Predictors and correlates for weight changes in patients co-treated with olanzapine and weight mitigating agents; a post-hoc analysis

Patients

From the pool of participants who were enrolled in these 3 clinical studies, patients with a baseline BMI ≥ 25 kg/m² who were receiving treatment with olanzapine and were randomized to 1 of the adjunctive weight-mitigating agents or placebo were included in these analyses. In addition, the 16-week time point was used as a common endpoint. Detailed study design information, including inclusion and exclusion criteria, can be found in the primary publications for the nizatidine and amantadine studies [13, 14] and in the CTR summary for the sibutramine study (http://www.lillytrials.com/results/by_product/results_zyprexa.html, Trial ID: 5102). The sibutramine and the amantadine studies were designed as weight-reduction studies (i.e., patients had already experienced a pre-specified threshold of weight gain while receiving olanzapine treatment), while the nizatidine study evaluated weight gain prevention after initiation of olanzapine treatment.

Definition of Outcomes

For the purpose of these analyses, we defined a priori successful outcome as the occurrence of ≥ 2 kg weight loss and unsuccessful outcome as ≥ 1 kg weight gain. We discriminated between weight loss and weight gain at any time during the study versus weight loss and weight gain sustained to the 16-week endpoint or to study discontinuation.

Eating Behavior Assessment

Outcome measures included 3 eating assessment scales: the Eating Inventory (EI), [15] the Eating Behavior Assessment (EBA, a Lilly-developed scale, not validated), and the Visual Analog Scale (VAS). Since the focus of our analyses was on predictors and correlates rather than treatment comparisons, clinically meaningful, non-validated scales are acceptable explanatory variables for use in a Cox proportional hazards regression. The EI is a 51-item questionnaire that measures 3 factors: cognitive restraint of eating, disinhibition of eating, and susceptibility to hunger. The EBA consists of 9 items and is used to determine eating behavior during the previous week, rated from 0 ("not at all," meaning the patient reported not experiencing the behavior/feeling at all) to 4 ("extremely," meaning the patient reported exceedingly experiencing the behavior/feeling). The VAS is used to determine eating behavior during the previous 24 hours and consists of 3 items (hunger, interest in food, and appetite) rated from 0 ("not at all," meaning the patient reported not experiencing the behavior/feeling at all) to 10 ("extremely," meaning the patient reported exceedingly experiencing the behavior/feeling). Weight, VAS score, and the EBA were measured at baseline and at Weeks 1–6, 8, 12, and 16. The EI was assessed at baseline and at Weeks 4, 8, 12, and 16.

Statistical Analysis

For each study individually, the overall time to weight loss or weight gain was evaluated with a Kaplan-Meier product-limit estimator. To examine associations between measures of craving, eating factors, and eating behaviors and subsequent or concurrent weight loss or weight gain, a proportional hazards Cox regression with study-specific baseline hazard functions and time-varying covariates was employed, with disease (psychiatric diagnosis) as one of the baseline covariates in the model.

Important predictors and correlates were identified using stepwise variable selection in a Cox proportional hazards regression model. The original set of variables included changes from baseline and baseline values for eating scales, BMI, ethnicity, gender, and age. Only the results for the final models selected are reported. No subgroup analyses were performed discriminating between patients with schizophrenia and those with bipolar disorder, as the resulting sample sizes would be too small to produce meaningful results. All statistical analyses are reported with a significance level of p < .05.

Patients

Eating Inventory and EBA

To better understand the relationships among the different measures of eating behaviors and attitudes, we computed pairwise correlations between the 3 factors of the EI as well as correlations between these factors and the items from EBA. At baseline, the 3 factors of the EI had the following correlations within the pooled data (N = 157; 1 patient had missing data at baseline): r = .222 (p = .0051) between factor I (Cognitive Restraint) and factor II (Disinhibition); r = .0025 (p = .9753) between factor I and factor III (Hunger); and r = .675 (p < .0001) between factor II and factor III. Note that if factor I and factor II measured opposite items as might be assumed in a model in which patients with more cognitive restraint have less disinhibition, one would expect a negative correlation between these factors, whereas our results showed a mild positive correlation (see Discussion pp.17–18).

