Association Between Low-Carbohydrate Diet Scores and Incidence of Hypertension Among the Middle-Aged Japanese Population: The Toon Health Study

Abstract

This study investigates the association between Low-Carbohydrate Diet (LCD) score and the incidence of hypertension in the Japanese population. This cohort study was conducted in an ongoing longitudinal study, the Toon Health Study. Hypertension was defined by measured blood pressure levels and self-reported treatment information, and dietary intake was measured through a validated food frequency questionnaire. Total, animal and plant-based LCD scores were calculated. Participants were divided into sex-specific tertiles of LCD scores. A multivariable-adjusted logistic regression model was used to calculate odds ratio (OR) and 95% confidence interval (95% CI) of incidence of hypertension. Total and plant-based LCD scores showed a borderline negative association with incidence of hypertension. The ORs (95% CI) for the highest versus lowest tertile of the total LCD score were 0.61(0.37–1.03, p for trend = 0.07) and 0.65(0.39–1.10, p for trend = 0.09) for the plant-based LCD score. Stratified analysis by drinking status showed significant negative association among alcohol drinkers, multivariable ORs (95% CI) for the highest versus lowest tertile of total LCD score was 0.38 (0.17–0.83, p for trend = 0.01) and for the plant-based LCD score was 0.39 (0.17–0.90, p for trend = 0.01). No significant association observed in non-drinkers. In conclusion, an increased LCD score was associated with the lower incidence of hypertension, especially in the drinking population.

1. Introduction

Hypertension is associated with several health complications, i.e., coronary heart disease (CHD), cardiovascular disease (CVD), peripheral arterial disease (PAD), renal dysfunction, arterial fibrillation, heart failure, stroke, and dementia [1,2]. High blood pressure levels were assessed as the second largest health risk factor globally, as 19.2% (16.9–21.3%) of total deaths, or 10.8 million, were caused by hypertension in 2019 [3]. The global burden of hypertension doubled from 650 million to 1.3 billion in just twenty years from 1990 to 2019, according to the WHO 2019 [4]. It is estimated that 10% of global health care spending is directly related to high BP levels and their related complications, and this spending reaches around 25% of total health care spending in Central Asian and Eastern European countries [5]. However, the control of hypertension is gradually improving, but it is still a major quality-of-life challenge.

Dietary patterns play a significant role in blood pressure regulation [6]. In Asian dietary patterns, most dietary requirements are driven by carbohydrate sources, accounting for approximately 60% of the total energy intake [7]. Low-Carbohydrate Diet Score (LCD score) is a comparatively new approach that targets macronutrients and explains the association between macronutrient intake and cardiovascular diseases [8]. This approach was introduced by Halton et al. [9], which provides a cumulative score of micronutrients in diet according to the relative intake of fats, proteins, and carbohydrates. Some previous studies showed the association between LCD and hypertension, as a cross-sectional study showed that LCD is significantly associated with lowering blood pressure (BP) levels [10], and the other cross-sectional study also demonstrated the association of LCD in reducing BP levels [11]. On the other hand, alcohol drinking and having high blood pressure are the major modifiable risk factors for the burden of non-communicable diseases [12], including hypertension [13]. Additionally, alcohol drinking may affect the dietary patterns, especially carbohydrate intakes [14,15,16,17]. The evidence of LCD scores and the incidence of hypertension examined by the cohort design is limited; moreover, there are limited studies that explain the potential modifying effect of alcohol consumption. Thus, this study primarily aims to examine the association between LCD scores and hypertension, and additionally, to observe the differential effect of alcohol drinking among the Japanese population.

