Association of Dietary Pattern with Cardiovascular Risk Factors among Postmenopausal Women in Taiwan: A Cross-Sectional Study from 2001 to 2015

Unhealthy diet and inappropriate lifestyle contribute to an imbalance in cardiometabolic profiles among postmenopausal women. This research aimed to analyze the association between dietary pattern and changes in cardiovascular risk factors among postmenopausal Taiwanese women using binary logistic regression. This cross-sectional study involved 5689 postmenopausal Taiwanese women aged 45 years and above, and the data were obtained from Mei Jau Health Management Institution database between 2001 and 2015. The cardiovascular risk dietary pattern characterized by high intakes of processed food, rice/flour products, organ meat, and sauce was derived by reduced rank regression. Participants in the highest quartile of the cardiovascular risk dietary pattern were more likely to have high levels of systolic blood pressure (OR = 1.29, 95% CI 1.08–1.53), diastolic blood pressure (OR = 1.28, 95% CI 1.01–1.62), atherogenic index of plasma (OR = 1.26, 95% CI 1.06–1.49), triglycerides (OR = 1.38, 95% CI 1.17–1.62), and fasting blood glucose (Q3: OR = 1.45, 95% CI 1.07–1.97). However, this dietary pattern was not correlated with total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, and C-reactive protein. Therefore, adherence to the cardiovascular risk dietary pattern increases the risk of having higher levels of blood pressure, triglycerides, fasting blood glucose in postmenopausal Taiwanese women.


Introduction
Menopause is defined as the cessation of menstruation owing to a decrease in ovarian follicles and the further reduction of estradiol levels. It occurs mostly at a median age of 51 years [1]. The diagnosis of menopause is based on no menstrual period for 12 consecutive months in women [2,3]. Around 467 million postmenopausal women were registered in the world in the 1990s, and by 2030 the number of postmenopausal women is expected to be 1.2 billion with 47 million new postmenopausal women each year [4]. Several studies revealed that postmenopause was associated with increased inflammatory markers such as C-reactive protein (CRP), interleukin-1α (IL-1α), and tumor necrosis factor-α (TNF-α) and an imbalance in cardiometabolic profiles such as low levels of high-density lipoprotein cholesterol (HDL-C) and elevated levels of total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), triglycerides (TG), visceral fat, and blood glucose [5][6][7][8]. These imbalanced cardiometabolic profiles were favorable for the progression of atherosclerosis and an increased risk of cardiovascular disease (CVD) [9]. Cardiovascular disease was naires. Blood samples were analyzed for biochemical parameters. The study was approved by the Joint Institutional Review Board of Taipei Medical University (TMU-JIRB N202007075). There were 377,124 subjects who visited the MJ Health Screening Center between 2005 and 2015. We included women aged ≥45 years who self-reported menopausal status after missing their menstrual period for at least 12 consecutive months using a questionnaire. We excluded 299,450 participants who were male, had disease conditions such as cancer, cystic fibrosis, lung disease, cirrhosis, kidney disease, or infectious disease, or used any forms of lipid-lowering drugs. In addition, we excluded 68,985 women who were non-postmenopausal, aged less than 45 years, or failed to complete the questionnaire about their dietary habits. After excluding 3000 participants who had multiple entries between 2005 and 2015, a total 5689 postmenopausal women were retained in this study ( Figure 1).

