A novel application of the Intent to Attend assessment to reduce bias due to missing data in a randomized controlled clinical trial.

BACKGROUND: Missing data are unavoidable in most randomized controlled clinical trials, especially when measurements are taken repeatedly. If strong assumptions about the missing data are not accurate, crude statistical analyses are biased and can lead to false inferences. Furthermore, if we fail to measure all predictors of missing data, we may not be able to model the missing data process sufficiently. In longitudinal randomized trials, measuring a patient's intent to attend future study visits may help to address both of these problems. Leon et al. developed and included the Intent to Attend assessment in the Lithium Treatment - Moderate dose Use Study (LiTMUS), aiming to remove bias due to missing data from the primary study hypothesis. PURPOSE: The purpose of this study is to assess the performance of the Intent to Attend assessment with regard to its use in a sensitivity analysis of missing data. METHODS: We fit marginal models to assess whether a patient's self-rated intent predicted actual study adherence. We applied inverse probability of attrition weighting (IPAW) coupled with patient intent to assess whether there existed treatment group differences in response over time. We compared the IPAW results to those obtained using other methods. RESULTS: Patient-rated intent predicted missed study visits, even when adjusting for other predictors of missing data. On average, the hazard of retention increased by 19% for every one-point increase in intent. We also found that more severe mania, male gender, and a previously missed visit predicted subsequent absence. Although we found no difference in response between the randomized treatment groups, IPAW increased the estimated group difference over time. LIMITATIONS: LiTMUS was designed to limit missed study visits, which may have attenuated the effects of adjusting for missing data. Additionally, IPAW can be less efficient and less powerful than maximum likelihood or Bayesian estimators, given that the parametric model is well specified. CONCLUSIONS: In LiTMUS, the Intent to Attend assessment predicted missed study visits. This item was incorporated into our IPAW models and helped reduce bias due to informative missing data. This analysis should both encourage and facilitate future use of the Intent to Attend assessment along with IPAW to address missing data in a randomized trial.

Clinical and Health Outcomes Initiative in Comparative Effectiveness for Bipolar Disorder (Bipolar CHOICE): a pragmatic trial of complex treatment for a complex disorder.

BACKGROUND: Classic and second-generation antipsychotic mood stabilizers are recommended for treatment of bipolar disorder, yet there are no randomized comparative effectiveness studies that have examined the 'real-world' advantages and disadvantages of these medications. PURPOSE: We describe the strategic decisions in the design of the Clinical and Health Outcomes Initiative in Comparative Effectiveness for Bipolar Disorder (Bipolar CHOICE). This article outlines the key issues and solutions the investigators faced in designing a clinical trial that would maximize generalizability and inform real-world clinical treatment of bipolar disorder. METHODS: Bipolar CHOICE was a 6-month, multi-site, prospective, randomized clinical trial of outpatients with bipolar disorder. This study compares the effectiveness of quetiapine versus lithium, each with adjunctive personalized treatments (APTs). The co-primary outcomes selected are the overall benefits and harms of the study medications (as measured by the Clinical Global Impression-Efficacy Index) and the Necessary Clinical Adjustments (a measure of the number of medication changes). Secondary outcomes are continuous measures of mood, the Framingham General Cardiovascular Risk Score, and the Longitudinal Interval Follow up Evaluation Range of Impaired Functioning Tool (LIFE-RIFT). RESULTS: The final study design consisted of a single-blind, randomized comparative effectiveness trial of quetiapine versus lithium, plus APT, across 10 sites. Other important study considerations included limited exclusion criteria to maximize generalizability, flexible dosing of APT medications to mimic real-world treatment, and an intent-to-treat analysis plan. In all, 482 participants were randomized to the study, and 364 completed the study. LIMITATIONS: The potential limitations of the study include the heterogeneity of APT, selection of study medications, lack of a placebo-control group, and participants' ability to pay for study medications. CONCLUSION: We expect that this study will inform our understanding of the benefits and harms of lithium, a classic mood stabilizer, compared to quetiapine, a second-generation antipsychotic with broad-spectrum activity in bipolar disorder, and will provide an example of a well-designed and well-conducted randomized comparative effectiveness clinical trial.

Subject-level matching for imbalance in cluster randomized trials with a small number of clusters.

