Dietary pattern and health-related quality of life among breast cancer survivors

Breast cancer survivors were recruited between September 2012 and November 2015 from three large hospitals. A total of 286 patients, aged 21-79 years, who had been diagnosed with primary breast cancer were enrolled in our study. Among the study participants, we excluded patients who had been diagnosed with stage 0 breast cancer or had missing data in their medical records (n = 29), those with other cancers before the breast cancer diagnosis (n = 7) or those who had breast cancer surgery within 6 months before recruitment (n = 12). In addition, we excluded women who did not completely record their dietary data (n = 5) and who reported implausible energy intake (above the log-transformed mean ± 3 sd, n = 1). As a result, we included a total of 232 breast cancer survivors who had been diagnosed with primary breast cancer (stage I-III) more than 6 months after breast cancer surgery. All participants gave written informed consent at enrollment. This study was approved by the Institutional Review Board at Seoul National University Hospital, the National Cancer Center and Soonchunhyang University Hospital in Korea.

A trained research nurse asked the participants about their demographic information, physical activity, alcohol consumption, smoking status, clinical information, parity, supplement use, health-related quality of life, family history, sun exposure and dietary principles for breast cancer survivors using a structured questionnaire. The score of physical activity was calculated as metabolic equivalent task (MET)-hour per week.

Through medical record data, we obtained information regarding weight and height at diagnosis of breast cancer and clinical information, including other incidences of cancer before their breast cancer diagnosis, the diagnosis date of primary breast cancer, menopausal status and menopausal hormone use at diagnosis, recurrence and metastasis after diagnosis, hormone receptor status, cancer stage, tumor size, date of surgery and initial treatment type. Body mass index (BMI, kg/m²) at diagnosis was calculated by dividing a participant’s weight (kg) by his/her height (m²).

Participants completed a non-consecutive 3-day dietary record on 2 weekdays and 1 weekend day. All foods and beverages that they consumed were recorded. To help with recording, we provided additional food photograph booklets originally designed by our study team. Dietary intake of nutrients and energy were calculated using the Computer-Aided Nutritional Analysis Program (CAN-Pro) version 4.0 (Korean Nutrition Information Center, Seoul, Korea). We grouped a total of 1150 food items one-by-one into 39 food groups based on the similarities of their nutrient profiles. Next, we assigned new food codes to all food items for analysis and combined the intake amount for each food group per person. Finally, we calculated the daily amount (g/day) of each food group.

We asked participants to answer the questions about HRQoL using a validated Korean version of the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire Core 30 (EORTC QLQ-C30) version 3.0 and the Quality of Life Questionnaire Breast Cancer Module 23 (QLQ-BR23) [14, 15]. The 30-item QLQ-C30 is used to assess the HRQoL of cancer survivors and is composed of the following categories: (1) global health status/quality of life scale, (2) functional scales (physical, role, emotional, cognitive, and social) and (3) symptom scales (fatigue, nausea and vomiting, pain, dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulty). The QLQ-BR23, is composed of functioning and symptom scales and is categorized into the following domains: (1) functional scales (body image, sexual functioning, sexual enjoyment, and future perspective) and (2) symptom scales (systemic therapy side effects, breast symptoms, arm symptoms, and upset by hair loss). In our study, we did not included ‘sexual enjoyment’, because 70% of the participants did not respond to this scale. We transformed the raw scores of the 4-point or 7-point scales to a 0 to 100 scale based on the EORTC scoring manual. A higher score reflects a higher quality of life in functioning and global health status/quality of life and a lower quality of life in symptoms.