Dietary Assessment and Other Covariates
A semi-quantitative food frequency questionnaire (FFQ) was developed, standardized, and validated by the MJ Health Management Institution, and used to assess dietary habits of the subjects. The FFQ questionnaire contained the closed-ended questions about 22 non-overlapping food groups with a total of 85 individual food items consumed by the participants in the past month [22]. The intake frequency was assessed in accordance with daily and weekly consumption. Each question was given the definition about one serving size of the food item, and presented 5 frequency response options as described previously [22]. Dietary data were collected for further frequency response options as described previously [22]. Dietary data were collected for further analysis to derive the dietary pattern using a reduced rank regression (RRR) model. The RRR model as a multivariable linear function was performed to derive the dietary pattern related to the disease of interest by a priori and a posteriori approaches based on the response variables for identifying a linear combination of the predictor variables [23].
Demographic data such as age, education (≤high school or >high school), and marital status (never married, married, or divorced/widowed) were collected. We also evaluated lifestyle data including smoking status (no or yes), drinking alcohol (no or yes), physical activity frequency (<150 min/week or ≥150 min/week), and sleep duration (<6 h, 6-8 h, or >8 h). Medical history regarding hypertension, diabetes mellitus, and CVD was recorded. All covariates were assessed using a self-reported questionnaire.

Dietary Assessment and Other Covariates
A semi-quantitative food frequency questionnaire (FFQ) was developed, standardized, and validated by the MJ Health Management Institution, and used to assess dietary habits of the subjects. The FFQ questionnaire contained the closed-ended questions about 22 non-overlapping food groups with a total of 85 individual food items consumed by the participants in the past month [22]. The intake frequency was assessed in accordance with daily and weekly consumption. Each question was given the definition about one serving size of the food item, and presented 5 frequency response options as described previously [22]. Dietary data were collected for further frequency response options as described previously [22]. Dietary data were collected for further analysis to derive the dietary pattern using a reduced rank regression (RRR) model. The RRR model as a multivariable linear function was performed to derive the dietary pattern related to the disease of interest by a priori and a posteriori approaches based on the response variables for identifying a linear combination of the predictor variables [23].
Demographic data such as age, education (≤high school or >high school), and marital status (never married, married, or divorced/widowed) were collected. We also evaluated lifestyle data including smoking status (no or yes), drinking alcohol (no or yes), physical activity frequency (<150 min/week or ≥150 min/week), and sleep duration (<6 h, 6-8 h, or >8 h). Medical history regarding hypertension, diabetes mellitus, and CVD was recorded. All covariates were assessed using a self-reported questionnaire.

