Demographic and clinical characteristics of the participants are presented in . Mean lamotrigine dose was 150.9 mg/d, range 25 to 200 mg/d. Concomitant medications included lithium(n = 6, 10.5%), divalproex/valproate (n= 9, 15.8%), any anticonvulsant (n = 19, 33.3%), antipsychotics (n = 18, 31.6%), and antidepressants (n = 34, 59.7%). There was significant improvement from baseline in MADRS (df = 8, χ2 = 43.37, P < .0001). Additionally, there was significant improvement in GRID HAM-D (df = 8, χ2 = 46.10, P < .0001).
Baseline Demographic and Clinical Characteristics of 57 Older Adult Participants With Bipolar Depression
Baseline Clinical Variables and BD Depression Treatment Outcomes
Significant associations were found between cardiometabolic risk variables and BD depression treatment outcomes. Using a regression model to examine treatment outcome, greater reduction in the MADRS from baseline was associated with higher baseline cholesterol levels and higher baseline endocrine/ metabolic burden (CIRS-G item #13) at weeks 6 and 9, and also lower baseline YMRS scores (F(1, 30) = 4.28; P = .047) at week 9 (see for parameter estimates, standard errors, and P values). However, at 12 weeks, improvement in the MADRS from baseline was no longer significantly related to baseline cholesterol (F(1, 30) = 2.30; P = .140), CIRS-G item #13 (F(1, 30) = 1.24; P = .275), or baseline YMRS scores (F(1, 30) = 3.03; P = .092). We found no relationship between treatment response at 6, 9, or 12 weeks between blood pressure and BMI.
Parameter Estimates for Predictors of Change in Depression (MADRS) Scores at 6 and 9 Weeks in 57 Older Adults Treated With Lamotrigine
Longitudinal mixed modeling corroborated the regression models with the finding of a significant time-by-baseline cholesterol interaction (F(7, 165)= 2.16, P = .040). An AR(1) autoregressive correlation structure was estimated, with participant-level random intercept effects, and time and CIRS-G item #13 viewed as a categorical variables, and baseline YMRS and total cholesterol as covariates. At 9 weeks, estimated interactions indicated relatively lower expected MADRS scores in individuals with higher baseline cholesterol; while at 12 weeks, relatively higher MADRS scores were expected. For example, in participants with the same YMRS baseline and CIRS-G item #13 scores, but cholesterol level of 100 mg/dL ormore, would lead to a lower MADRS score of 2.35 at 9 weeks. At 12 weeks, the same assumptions would lead to a higher MADRS score of 1.22; while at 6 weeks, there would be essentially no expected difference (0.03 MADRS score).
Executive Function and BD Depression Treatment Outcomes
For the following analyses, participants were considered at higher cardiometabolic risk if either their baseline cholesterol level was greater than or equal to 224 mg/dL or their baseline CIRS-G item #13 score was equal to 2 or 3 (32 [56%] out of 57 depressed elders with BD). Participants were considered to have higher or lower neurocognitive status if their ratio of TRAILS B/TRAILS A time was below or above the median of 2.77, respectively. These groups were then fit to a regression model as covariates with reduction in MADRS score from baseline as the dependent variable. Greater reductions from baseline in MADRS scores at week 9 were associated with higher neurocognitive status (n = 16, t = 2.847, P = .014) and higher cardiometabolic risk (n = 16; t = −2.493, P = .027), meaning better executive function was associated with greater improvement in mood (in the context of higher cardiometabolic disease). However, we found no relationship between neurocognitive status and mood at 6 or 12 weeks.
Predictors of Treatment Termination
Dropout at 6, 9, and 12 weeks was 22.8% (n = 44 remaining), 31.6% (n = 39 remaining), and 33.3% (n = 38 remaining), respectively. There was a significant positive relationship between YMRS scores and the likelihood of treatment termination or dropout. A binary logistic regression model of dropout that included YMRS scores, sex, and cholesterol levels as explanatory variables revealed that higher YMRS scores were a significant predictor of dropout (df = 1, χ2 = 5.49; P = .019). Further supporting this finding, a Cox regression model of time until dropout that included YMRS scores, sex, and CIRS-G item #13 as explanatory variables revealed that higher YMRS scores were associated with a higher hazard rate or less time until dropout (df = 1, χ2 = 6.57; P = .010). The other explanatory variables in these 2 regression models were not significant predictors of dropout or hazard rate.