A prospective cohort study named the Hong Kong NTEC-KWC Breast Cancer Survival Study (HKNKBCSS) was initiated in 2011. This enrolled Chinese women with early stage breast cancer, and aimed to investigate whether soy isoflavones and other lifestyle factors were related to breast cancer outcomes.[11-14] The eligibility criteria were defined as follow: histologically confirmed breast cancer with American joint Committee on Cancer (AJCC) stage 0-III diagnosed within 1 year before study entry, any age, female gender, Chinese ethnicity, able to read Chinese, and had no prior history of breast or other cancers.
In total, the study cohort recruited 1462 patients from two regional cancer centers in Hong Kong. Each participant provided written informed consent. Enrolled patients were interviewed at four time-points: baseline at study entry, 18-month, 36-month and 60-month follow-up post-diagnosis. The study was approved by the Joint CUHK-NTEC Clinical Research Ethics Committee and the KWC Research Ethics Committee of the Chinese University of Hong Kong and the Hong Kong Hospital Authority.
The 18-month follow-up was conducted between 12 and 24 months after breast cancer diagnosis, during which 1310 patients in the study cohort completed assessment at that time-point. Participants who had incompletely filled questionnaire for QoL (n=4), who reported implausible dietary intake (energy intake estimates <500 or >4000 kcal per day; n=3) or who had carcinoma in-situ (n=77) were excluded, resulting in 1226 patients for the present analysis.
Data collection at each assessment
Trained personnel conducted face-to-face interview at baseline and each follow-up. During each interview, socio-demographic data were collected by structured questionnaires, and included age, marital status, occupation, education level, family income, menopausal status and prior medical history. Validated questionnaires were used to collect information about lifestyle factors, such as dietary intake and level of physical activity. At each assessment, body weight and height were also measured. Body mass index (BMI) classification was based on criteria adopted in the Asia-Pacific region, which consisted of four groups, as follow: underweight <18.5 kg/m2, normal 18.5-22.9 kg/m2, overweight 23-24.9 kg/m2 and obese ≥25 kg/m2. Patients’ clinical information on breast cancer was retrieved by reviewing hospital medical records.
Dietary intake and dietary pattern assessment
Patients recalled their dietary intake over the proceeding 12 months at 18-month follow-up assessment. Dietary intake was collected by filling a validated food frequency questionnaires (FFQ) during face-to-face interview. The FFQ contained 109 food items that were commonly consumed in Hong Kong population and a specific proper portion size was used for quantification of each food item. Based on the design of FFQ, patients should report the frequency of consumption and average amount of intake at each time for each food item. Interviewers would provide a food photographs with individual food portions during the interview, which was useful for more clear estimation. Based on data collected on FFQ, the daily total energy intake and other nutrients can be calculated according to Chinese Food Composition Table.
After excluding uncommon dietary items (with average intake < 2 serving/month), the 109 food items in the FFQ were grouped into 17 food groups based on the similarity of food type, nutrient profiles as well as local eating habits in local Chinese population (Supplementary table 1). This approach has also been adopted by previous studies, and it could reduce the complexity of dietary data.[6-8] Based on the daily consumption of the 17 food groups that was not energy adjustment, principal component analysis was used to derive dietary patterns. The food groups (factors) were rotated by orthogonal transformation (Varimax rotation function), resulting in uncorrelated, independent factors. Major factors retained were based on eigenvalue greater than 1.0, the scree plot (supplementary figure 1), and factor interpretability, which were commonly used in breast cancer studies.
Factor loadings represent correlation coefficients between the food groups and the dietary pattern. The derived factors (patterns) were labeled as major food groups with higher factor loading as well as the interpretation of the data. This study identified two primary factors, which were labeled “grain and animal food pattern” and “vegetables and fruits pattern”. The factor score for each dietary pattern was calculated for each participant by summing intakes of food groups weighted by their factor loading. Each individual was assigned a score for each identified dietary pattern, which reflect one’s conformity with that pattern; a higher factor score suggested better conformity with that pattern.
QoL assessment using EORTC QLQ-C30
EORTC QLQ-C30 was used to measure patients’ QoL at the time of 18-month follow-up assessment. EORTC QLQ-C30 was designed to assess a range of cancer-specific QoL issues relevant to a broad spectrum of cancer patients. This questionnaire consisted of 30 cancer-specific questions with multiple-point scales, including a global health status/QoL scale, five functional scales (physical, role, emotional, cognitive and social), nine symptom scales (fatigue, nausea and vomiting and pain, dyspnea, insomnia, appetite loss, constipation, diarrhea and financial difficulty). According to the EORTC QLQ C-30 Scoring Manual, multiple-point scales were transformed into standard scores (from 0 to 100) in the analysis. High scores on global health status/QoL and functioning scales represented good QoL, while high scores on the symptom scales indicated more severe symptoms.
For comparing the characteristics of participants, they were categorized into three groups (tertile 1, tertile 2 and tertile 3) according to the tertiles of factor scores in each dietary pattern. Analysis of variance was used for continuous data and chi-square test for categorical data. Multivariable linear regression models were used to investigate the association between dietary pattern and QoL items. QoL items were log10 transformed to fulfil the normal distribution assumption. The potential confounders were introduced in models using the enter method, which were identified based on the theoretical considerations, the previous literature, and the results of univariate analyses. Collinearity diagnostics were applied to exclude multicollinearity of variables. The following covariables were adjusted in the first model age at 18-month follow-up (years; continuous), education level (high school or below, college or above), household income (<15,000, 15,000-30,000, ≥30,000HKD/month), total number of comorbidities (0, 1, ≥2), menopausal status at 18-month follow-up (pre-menopausal, post-menopausal), AJCC stage (I, II, III, I-III without detail), ER status (positive, negative, missing), PR status (positive, negative, missing), current adjuvant hormonal therapy usage (yes, no), BMI at 18-month follow-up (kg/m2; continuous), level of physical activity (MET-hours/week; continuous), total and energy intake (kcal/day; continuous). In the second model, factor scores of the other dietary pattern (when one of the two dietary patterns were being analyzed; continuous) were further adjusted. All analyses were performed using SPSS 26.0; and P value <0.05 based on two-sided analysis were considered statistically significant.