Study population

This cross-sectional study extracted participant results from the Korea National Health and Nutrition Examination Surveys (KNHANES) 2008–2011. As previously described in detail, the KNHANES is a nationwide, population-based, cross-sectional health examination and annual survey. The survey is conducted by the Division of Chronic Disease Surveillance of the Korea Centers for Disease Control and Prevention in the Ministry of Health and Welfare to monitor the public health and nutrition in South Korea [13]. Each KNHANES is composed of independent data sets from the general population of Korea. As described in Fig. 1, of 37,753 participants from KNHANES 2008–2011, we initially selected those aged 40–79 without prior history of CVD. Subjects excluded were those with missing data for dietary intake and CVD risk assessment components. Ultimately, 11,601 subjects (4813 men and 6788 women) were included in the analysis. All participants provided written informed consent. The survey protocol was approved by the institutional review board of the Korean Centers for Disease Control and Prevention (2008-04EXP-01-C, 2009-07CON-03-2C, 2010-02CON-21-C, and 2011-02CON-06C).

Measurement of clinical and laboratory parameters

KNHANES data includes a 3-part medical history, nutritional status, and laboratory tests. Subjects’ medical history was evaluated, including smoking, alcohol consumption, exercise, and disease diagnosis or treatment, based on direct interviews and self-reporting. Regular exercise was defined as more than 20 min per session and at least 3 times per week, and heavy drinkers were defined as those whose alcohol consumption was >140 g/week for men and >70 g/week for women. Blood pressure was manually measured using mercury sphygmomanometers (Baumanometer; W.A. Baum, Copiague, NY) three times on the right arms of people in resting, seated positions, and final blood pressure values were assessed by averaging the second and third blood pressure readings. Appendicular skeletal muscle (ASM) was measured using dual-energy X-ray absorptiometry (DXA, QDR 4800A; Hologic Inc., Bedford, MA, USA) and was limited in 8690 subjects. Overnight (8 h) fasting blood and spot urine samples were collected, refrigerated, and transported to the central laboratory institute (NeoDin Medical Institute, Seoul, South Korea) within 24 h. All biochemical samples were measured as previously described [14]. Estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology collaboration (CKD-EPI) equation [15]. The homeostasis model assessment of insulin resistance (HOMA-IR) was subtracted from the formula as previously described [16].

Cardiovascular disease definition and assessment

CVD risk was evaluated using various approaches, including the 10 year atherosclerotic vascular disease (ASCVD) risk score from the 2013 American College of Cardiology (ACC)/AHA guidelines, as well as the Framingham 10 year CVD risk score [17, 18]. In addition, individuals exhibiting >10 % of ACC/AHA ASCVD 10 year risk and >20 % of Framingham CVD risk were classified into the high-risk group [17, 18]. Impaired fasting glucose was defined as fasting plasma glucose ranging from 100 to 125 mg/dL without prior diabetes diagnosis. Diabetes was defined in subjects who were using oral anti-diabetic agents, or exhibited fasting plasma glucose ≥126 mg/dL or glycated hemoglobin (HbA1c) ≥6.5 %. Subjects were considered to have hypertension if their systolic blood pressure was ≥140 mmHg, or their diastolic blood pressure was ≥90 mmHg, or if they were taking anti-hypertensive medications. Body mass index (BMI) was evaluated using the ratio of weight to height (kg/m²), and overweight was defined based on the criteria in the Asians (BMI ≥23 kg/m²) [19]. Chronic kidney disease (CKD) was characterized in subject with less than 60 mL/min/1.73 m², accordance with CKD guideline [20]. Sarcopenia was defined as an ASM divided by BMI (ASM/BMI) <0.789 for men and <0.512 for women, as recommended by the international consensus meeting from the Foundation for the National Institutes of Health (FNIH) [21]. Because a standard definition of sarcopenia has not yet been established, ASM/height² and ASM/weight estimations were applied [22, 23]. Sarcopenia was defined as <2 standard deviations (SD) below the sex-specific average for a young (age 18–39), healthy reference population from the datasets (2364 men and 3093 women) [22–24]. The ASM/height² cutoff was 6.44 kg/m² for men and 4.48 kg/m² for women, whereas the ASM/weight cutoff was 28.80 and 22.72 % for men and women, respectively [25]. An age cutoff point to define younger or older adults was 55, the median value.

