of total drinks, number of drinking consequences) with repeated assessments (e.g., longitudinal follow-up after intervention or daily diary data) present challenges for data analyses. The current article provides a tutorial on methods for analyzing longitudinal substance use data, focusing on Poisson, zero-inflated, and hurdle mixed models, which are types of hierarchical or multilevel models. Two example datasets are used throughout, focusing on drinking-related consequences following an intervention and daily drinking over the past 30 days, respectively. Both datasets as well as R, SAS, Mplus, Stata, and SPSS code showing how to fit the models are available on a supplemental website.

Assessing Clinical Significance: Does it Matter which Method we Use?.

Measures of clinical significance are frequently used to evaluate client change during therapy. Several alternatives to the original method devised by N. S. Jacobson, W. C. Follette, & D. Revenstorf (1984) have been proposed, each purporting to increase accuracy. However, researchers have had little systematic guidance in choosing among alternatives. In this simulation study, the authors systematically explored data parameters (e.g., reliability of measurement, pre-post effect size, and pre-post correlation) that might yield differing results among the most widely considered clinical significance methods. Results indicated that classification across methods was far more similar than different, especially at greater levels of reliability. As such, the existing methods of clinical significance appear highly comparable; future directions for clinical significance use and research are discussed.

Marital Status and Satisfaction Five Years Following a Randomized Clinical Trial Comparing Traditional versus Integrative Behavioral Couple Therapy.

OBJECTIVE To follow distressed married couples for 5 years after their participation in a randomized clinical trial. METHOD A total of 134 chronically and seriously distressed married couples were randomly assigned to approximately 8 months of either traditional behavioral couple therapy (TBCT; Jacobson & Margolin, 1979) or integrative behavioral couple therapy (IBCT; Jacobson & Christensen, 1998). Marital status and satisfaction were assessed approximately every 3 months during treatment and every 6 months for 5 years after treatment. RESULTS Pre- to posttreatment effect sizes on marital satisfaction were d = 0.90 for IBCT and d = 0.71 for TBCT, which were not significantly different. However, data through 2-year follow-ups revealed statistically significant superiority of IBCT over TBCT in relationship satisfaction, but subsequent data showed increasing similarity and nonsignificant differences in outcome. At 5-year follow-up for marital satisfaction relative to pretreatment, effect sizes were d = 1.03 for IBCT and d = 0.92 for TBCT; 50.0% of IBCT couples and 45.9% of TBCT couples showed clinically significant improvement. Relationship status, obtained on all 134 couples, revealed that 25.7% of IBCT couples and 27.9% of TBCT couples were separated or divorced. These follow-up data compared favorably to other, long-term results of couple therapy. CONCLUSION TBCT and IBCT both produced substantial effect sizes in even seriously and chronically distressed couples. IBCT produced significantly but not dramatically superior outcomes through the first 2 years after treatment termination but without further intervention; outcomes for the 2 treatments converged over longer follow-up periods.