A randomized trial of a lifestyle intervention in obese endometrial cancer survivors: quality of life outcomes and mediators of behavior change

Study Design & Patient Recruitment

The study was a prospective randomized controlled trial comparing a lifestyle intervention (LI) consisting of exercise and nutritional counseling with cognitive-behavior modification to usual care (UC) in overweight and obese endometrial cancer survivors. An invitation letter was sent to women included in the cancer registry from the Ireland Cancer Center at University Hospitals Case Medical Center diagnosed with stage I-II endometrial cancer from 2001–2004. The 6 month intervention was delivered to the LI group while the UC group received only an informational brochure outlining the benefits of proper nutrition and physical activity. Institutional review board approval was granted and informed consent and authorization to use protected health information (HIPAA) was obtained from all patients prior to study procedures. Feasibility, eligibility criteria, and details of the intervention have been published elsewhere [11]. Randomization assignment was stratified by BMI (<40 versus ≥ 40). Prior analysis revealed that demographics and clinical characteristics were equivalent between groups [11].

Intervention Program

The counseling protocol was administered by the study registered dietitian (RD), primary investigator (PI), and psychologist. It followed a stepwise, phased approach using strategies outlined by social cognitive theory, indicating that the optimal intervention for a major behavior change should focus on establishing short-term goals, and enabling the person to build self-efficacy [12-16]. The intervention covered topics related to nutrition and exercise and met weekly for six weeks, biweekly for one month, and monthly for 3 months. Participants were contacted by the RD either by phone, email, or newsletter every week that the group did not meet. The psychologist met with the group during 2 of the 11 sessions and topics included cognitive and behavioral self-management strategies for weight loss and stress management. The PI met with both LI and UC participants at 3, 6 and 12 months.

Measures

Patient demographic and clinical data was obtained at baseline and prior to randomization. QOL and self-efficacy were assessed at baseline and at 3, 6, and 12 months. Eating behavior and depression was assessed only at baseline and 12 months. QOL was measured by the Functional Assessment of Cancer Therapy-General (FACT-G), a valid and reliable questionnaire evaluating physical, functional, family-social, and emotional well-being domains [17,18]. A fatigue subscale (-F) and an endometrial symptom subscale (-En) were also used [19,20]. Functional status was measured with the SF-36, a comprehensive short-form survey designed to measure functional health and well-being [21].

Depression was measured by the Beck Depression Inventory (BDI) which is well-validated and frequencly used in lifestyle research studies [22]. Self-efficacy, an individual's judgement regarding his/her ability to perform certain behaviors, was measured with the Weight Efficacy Life-Style (WEL) questionnaire. This scale specifically evaluates self-efficacy judgments specific to eating behaviors in five situational factors: negative emotions, food availability, social pressure, physical discomfort, and positive activities [23,24]. This measure has evidenced adequate psychometric properties, including internal consistency coefficients ranging from 0.76 to 0.90 [23].

Eating behavior was measured by the Three-Factor Eating Questionnaire (TFEQ), a self-assessment questionnaire developed to measure cognitive and behavioral components of eating [25,26]. Responses are aggregated into three scales, cognitive restraint, disinhibition and hunger. Cognitive restraint is designed to measure dietary restraint, that is, control over food intake in order to influence body weight and body shape. Disinhibition measures episodes of loss of control over eating, while the hunger scale is concerned with subjective feelings of hunger and food cravings. Higher numbers reflect increased restraint, decreased tendency to overeat in the presence of disinhibitors (i.e. stress, mood, alcohol) and decreased perception of hunger. All three subscales have demonstrated high internal consistency and reliability [27,28].

Patient anthropometric data including weight, waist circumference, and BMI were measured at all time points manually by the study dietitian. BMI was computed as patient weight in kilograms divided by the square of their height in meters. Participants were categorized according to World Health Organization guidelines: < 18.5 (underweight), 18.5 to 24.9 (healthy weight), 25.0 to 29.9 (overweight), Class I – 30.0 to 34.9, Class II – 35.0–39.9, and Class III or morbid obesity ≥ 40.0 kg/m² [29].