In a cluster randomized controlled trial (RCT), the number of randomized units is typically considerably smaller than in trials where the unit of randomization is the patient. If the number of randomized clusters is small, there is a reasonable chance of baseline imbalance between the experimental and control groups. This imbalance threatens the validity of inferences regarding post-treatment intervention effects unless an appropriate statistical adjustment is used. Here, we consider application of the propensity score adjustment for cluster RCTs. For the purpose of illustration, we apply the propensity adjustment to a cluster RCT that evaluated an intervention to reduce suicidal ideation and depression. This approach to adjusting imbalance had considerable bearing on the interpretation of results. A simulation study demonstrates that the propensity adjustment reduced well over 90% of the bias seen in unadjusted models for the specifications examined.

Effects of anxiety on the long-term course of depressive disorders.

BACKGROUND: It is well established that the presence of prominent anxiety within depressive episodes portends poorer outcomes. Important questions remain as to which anxiety features are important to outcome and how sustained their prognostic effects are over time. AIMS: To examine the relative prognostic importance of specific anxiety features and to determine whether their effects persist over decades and apply to both unipolar and bipolar conditions. METHOD: Participants with unipolar (n = 476) or bipolar (n = 335) depressive disorders were intensively followed for a mean of 16.7 years (s.d. = 8.5). RESULTS: The number and severity of anxiety symptoms, but not the presence of pre-existing anxiety disorders, showed a robust and continuous relationship to the subsequent time spent in depressive episodes in both unipolar and bipolar depressive disorder. The strength of this relationship changed little over five successive 5-year periods. CONCLUSIONS: The severity of current anxiety symptoms within depressive episodes correlates strongly with the persistence of subsequent depressive symptoms and this relationship is stable over decades.

Two propensity score-based strategies for a three-decade observational study: investigating psychotropic medications and suicide risk.

The US Food and Drug Administration issued separate warnings for suicidality with antidepressants and antiepileptic drugs in the past 5 years. This study describes methods for examining the association of these agents with suicide attempts and suicide deaths in more broadly generalizable samples than examined by the US Food and Drug Administration. An observational study of mood disorders was examined that includes three decades of prospective assessments. Because of sample size differences, two distinct longitudinal implementations of the propensity adjustment are used in separate analyses of antidepressants and antiepileptic drugs. Propensity score quintile-stratified safety analyses were used with the large antidepressant data set; whereas, propensity score matched safety analyses were used with the smaller antiepileptic drug data because stratification was not feasible. In each case, mixed-effects survival models compared the safety of participants when receiving the respective class of medication to periods when they did not receive that medication. When participants were more severely ill, they were significantly more likely to receive either class of psychotropics. Propensity quintile-stratified safety analyses found that risk of suicide attempts or suicides was significantly reduced when participants received antidepressants. In contrast, propensity score matched safety analyses found neither significant risk nor protection from suicidality among participants receiving antiepileptics.

Performance of a propensity score adjustment in longitudinal studies with covariate-dependent representation.

Longitudinal observational studies provide rich opportunities to examine treatment effectiveness during the course of a chronic illness. However, there are threats to the validity of observational inferences. For instance, clinician judgment and self-selection play key roles in treatment assignment. To account for this, an adjustment such as the propensity score can be used if certain assumptions are fulfilled. Here, we consider a problem that could surface in a longitudinal observational study and has been largely overlooked. It can occur when subjects have a varying number of distinct periods of therapeutic intervention. We evaluate the implications of baseline variables in the propensity model being associated with the number of post baseline observations per subject and refer to it as 'covariate-dependent representation'. An observational study of antidepressant treatment effectiveness serves as a motivating example. The analyses examine the first 20 years of follow-up data from the National Institute of Mental Health Collaborative Depression Study, a longitudinal, observational study. A simulation study evaluates the consequences of covariate-dependent representation in longitudinal observational studies of treatment effectiveness under a range of data specifications.The simulations found that estimates were adversely affected by underrepresentation when there was lower ICC among repeated doses and among repeated outcomes.

Methods to limit attrition in longitudinal comparative effectiveness trials: lessons from the Lithium Treatment - Moderate dose Use Study (LiTMUS) for bipolar disorder.

