The Community Healthy Activities Models Program for Seniors is a validated and sensitive assessment of older adults' physical activity that has been adapted for telephone use.24-26 Mean weekly minutes of moderate-to-vigorous exercise were computed for the current research.

Overall diet quality was assessed from 2 unannounced, 24-hour recalls performed by trained interviewers. The 24-hour dietary recall represents 1 of the most widely used methods to collect dietary data, and 2-day recalls have been performed for a variety of major studies, such as the National Health and Nutrition Examination Survey27 and the Women's Intervention Nutrition Study.28 Interviewers obtained dietary intake data using revised interactive Nutrition Data System software (Nutrition Coordinating Center [NCC] Food and Nutrient Database System, version 2006; NCC, University of Minnesota, Minneapolis, Minn). Data obtained from recalls were averaged over the 2-day period and used to calculate the Healthy Eating Index 05 (HEI05),29 which ranges from 0 (worst) to 100 (best) with scores >80 indicating good diet quality.

Self-reported height and weight were used to compute BMI (kg/m²).

The Medical Outcomes Survey (MOS) Short Form-36 (SF-36)30 is a brief, reliable, and valid 36-item QoL measure that has proven reliability among older adults and cancer survivors.7, 31 The SF-36 was used to assess the following 8 domains of QoL: physical and social functioning, role limitations caused by physical and emotional problems, mental health, vitality, pain, and general health perceptions. These 8 subscales provide the basis for calculating 2 summary measures, the Physical Component Summary and the Mental Component Summary. Higher scores represent better functioning on the 2 summary measures and the 8 subscales. Norm-based scaling is used for both summary scores, so that a score of 50 represents the US national average. A score 10 points above or below the mean score of 50 represents a difference of 1 standard deviation from the national average.

Six medical conditions (eg, arthritis, heart conditions) were assessed using a measure that was developed previously by our research center.32

Descriptive statistics were used to characterize the demographics and health behaviors of the study sample. Omnibus tests with 3 degrees of freedom were conducted using ordinary least squares (OLS) regression to determine whether exercise, diet quality, BMI, and physical and mental QoL varied by sex and cancer type. For the non-normally distributed measures (ie, minutes of exercise and some SF-36 subscales), we conducted parallel, unadjusted nonparametric and adjusted ordinal logistic regression analyses, which delivered a set of inferences similar to that delivered by using OLS. For simplicity, we present the OLS parametric measures of means, standard deviations, and P values. In addition, post hoc pairwise comparisons among the 4 sex/cancer groups were conducted using the Tukey honestly significant differences (HSD) procedure. Next, Pearson (or, where appropriate for non-normally distributed measures, Spearman) correlations were computed to examine the associations between diet quality and BMI and mental and physical QoL. All analyses were conducted with and without control for demographic and medical factors that were associated significantly with study variables. With the exception of using the Tukey HSD procedure for pairwise comparisons of sex/cancer groups, no correction for Type I error was conducted for these exploratory analyses. All reported P values were 2-sided, and a value of P < .05 was considered statistically significant. Data were analyzed with SAS statistical software (version 9.1; SAS Institute Inc, Cary, NC).

Complete descriptions of the sample and statistical analyses of the accrual procedures have been reported previously.22 To briefly summarize, 20,015 cancer survivors were mailed a letter inviting their participation, and a preliminary response was obtained from 2156 survivors who called for more information (11% response rate). After receiving a consent form and additional study information, 1208 survivors completed screening and consent forms (6% overall response rate). Respondents (n = 1208) differed significantly from nonrespondents (n = 18,807) with respect to age (mean age, 73 years vs 76 years, respectively; P < .0001), race (13% vs 17% minority, respectively; P < .0001), sex (50% vs 45% women, respectively; P = .0004), and time since diagnosis (mean elapsed time, 9 years vs 10 years postdiagnosis, respectively; P < .0001). Of the 1208 survivors who provided informed consent, 455 were ineligible based on responses to the written screener and were not considered for further evaluation. Reasons for ineligibility included medical conditions (n = 207), BMI <25 kg/m² (n = 138), >150 minutes of exercise per week (n = 84), and morbid obesity (n = 26). Thus, 753 survivors who completed 2 baseline telephone interviews were included in the current analyses. Because some individuals were ineligible for the study intervention based on further assessment of their BMI or weekly minutes of exercise, only 641 of these survivors ultimately were enrolled in the intervention phase of the trial.