2. Materials and Methods

• Study Participants:

This is a cohort study, part of an ongoing longitudinal epidemiological study, the Toon Health study, based on the population of Toon City, in Ehime prefecture, Japan. The total population of this city was around 35,000 individuals by 2009. The participants for this study joined voluntarily and were invited through newspapers, poster advertisements, and direct invitations. The baseline data was taken from the survey from 2009 to 2012, a total of 2032 (726 males and 1306 females) aged 30 to 79 years old participated in this survey. Among those, 1461 participants participated in the follow-up data was taken in 2014–2018. We excluded the persons who were defined as hypertensive in the baseline survey. Finally, the remaining participants were 912 (259 males and 653 females). A written informed consent was provided to every participant, and the study procedure (170511) was approved by the Institutional Review Board of Ehime University Hospital, Approval Code: #170511, Approval Date: 22 May 2017.

• Measurement of blood pressure levels and definition of hypertension:

Two blood pressure readings were taken from participants in a seated position after a minimum of 5 min of rest using an automatic sphygmomanometer (BP-103iII; OMRON Colin Co., Ltd., Tokyo, Japan). The mean of the two measurements was used for analysis. The measurements were taken between 8:00 a.m. and 11:00 a.m. In this study, hypertension was defined as having a systolic blood pressure of 140 mmHg or higher, and/or a diastolic blood pressure of 90 mmHg or higher, and/or being on hypertension medication. Normally, the participants arrived without having breakfast, and in most cases, the measurements were conducted before any meal intake. Therefore, the blood pressure readings can be considered as taken in a pre-prandial state.

• Dietary assessment and LCD score calculation:

A reasonably validated food frequency questionnaire (FFQ) was used to assess the dietary intake of participants [18]. Validity of the FFQ was assessed in a subset of participants of this study, comprising 21 females and 14 male individuals. Following natural logarithmic transformation and residual method adjustment for energy intake, the Spearman rank correlation coefficients between the FFQ and the 7-day dietary records for energy and main nutrient intakes were as follows: 0.62 for protein, 0.56 for fat, 0.49 for sodium, 0.45 for carbohydrate, 0.35 for total dietary fiber, 0.27 for energy, 0.21 for fruits, 0.14 for green vegetables, and 0.40 for other vegetables. Among our participants, protein, fat, sodium, carbohydrate, and other vegetables showed reasonable validity, but energy, fruit, and green vegetables showed low-to-middle validity. We estimated energy and nutrient intakes using Japan’s Standard Tables of Food Composition (sixth revised version) [19]. All nutrients and dietary groups were adjusted for energy consumption by using the density method.

Three LCD scores were calculated according to the approach introduced by Nakamura et al. [20]. This approach is a modified version of a method introduced by Halton et al. [9]. Using the density method, the total LCD score divides the energy-adjusted consumption of protein, fat, and carbohydrates into deciles. A final LCD score ranging from 0 to 30 points was obtained by adding the energy-adjusted carbohydrate intake in descending deciles and the energy-adjusted fat and protein intake in ascending deciles, which were labeled 1–10. Eleven strata were identified based on the proportions of energy that were consumed by fat, protein, and carbohydrates. The protein and fat scores, which range from 0 to 10, rise with the strata level. With a range of 10–0, carbohydrates show the opposite tendency. A total score between 0 and 30 was obtained by adding the final three macronutrient scores. Subjects with higher scores ate less carbohydrates and more fat and protein. Furthermore, the study also estimated two additional LCD scores: the plant-based LCD score and the animal-based LCD score. The percentage of energy obtained from saturated fatty acids (SFA), animal proteins, and carbohydrates was calculated to determine the Animal-based LCD score. The percentage of energy obtained from carbohydrates, plant protein, and monounsaturated fatty acids + polyunsaturated fatty acids (MUFA + PUFA) was calculated to determine the plant-based LCD score.