Statistical Analysis
Statistical analysis was performed using SAS version 9.4 (SAS Institute Inc., Chicago, IL, USA) and IBM SPSS 20 (IBM Corp., Armonk, NY, USA). Kolmogorov-Smirnov test was used to determine the distribution of the data. To compare the differences between two groups, Mann-Whitney U test and chi-square test were used for categorical data. To compare data among multiple groups, one-way analysis of variance (ANOVA) and Kruskal-Wallis test were performed. We used binary logistic regression expressed as odds ratios (ORs) and 95% confidence intervals (CIs) to determine the association between the dietary pattern and cardiovascular risk factors. The dietary pattern was derived by RRR using PROC PLS function in SAS 9.4, and 22 food groups were considered as the predictors. After performing Pearson's correlation coefficient, triglycerides, systolic blood pressure, fasting blood glucose, and AIP were retained as the response variables ( Figure 2). In compliance with previous investigation, to obtain the dietary pattern linked to CVD risk, the value of factor loading was set at ≥0.20 [31]. The dietary factor score for each food group was calculated by summing food frequency intake weighed by their factor loadings. Finally, we only retained the first dietary factor for further analysis because it explained the maximum variation of the response variables. The derived dietary pattern was then divided into quartiles according to the dietary factor score. The reference group for the cardiovascular risk dietary pattern was quartile 1 (Q1) which was the lowest quartile of the dietary factor score, and quartile 4 (Q4) represented the highest quartile of the dietary factor score. In binary logistic regression analysis, model 1 was unadjusted, model 2 was adjusted for age, BMI, WC, and WHR, and model 3 was adjusted for model 2 variables plus education, family income, smoking, drinking alcohol, physical activity frequency, and sleep duration. The p-value < 0.05 was considered statistically significant.  Table 1 presents the demographic and lifestyle characteristics of the participants. The majority of postmenopausal women in this study had education below high school (81.5%), non-professional occupation (63.9%), low annual income (<NTD800,000: 69.1%), married status (70.3%), no smoking (97.9%), no drinking alcohol (95.5%), less physical activity frequency (<150 min/week: 55.5%), and sleep duration for 6-8 h (58.6%). The anthropometric, clinical, and biochemical data are shown in Table 2. The majority of postmenopausal women had normal BMI (44.2%), waist circumference (56.1%), and waist-to-hip ratio (68.9%). However, 31.3% postmenopausal women were overweight, 22.7% subjects were obese, 43.9% subjects had central obesity, and 31.1% subjects had abnormal waistto-hip ratio. The prevalence of hypertension, diabetes, and CVD was 11.3%, 17.5%, and 10.7%, respectively. The mean value of AIP (0.3 ± 0.3) was higher than 0.24 defined as a CVD risk factor. The mean values of TC (5.9 ± 0.8 mmol/L) and LDL-C (3.7 ± 0.8 mmol/L) were abnormal among postmenopausal women. Among 5689 participants, only 7.3% subjects had normal FBG level (<7.0 mmol/L, data not shown).   Table 1 presents the demographic and lifestyle characteristics of the participants. The majority of postmenopausal women in this study had education below high school (81.5%), non-professional occupation (63.9%), low annual income (<NTD800,000: 69.1%), married status (70.3%), no smoking (97.9%), no drinking alcohol (95.5%), less physical activity frequency (<150 min/week: 55.5%), and sleep duration for 6-8 h (58.6%). The anthropometric, clinical, and biochemical data are shown in Table 2. The majority of postmenopausal women had normal BMI (44.2%), waist circumference (56.1%), and waistto-hip ratio (68.9%). However, 31.3% postmenopausal women were overweight, 22.7% subjects were obese, 43.9% subjects had central obesity, and 31.1% subjects had abnormal waist-to-hip ratio. The prevalence of hypertension, diabetes, and CVD was 11.3%, 17.5%, and 10.7%, respectively. The mean value of AIP (0.3 ± 0.3) was higher than 0.24 defined as a CVD risk factor. The mean values of TC (5.9 ± 0.8 mmol/L) and LDL-C (3.7 ± 0.8 mmol/L) were abnormal among postmenopausal women. Among 5689 participants, only 7.3% subjects had normal FBG level (<7.0 mmol/L, data not shown).

Cardiovascular Risk Dietary Pattern
A dietary pattern identified as a "cardiovascular risk dietary pattern" was derived using the RRR model. Four food groups including processed food, rice/flour products, organ meat, and sauce showed a positive correlation (factor loading ≥0.20) with the cardiovascular risk dietary pattern, meanwhile food groups such as dairy products, fruits, whole grains, and sweet bread had a negative correlation with this dietary pattern (factor loading ≤−0.20) (Figure 3). The cardiovascular risk dietary pattern explained 6.6% cumulative percentage of variation and 1.7% of the total variation for the four response variables. The explained variation was 1.5% for TG, 1.6% for AIP, and 1.8% for both SBP and FBG. organ meat, and sauce showed a positive correlation (factor loading ≥ 0.20) with the cardiovascular risk dietary pattern, meanwhile food groups such as dairy products, fruits, whole grains, and sweet bread had a negative correlation with this dietary pattern (factor loading ≤ −0.20) (Figure 3). The cardiovascular risk dietary pattern explained 6.6% cumulative percentage of variation and 1.7% of the total variation for the four response variables. The explained variation was 1.5% for TG, 1.6% for AIP, and 1.8% for both SBP and FBG.