BACKGROUND: High attrition rates, which occur frequently in longitudinal clinical trials of interventions for bipolar disorder, limit the interpretation of results. PURPOSE: The aim of this article is to present design approaches that limited attrition in the Lithium Treatment - Moderate dose Use Study (LiTMUS) for bipolar disorder. METHODS: LiTMUS was a 6-month randomized, longitudinal multisite comparative effectiveness trial that enrolled bipolar participants who were at least mildly ill. Participants were randomized to either low to moderate doses of lithium or no lithium; other treatments needed for mood stabilization were administered in a guideline-informed, empirically supported, and personalized fashion to participants in both treatment arms. RESULTS: Components of the study design that may have contributed to low attrition (16%) among 283 participants randomized included the use of (1) an intent-to-treat design, (2) a randomized adjunctive single-blind design, (3) participant reimbursement, (4) assessment of intent to attend the next study visit (included a discussion of attendance obstacles when intention was low), (5) quality care with limited participant burden, and (6) target windows for study visits. LIMITATIONS: The relationships between attrition and effectiveness and tolerability of treatment have not been analyzed yet. CONCLUSIONS: These components of the LiTMUS design may have limited attrition and may inform the design of future randomized comparative effectiveness trials among similar patients and those from other difficult-to-follow populations.

Suicidal overdoses of psychotropic drugs by elderly in New York City: comparison with younger adults.

We compared New York City suicide victims aged 18-59 with those 60+ according to rates by which psychotropic/analgesic drugs and ethanol contributed to death. Barbiturates were more frequent in the elderly, while antidepressants were more frequent in younger adults. Addressing the potential for overdose with barbiturates may aid suicide prevention in the elderly.

A new proposal for randomized start design to investigate disease-modifying therapies for Alzheimer disease.

BACKGROUND: The increasing prevalence of Alzheimer disease (AD) and lack of effective agents to attenuate progression have accelerated research and development of disease modifying (DM) therapies. The traditional parallel group design and single time point analysis used in the support of past AD drug approvals address symptomatic benefit over relatively short treatment durations. More recent trials investigating disease modification are by necessity longer in duration and require larger sample sizes. Nevertheless, trial design and analysis remain mostly unchanged and may not be adequate to meet the objective of demonstrating disease modification. Randomized start design (RSD) has been proposed as an option to study DM effects, but its application in AD trials may have been hampered by certain methodological challenges. PURPOSE: To address the methodological issues that have impeded more extensive use of RSD in AD trial and to encourage other researchers to develop novel design and analysis methodologies to better ascertain DM effects for the next generation of AD therapies, we propose a stepwise testing procedure to evaluate potential DM effects of novel AD therapies. METHODS: Alzheimer Disease Assessment Scale-Cognitive Subscale (ADAS-cog) is used for illustration. We propose to test three hypotheses in a stepwise sequence. The three tests pertain to treatment difference at two separate time points and a difference in the rate of change. Estimation is facilitated by the Mixed-effects Model for Repeated Measures approach. The required sample size is estimated using Monte Carlo simulations and by modeling ADAS-cog data from prior longitudinal AD studies. RESULTS: The greatest advantage of the RSD proposed in this article is its ability to critically address the question on a DM effect. The AD trial using the new approach would be longer (12-month placebo period plus 12-month delay-start period; total 24-month duration) and require more subjects (about 1000 subjects per arm for the non-inferiority margin chosen in the illustration). It would also require additional evaluations to estimate the rate of ADAS-cog change toward the end of the trial. LIMITATIONS: A regulatory claim of disease modification for any compound will likely require additional verification of a drug's effect on a validated biomarker of Alzheimer's pathology. CONCLUSIONS: Incorporation of the RSD in AD trials is feasible. With proper trial setup and statistical procedures, this design could support the detection of a disease-modifying effect. In our opinion, a two-phase RSD with a stepwise hypothesis testing procedure could be a reasonable option for future studies.

Evaluation of psychiatric interventions in an observational study: issues in design and analysis.