• The other measurements:

The participant’s weight in light clothing and height in stocking feet were measured. BMI was calculated as measured weight (kg)/height (m)². Participants were questioned by doctors about their medical history, including whether they had hypertension. The trained staff utilized a self-administered questionnaire to ask about alcohol and smoking habits. The Japan Arteriosclerosis Longitudinal Study Physical Activity Questionnaire (JALSPAQ), which was validated by Ishikawa-Takata et al. [21], is a frequently utilized questionnaire with 14 items about occupation, locomotion, housework, sleep duration, and leisure time spent engaging in physical activity. It was used to calculate the level of physical activity. The answers were converted into evaluations of each physical activity’s intensity, which were then shown as metabolic equivalents (METs) and given as METs per hour per day (METs h/day). Alcohol intakes of equal to 1 g or over per week were defined as regular alcohol drinkers. Based on average ethanol consumption, alcohol drinkers were further divided into two groups: ethanol intake <21 g/day or ≥21 g/day [22]. Diabetes was defined as: if a fasting serum glucose level ≥126 mg/dL, or a 2 h postprandial serum glucose level of 75 g glucose tolerance test ≥200 mg/dL, or the participant was taking anti-hyperglycemic medicines. Dyslipidemia was defined as a fasting serum triglyceride level ≥150 mg/dL, HDL-cholesterol level <40 mg/dL, LDL-cholesterol level ≥140 mg/dL, or if taking cholesterol control medication.

• Statistical analysis:

ANCOVA was used to estimate the age and sex-adjusted means and proportions of participants’ characteristics based on sex-specific tertiles of three LCD scores. Regression analysis was used to evaluate the linear trend by using the median value of each tertiles after age and sex adjustment. After adjusting for age, sex, physical activity, BMI, smoking, alcohol intake, and intakes of energy, sodium, and dietary fiber, the odds ratio (OR) and 95% confidence intervals (95% CIs) of hypertension were estimated by using a logistic regression model according to the tertiles of the three LCD scores. Stratified analyses were also performed by alcohol consumption (yes/no). The effect modification by drinking status on the association between LCD scores and hypertension was tested by using the multivariable-adjusted logistic regression model. p value < 0.05 was considered statistically significant, and probability values for statistical tests were two-tailed. SAS Institute Inc., located in Cary, NC, USA, produced the SAS version 9.4 software program, which was used for all statistical analyses.

3. Results

Table 1. Subject characteristics according to the tertiles of LCD score at baseline.

	T1 (Low)	T2	T3 (High)	p for Trend
Number of subjects	291	307	314
Age (Years)	56.2	55.1	52.9	<0.01
Men (%)	28.8	28.7	27.7	-
BMI (kg/m2)	22.5	22.7	22.7	0.36
Current alcohol drinkers (%)	62.7	63.2	56.5	0.08
Ethanol Intake (g/day)	10.8	28.6	27.7	0.16
Current smokers (%)	12.9	13	10.4	0.21
Physical activity (METs h/day)	35.5	35.9	35.5	0.92
Dyslipidemia (%)	35.8	41.7	37.2	0.83
Diabetes (%)	6.7	4.1	6.0	0.83
Dietary intakes
Total energy (kcal)	1771.3	1912.4	2030.3	<0.01
Energy from Carbohydrates (%E)	63.1	57.4	51.9	<0.01
Energy from Protein (%E)	12.4	13.9	15.2	<0.01
Energy from Fat (%E)	24.5	28.6	32.9	<0.01
Energy from MUFA and PUFA (%E)	13.7	16	18.4	<0.01
Energy from SFA (%E)	7.6	8.9	10.2	<0.01
Dietary Fiber (g/1000 kcal)	6.9	7.1	7	0.40
Sodium (mg/1000 kcal)	1879.7	1955.3	1944.5	0.16
Plant-based LCD score	8.8	15.0	20.8	<0.01
Animal-based LCD score	6.4	15.0	23.1	<0.01

Age and sex adjusted means or proportions.