Association between the Dietary Pattern and Cardiovascular Risk Factors
The unadjusted and adjusted models for the association of the cardiovascular risk dietary pattern with SBP, DBP, and AIP among postmenopausal Taiwanese women are presented in Table 3. Model 1 was unadjusted, model 2 was adjusted for age, BMI, WC, and WHR, and model 3 was adjusted for model 2 variables plus education, family income, smoking, drinking alcohol, physical activity frequency, and sleep duration. The results showed that participants in the higher quartiles (Q3 and Q4) of the cardiovascular risk dietary pattern were more likely to increase the odds of having high SBP (OR = 1. The association of the cardiovascular risk dietary pattern with TC, LDL-C, and HDL-C among postmenopausal Taiwanese women in the unadjusted and adjusted models is shown in Table 4. The cardiovascular risk dietary pattern was not correlated with the odds of high TC in all models. Participants in the Q2 quartile of the cardiovascular risk dietary pattern were more likely to decrease the odds of having high LDL-C in all models (model 1: OR = 0.63, 95% CI 0.47-0.83, model 2: OR = 0.68, 95% CI 0.51-0.91, model 3: OR = 0.71, 95% CI 0.53-0.94) compared to those in the Q1 quartile. Participants in the higher quartiles (Q3 and Q4) of the cardiovascular risk dietary pattern were more likely to decrease the odds of having low HDL-C in the unadjusted model; however, no association was found after adjustment in models 2 and 3. Table 5 demonstrates the association of the cardiovascular risk dietary pattern with TG, FBG, and CRP in the unadjusted and adjusted models among postmenopausal Taiwanese women. Participants in the higher quartiles (Q2-Q4) of the cardiovascular risk dietary pattern were more likely to increase the odds of having high TG in all models

Association between the Dietary Pattern and Cardiovascular Risk Factors
The unadjusted and adjusted models for the association of the cardiovascular risk dietary pattern with SBP, DBP, and AIP among postmenopausal Taiwanese women are presented in Table 3. Model 1 was unadjusted, model 2 was adjusted for age, BMI, WC, and WHR, and model 3 was adjusted for model 2 variables plus education, family income, smoking, drinking alcohol, physical activity frequency, and sleep duration. The results showed that participants in the higher quartiles (Q3 and Q4) of the cardiovascular risk dietary pattern were more likely to increase the odds of having high SBP (OR = 1. The association of the cardiovascular risk dietary pattern with TC, LDL-C, and HDL-C among postmenopausal Taiwanese women in the unadjusted and adjusted models is shown in Table 4. The cardiovascular risk dietary pattern was not correlated with the odds of high TC in all models. Participants in the Q2 quartile of the cardiovascular risk dietary pattern were more likely to decrease the odds of having high LDL-C in all models (model 1: OR = 0.63, 95% CI 0.47-0.83, model 2: OR = 0.68, 95% CI 0.51-0.91, model 3: OR = 0.71, 95% CI 0.53-0.94) compared to those in the Q1 quartile. Participants in the higher quartiles (Q3 and Q4) of the cardiovascular risk dietary pattern were more likely to decrease the odds of having low HDL-C in the unadjusted model; however, no association was found after adjustment in models 2 and 3. Table 5 demonstrates the association of the cardiovascular risk dietary pattern with TG, FBG, and CRP in the unadjusted and adjusted models among postmenopausal Taiwanese women. Participants in the higher quartiles (Q2-Q4) of the cardiovascular risk dietary pattern were more likely to increase the odds of having high TG in all models (model 1: OR = 1. 39

