Plasma boron concentrations in the general population: a cross-sectional analysis of cardio-metabolic and dietary correlates

The “PopGen controls” are a prospective population-based cohort, which was originally established as a reference (“control”) sample for genetic association analyses by the PopGen Biobank in Kiel, Germany. We recruited a total of 1317 residents of the city of Kiel and its surrounding communities in Northern Germany from a sample of individuals randomly selected from local population registries as well as from blood donors from the University Hospital in Kiel between 2005 and 2007. The age range of participants at the time of recruitment was between 19 and 77 years. As expected, participants randomly recruited from population registries were older and had higher prevalence of cardio-metabolic disease, but similar body mass index (BMI) as compared to individuals recruited from blood donors (n = 570) [10, 11]. The present cross-sectional analyses were based on data from the first follow-up examination, which was conducted between 2010 and 2012 and included physical examinations, laboratory analyses, and standardized questionnaires on demographics, lifestyle factors (including diet, physical activity, education, and smoking status), and medical history [12, 13]. For all of the 929 individuals participating in this follow-up examination, plasma boron concentrations were available. After exclusion of 30 individuals with missing data on e.g. dietary intake, physical activity, education, or smoking status, a total of 899 individuals were eligible for analyses on the associations of plasma boron concentrations with patterns of food intake and on the metabolic correlates of boron (Fig. 1).

The study was conducted in accordance with the Declaration of Helsinki. The study protocol was approved by the Ethics Committee of the Medical Faculty of Kiel University (Project identification code A 156/03; P2N reference numbers 2020-013, 2020-041). All participants provided written informed consent prior to their inclusion in the study.

Trained personnel performed anthropometric measurements in a standardized manner [14, 15]. The systolic and diastolic blood pressure values were calculated as the arithmetic mean of two blood pressure measurements, conducted using a sphygmomanometer with the participants sitting for at least 5 min and with at least 2 min between the measurements. Hypertension was defined as systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg or use of antihypertensive medication [12, 14, 16].

Information about participant characteristics such as age, sex, years of education, smoking habits, and medical history were obtained using self-administered questionnaires [14, 15]. Participants additionally reported their habitual time (hours) spent in physical activity (walking, cycling, sports, gardening, housework, home repair, stair climbing) per week during the last 12 months [14]. From these questions, total physical activity was derived by summing up the metabolic equivalent of task (MET) values assigned to each corresponding activity [17].

Whole-blood and lithium heparin plasma samples were drawn for biochemical assessment from the participants in a sitting position. Concentrations of HbA1c, plasma glucose, CRP, plasma lipids (triglycerides, high-density lipoprotein [HDL] cholesterol, low-density lipoprotein [LDL] cholesterol), were analyzed on the same day in unfrozen blood samples with HbA1c being measured by high performance liquid chromatography (HPLC) with photometric detection (D-10 HPLC analyzer; Bio-Rad Laboratories, Munich, Germany), glucose using an enzymatic UV test (hexokinase method, Hitachi Modular; Roche Diagnostics, Mannheim, Germany), CRP by immunoturbidimetry, and lipids using enzymatic colorimetry (Hitachi Modular; Roche Diagnostics) [14, 15, 18]. Participants having CRP concentrations below the detection limit of 0.9 mg/L (n = 345) were assigned half of the detection limit as their CRP value. Prevalent diabetes was defined as HbA1c ≥ 6.5% (48 mmol/mol) or fasting plasma glucose ≥ 126 mg/dL or use of anti-diabetic medication [12]. Plasma creatinine concentrations were measured enzymatically on Roche/Hitachi cobas c systems (Roche Diagnostics). Estimated glomerular filtration rate (eGFR) was derived using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [19].