Weight Outcomes

Analysis of weight outcomes within the individual studies revealed that the highest percentage of patients who experienced successful weight loss at any time (42.9%; 18/42) was in the sibutramine study, while the highest percentage of patients who showed successful weight loss sustained to endpoint (33.3%; 16/48) was in the amantadine study. The highest percentages of weight gain were observed in the nizatidine study, with 70.2% (47/67) of patients showing weight gain at any time and 59.7% (40/67) whose weight gain was sustained to endpoint (Table 3). Figure 1 illustrates the time to weight loss (Figure 1a) and to weight gain (Figure 1b) in the individual study populations. While Figure 1 summarizes the results, it is not intended to suggest direct comparisons of the efficacies of the different weight mitigating agents used in our analyses.

	Weight Loss ≥ 2 kg		Weight gain ≥ 1 kg
	At any time	Sustained to endpoint	At any time	Sustained to endpoint
Nizatidine study (n = 67)	20.9% (n = 14)	7.5% (n = 5)	70.2% (n = 47)	59.7% (n = 40)
Sibutramine study(n = 42)	42.9% (n = 18)	26.2% (n = 11)	42.9% (n = 18)	19.1% (n = 8)
Amantadine study(n = 48)	39.6% (n = 19)	33.3% (n = 16)	56.3% (n = 27)	37.5% (n = 18)

Only patients with at least 1 post-baseline measurement are included.

Significant Weight Predictors

Predictors for Weight Loss and Weight Gain

Five variables were identified as predictors for weight loss in patients treated with olanzapine and weight mitigating agents, 2 of which were baseline variables that existed in patients treated with olanzapine before the introduction of any weight-mitigating agent. These baseline predictors were: higher BMI and less interest in food. The other 3 predictors for weight loss were time-dependent variables that gained significance after initiation of treatment with weight-mitigating agents. Patients who experienced decreases in appetite, cravings for carbohydrates, or hunger were more likely to lose weight.

Interestingly, all significant predictors for weight gain in patients treated with olanzapine and weight mitigating agents were time-dependent variables, measured concurrently with weight gain after initiation of treatment. Patients who experienced a decrease in cognitive restraint and/or an increase in hunger and/or an increase in overeating were more likely to gain weight. The fact that none of the variables available prior to treatment (e.g. patient characteristics) were significant predictors of weight gain, suggests that identifying such patients prior to treatment with weight-mitigating agents may be a challenging task, and underscores the importance of regular patient monitoring during treatment.

To test the robustness of our predictor analysis, we repeated the analysis using a subset of the original patient population, excluding all patients from the nizatidine group. We chose to exclude the nizatidine group for this test, as it is the only weight gain prevention study among the 3 studies and therefore the trial with the most potential for bias. The results confirmed BMI and interest in food as baseline predictors for weight loss, as well as decreases in cravings for carbohydrates and/or hunger as time-dependent predictors for weight loss. Similarly, a decrease in cognitive restraint and an increase in overeating were identified as predictors for weight gain. This subgroup analysis did not yield the previously identified decrease in appetite as a predictor for weight loss and increase in hunger as a predictor for weight gain. Those factors seem to be of importance for potential weight changes in patients that have not yet experienced weight gain, as is the case with the nizatidine group. It appears that in patients that have already gained weight, as is the case for patients in the amantadine and sibutramine groups, these factors were not predictive for potential weight loss or weight gain in response to the addition of a weight mitigating agent. One explanation might be that changes in appetite and hunger play a more important role as predictors/correlates of changes in weight for patients at earlier stages of weight gain. Once patients have gained substantial weight, further fluctuations in appetite or hunger do not predict weight changes, but cognitive restraint and actual eating behavior seem to remain important predictive factors. However, more comprehensive analyses of larger populations of patients are needed to validate this hypothesis.