Association between the Dietary Pattern and Cardiovascular Risk Factors
In this cross-sectional study of 5689 postmenopausal Taiwanese women, we derived the cardiovascular risk dietary pattern and found a positive association with several CVD risk factors such as SBP, DBP, AIP, TG, and FBG. Among the participants in the highest quartile of the cardiovascular risk dietary pattern, 60.4% of postmenopausal women were overweight or obese, and 64.6% were physically inactive (<150 min/week) (data not shown). The cardiovascular risk dietary pattern was recognized by high consumption of processed food, rice/flour products, organ meat, and sauce, but low intakes of dairy products, fruit, whole grains, and sweet bread. The cardiovascular risk dietary pattern reflected similar characteristics as the western dietary pattern recognized by high intakes of processed food, meat, organ meat, rice/flour products, but low consumption of fruit, dark-colored vegetables, bread, and legume/soy products among Taiwanese middle-aged and elderly with chronic kidney disease [23]. Processed food and organ meat are often rich in calories, cholesterol, and/or saturated fat, and all of which could contribute to excessive energy consumption [23].
Low fiber and excessive salt and/or sugar in processed food as well as unbalanced saturated and unsaturated fats in animal food could be correlated with abnormal blood pressure, blood lipids, and blood glucose among Taiwanese middle-aged adults and elderly [23]. Highly refined carbohydrate in rice/flour products, a dietary component for high intake in the cardiovascular risk dietary pattern, could be associated with increases in cardiovascular risk and the development of atherosclerosis among middle-aged adults [32].
Our results revealed that the cardiovascular risk dietary pattern was positively associated with blood pressure. We found that the prevalence of hypertension was only 11.3% among 5689 postmenopausal Taiwanese women. Unlike our results, the previous studies conducted among postmenopausal women reported that the prevalence of hypertension was 31.6% and 56.0% in Brazilian and Chinese postmenopausal women, respectively [20,33]. Weight gain and increased sensitivity to salt in the diet might occur due to hormonal changes after menopause and age-associated metabolic changes, which could lead to a raise in blood pressure [34]. Weight status and physical activity could also contribute to abnormal blood pressure. Postmenopausal women aged <65 years with overweight (33.3%) or obesity (42.9%) also had higher prevalence of high blood pressure (130 mmHg/85 mmHg) compared to those who had normal weight (18.8%), and those who did not do aerobic exercise tended to have higher prevalence of high blood pressure compared to those who did aerobic exercise actively (44.0% vs. 14.3%, p = 0.06) [20].
Our findings showed that the cardiovascular risk dietary pattern was correlated with an increase in AIP among postmenopausal women. Numerous studies demonstrated that AIP was an important cardiovascular risk factor and a better predictor for CVD [33,35,36]. The previous studies have reported that AIP was a better predictor of the fractional esterification rate of HDL-C which is a powerful predictor of CVD [35], and a more sensitive diagnostic marker for studies of CVD [35], and a more sensitive diagnostic marker for CVD among postmenopausal women compared to traditional lipid parameters [35,36].
Our results revealed that participants in the highest quartile (Q4) of the cardiovascular risk dietary pattern were more likely to increase the odds of having high CRP before adjustment, even the association between dietary pattern and CRP was not significant after adjustment. A previous study conducted in Southern Brazil among postmenopausal women also observed that participants with high CRP were positively correlated with BMI, WC, body fat, TG, glucose, sedentary lifestyle, and excessive dietary carbohydrate intake (>55% of total energy) [37].
Although the association between aberrant lipid profiles and certain nutrients or food groups has been established, few have demonstrated the association between dietary pattern or quality and blood lipids in postmenopausal women [32]. We found that the cardiovascular risk dietary pattern was positively correlated with increased odds of high TG among postmenopausal women after full adjustment. Brazilian postmenopausal women with a low-quality inadequate diet characterized by an excessive intake of sodium (>2400 mg/day) had increases in the prevalence of high TC and high LDL-C known as cardiovascular risk factors [20]. However, Tardivo and co-workers [32] showed that there was no significant association between diet quality determined by healthy eating index scores and blood lipids in Brazilian postmenopausal women. A study conducted among Korean women showed that postmenopausal women who consumed the western dietary pattern with high intakes of oil and fats, meat, eggs, fast food, and sweets but low intake of grains were correlated with hyper LDL-C [38]. Other studies conducted among Chinese women and Japanese women consuming a western dietary pattern with high intakes of milk, dairy products, and fast food but a low intake of rice or vegetables revealed an imbalance in lipid profile, especially increases in TC and LDL-C [39,40]. The abnormality of various serum lipids was linked to hormonal changes, such as the rise in circulating androgen and the reduction in estrogen, during the menopausal transition period [39].
Adherence to a western type dietary pattern could be associated with the status of being overweight or obese and having high WC, which might contribute to metabolic alteration. The metabolic changes in postmenopausal women could explain the imbalance of CVD-related biochemical variables [41]. Because of estrogen deficiency, postmenopausal women could increase CVD risk factors including central obesity, elevated blood pressure, increased blood lipids, decreased glucose tolerance, and increased vascular inflammation [42]. Compared to premenopausal women, postmenopausal women were more prone to increase blood lipids, which could lead to increase the risk for the development of atherogenesis [16]. In addition, the dietary components could be correlated with abnormal CVD-related biochemical variables in postmenopausal women. High consumption of energy [43], saturated fatty acids [44,45], trans fats [45], cholesterol [46], and eggs [46] was associated with an increased risk of CVD or abnormal CVD-related biochemical variables among postmenopausal women. In contrast, a low-fat dietary pattern [45] or the dietary pattern with high consumption of plant food such as whole grains, vegetables, fruits, legumes, and nuts or seeds, but low intakes of processed food, red meat, sugar, and sodium [47] were correlated with a reduced risk of CVD among postmenopausal women. The cardiovascular risk dietary pattern identified in our study was characterized by high intakes of processed food, rice/flour products, organ meat, and sauce which were accompanied by a high amount of energy, saturated fats, trans fats, cholesterol, added sugar, and sodium. Although the underlying mechanism for the effects of dietary patterns or dietary components on CVD risk factors among postmenopausal women has not been fully understood, changes in lipid metabolism and the increased accumulation of visceral fat related to estrogen deficiency in postmenopausal women could partially contribute to the effect of the dietary pattern on CVD risk factors.