Blood samples for biomarker analyses in plasma were drawn into plasma separator tubes (Sarstedt AG, Nümbrecht, Germany), centrifuged, aliquoted, and stored at − 80 °C until analyses [12]. Plasma boron, phosphate, magnesium, and calcium concentrations were measured via an inductively coupled plasma-mass spectrometry (ICPMS) ICAP Q instrument (Thermo Fisher Scientific, Waltham, MA, USA) at Synalb (Jena, Germany). Measurements were conducted in accordance with DIN EN ISO 17294-2: 2017-01. Samples were decomposed with a mixture of nitric acid and hydrogen peroxide (4:1) using microwave pressure digestion. For calibration, a multi-element standard was used. To minimize spectroscopic interferences, a collision/reaction cell was used for detection. Rhodium (2 µg/L) was added as the internal standard.

Dietary assessment within the PopGen cohort has been described in detail before [12‐15, 18]. In brief, habitual dietary intake over the past 12 months was assessed using a self-administered 112-item food frequency questionnaire, which has been validated within the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam study [20]. On the basis of the intake frequencies and quantities of foods consumed, food group intake, nutrient content, and energy intake were calculated using the German Food Code and Nutrient Database (version II.3) [20, 21].

Categorical variables were reported as absolute numbers and percentages, normally distributed continuous variables as mean and standard deviation (SD), and continuous skewed variables as median and interquartile range (IQR). Study participants’ characteristics were compared across tertiles of plasma boron concentration using chi-square test for categorical variables, general linear models for continuous normally distributed variables, and Kruskal–Wallis test for continuous skewed variables (Table 1).

For the assessment of seasonal variations in plasma boron concentrations, categories corresponding the four seasons and based on the month of study examination (winter, January–March; spring, April–June; summer, July–September; fall, October–December) were created [12]. Plasma boron concentrations were presented as median and IQR with the P value for comparison across the seven seasons calculated using Kruskall–Wallis test (Table 2).

For presentation of dietary intake according to tertiles of plasma boron concentrations, the 112 food items of the FFQ were assigned to 42 predefined food group categories according to similarity of nutrient characteristics and/or culinary usage for simplification of data interpretation and minimization of within-person variations in individual food intakes [12, 14] in a first step. Food group intakes were then ln-transformed prior to analysis due to their skewed distribution and presented as re-transformed least square means and 95% confidence interval (CI) adjusted for age and sex. To allow ln-transformation of food groups including non-consumers (i.e. beer, poultry, beef), a positive constant of 0.1 was added to the respective food groups. Age- and sex-adjusted differences in food group intakes by tertiles of plasma boron concentrations were calculated using general linear models (Supplemental Table 1).

Of the 899 participants eligible for analyses (Fig. 1), 57% were men and the mean age of this elderly sample was 61 years (Table 1). Overall, the average mean BMI of the participants was in the overweight range and hypertension was present in almost two-thirds of the study sample. The median plasma boron concentration of the total study sample was within the range of previously reported mean plasma boron concentrations of about 17–80 µg/L [6]. Compared to the bottom tertile, individuals in higher tertiles of plasma boron concentrations were older, had a higher daily alcohol intake, higher plasma glucose concentrations, and a lower eFGR, with only individuals within the lowest tertile of plasma boron concentrations having on average a normal kidney function [28]. However, individuals in the highest compared to lower tertiles of plasma boron concentrations had a lower BMI, a higher level of physical activity, and were less likely current smokers (Table 1). Circulating boron varied according to the four seasons and were highest during summer and lowest during winter (Table 2).

Individuals in the highest compared to the lowest tertile of plasma boron concentrations had higher age- and sex-adjusted daily mean intakes of the following plant-based food groups: leafy and root vegetables, fruits, nuts and seeds, cereals other than bread/pasta/rice, vegetable oils, and soya products and lower intakes of the following animal-based food groups: pork, poultry, processed meat, and other fats. Regarding mean beverage intake, individuals in the highest compared to the lowest tertile of plasma boron concentrations had lower intakes of soft drinks and higher intakes of non-alcoholic beverages, tea, and wine (Supplemental Table 1).

Additionally, RRR analysis identified a dietary pattern explaining about 30% of the variation in plasma boron concentrations. This RRR-derived dietary pattern was characterized by high intakes of fruits, nuts and seeds, tea and wine and low intakes of bread, poultry, processed meat, margarine, chocolate and sweets, soft drinks, sauces, and snacks (Table 3). As expected, the RRR pattern was strongly and independently associated with plasma boron concentrations: a 1-unit increase in the RRR pattern was associated with a 2.7% (95% CI: 2.1; 3.3), P < 0.001 (model 3) increase in circulating plasma boron concentrations.

