Comparison of Low and Moderate Dosages of Extended-Release Quetiapine in Borderline Personality Disorder: A Randomized, Double-Blind, Placebo-Controlled Trial

Persons 18–45 years of age with moodiness, impulsivity, distrustfulness, and difficult relationships were recruited through referral, advertisements, and word of mouth. After screening with the Diagnostic Interview for DSM-IV Personality Disorders (19) to confirm the presence of DSM-IV borderline personality disorder, we administered the Structured Clinical Interview for DSM-IV to assess comorbid disorders (20). Participants had to meet Revised Diagnostic Interview for Borderlines criteria (21) for borderline personality disorder and could not meet current criteria for major depressive disorder, posttraumatic stress disorder, panic disorder, or obsessive-compulsive disorder. They were required to have a total score ≥9 on the Zanarini Rating Scale for Borderline Personality Disorder (“Zanarini scale”) (22) at visit 2. Individuals were excluded if they had ever met criteria for a psychotic disorder, had a primary neurological condition, or were cognitively impaired; had current substance dependence or had recently abused opiates, amphetamine, barbiturates, cocaine, or hallucinogens; were medically unstable; had a history of lack of response to an atypical antipsychotic; were pregnant or lactating; or were acutely suicidal.

Participants entering the study could not begin any type of psychotherapy during the study. Concomitant use of benzodiazepines was allowed at dosages equivalent to ≤1.0 mg/day of lorazepam; only one participant (in the placebo group) took benzodiazepines during the study. Episodic use of anticholinergics was allowed for sleep. No other psychotropic medication was permitted.

Rater-administered scales included the Zanarini scale (22), the Montgomery-Åsberg Depression Rating Scale (MADRS) (23), the Modified Overt Aggression Scale (24), the Young Mania Rating Scale (25), and the Global Assessment of Functioning Scale (GAF). Self-administered measures included a self-report version of the Zanarini scale (26), the Borderline Evaluation of Severity Over Time (27), the Barratt Impulsiveness Scale (28), the Symptom Checklist–90–Revised (SCL-90-R) (29), and the Sheehan Disability Scale (30).

The primary outcome measure was the Zanarini scale total score. This semistructured interview has anchored ratings (0=no symptoms, 4=severe symptoms) on nine items that correspond to the DSM-IV borderline personality disorder criteria. Its subscales were considered secondary efficacy measures. Other secondary measures included the self-rated version of the Zanarini scale, the MADRS, the Borderline Evaluation of Severity Over Time, the Modified Overt Aggression Scale, the GAF, the Barratt Impulsiveness Scale, the SCL-90-R, the Young Mania Rating Scale, and the Sheehan Disability Scale.

Adverse events, vital signs, electrocardiogram findings, laboratory values, and extrapyramidal symptoms were assessed. Laboratory tests included clinical chemistry, electrolyte levels, lipid profile, prolactin level, and hematology panels. A urine drug screen and pregnancy test were performed. These tests were performed at protocol-specified time points and when clinically indicated. Extrapyramidal symptoms were assessed with the Simpson-Angus Rating Scale (31), the Barnes Akathisia Scale (32), and the Abnormal Involuntary Movement Scale (33).

A sample size of 33 in each group was considered sufficient for detecting an effect size greater than 0.68 with a power of 0.80, an alpha of 0.05 (two-tailed), and a 15% dropout rate. The numbers of participants assigned to 150 mg/day of quetiapine (N=33), 300 mg/day of quetiapine (N=33), and placebo (N=29) were close to those planned.

The analyses included comparisons of baseline (visit 2) severity and background variables, time to adverse events, time to study discontinuation, and treatment response. All participants who were assigned to a treatment group and met study inclusion criteria were included in each analysis. Statistical tests were performed using a two-sided alpha of 0.05.

Pearson’s chi-square test (or Fisher’s exact test) was used to test for categorical baseline group differences. The Kruskal-Wallis test was used to test for dimensional baseline group differences. Mean severity scores from the baseline visit were compared using analysis of variance.