Our finding of a decrease in cognitive restraint (the cognitive control of eating) as a significant predictor for weight gain is especially interesting considering the findings of Khazaal and colleagues, who showed that patients with schizophrenia present with cognitive distortions (in their thinking about weight regulation and self control) regarding weight gain when compared with control individuals. [16] Additionally, they also demonstrated that cognitive behavioral therapy improved binge eating symptomatology and weight-related cognitive thinking in patients who had gained weight during treatment with antipsychotic drugs. These patients experienced more progressive weight loss after cognitive behavioral therapy than control patients. [17] To address and utilize the importance of cognitive restraint as a predictive factor, clinically relevant measures to monitor cognitive restraint in a given patient need to be developed.

Recently, Lipkovich and colleagues [18] reported early predictors of substantial weight gain in patients who were treated with olanzapine. Their analysis revealed that patients with bipolar disorder treated with olanzapine who had gained 2 to 3 kg in the first 3 weeks after initiation of treatment, were at higher risk for substantial weight gain after 30 weeks of treatment. These results are in agreement with another study that showed patients with early weight gain while receiving olanzapine treatment for schizophrenia, schizophreniform, or schizoaffective disorder (increases of at least 2 kg in the first 3 weeks of treatment) were more likely to develop substantial weight gain (>10 kg after 30 weeks) over the course of treatment. [19] Additionally, patients who did not experience this amount of early weight gain, but had a BMI ≥ 27 kg/m² at treatment initiation, were also at higher risk for substantial weight gain (>10 kg) after 30 weeks of treatment. [18] Similarly, lower BMI at baseline was identified as a predictor for weight gain during treatment with antipsychotics in additional studies. [20, 21] Identification of the discussed risk factors may help clinicians to better focus weight management efforts on susceptible patients.

Eating Behavior Assessment

In the analyses presented here, we utilized the 3-factor EI developed by Stunkard and Messick. [15] Their original work utilized 2 main cohorts, "dieters" and "free eaters", to validate and optimize the questionnaire and factor analysis. Comparison of our results with theirs revealed that correlations between eating factors evaluated in our patient sample were very similar to those in their cohort of "free eaters" (factor I to factor II: r = .222 versus r = .19; factor I to factor III: r = .0025 versus r = -.06; factor II to factor III: r = .675 versus r = .73). The majority of our patients are probably "free eaters", meaning that they do not routinely restrict their dietary intake in order to control their body weight.

Analysis of correlations between the 3-factor EI and the EBA yielded interesting results. While EI factor II (Disinhibition) and factor III (Hunger) showed highly significant correlations to EBA items, EI factor I (Cognitive Restraint) was only mildly correlated to 2 EBA items ("When you finished a meal, have you felt full or satisfied?" and "Have you been thinking about food?"). It appears that the dimension captured in factor I (Cognitive Restraint) of the EI is poorly represented on the EBA scale. At least in the patient population analyzed in the current study, factor I (Cognitive Restraint) forms a complementary dimension, it is not the opposite of factor II (Disinhibition), as the labeling of those dimensions might suggest. The EI and the EBA scales appear to complement one another in the evaluation of different dimensions of eating behavior. While it is a limitation of the current analyses that both EBA and VAS are non-validated scales, their usage within an exploratory analysis as presented here is appropriate as we did not attempt to examine treatment differences with those scales. Instead, they were used in conjunction with EI, a validated scale, to examine correlations between patient characteristics and susceptibility to weight change during treatment with olanzapine and weight mitigating agents.

Categorical Weight Gain and Weight Loss

In this analysis, we chose the occurrence of ≥ 2 kg weight loss as an indication of a successful outcome and the occurrence of ≥ 1 kg weight gain as a sign for an unsuccessful outcome. While those cutoff points were defined without prior evidence for their validity, we believe that this categorical evaluation of the study populations allows valuable insights. No established cutoff criteria were available to serve as an alternative for our arbitrary cutoff points.