Characteristics of randomized controlled clinical trials (RCTs) and observational studies of psychiatric intervention effectiveness are contrasted. Randomization drives treatment assignment in an RCT, whereas clinician and patient selection determine treatment in an observational study. Strengths and weaknesses of randomized and observational designs are considered. The propensity adjustment, a statistical approach that allows for intervention evaluation in a nonrandomized observational study, is described here. The plausibility of propensity adjustment assumptions must be carefully evaluated. This data analytic strategy is illustrated with the longitudinal observational data from the National Institute of Mental Health Collaborative Depression Study. Evaluations presented here examine acute and maintenance antidepressant effectiveness and demonstrate effectiveness of the higher categorical doses.

Site versus centralized raters in a clinical depression trial: impact on patient selection and placebo response.

The use of centralized raters who are remotely linked to sites and interview patients via videoconferencing or teleconferencing has been suggested as a way to improve interrater reliability and interview quality. This study compared the effect of site-based and centralized ratings on patient selection and placebo response in subjects with major depressive disorder. Subjects in a 2-center placebo and active comparator controlled depression trial were interviewed twice at each of 3 time points: baseline, 1-week postbaseline, and end point--once by the site rater and once remotely via videoconference by a centralized rater. Raters were blind to each others' scores. A site-based score of greater than 17 on the 17-item Hamilton Depression Rating Scale (HDRS-17) was required for study entry. When examining all subjects entering the study, site-based raters' HDRS-17 scores were significantly higher than centralized raters' at baseline and postbaseline but not at end point. At baseline, 35% of subjects given an HDRS-17 total score of greater than 17 by a site rater were given an HDRS total score of lower than 17 by a centralized rater and would have been ineligible to enter the study if the centralized rater's score was used to determine study entry. The mean placebo change for site raters (7.52) was significantly greater than the mean placebo change for centralized raters (3.18, P < 0.001). Twenty-eight percent were placebo responders (>50% reduction in HDRS) based on site ratings versus 14% for central ratings (P < 0.001). When examining data only from those subjects whom site and centralized raters agreed were eligible for the study, there was no significant difference in the HDRS-17 scores. Findings suggest that the use of centralized raters could significantly change the study sample in a major depressive disorder trial and lead to significantly less change in mood ratings among those randomized to placebo.

The interactive computer interview for mania.

OBJECTIVES: The Interactive Computer Interview for Mania (ICI-M) is a computer-administered interview that presents probes to assess symptom severity and utilizes a scoring algorithm to select follow-up questions and rate subject responses in accordance with rating scale anchor points. The current study examines the acceptability, feasibility, and reliability of the ICI-M as a potential method for evaluating the performance of human raters. METHODS: Participants with a diagnosis of bipolar I or II disorder completed both a live interview of the Young Mania Rating Scale with a human rater (LR) and the ICI-M. A panel of three expert raters reviewed each videotaped LR and assigned a consensus rating (CR). Participants completed a modified version of the Client Satisfaction Questionnaire to assess each method. RESULTS: Intraclass correlation coefficients were 0.91 between the ICI-M and CR and 0.97 between the LR and CR (n = 100), providing empirical support for the inter-rater reliability of each approach. Coefficient alphas indicated comparable internal consistency reliability: ICI-M = 0.82, LR = 0.83, and CR = 0.84. The ICI-M was significantly more sensitive in detecting symptomatology than the LR (p < 0.001) and the CR (p < 0.001), and resulted in significantly higher ratings than CR on mood, speech, psychotic content, and disruptive-aggressive behavior. While participants endorsed significantly higher overall satisfaction with LR, no significant differences emerged between ICI-M and LR regarding willingness to participate again or ability to understand the questions. CONCLUSIONS: The ICI-M is a well-accepted and reliable method for assessing manic symptoms. The ICI-M is a tool with adequate sensitivity to elicit symptoms and rate severity and is recommended as a tool to monitor and improve rater performance, not as a replacement of a human rater.

A new scale for assessing the quality of randomized clinical trials of psychotherapy.