We conducted a factor analysis to determine dietary patterns for breast cancer survivors. The dietary patterns were generated by a principal component analysis using an orthogonal rotation procedure [16]. We determined two factors based on eigenvalue, screening test, and our interpretations. We labeled these two factors as “Healthy dietary pattern” and “Western dietary pattern”. The factor score for each pattern was calculated by summing the intakes of all the food groups weighted by their factor loadings. Each dietary pattern score was categorized by quartiles for all participants. Individuals with high dietary pattern scores had a greater tendency to follow the pattern, compared to individuals with low scores. The association between HRQoL scores and each dietary pattern was estimated from the least squares mean (LS mean) and 95% confidence interval (95% CI) using a generalized linear model (GLM). In addition, we compared the top quartile with the bottom quartile of the factor scores using the contrast command. All models were adjusted for age at diagnosis (years; continuous), BMI (kg/m²; continuous) at diagnosis, energy intake (kcal/day; continuous), marital status (married/cohabitation and unmarried/divorced/widowed), education level (high school or below and college or above), breast cancer stage at diagnosis (I, II, and III), physical activity (MET-h/week; continuous), time since surgery (6 months- < 1 year, 1 year- < 5 years, and ≥ 5 years) and menopausal status at diagnosis (pre- or postmenopausal status). To test for trends across categories, we assigned the median value to the quartile category of dietary pattern being used as a continuous variable. If we had missing variables (the proportion for each < 1%), we assigned the participants to the median or the common category. We used the log-transformed variables of the HRQoL scores to meet normality, and these variables were exponentiated. If there was no information (unknown) regarding menopausal status at diagnosis in the medical records, we regarded the participant as postmenopausal at diagnosis if their age at menopause on the questionnaire was younger than the age at diagnosis. We regarded the participant as premenopausal at diagnosis if their age at menopause on the questionnaire was older than the age at diagnosis. When information regarding age at menopause was missing (n = 10), we considered the participant postmenopausal if they had been diagnosed after the age of 50, which was the median age of menopause in Korean women in 2013 [17]. We conducted a subgroup analysis by stage (I and II or III), menopausal status at diagnosis (pre or postmenopausal), and time since surgery (< 2 or ≥ 2 years). All statistical analyses were performed using SAS software 9.4 (SAS Institute Inc., Cary, NC). Statistical significance was defined as a P-value < 0.05 in a two-sided test. The resulting raw p-values were adjusted for multiple comparisons using the False Discovery Rate (FDR) method [18].

We identified two dietary patterns: “Healthy dietary pattern” and “Western dietary pattern”. We presented a factor loading of more than 0.20 or less than − 0.20 (Table 1). The Healthy dietary pattern explained 2.88% of the variance, and the Western dietary pattern explained 2.01%. The Healthy dietary pattern was characterized by higher intake of vegetables, whole grains, soy, potatoes, fish, fruits, yogurt, kimchi, mushrooms, seasonings, dressings and eggs, and lower intake of cakes or snacks, alcoholic drinks, pork, rice rolls, ice cream, beverages, hamburgers or pizza, noodles, refined grains, and coffee. The Western dietary pattern was characterized by higher intake of salad, seasonings, dressings, mixed rice, pancakes, eggs, processed seafood, chicken or duck meat, and beef and lower intake of fruits, nuts and seaweed.

We compared the general characteristics of the participants according to the quartiles of the factor scores for each dietary pattern (Table 2). Breast cancer survivors with higher Healthy dietary pattern scores were older and had a lower proportion of college education or above, compared to breast cancer survivors with lower Healthy dietary pattern scores. In addition, higher Healthy dietary pattern scores were associated with higher levels of physical activity and energy intake. Meanwhile, breast cancer survivors with higher Western dietary pattern scores were younger and had a higher proportion of college education or above, and they had lower levels of physical activity but higher levels of total energy intake, compared to breast cancer survivors with lower Western dietary pattern scores.

Table 3 shows the associations between the Healthy dietary pattern and HRQoL levels in the multivariable models. Breast cancer survivors who had higher Healthy dietary pattern scores tended to have lower dyspnea scores but higher insomnia scores, compared to breast cancer survivors with lower Healthy dietary pattern scores. For dyspnea, the LS mean (95% CI) was 8.86 (5.05-15.52) in the bottom quartile and 2.87 (1.62-5.08) in the top quartile (p for trend = 0.005). For insomnia, the LS mean (95% CI) was 11.46 (6.59-19.95) in the bottom quartile and 29.77 (16.87-52.51) in the top quartile (p for trend = 0.005). When we adjusted for multiple comparisons, statistically significant associations remained at an FDR of 0.1 (FDR = 0.055 for both dyspnea and insomnia). When we compared the top quartile with the bottom quartile for the Healthy dietary pattern scores, we also found statistically significant differences for dyspnea and insomnia (p-contrast = 0.003 for dyspnea and p-contrast = 0.01 for insomnia).