Analysis of categorical weight gain and weight loss within each of the individual study populations revealed that patients receiving adjunctive treatment with nizatidine showed the poorest weight control performance. This study had the lowest percentage of patients who experienced successful weight loss and the highest percentage of patients who gained weight throughout the course of the study. The patients treated with nizatidine had a slightly lower baseline BMI in comparison to the other 2 study populations, which might be an explanation for the lower proportion of patients who experienced weight loss. Also of note, the inclusion criteria varied between the 3 studies. Both the amantadine and sibutramine studies required an initial 5% to 7% weight gain while receiving olanzapine treatment prior to study entry, but the nizatidine study did not. The principal objective of the nizatidine study was weight prevention when initiating olanzapine treatment. It has been reported that patients initiated on olanzapine may experience most of their weight gain during the first 6 to 9 months of treatment. [22] Therefore, the previously mentioned inclusion criterion in the amantadine and sibutramine studies selected for patients that had potentially already experienced initial weight gain, while the patient population enrolled in the nizatidine study included patients who were just initiated on olanzapine treatment. Finally, we cannot exclude the possibility that nizatidine is less effective than amantadine or sibutramine as a weight-mitigating agent. Since the analysis presented here does not represent a direct statistical comparison of all 3 adjunctive treatments, further studies are needed. However, our result is in agreement with reports in the literature that did not show strong weight loss properties of nizatidine in double-blind, placebo-controlled studies. [11, 23]

Intervention Strategies

A limitation of the analyses presented here is that only patients with a BMI of ≥ 25 kg/m² were included, therefore the results cannot be generalized to patients with a BMI of <25 kg/m². Additionally, the enrollment criteria for the studies we utilized did not take into account the individual patient's stage of treatment with olanzapine. While patients initiating olanzapine treatment may need to focus on weight gain prevention, those patients who have already experienced weight gain during olanzapine treatment may need to focus on weight reduction.

Generally, early intervention with the goal of weight gain prevention seems to be the more promising approach. Several studies suggest that lifestyle modifications resulting in reduced caloric intake and enhanced physical activity are helpful in minimizing weight gain during treatment with olanzapine in patients susceptible to weight gain, [8] while these measures are not effective in all patients. [24, 25] Additionally, there are promising reports of successful prevention of weight gain with weight-mitigating agents in some patients. [14, 26–29] Recently, Wu and colleagues demonstrated the efficacy of metformin in preventing weight gain temporally associated with olanzapine treatment in a randomized, placebo-controlled trial in drug-naïve, first-episode patients with schizophrenia. [30] However, for many patients, weight gain has already occurred and weight reduction is needed. This goal is more difficult to achieve. Some promising results point to the effectiveness of behavioral therapy and adjunctive pharmacotherapy in helping to achieve weight reduction. [8, 23, 31–40] In a randomized, placebo-controlled trial, metformin plus lifestyle intervention showed the best effect on weight loss. [41] This is some of the best empirical evidence to date for the efficacy of treating weight gain during treatment with antipsychotics. Overall, both weight gain prevention and weight reduction are important therapeutic goals in patients receiving olanzapine treatment.

To the best of our knowledge, this is the first analysis examining predictors for weight loss and weight gain in patients treated with olanzapine and weight-mitigating agents. Interestingly, baseline BMI has been identified in the past as a predictor for weight gain with olanzapine treatment, with patients with lower BMI tending to gain more weight during treatment with olanzapine. [21, 42] The results of our analysis indicate that weight-mitigating agents (in particular nizatidine, amantadine and sibutramine) as adjunctive treatment to olanzapine therapy do not appear to be beneficial for all patients, but might have therapeutic potential for some patients. Prospective studies are needed to better identify patients who will benefit from such treatments. In addition, we suggest that this analytic approach may be beneficial as a secondary or post-hoc analysis of ongoing studies with potential pharmacologic agents to gain a better understanding of patients who may or may not respond to a particular strategy.

Clinicians can help mitigate the potential weight gain temporally associated with olanzapine treatment by being aware of the patient's characteristics like baseline BMI and baseline level of interest in food, and monitoring the patient early in treatment for weight gain, changes in appetite or hunger, or cravings for carbohydrates, and for reduced cognitive restraint. These factors can help the clinician determine when to intervene.