Strengths and Limitations
To our knowledge, the present study is the first one to identify the cardiovascular risk dietary pattern in postmenopausal Taiwanese women using the RRR model as a novel and powerful method. Additionally, the RRR model gave more explanation about the association between the dietary pattern and the disease of interest. Since the RRRderived dietary pattern was generated by a disease-specific response, the response variables were correlated to the disease of concern [48]. Instead of explaining the variation in significant biomarkers, principal component analysis only provided the explanation of the overall variation in food group intake [48]. Meanwhile, by maximizing the explained variation in the biomarkers for diet-related disorders, the RRR model could be able to predict dietary pattern scores. Researchers can also determine the percentage variance using the RRR approach from the predictor variables and response variables, and both of which contributed to the dietary component [48]. Both the corresponding response scores and the explained variation in the predictor variables could be used to evaluate the extracted factor scores [48]. The large study population collected for 15 years could be representative of postmenopausal Taiwanese women. We also included demographic, anthropometric, clinical, biochemical, and dietary data to explore the association between these variables. However, a number of methodological limitations need to be addressed. First, our study was a cross-sectional study which provided features of eating habits and other characteristics at a specific time point and could raise the possibility of reverse causation bias. Second, the information for FFQ used to identify dietary habits could have self-reported bias. Additionally, the FFQ could be used for an estimate of habitual food intake but not for actual nutrient consumption. Even though the analysis was adjusted for the majority of known confounding variables, the residual confounding bias due to unknown or unmeasured covariates could not be completely ruled out. A longitudinal study is needed to explore the association between dietary patterns and CVD risk factors among postmenopausal Taiwanese women. Further research should be conducted to compare the association in premenopausal versus postmenopausal Taiwanese women.

Conclusions
The cardiovascular risk dietary pattern with a high intake of processed food, rice/flour products, organ meat, and sauce is associated with increased odds of high blood pressure, AIP, TG, and FBG among postmenopausal women. Our study suggests that choosing a healthier dietary pattern with a lower intake of processed food, rice/flour products, organ meat, and sauce could reduce the risk of CVD in postmenopausal Taiwanese women.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki, and approved by the Taipei Medical University-Joint Institutional Review Board (N202007075 and date of approval 11 September 2020).

Informed Consent Statement:
All the participants signed the consent form authorized by the Mei Jau Health Management Institution.

Data Availability Statement:
The data that support the findings of this study are available from the Mei Jau Health Management Institution, but restricted for research use only. The data are not publicly available.