shows the means and proportions of subject characteristics adjusted for age and sex according to the tertiles of the total LCD score. The total LCD score was negatively associated with age and energy from carbohydrates and was positively associated with intakes of energy from protein, fat, MUFA, PUFA, and SFAs, total energy, and the mean of animal and plant-based LCD scores. No significant association was found between LCD score and BMI, current smokers, current alcohol drinkers, ethanol intake, dyslipidemia, diabetes, dietary fiber, sodium, and physical activity. Table S1 shows the means and proportions of subject characteristics adjusted for age and sex according to the tertiles of animal-based LCD score. The animal-based LCD score was negatively associated with age, current smokers, dietary fiber, and energy from carbohydrates, and the animal-based LCD score was positively associated with intakes of energy from protein, fat, MUFA, PUFA, and SFAs, and total energy. No significant association was found between animal-based LCD score and BMI, current alcohol drinkers, ethanol intake, physical activity, dyslipidemia, diabetes, and sodium. Table S2 shows the means and proportions of subject characteristics adjusted for age and sex according to the tertiles of the plant-based LCD score. The plant-based LCD score was negatively associated with age, current alcohol drinkers, ethanol intake, and energy from carbohydrates, and the plant-based LCD score was positively associated with intakes of energy from protein, fat, MUFA, PUFA, and SFAs, sodium, dietary fiber, and total energy. No significant association was found between plant-based LCD score and BMI, current smokers, physical activity, dyslipidemia, and diabetes.

Multivariable adjusted OR and 95% CI for the incidence of hypertension according to LCD scores are shown in

Table 2. Multivariable adjusted odds ratios (OR) and 95% confidence intervals (95% CI) for incidence of hypertension according to the tertiles of LCD scores.

	T1 (Low)	T2	T3 (High)	p for Trend
Total LCD
Cases of hypertension /number of subjects	44/291 (15.1%)	29/307 (9.4%)	28/314 (8.9%)
Age and sex-adjusted OR (95% CI)	1.00	0.60 (0.36–0.99)	0.61 (0.37–1.03)	0.07
Multivariable OR (95% CI) *	1.00	0.62 (0.37–1.06)	0.67 (0.38–1.17)	0.17
Animal-based LCD
Cases of hypertension /number of subjects	37/307 (12%)	37/314 (11.8%)	27/291 (9.2%)
Age and sex-adjusted OR (95% CI)	1.00	0.98 (0.60–1.61)	0.84 (0.49–1.44)	0.50
Multivariable OR (95% CI) *	1.00	1.04 (0.62–1.74)	0.88 (0.50–1.57)	0.65
Plant-based LCD
Cases of hypertension /number of subjects	42/289(14.5%)	32/321(9.9%)	27/302(8.9%)
Age and sex-adjusted OR (95% CI)	1.00	0.70(0.42–1.15)	0.65(0.39–1.10)	0.09
Multivariable OR (95% CI) *	1.00	0.79(0.47–1.33)	0.83(0.47–1.46)	0.45

* Adjusted for sex, age, BMI, physical activity, smoking, ethanol intake and intakes of energy, sodium, fiber, diabetes and dyslipidemia at baseline.

. A borderline association between total LCD score and hypertension was found in the age and sex-adjusted model the age and sex-adjusted OR (95% CI) of hypertension for highest tertile of total LCD score in comparison with lowest was 0.61(0.37–1.03, p for trend = 0.07), but after adjusting for confounding factors, the associations became weakened and non-significant. Furthermore, there were no significant associations between animal-based and plant-based LCDs and the incidence of hypertension.

Table 3. Multivariable adjusted odds ratios (OR) and 95% confidence intervals (95% CI) for incidence of hypertension according to the tertiles of LCD scores stratified by drinking status.