Increases in PDI and hPDI were associated with increases in circulating boron concentrations, while an increase in uPDI was related to a reduction in circulating boron (Table 4). RCS analysis did not reveal any statistically significant nonlinear associations between plant-based diet indices and plasma boron concentrations (all P > 0.05). The test for differences in slopes revealed differences between hPDI and uPDI in their association with plasma boron concentrations (P = 0.002).

In a forward selection process, the variables displayed in Table 5 were identified as statistically significant correlates of plasma boron concentrations explaining 31% of the inter-individual variation in plasma boron concentrations. BMI, CRP, and plasma lipid concentrations displayed inverse associations, whereas age and plasma phosphate concentrations were positively associated with boron levels. Circulating boron concentrations displayed a nonlinear, U-shaped association with eGFR (Table 5). Below a threshold of about 97 mL/min/1.73 m², eGFR was inversely associated with plasma boron concentrations, while above this threshold, higher eGFR was associated with higher plasma boron concentrations (Fig. 2). Also, plasma boron concentrations were higher in individuals with high education level and medium education level vs. low education level, in summer vs. winter, and in individuals taking antihypertensive medication vs. those taking no antihypertensive medication. Anthropometric and cardio-metabolic traits which were not identified as significant correlates of plasma boron concentrations were gender, WHR, HbA1c, systolic blood pressure, plasma magnesium and calcium concentrations, physical activity level, smoking status, and glucose- and lipid-lowering medication.

Food group intake across tertiles of circulating boron within this study largely mirrored the known categorization of foods into good (i.e. fibre- and protein-rich plant foods, wine) and poor (i.e. animal foods) sources of boron. Only the higher intakes of dairy products, cheese, and fish in the highest compared to the lowest tertile of plasma boron concentrations contrasts prior evidence [3, 4, 6], but appear plausible in the light of the plant-derived boron intake of cows, which is transferred into the cows’ milk dose-dependently with the cows’ boron ingestion [29, 30]. Higher plasma boron concentrations in our study were not only associated with a closer adherence to the PDI and hPDI, but additionally with a more favorable lifestyle in general as participants in a higher compared to a lower tertile of circulating boron were less likely to be current smokers and more likely to be physically active. Also, individuals with high or medium vs. low education level showed higher circulating boron, which might be also mediated by healthier food choices.

We identified a multivariable-adjusted model that explained 31% of the inter-individual variation in plasma boron concentrations. This model included age, BMI, CRP, eGFR, plasma lipid and phosphate concentrations, education level, season, and antihypertensive medication.

Our observation of an inverse association of BMI with circulating boron is comparable with results from experimental studies showing a reduction in body weight and obesity with boron administration in animal models [31, 32]. Proposed mechanisms are the overexpression of thermogenic proteins in murine adipose and skeletal muscle tissue, resulting in accelerated lipolysis and body weight loss [33]. Studies in cell models suggested that boron might suppress adipogenesis [34, 35]. In agreement with this experimental evidence and with our observations, a small intervention study in 13 healthy women reported that an increase in dietary boron intake for one month by means of boron-rich foods decreased body weight and BMI [36].

Boron has been also suggested to reduce systemic inflammation, as indicated by associations of lower concentrations of inflammatory biomarkers such as CRP with higher boron concentrations in our sample and as reviewed elsewhere [4, 7]. Similar observations have been reported in a further human study, where CRP concentrations decreased after supplementation with calcium fructoborate for 30 days in 78 healthy individuals [37]. The dosages of boron used in these intervention studies ranged from 3.0 to 12.0 mg boron/day [4, 7, 37], which is above the 1.5–3.0 mg/day of dietary boron intake reported to be achievable by consuming a diverse, plant-food-rich diet, but clearly below the tolerable upper intake level of 20 mg boron/day for adults [4, 7].