Dietary acid load score assessment

Dietary intake information was collected by 1 day face-to-face 24 h recall method by well-trained dietary interviewers aided by various measuring devices (a common size container, standard measuring cups and spoons, a three-dimensional food model, or two-dimensional aids). Moreover, the quality control on interview at filed was conducted throughout the survey by the Center for Nutrition Policy and Promotion at the Korea Health Industry Development Institute with the checklist of the control manual. The 24 h recall method provides high-quality nutrient data and has demonstrated its accuracy in other studies [26, 27]. Detailed information and data resource profile of the survey has been provided elsewhere [28]. To calculate PRAL, intake data were needed for protein, phosphorus, potassium, calcium, and magnesium. Because magnesium intake data were not available in KNHANES, daily magnesium intake values were estimated using an external magnesium content database developed by another research group [29]. We matched the 5105 food items contained in the KNHANES 2008–2011 to the magnesium content database of 366 foods commonly consumed by Koreans. Foods not included in the database were indirectly matched with similar foods in the database. Consumption frequencies of foods by this matching process contributed 65.4 % of the total consumption frequency in KNHANES. PRAL and DAL scores were derived from the equations of nutrient intakes and tertiles of the scores were used for statistical analysis: PRAL (mEq/day) = (0.49 × protein [g/day]) + (0.037 × phosphorus [mg/day]) − (0.021 × potassium [mg/day]) − (0.026 × magnesium [mg/day]) − (0.013 × calcium [mg/day]) [30], and DAL (mEq/day) = PRAL + (body surface area [m²] × 41 [mEq/day]/1.73 m²) [10]. Both PRAL and DAL score was categorized into sex-specific tertiles. Body surface area was calculated by the Du Bois formula: 0.007184 × height^0.725 × weight^0.425 [31, 32]. Because data on body surface area were missing for 36 subjects, 11,565 subjects with DAL scores were newly categorized according to DAL tertiles.

Statistical analysis

Data are presented as the mean ± SD for continuous variables, and presented as number (N), or percent (%) for categorical variables. We analyzed the study participants’ characteristics according to PRAL tertiles, using one-way analysis of variance (ANOVA) to compare continuous variables, χ² tests for categorical variables followed by post hoc analyses with the Bonferroni method. To evaluate the association between acid load and CVD risk, the effects of comorbidities should be minimized. Subjects were divided according to the presence of obesity, sarcopenia, and regular exercise, and then χ² tests were applied for each group. The higher PRAL group included both the second and highest PRAL tertiles, and this group was compared against the lowest PRAL tertile. Multiple logistic regression analysis was used to assess the independent association between acid load and high CVD risk (>10 % of ACC/AHA ASCVD 10 year risk score or >20 % of Framingham 10 year risk score) including other covariates. Fasting insulin, HOMA-IR, total cholesterol, triglyceride, HDL cholesterol, and LDL cholesterol values were not normally distributed; analyses were performed using log- and back-transformed data. Statistical analyses were performed using IBM SPSS version 20.0 for Windows (IBM Corp., Armonk, NY, USA); P < 0.05 was considered statistically significant.