	T1 (Low)	T2	T3 (High)	p for Trend	p for Interaction
Total LCD
Alcohol drinkers
Cases of hypertension/number of subjects	28/160 (17.5%)	19/173 (10.9%)	11/160 (6.8%)		0.04
Multivariable OR (95% CI)	1.00	0.61(0.31–1.19)	0.38(0.17–0.83)	0.01
Non-Drinkers
Cases of hypertension/number of subjects	16/131 (12.2%)	10/134 (7.5%)	17/154 (11.0%)
Multivariable OR (95% CI)	1.00	0.62 (0.25–1.49)	1.29 (0.56–2.98)	0.47
Animal-based LCD
Alcohol drinkers
Cases of hypertension/number of subjects	24/171 (14.0%)	20/166 (12.0%)	14/156 (8.9%)		0.20
Multivariable OR (95% CI)	1.00	0.87 (0.44–1.72)	0.65 (0.30–1.37)	0.26
Non-Drinkers
Cases of hypertension/number of subjects	13/136 (9.5%)	17/148 (11.5%)	13/135 (9.6%)
Multivariable OR (95% CI)	1.00	1.25 (0.55–2.82)	1.32 (0.52–3.33)	0.55
Plant-based LCD
Alcohol drinkers
Cases of hypertension/number of subjects	29/166 (12.6%)	19/176 (10.8%)	10/151 (6.6%)		0.03
Multivariable OR (95% CI)	1.00	0.65 (0.33–1.25)	0.39 (0.17–0.90)	0.01
Non-Drinkers
Cases of hypertension/number of subjects	13/123 (10.6%)	13/145 (8.9%)	17/151 (11.2%)
Multivariable OR (95% CI)	1.00	1.02 (0.43–2.42)	1.88 (0.78–4.52)	0.17

Adjusted for sex, age, BMI, physical activity, smoking, intakes of energy, sodium, fiber, diabetes, and dyslipidemia at baseline.

shows the multivariable-adjusted OR and 95% CIs for hypertension across tertiles of total, plant-based, and animal-based LCD scores stratified by alcohol consumption (current alcohol drinkers and non-drinkers). There was a significant association between total, plant-based LCD score and the incidence of hypertension among current alcohol drinkers. The multivariable OR (95% CI) for the highest tertile of total LCD score in comparison with the lowest tertile was 0.38 (0.17–0.83, p for trend = 0.01), and the respective OR (95%CI) for the highest tertile of plant-based LCD score was 0.39 (0.17–0.90, p for trend = 0.01). However, there was no significant association of respective LCD scores with the incidence of hypertension found for the non-drinkers. We also observed significant effect modification by alcohol consumption on the associations of total and plant-based LCD score and incidence of hypertension (respective p for interaction = 0.04 and 0.03). On the other hand, a non-significant association was found between animal-based LCD scores and incidence of hypertension, with the multivariable OR (95% CI) for the highest tertile in comparison with the lowest tertile was 0.65 (0.30–1.37, p for trend = 0.26).

4. Discussion

This study examined the association between LCD scores by their dietary sources (total, animal-based, and plant-based) and the incidence of hypertension. To the best of our knowledge, this is the first Asian population-based cohort study that examines the incidence of hypertension according to the LCD scores by their sources. We found that higher total and plant-based LCD scores were significantly associated with lower hypertension in alcohol drinkers, but we could not find significant associations in the total population and non-drinkers. We found that alcohol consumption could modify the associations of total and plant-based LCD scores with hypertension.

Our study showed an insignificant association between LCD scores and hypertension in the total population. It was observed that the energy consumption in the highest tertile of the LCD score is not as low as typical low-carbohydrate diets. It was observed that the percentage of energy from carbohydrates in the highest LCD score tertile was 51.9% which might not be considered typical LCD. This relatively high carbohydrate intake might have limited the associations of LCD scores with hypertension, and the association could not reach the level of significance for the total population. Our LCD score approach is not based on any kind of strict cut-off value of carbohydrate consumption, but it is a relative approach that seems appropriate to apply for observational studies. It might be one of the potential limitations of our dataset. However, in the present study, an inverse association between the LCD score and hypertension was observed among alcohol drinkers. In Japan, there has been little improvement in the proportion of individuals who consume alcohol at levels that increase the risk of lifestyle-related diseases, including hypertension [23]. Therefore, identifying dietary habits that may reduce the risk of hypertension among alcohol drinkers has potential value, particularly in Japan, where a substantial proportion of the population engages in such high-risk alcohol drinking behavior.