Statistical Analyses

Patient demographic, clinical variables and baseline values were compared between groups and by BMI (25.0 – 39.9 versus ≥ 40) by use of independent samples t-test or chi-square test for proportions. Primary analysis used repeated measures analysis of variance (ANOVA) with the 3, 6 and 12 month data as outcomes and the appropriate baseline measurement as a covariate to test for the main effect of group (LI versus UC, intention-to-treat analyses) on QOL outcome measures (FACT-G; physical, functional, social, emotional well-being, fatigue and endometrial subscales) and self-efficacy (WEL; negative emotions, food availability, social pressure, positive activities, physical discomfort). Change (from baseline to 12 months) in eating behavior (TFEQ) and depression measures were compared using paired samples t-test for each group and independent samples t-test for the difference in change between groups. Missing data for participants who did not complete all study assessments was handled according to the last and next method or previous row mean method as recommended by Engels et al [30]. Imputation was done on between 15–19% of values for the various QOL and eating behavior measures. The percentage of missing data for these measures was less than that imputed for weight values as some patients opted to mail these surveys back as data were self-report. Data were also analyzed using last observation carried forward and completers only approaches. There were no substantive differences among the three approaches and thus we present results of the first approach only.

QOL outcomes were also examined according to whether patients lost weight or their weight remained stable/gained over the course of the year as an ancillary stratified analyses. We were interested in examining if QOL outcomes differed between these two groups as there were some women in the UC group who lost weight and likewise women in the LI group who did not lose weight [31]. SPSS for Windows (version 14.0) was used for statistical analyses (SPSS Inc., Chicago, IL).

Approximately 25 patients per group were needed to provide 80% power to detect a difference between groups in mean weight change from baseline to twelve months of 5 kg (11 pounds) or greater, representing approximately 5% weight loss for an obese female (alpha = 0.05, two-sided, SD = 5.0) [32]. Five percent weight change is considered clinically relevant and a recommended goal for weight loss over 6 months [33,34]. Weight change was the primary endpoint of this feasibility trial and the study was powered based on this endpoint. In addition, 25 patients/group were needed to detect a large standardized effect size (d = 0.80) with power of 0.80 and a two-tailed alpha of 0.05. A large standardized effect size on the FACT-G is approximately 10 points based on a standard deviation of 12 and exceeds the 7 point MID identified for this scale [18,35].

Baseline differences based on body mass index

Baseline QOL (FACT-G, SF-36), BDI, WEL and TFEQ scores did not differ between the LI and UC groups. However significant differences were observed when patients were categorized according to BMI (Table 1). Total QOL (FACT-G) was not significantly lower in morbidly obese women [BMI ≥ 40: 78.0 (SD = 14.4) vs. BMI < 40: 83.7 (SD = 11.4); p = 0.14], however the difference was consistent with minimally important differences (5–6 points) [18]. SF-36 physical composite score was decreased in morbidly obese women (p = 0.04). At baseline 7/45 (15%) patients reported mild depressive symptoms, and 4/45 (9%) reported moderate depressive symptomatology. Beck depression score did not differ according to BMI (Table 1).

In regards to eating patterns, morbidly obese patients had significantly lower self-efficacy when feeling physical discomfort and decreased total self-efficacy (WEL) score. Restraint on the TFEQ questionnaire was decreased in patients with BMI > 40 as compared to BMI < 40.

Effects of the lifestyle intervention (Intention-to-treat analysis)

Repeated measures ANOVA with baseline measurements as a covariate and 3, 6 and 12 months as outcomes revealed no group (LI versus UC) effects for QOL outcomes (Table 2). There was a significant group effect for self-efficacy related to social pressure and restraint improved in the LI group. The UC group had a significant change in depression from baseline. Change scores for depression or TFEQ, compared by independent samples t-test, however did not differ by group.

Association between weight loss and outcomes (exploratory analysis)

Table 3 presents an ancillary analysis for QOL, self-efficacy, depression and eating behavior outcomes. For this analysis we compared women who lost weight (WL; n = 21) versus those whose weight was the same or who gained weight (WG; n = 24) from baseline to twelve months. There was a significant effect (WL versus WG) for emotional well-being QOL, self-efficacy related to negative emotions, food availability, and physical discomfort as the WL group had higher scores. For self-efficacy related to negative emotions, there was a mean increase of 8.9 in women who lost weight versus 0.6 in those whose weight was stable/gained. Similarly, for self-efficacy related to food availability and physical discomfort, the weight loss group had an increase of 7.9 and 5.3 versus 1.9 and 1.0 in women who did not lose weight. Women who lost weight had improvement in depression (2.3 versus 1.2) and TFEQ restraint score (2.8 versus 0.1). Women who lost weight also had a lower disinhibition score at 12 months. Group differences for change in TFEQ restraint score were observed using independent samples t-test for comparison of change scores between WL and WG (p = .01), but not for depression.