CONTEXT: In 2004, the American Psychiatric Association's Committee on Research on Psychiatric Treatments appointed a subcommittee to investigate the status of empirical evidence with regard to psychodynamic psychotherapy. OBJECTIVE: As a part of this effort, the committee developed a rating scale designed to assess the quality of randomized controlled trials (RCTs) of psychotherapy. DATA SOURCES: A 25-item RCT of Psychotherapy Quality Rating Scale was generated by expert consensus. Interrater reliability, internal consistency, and validity testing were undertaken using 7 trained raters. STUDY SELECTION: A PubMed search was conducted to locate all RCTs of psychotherapies identified by their authors as being "psychodynamic" or "psychoanalytic" in origin and implementation. DATA EXTRACTION: A total of 69 RCTs were independently rated by 2 raters. DATA SYNTHESIS: The scale was found to have good interrater reliability (total score intraclass correlation = 0.76), internal consistency (Cronbach alpha = .87), and external validity. CONCLUSIONS: This scale establishes a new standard for the design and execution of psychotherapy RCTs and provides a systematic empirical method for evaluating the quality of published RCTs.

Sample sizes required to detect two-way and three-way interactions involving slope differences in mixed-effects linear models.

Based on maximum likelihood estimates obtained from mixed-effects linear models, closed-form power functions are derived to detect two-way and three-way interactions that involve longitudinal course of outcome over time in clinical trials. Sample size estimates are shown to decrease with increasing within-subject correlations. It is further shown that when clinical trial designs are balanced in group sizes, the sample size required to detect an effect size for a three-way interaction is exactly fourfold that required to detect the same effect size of a two-way interaction. Furthermore, this fourfold relationship virtually holds for unbalanced allocations of subjects if one factor is balanced in the three-way interaction model. Simulations are presented that verify the sample size estimates for two-way and three-way interactions.

Manic/hypomanic symptom burden and cardiovascular mortality in bipolar disorder.

OBJECTIVES: To compare the risk for cardiovascular mortality between bipolar I and bipolar II subtypes and determine correlates of cardiovascular mortality. Bipolar disorder conveys an increased risk of cardiovascular mortality. METHODS: Participants with major affective disorders were recruited for the National Institute of Mental Health Collaborative Depression Study and followed prospectively for up to 25 years. A total of 435 participants met the diagnostic criteria for bipolar I (n = 288) or bipolar II (n = 147) disorder based on Research Diagnostic Criteria at intake and measures of psychiatric symptoms during follow-up. Diagnostic subtypes were contrasted by cardiovascular mortality risk using Cox proportional hazards regression. Affective symptom burden (the proportion of time with clinically significant manic/hypomanic or depressive symptoms) and treatment exposure were additionally included in the models. RESULTS: Thirty-three participants died from cardiovascular causes. Participants with bipolar I disorder had more than double the cardiovascular mortality risk of those with bipolar II disorder, after controlling for age and gender (hazard ratio = 2.35, 95% Confidence Interval = 1.04-5.33; p = .04). The observed difference in cardiovascular mortality between these subtypes was at least partially confounded by the burden of clinically significant manic/hypomanic symptoms which predicted cardiovascular mortality independent of diagnosis, treatment exposure, age, gender, and cardiovascular risk factors at intake. Selective serotonin uptake inhibitors seemed protective although they were introduced late in follow-up. Depressive symptom burden was not related to cardiovascular mortality. CONCLUSIONS: Participants with bipolar I disorder may face a greater risk of cardiovascular mortality than those with bipolar II disorder. This difference in cardiovascular mortality risk may reflect manic/hypomanic symptom burden.

Empirical typology of bipolar I mood episodes.

BACKGROUND: Much remains unknown about the phenomenology of bipolar I disorder. AIMS: To determine the type of bipolar I mood episodes that occur over time, and their relative frequency. METHOD: A total of 219 individuals with Research Diagnostic Criteria bipolar I disorder were prospectively followed for up to 25 years (median 20 years). Psychopathology was assessed with the Longitudinal Interval Follow-up Evaluation. RESULTS: Overall, 1208 mood episodes were prospectively observed. The episodes were empirically classified as follows: major depression, 30.9% (n = 373); minor depression, 13.0% (n = 157); mania, 20.4% (n = 246); hypomania, 10.4% (n = 126); cycling, 17.3% (n = 210); cycling plus mixed state, 7.8% (n = 94); and mixed, 0.2% (n = 2). CONCLUSIONS: Cycling episodes constituted 25% of all episodes. Work groups revising ICD-10 and DSM-IV should add a category for bipolar I cycling episode.