We observed a relatively pronounced inverse association of eGFR with plasma boron concentrations for individuals with eGFR <97 mL/min/1.73 m². The most likely explanation is that boron is mainly excreted through the kidney [3]. Indeed, the above-mentioned human study by Kuru et al. reported a 6.0-fold increase in urinary boron concentrations with rising dietary boron intake [36]. As comprehensively summarized by Pahl et al., chronic exposure to nonlethal doses of boric acid in humans does not seem to be associated with renal abnormalities [38]. Therefore, the present results might rather reflect a reduced renal clearance, thus higher concentrations of boron within the body in individuals with slightly reduced kidney function, than pointing towards a direct detrimental effect of boron on renal function.

In animal studies, boron supplementation reduced circulating concentrations of triglycerides and cholesterol and, thus, improved plasma lipid profiles [3, 7]. These positive effects of boron appear to be also present in humans as we observed inverse associations of circulating boron with circulating triglycerides and LDL-cholesterol concentrations. However, circulating boron was additionally inversely related to HDL-cholesterol concentrations in our sample.

So far, the cardio-protective benefits of (red) wine consumption have been commonly attributed to its polyphenol and/or alcohol content [39]. However, taking together our associative findings of higher intakes of wine in the highest (least square mean [95% CI]: 55.1 g/day [47.3; 64.2]) compared to the lowest (least square mean [95% CI]: 20.8 g/day [17.1; 24.3]) tertile of plasma boron concentrations (see also Supplemental Table 1) and the cardio-metabolic correlates of plasma boron concentrations, these cardio-protective benefits might be (at least partially) mediated by boron.

Our observed seasonal variations of plasma boron concentrations might reflect its previously described interactions with vitamin D metabolism [1, 4]. It has been hypothesized that boron supplementation increases the biological half-live and bioavailability of vitamin D [4]. A bi-directional regulation of boron and vitamin D or a lower intake of plant-based, thus boron-rich foods, during winter compared to summer might possibly explain our observations. Also, experimental and human evidence suggested that boron supplementation improves absorption and reduces excretion of phosphate, calcium, and magnesium [1, 3, 4], which is reflected by direct associations of circulating phosphate with circulating boron in our study. Of note, the hypothesis of boron influencing renal phosphate excretion is additionally supported by a partial tubular reabsorption of boron as observed in humans and animals [38]. Through boron’s interplay with phosphate and vitamin D, it is likely to exert essential functions on bone metabolism and prevention of osteoporosis [1, 3, 4, 6, 7].

Strengths of our study include the moderate-sized population-based sample and the comprehensive assessment of cardio-metabolic traits. Furthermore, we assessed dietary intake using a validated FFQ and derived both, an explorative dietary pattern explaining about 30% of the variation in plasma boron concentrations and three predefined plant-based diet indices using continuous scores. Also, we applied RCS regression to model non-/linear relationships. The following limitations merit consideration. First, boron has several postulated biological functions [1, 3, 4, 6‐8], of which we were able to investigate many (e.g. interaction with minerals, biomarkers of systemic inflammation, and kidney function), but we had no sufficient data on e.g. arthritis, osteoporosis, and activity of antioxidant enzymes available. Second, dietary boron intake could not be directly calculated, as robust databases for boron amounts in foods are lacking [3]. We thus used the diet indices/pattern as surrogate parameters of dietary boron intake. However, plasma boron concentrations and FFQ-derived proxies for boron intake refer to different time intervals (i.e. the FFQ covered food intake during the last year [20], half-life of circulating boron after dietary intake is about 21 h [7]). Third, a boron-rich diet seems to be characterized by high intakes of foods presumed to be healthy [36]. Although organic boron-containing compounds might itself contribute to a reduced risk of cardiovascular diseases [40, 41], residual confounding by other ingredients cannot be ruled out [36]. Importantly, boron is also contained in pesticides and inorganic fertilizers [42] and might be ingested from these sources via conventionally grown fruits and vegetables. In our study, we cannot distinguish between organically versus conventionally grown fruits and vegetables to investigate the impact of pesticide-derived boron intake. Thus, further studies are needed regarding the influence of the farming method on the boron content in fruits and vegetables.