Diet-induced acid load and baseline characteristics

Diet-induced acid load and cardiovascular disease risks

The average 10 year ACC/AHA ASCVD risk score increased as PRAL increased (from 8.5 to 9.6 %, P < 0.001, Fig. 2a), and the discrepancy among the tertiles was greater in the older age group (≥55 years) (Fig. 2b). Overall, the proportion of patients in the high-risk group (predicted 10 year risk of hard ACC/AHA ASCVD event >10 %) in our population was 32.6 %, and the distribution of high-risk group members and PRAL tertiles showed the same trend as ACC/AHA ASCVD scores (29.5 and 34.0 % for lowest and highest tertiles, respectively, P < 0.001, Fig. 2c). The link between higher PRAL score and high-risk group classification was observed in older age group (Fig. 2d). Similarly, higher PRAL scores were associated with CVD risk that was estimated by a different method, the Framingham 10 year risk score (Fig. 3). Regarding DAL score, the correlation between higher DAL score and increased ACC/AHA ASCVD score was only observed in young age group (Fig. 4). As CVD risk gradually increases according to age, we stratified subjects using 10 year units, and evaluated PRAL scores and ASCVD risk scores (Fig. 5). The proportion of individuals with higher ASCVD risk was increased in the higher PRAL population, especially among middle-aged individuals (50–59 years old). A similar trend was also observed using DAL scores.

Diet-induced acid load is linked with CVD risk independent of obesity, exercise, and insulin resistance, but not sarcopenia

As overweight and low exercise are considered important CVD risk factors, we stratified the study population according to these variables. Higher PRAL scores raised the proportion of high ACC/AHA ASCVD risk regardless of BMI (Odds Ratio [OR] ranged from 1.19 to 1.37, P < 0.05 for all groups), and this association was stronger in the lean group (Fig. 6a). Interestingly, lean subjects with higher PRAL scores exhibited a similar proportion of higher CVD risk compared to those in the obese group. Although it was not statistically significant, more lean subjects with higher PRAL scores were in the ACC/AHA ASCVD high-risk group than obese subjects with lower PRAL scores. A high-PRAL diet increased ACC/AHA ASCVD risk group distribution independent of exercise status (OR ranged from 1.23 to 1.27, P < 0.05 for all groups, Fig. 6b). Moreover, the lowest PRAL individuals who did not exercise showed comparable ACC/AHA ASCVD risk with those in the higher PRAL group who regularly exercised. Increased ACC/AHA ASCVD risks in the higher PRAL group were also observed regardless of insulin resistance status assessed by HOMA-IR (Fig. 6c). Regarding skeletal muscle mass, the association between PRAL and high-risk ACC/AHA ASCVD was stronger among subjects with preserved skeletal muscle mass (OR 1.17, 95 % CI 1.06–1.30, Fig. 6d); this association was not observed among sarcopenic subjects (OR 1.17, 95 % CI 0.81–1.69). This trend was consistently observed when other sarcopenia definitions were applied (Fig. 7).

Higher PRAL and DAL scores are associated with elevated predicted risk of CVD regardless of hypertension or diabetes

Diet-induced acid load, cardiovascular diseases and possible mechanisms

The mechanism linking diet-induced acid load and metabolic disease incidence is mainly reported as insulin resistance [7, 9]. Toward that end, insulin binding affinity to its receptor was markedly decreased in individuals with metabolic acidosis [37]. Thus, even in healthy individuals, a very low degree of metabolic acidosis could lead to insulin resistance, resulting hyperglycemia [38]. Moreover, a previous study reported that attenuating metabolic acidosis could increase insulin sensitivity [39]. In accordance with this evidence, our study results demonstrated that the highest DAL tertile was associated with increased values of HOMA-IR. Regarding the effect of diet-induced acid load on blood pressure, several mechanisms have been suggested. A diet depleted of potassium could affect vasodilatation and be toxic to the blood vessels [40]. Potassium restriction results in intracellular potassium deficits and causes compensatory sodium gains in cells to maintain tonicity and volume [40]. In a human study, even a 10 day period of low potassium intake increased systolic blood pressure by 5 mmHg (P < 0.02) and modified salt sensitivity, resulting in exacerbation of hypertension [41]. Conversely, hypertensive rats fed a potassium-rich diet showed decreases in blood pressure and stroke development [42, 43]. Likewise, our study demonstrated that PRAL and DAL scores were positively associated with both systolic and diastolic blood pressure values, as well as hypertension prevalence (OR 1.19, 95 % CI 1.09‒1.31). In addition, systemic metabolic acidosis, caused by the Western diet, is associated with excessive cortisol levels and leads to ammoniagenesis, which may lead to loss of renal function [12, 44]. Diet-induced insulin resistance could autonomously promote cardiovascular disease in various pathways; impair coronary microcirculatory function [45], stimulate conduction dysfunction and increase arrhythmogenesis [46].