Early adversity in chronic depression: clinical correlates and response to pharmacotherapy.

BACKGROUND: There is growing evidence suggesting that early adversity may be a marker for a distinct pathway to major depressive disorder (MDD). We examined associations between childhood adversity and a broad variety of clinical characteristics and response to pharmacotherapy in a large sample of patients with chronic forms of MDD. METHODS: Subjects included 808 patients with chronic forms of MDD (chronic MDD, double depression, or recurrent MDD with incomplete recovery between episodes and a total continuous duration of >2 years) who were enrolled in a 12-week open-label trial of algorithm-guided pharmacotherapy. Baseline assessments included a semi-structured diagnostic interview, and clinician- and self-rated measures of depressive symptoms, social functioning, depressotypic cognitions, and personality traits, and childhood adversity. Patients were re-evaluated every 2 weeks. RESULTS: A longer duration of illness; earlier onset; greater number of episodes, symptom severity, self-rated functional impairment, suicidality, and comorbid anxiety disorder; and higher levels of dysfunctional attitudes and self-criticism were each associated with multiple forms of childhood adversity. A history of maternal overcontrol, paternal abuse, paternal indifference, sexual abuse, and an index of clinically significant abuse each predicted a lower probability of remission. Among patients completing the 12-week trial, 32% with a history of clinically significant abuse, compared to 44% without such a history, achieved remission. CONCLUSIONS: These findings indicate that a history of childhood adversity is associated with an especially chronic form of MDD that is less responsive to antidepressant pharmacotherapy.

The NimStim set of facial expressions: judgments from untrained research participants.

A set of face stimuli called the NimStim Set of Facial Expressions is described. The goal in creating this set was to provide facial expressions that untrained individuals, characteristic of research participants, would recognize. This set is large in number, multiracial, and available to the scientific community online. The results of psychometric evaluations of these stimuli are presented. The results lend empirical support for the validity and reliability of this set of facial expressions as determined by accurate identification of expressions and high intra-participant agreement across two testing sessions, respectively.

Lithium treatment -- moderate dose use study (LiTMUS) for bipolar disorder: rationale and design.

BACKGROUND: Recent data indicate that lithium use for bipolar disorder has declined over the last decade and that lithium largely has been replaced with alternate, commercially promoted medications that may or may not result in better outcomes. PURPOSE: This article describes the rationale and study design of LiTMUS, a multi-site, prospective, randomized clinical trial of outpatients with bipolar disorder. LiTMUS seeks to address whether initiating therapy at lower doses of lithium as part of optimized treatment (OPT, guideline-informed, evidence-based, and personalized pharmacotherapy) improves outcomes and decreases the need for other medication changes across 6 months of therapy. METHODS: LiTMUS will randomize 284 adults with bipolar disorder (Type I or II) across 6 study sites. The co-primary outcomes are overall illness severity on clinical global improvement scale for bipolar disorder and a novel measure, necessary clinical adjustments. This metric provides a composite that reflects both clinical response and tolerability. Other relevant outcomes include full symptomatic recovery, quality of life, suicidal behaviors, and moderators of suicidality. RESULTS: As of August 28th, 2009, we have consented 338 patients and randomized 281 for this study. LIMITATIONS: The potential limitations of the study include an arbitrary definition of 'low, but effective' doses of lithium, lack of a placebo-controlled group, open treatment, and use of a new outcome measure (i.e., necessary clinical adjustments). CONCLUSION: We expect that this study will inform our understanding of the effectiveness of low to moderate doses of lithium therapy for individuals with bipolar disorder.

Enhancing clinical trial design of interventions for posttraumatic stress disorder.

The 2008 Institute of Medicine review of interventions research for posttraumatic stress disorder (PTSD) concluded that new, well-designed studies are needed to evaluate the efficacy of treatments for PTSD. The Department of Veterans Affairs (VA), the Department of Defense, and the National Institute of Mental Health convened a meeting on research methodology and the VA issued recommendations for design and analysis of randomized controlled clinical trials (RCTs) for PTSD. The rationale that formed the basis for several of the components of the recommendations is discussed here. Fundamental goals of RCT design are described. Strategies in design and analysis that contribute to the goals of an RCT and thereby enhance the likelihood of signal detection are considered.