Adverse events were recorded as mild, moderate, or severe, with additional notations if an adverse event was thought to be related to the study medication. Time to adverse event was defined as the first postscreening visit with an adverse event reported, with censoring occurring if an event was not reported by the time of study discontinuation or study completion. In addition, we considered time to different types of adverse events (e.g., appetite change, bodily pain). Cox proportional hazards regression analysis was performed to compare the treatment groups on time to each definition of adverse event.

Time to discontinuation was defined as the last postscreening visit attended during the treatment phase, with censoring occurring if the participant attended visit 10. The analysis examined predictors of time to discontinuation. First, the three groups were compared with no covariates. Next, we examined severity of adverse events as time-dependent predictors of discontinuation. The analysis included any event, as well as the different types of adverse events. In addition, we examined illness severity (based on Zanarini scale score) as a time-dependent predictor. This was done to provide a better understanding of factors related to discontinuation. Cox proportional hazards regression analysis was used to model time to discontinuation.

For the primary efficacy variable, a mixed-effects model was used that included terms for treatment group, linear time effect (weeks since visit 2 [baseline]), quadratic time effect, site, and treatment-by-time interactions. (Because one site enrolled only three participants, those participants were pooled with the smaller of the two remaining sites.) Each participant’s outcome profile during the 8-week treatment phase (visit 2 to visit 10) was summarized with a subject-specific intercept and slope. We fitted a shared parameter model, where the treatment-response model (the mixed-effects model) and a time-to-discontinuation model were fitted simultaneously, with the subject-specific intercepts and slopes from the response model used as predictors of discontinuation (34). The shared-parameter model reduces bias in treatment effects caused by informative dropout (35). The NLMixed procedure in SAS was used to fit the shared parameter model (36, 37). The same model was also used for all other response measures for which data were collected at each visit. The mixed-effects model was used for measures collected at visits 2, 6, and 10.

The statistical tests for the null hypothesis were performed by testing differences in mean change from baseline for the groups receiving 150 mg/day or 300 mg/day of quetiapine compared with the placebo group. For each group, mean change from baseline (visit 2) to week 8 of treatment (visit 10) was calculated as a function of the group’s linear and quadratic coefficients. Effect size (d) was defined as the group difference from placebo in change from visit 2 to visit 10, expressed in standard deviation units. The same procedure was applied to secondary efficacy variables and measures of extrapyramidal symptoms, as well as to evaluate changes in blood pressure and heart rate. Participants were classified as responders if they had a reduction of ≥50% in Zanarini scale total score; the Cox proportional hazards regression model was used to model time to response. Among participants who completed the study, number needed to treat was calculated based on the proportion of participants whose Zanarini scale total score at visit 10 represented a reduction of ≥50% relative to visit 2. Levels of serum electrolytes, glucose, lipids, and prolactin were assessed at visit 2 and visit 10, and group differences in changes were tested using analysis of variance.

The numbers of participants reporting adverse events are presented in Table 3. Overall, 88% reported at least one adverse event, with 82% reporting an adverse event thought to be related to the study medication. Among all adverse events, 68% were suspected to be related to the study drug. No serious adverse events occurred.

Cox proportional hazards regression models were used to test for group differences in time to adverse events. Risk of any type of adverse event was not significantly different for the quetiapine groups compared with the placebo group. Risk of an adverse event thought to be related to the study medication was higher for the moderate-dosage quetiapine group (hazard ratio=1.73), but the result was not significant (p=0.074). For the moderate-dosage quetiapine group, risk was elevated for sedation (hazard ratio=2.16, p=0.021), change in appetite (hazard ratio=3.89, p=0.018), and dry mouth (hazard ratio=16.77, p=0.007). For the low-dosage quetiapine group, risk was higher for dry mouth (hazard ratio=9.32, p=0.034). The hazard ratios suggest that sedation, change in appetite, dry mouth, and dizziness are more likely with higher dosages of quetiapine.