A systematic review and meta-analysis of 12 clinical trials also showed the reverse (negative) association between a low-carbohydrate diet and overall blood pressure levels [24]. The low-carbohydrate diet significantly improves endothelial function by lowering insulin levels and oxidative stress, as well as limiting the production of reactive aldehydes and advanced glycation end products (AGEs), which helps to maintain the inner lining of blood vessels by improving the production of nitric oxide (NO), which smoothen the blood vessels and results in maintaining blood pressure levels, as this study [25] explains that the LCD is associated with blood pressure levels and also reduces body weight but the study did not explained the influence of body weight on blood pressure levels, as per our current investigation, the association remained even after adjusting for BMI. This shows that while body size may play an essential role, but according to the results of the current study, we highlighted the discussion of the reasons that are independent of body size.

This study showed that plant-based LCD scores were significantly associated with lower hypertension in alcohol drinkers. A recent systematic review and meta-analysis of the 39 clinical trials [26] proposed that plant-based diets significantly decrease blood pressure levels. Plant-based proteins are rich in glutamic acid, which has glutathione molecules, and play a role in improving blood pressure homeostasis. Another cross-sectional epidemiological study and a review article [27] further supported the effectiveness of Glutamic Acid, which is high in plant-based proteins, but a cross-sectional study on the older population [28] did not support the association between glutamic acid and blood pressure levels. Plant proteins are a source of Angiotensin Converting Enzyme (ACE) inhibitors, which are effective in blocking the enzyme that converts angiotensin I to angiotensin II, which are the substances that cause narrow blood vessels which resulting in high blood pressure levels [29]. Therefore, the findings in our study could be reasonable.

This study showed the significant associations of higher LCD scores with lower incidence of hypertension in the drinking population; according to the meta-analysis, it is also suggested that moderate alcohol consumption, 5 to 25 g/day, is associated with lowering the risk of cardiovascular diseases [30]. According to a systematic review of prospective cohort studies, light and moderate intake of alcohol is associated with a reduced risk of multiple cardiovascular outcomes [31]. According to the subject characteristics of our study population, the mean consumption of ethanol is 27.7 g/day which is near the moderate level of alcohol consumption per day; a comparative LCD and moderate alcohol consumption might be one of the potential reasons that we found a significant association in the alcohol drinker’s population following the plant-based LCD score diet pattern.

Despite the findings, this study might have some limitations. Firstly, the dietary intake was assessed by using a validated FFQ, but there is still a probability of some recall bias. Secondly, we have already measured and adjusted for all major confounding factors that could affect the outcome, but still some confounding factors might not be measured, like psychological stress or genetic factors, but as the major confounders were adjusted so the unmeasured factors might not affect the findings. Thirdly, the nature of this study is observational, and it might prevent us from finalizing the causal relationship between LCD scores and the incidence of hypertension. One more potential limitation is that we cannot rule out the possibility that the calculated LCD score includes the contribution of alcohol, because the software used in this study to calculate nutrient intake from FFQ does not estimate alcohol consumption. Moreover, a separate questionnaire was used to assess alcohol consumption, which was not incorporated into the calculation of the LCD score. Therefore, a higher LCD score among the drinking population may not necessarily reflect the lower alcohol consumption, and we cannot entirely exclude this possibility. Lastly, the study was conducted in a specified Japanese population; the potential cultural, dietary, social, or food accessibility differences might affect the applicability of results in a generalized population.

5. Conclusions

In conclusion, our study suggests that the higher total and plant-based LCD scores tend to be associated with a lower risk of hypertension in the drinking population. In the future, the research needs to be performed in a more diverse population to explore the causal mechanisms and investigate these associations to validate our findings.