Demographic and medical characteristics of the sample appear in Table 1, and descriptive statistics for study variables appear in Table 2. Participants were primarily white and well educated survivors from North Carolina (92%) or other US states (7%). The participants ranged in age from 65 years to 87 years (median age = 73 years). The average time since diagnosis at the time of telephone screening was 9 years. The vast majority of participants (94%) were nonsmokers. Because of the weight criterion for study enrollment, 60% of participants were overweight and 38% were obese at baseline. In addition, a more comprehensive assessment of survivors' physical activities relative to the screening assessment indicated that almost half (47.5%) of participants did not engage in moderate or vigorous exercise each week, and most participants (86%) did not meet national exercise guidelines. Only 7% of the sample had HEI05 scores >80, indicative of good eating habits relative to national guidelines. Despite suboptimal health behaviors, the sample's mental and physical QoL exceeded US age-related norms.33

Study outcomes were correlated such that weekly minutes of exercise were associated with lower BMI (ρ = -.10; P = .0064) and better diet quality (ρ = .14; P = .0002). BMI was not associated with diet quality (r = -.05; P = .21).

Significant bivariate correlations between study outcomes and demographic and medical characteristics are shown in Table 3. Older age was associated with lower BMI and worse physical and social QoL outcomes. Higher levels of education were associated with more weekly minutes of exercise, better diet quality, lower BMI, and better physical QoL outcomes. White race was related to reduced BMI and vitality, and cancer stage was correlated positively with BMI. Finally, having more comorbidities was associated with fewer weekly minutes of exercise, better diet quality, lower BMI, and worse physical, social, and role functioning. None of the other correlations between demographic and medical factors (ie, age, race, education level, cancer stage, comorbidities, time since cancer diagnosis) and study outcomes were statistically significant.

Means, standard deviations, and P values for comparisons of study variables by sex and cancer type appear in Table 2. Results of the OLS regression models adjusted for age, race, education level, and comorbidities indicated that only weekly minutes of exercise, diet quality, health perceptions, and physical functioning varied across the 4 sex/cancer groups. Pairwise comparisons revealed that breast cancer survivors (women) reported fewer weekly minutes of exercise and better diet quality relative to men who were colorectal cancer survivors and prostate cancer survivors. Women who were colorectal cancer survivors reported better diet quality only relative to men who were colorectal cancer survivors. Finally, women who were colorectal cancer and breast cancer survivors reported worse physical functioning relative to prostate cancer survivors. By using the Tukey HSD method, none of the other pairwise comparisons yielded statistically significant differences as a function of sex or cancer type.

When the models described above were analyzed without adjusting for demographic covariates, the same pattern of results was obtained with 2 exceptions. In unadjusted analyses, pain varied significantly as a function of sex and cancer type (P = .02) such that women who were breast cancer survivors reported more pain than prostate cancer survivors. The pain levels of other sex/cancer groups did not significantly differ from one another. In addition, health perceptions did not vary as a function of sex or cancer type.

Associations between weekly minutes of exercise, diet quality, BMI, and QoL outcomes were examined (see Table 4). In analyses with and without adjustment for age, level of education, and comorbidities, greater weekly minutes of exercise were associated with better physical QoL, including less pain and better health perceptions, physical functioning, and vitality. More exercise also was correlated with better social functioning. Diet quality had a positive association with a range of physical QoL outcomes in analyses that were adjusted for age, level of education, and comorbidities. However, only physical functioning and vitality were correlated significantly and positively with diet quality in unadjusted analyses. Finally, greater BMI was associated with worse physical QoL, including greater pain and role limitations because of physical problems and worse health perceptions, physical functioning, and vitality. These associations remained significant when adjusting for age, race, level of education, cancer-type, and comorbidities.