Of all participants assigned to a treatment group, 64 (67%) completed the study, 23 of them in the placebo group, 22 in the low-dosage quetiapine group, and 19 in the moderate-dosage quetiapine group. Eight participants dropped out before a postbaseline assessment. Risk of discontinuation was higher among participants who received quetiapine, but the differences from placebo were not significant for either dosage group. Illness severity (measured by the Zanarini scale total score as a time-varying predictor) was not associated with discontinuation. Risk of discontinuation increased with severity of any adverse event (hazard ratio=1.74, p=0.018). Sedation was predictive of discontinuation (hazard ratio=1.77, p=0.025).

For all treatment groups, Zanarini scale total score improved during the treatment phase (Figure 1). For the quetiapine groups, the rate of improvement was highest from visit 2 to visit 6; after that, improvement plateaued. For this reason, the model included a quadratic effect for time. Efficacy results (Table 4) are based on the shared parameter model, which estimates each treatment group’s change from baseline, accounting for informative dropout. The difference in improvement between the low-dosage quetiapine group and the placebo group was statistically significant (d=−0.79, p=0.031); the difference between the moderate-dosage quetiapine group and the placebo group was not (d=−0.41, p=0.265). The mean score decreased by 1.22 points per week for the low-dosage quetiapine group, 0.99 points per week for the moderate-dosage quetiapine group, and 0.75 points per week for the placebo group. The difference between the two quetiapine groups was not significant. Time to treatment response was faster for the quetiapine groups relative to the placebo group (low-dosage group: hazard ratio=2.54, p=0.007; moderate-dosage group: hazard ratio=2.37, p=0.011). Including participants who discontinued before visit 10, the percentage classified as responders for at least one postbaseline visit was 82% for the low-dosage quetiapine group, 67% for the moderate-dosage quetiapine group, and 62% for the placebo group. (Some participants were classified as responders but were reclassified as nonresponders by their last visit.) Among the 31 participants who did not complete the study, 10 (32%) were responders. Among participants who completed the study, the visit 10 response rates were 82% for the low-dosage quetiapine group, 74% for the moderate-dosage quetiapine group, and 48% for the placebo group. From these rates, number needed to treat is estimated at 2.9 for 150 mg/day of quetiapine and 3.9 for 300 mg/day of quetiapine.

Both quetiapine dosages were superior to placebo for many secondary efficacy measures, including the self-rated Zanarini scale total score and subscale scores (except impulsivity); the Borderline Evaluation of Severity Over Time total score as well as the thoughts and feelings subscale; the Modified Overt Aggression Scale score; and work or school days lost because of symptoms (from the Sheehan Disability Scale). Neither quetiapine dosage was superior to placebo with regard to the Barratt Impulsiveness Score, but the moderate-dosage group had better scores on the Young Mania Rating Scale (d=−0.50, p=0.019) and the SCL-90-R general severity index (d=−0.62, p=0.033) (Table 4).

The moderate-dosage quetiapine group experienced a significant increase in heart rate (1.03 bpm/week). The low-dosage quetiapine group experienced a significant decrease in systolic blood pressure (0.53 mmHg/week). From visits 1 to 10, triglyceride levels increased 9.8 mg/dL on average (SD=52.1), with no significant group differences. There were no significant group differences for serum levels of electrolytes, glucose, lipids, and prolactin. There were no significant group differences with respect to changes in extrapyramidal symptoms from baseline to endpoint. Urine pregnancy tests were conducted at visits 1, 6, and 10. One woman tested positive at visit 6 and discontinued study participation; all other results were negative.

Among participants who completed the study, the mean weight gained between visit 1 and visit 10 was 1.0 lb (SD=4.4) for the placebo group, 1.0 lb (SD=7.2) for low-dosage quetiapine group, and 3.0 lb (SD=9.2) for moderate-dosage quetiapine group. Standard deviations for weight gain were greater for the quetiapine groups relative to the placebo group (p=0.028 for the low-dosage group, and p=0.001 for the moderate-dosage group). Median weight gain was not significantly different between groups.

Sixty participants (63%) attended visit 11, which occurred 1 week after study medication was discontinued. Twelve participants (20%) reported a new adverse event, with three reports of increased moodiness or irritability and three of trouble sleeping. All adverse events were rated as mild.