Diet-induced acid load in unhealthy states

Previously, the correlation between diet-acid load and metabolic disease was stronger in non-obese individuals, suggesting that it is independent of adiposity [7, 9]. We evaluated the PRAL score and ACC/AHA ASCVD risk according to BMI as well as insulin resistance. Although this association remained significant in both lean/overweight and insulin sensitive/resistant groups, more substantial correlations were observed in lean individuals (OR 1.37 vs 1.19) and in the insulin sensitive group (OR 1.34 vs 1.21). Moreover, there was no significant difference in risk of ACC/AHA ASCVD between lean subjects with higher PRAL scores and those in the overweight group in the lowest PRAL tertile, suggesting that CVD risk of lean individuals would increase if diet-induced acid loads were elevated. An epidemiologic study that reported that dietary patterns could increase the risk of obesity, even in lean individuals, suggests that diet affects metabolism in non-obese populations [47]. In addition, if sedentary individuals lowered their intake of dietary acid, they might exhibit comparable CVD risks with people who regularly exercised. While obesity and exercise are major contributing factors to CVD risk, diet-induced acid load might modify these effects. Regardless, the association between diet-induced acid load and CVD risk with respect to muscle mass seems compelling. There has been some evidence that sarcopenia could influence CVD, such as increased arterial stiffness and inflammatory markers in people with sarcopenia [48, 49]. Our results also demonstrated increased CVD risk in the sarcopenic group compared to the non-sarcopenic group; however, the association between PRAL score and ACC/AHA ASCVD risk did not remain significant among sarcopenic individuals. Furthermore, subjects with higher PRAL scores but preserved muscle mass had lower CVD risk than those in the sarcopenic group with lower PRAL scores. This might suggest that skeletal muscle mass is more closely associated with lower CVD risks compared to healthy diet patterns.

This study has some limitations. First, our cross-sectional study design could not conclude causality between diet-induced acid load and CVD risk. Second, PRAL and DAL scores were derived from self-reports using a 24 h recall method, which only confirmed the short term dietary intakes of the study participants; the daily variation in food consumption could not be considered. Third, we could not assess the quality of protein that was used to calculate PRAL and DAL scores. In addition, KHNASE did not collect each individual’s medication information, and not evaluate physical performance status, and muscle strength. Sarcopenic status was classified by muscle mass alone, not considering muscle strength or physical performance. The potential medication effect on CVD risk was not considered in our results.

Despite these limitations, the current study had several strengths. First, it was a large population study based on national data, guaranteeing the statistical reliability of results. As KNHANES represents the non-institutionalized general population, selection bias was minimized. Second, this investigation provided strong evidence of a close relationship between diet-induced acid load and CVD risk by adjusting for other confounding factors and conducting stratification analyses. To the best of our knowledge, this is the first large-scale study to estimate the CVD risk associated with diet-induced acid load. We assessed diet-induced acid load using variable methods: PRAL and DAL scores. The correlation between PRAL and DAL scores was high (Pearson correlation coefficient = 0.99, P < 0.001) in our study, which was consistent with previous studies [10]. Third, individual CVD risks were estimated using various equations, and comparable results were derived regardless of these different models. Fourth, although the current study was cross-sectional, we limited the study population by excluding those with prior CVD history to prevent reverse causality.

In the light of dietary pattern and CVD risk, our study results demonstrate that individuals with higher diet-induced acid load are more vulnerable to CVD risks. Despite advances in CVD diagnostic and treatment modalities, prevention is the most high-leverage action, and therefore, modifiable risk factors are important. Prospective, well-designed, longitudinal studies with sufficient laboratory and cardiovascular-imaging resources are warranted to elucidate the complex relationship between diet-induced acid load and CVD risk. If the causality between diet-induced acid load and CVD risk, as well as the cutoff value of diet-induced acid load are clearly investigated, then meaningful, comprehensive, and practical dietary recommendations would be provided to the public.