The Western dietary pattern was marginally significantly associated with the components of physical functioning and constipation among breast cancer survivors (Table 4). For physical functioning, the LS mean (95% CI) was 67.65 (60.64-75.46) in the bottom quartile and 77.41 (69.47-86.24) in the top quartile (p for trend = 0.06). For constipation, the LS mean (95% CI) was 4.44 (2.56-7.70) in the bottom quartile and 8.09 (4.69-13.94) in the top quartile (p for trend = 0.07). However, the multiple comparison test did not show statistical significance at an FDR of 0.1. In the contrasting comparison of the Western dietary pattern, physical functioning and constipation scores were higher in the top quartile than in the bottom quartile, with marginal significance (p-contrast = 0.05 for physical functioning and p-contrast = 0.08 for constipation).

When we examined the associations between the dietary pattern and HRQoL levels according to the stage at diagnosis, among stage I breast cancer survivors, we found that breast cancer survivors with higher Healthy dietary pattern scores tended to have lower global health status/QoL and dyspnea scores, compared to those with lower Healthy dietary pattern scores (p for trend = 0.01 for global health status/QoL and p for trend = 0.01 for dyspnea) (Additional file 1: Table S1). In addition, upset by hair loss scores were higher in the top quartile, compared to the bottom quartile (p-contrast = 0.01). The multiple comparison test showed significant contrast comparing the top with the bottom quartiles for global health status/QoL, dyspnea, and upset by hair loss at an FDR of 0.1. We observed that stage II or III breast cancer survivors with higher Healthy dietary pattern scores tended to have higher insomnia scores, compared to stage II or III breast cancer survivors with lower Healthy dietary pattern scores (p for trend = 0.01) (Additional file 1: Table S2).

Higher Western dietary pattern scores were associated with higher constipation scores among stage I breast cancer survivors (Additional file 1: Table S3). The LS mean (95% CI) was 4.72 (2.09-10.68) in the bottom quartile and 18.85 (8.86-40.10) in the top quartile (p for trend = 0.02). However, the Western dietary pattern was not associated with HRQoL levels among stage II or stage III breast cancer survivors (Additional file 1: Table S4).

We examined whether the association between dietary pattern and HRQoL levels varied by menopausal status. We did not observe any significant associations with the Healthy dietary pattern among premenopausal breast cancer survivors (Additional file 1: Table S5). However, among postmenopausal breast cancer survivors, higher Healthy dietary pattern scores were associated with decreasing levels of dyspnea (p for trend = 0.003) or diarrhea (p for trend = 0.06) and increasing levels of insomnia (p for trend = 0.02) (Additional file 1: Table S6). Dyspnea remained significant even after adjusting for multiple comparisons (FDR P value = 0.066). The Western dietary pattern was associated with role functioning among premenopausal breast cancer survivors (p for trend = 0.06) (Additional file 1: Table S7) and nausea and vomiting among postmenopausal breast cancer survivors (p for trend = 0.03, Additional file 1: Table S8).

When we examined the associations according to time since surgery, we observed that breast cancer survivors who had higher Healthy dietary pattern scores tended to have higher upset by hair loss scores, compared to breast cancer survivors with lower Healthy dietary pattern scores who were within 2 years of surgery (p for trend = 0.02 for upset by hair loss) (Additional file 1: Table S9). Among breast cancer survivors for whom it had been more than 2 years since their surgery, higher Healthy dietary pattern scores were associated with decreasing levels of dyspnea (p for trend = 0.03) but increasing levels of insomnia (p for trend = 0.03) and financial impact (p for trend = 0.01, Additional file 1: Table S10). We did not observe any statistically significant associations between the Western dietary pattern and HRQoL levels among breast cancer survivors whose surgery was less than 2 years ago and more than 2 years ago (Additional file 1: Tables S11 and S12).