Dietary index for gut microbiota and risk of hypertension: the mediating role of HOMA-IR in a five-year prospective cohort study in Iranian adults

Abstract

Background

Diet–microbiota interactions may influence blood pressure via metabolic pathways, but prospective evidence on microbiota-supportive dietary patterns and incident hypertension, particularly in Middle Eastern populations, is limited. We examined the association between a Dietary Index for Gut Microbiota (DI-GM) and incident hypertension, and assessed whether insulin resistance, measured by homeostatic model assessment for insulin resistance (HOMA-IR), mediates this relationship.

Methods

This prospective cohort included 5,185 Iranian adults free of hypertension at baseline. Usual dietary intake was assessed using a validated 125-item food-frequency questionnaire, and a 12-component DI-GM was computed. Incident hypertension over five years was defined according to JNC-7 criteria. Multivariable logistic regression estimated odds ratios (ORs) and 95% confidence intervals (CIs) for hypertension across DI-GM quartiles and per one-unit increment. Mediation by HOMA-IR was evaluated using bootstrapped mediation analysis with 10,000 resamples.

Results

During follow-up, 2,150 participants experienced incident hypertension (41.5%). In fully adjusted models, participants in the highest DI-GM quartile had 21% lower odds of hypertension than those in the lowest quartile (OR 0.79, 95% CI 0.66–0.95), and each one-unit higher DI-GM was associated with 4% lower odds (OR 0.96, 95% CI 0.93–0.99). Results were robust in multiple sensitivity analyses. Approximately 34.5% of the association between DI-GM and hypertension was mediated by HOMA-IR.

Conclusions

A microbiota-supportive dietary pattern is prospectively associated with a lower risk of incident hypertension, and about one-third of this relationship appears to operate through reduced insulin resistance. These findings support the promotion of gut-friendly dietary patterns in strategies for hypertension prevention.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12986-026-01121-4.

Introduction

Hypertension is a leading global health problem, contributing to cardiovascular disorders such as myocardial infarction, stroke, heart failure, and chronic kidney disease, which together account for a substantial proportion of global mortality [1–3]. Despite the availability of various medications, the global prevalence of hypertension is expected to rise by 2040 for men in 180 countries, women in 186 countries, and both sexes in 187 countries, particularly in many low- and middle-income countries [4]. Because pharmacological control remains suboptimal in a substantial proportion of patients, prevention through modification of lifestyle factors, particularly diet, is a central public health priority.

Accumulating evidence indicates that overall dietary patterns, rather than isolated nutrients, are strongly related to blood pressure [5]. Diets rich in fruits, vegetables, whole grains, legumes, and low-fat dairy products, and lower in sodium, refined grains, and red and processed meats, have been consistently associated with lower blood pressure and reduced risk of incident hypertension [6, 7]. These findings suggest that the quality and composition of habitual diet are key determinants of long-term blood pressure trajectories.

More recently, the gut microbiota has been proposed as an important biological pathway linking diet to blood pressure regulation [8]. Microbial composition and function appear to influence insulin sensitivity, systemic inflammation, and vascular tone, all of which are implicated in the risk of incident hypertension [8]. Among these pathways, we focused on insulin resistance as a key mediator, given its established role in hypertension and the ability to quantify it reliably using homeostatic model assessment for insulin resistance (HOMA-IR) in large cohorts. In this context, the dietary index for gut microbiota (DI-GM) has been established as a reliable tool to measure how dietary patterns affect gut microbiome diversity [9]. However, most available data on microbiota-supportive diets and hypertension are cross-sectional and derived largely from Western populations [10–12].

To the best of our knowledge, no prospective study has examined the association between DI-GM and incident hypertension, nor the potential mediating role of insulin resistance, in Middle Eastern populations. This evidence gap is particularly relevant in Iran, where a rapid nutritional transition from traditional, largely plant-based diets toward more Westernized, energy-dense and low-fiber diets has coincided with a rising burden of cardiometabolic disorders [13, 14]. Therefore, we aimed to investigate the association between DI-GM and risk of incident hypertension in Iranian adults free of hypertension at baseline, and to assess whether insulin resistance, estimated by HOMA-IR, mediates this relationship. We hypothesized that higher DI-GM scores would be associated with a lower risk of incident hypertension, partially mediated by lower insulin resistance.

Method

Study population

This prospective analysis leveraged data from the MMRT cohort, a Tehran-based study designed to identify determinants of non-communicable diseases [15].

Using a convenience sampling frame, adults aged 20–70 years who presented to a comprehensive health center were invited to enroll. During the initial recruitment window (January 2017–December 2018), 8,760 individuals were approached; 7,836 consented and entered the cohort (participation rate = 89.4%). Each participant provided written informed consent and received a unique identification code to protect confidentiality. Baseline information was obtained via standardized, face-to-face interviewer-administered questionnaires, clinical examinations, and laboratory sampling, capturing demographics, lifestyle behaviors, dietary intake, medical history, and anthropometric indices. A planned five-year re-evaluation was conducted during January 2022–December 2023. All baseline participants were invited to repeat the assessments; 7,600 completed follow-ups, yielding a participation rate exceeding 95% with no missing data for the variables analyzed. Figure 1 illustrates the participant flow. For the present analysis, we excluded individuals with hypertension at baseline (n = 1701), cardiovascular diseases (n = 395), cancer (n = 23), rheumatoid arthritis (n = 214), and chronic kidney disease (n = 79). These exclusions were applied because the selected conditions are strongly associated with blood pressure, dietary habits, or gut microbiota composition, and could confound the association between DI-GM and incident hypertension. Participants with other conditions, including endocrine or metabolic disorders, were retained in the primary analysis but addressed in sensitivity analyses. We also removed under- or over-reporters of total energy intake, defined as values beyond ± 3 standard deviations from the cohort mean (n = 2). The final analytical sample comprised 5185 participants. All procedures adhered to the Declaration of Helsinki and were approved by the Ethics Committee of Shahid Beheshti University of Medical Sciences (Approval code: IR.SBMU.NNFTRI.REC.1404.003).

Fig. 1.

Flow chart of the study

Dietary assessment

Usual dietary intake over the previous year was assessed with a 125-item, Willett-type Food Frequency Questionnaire (FFQ) adapted to Iranian dietary patterns. The instrument has been validated in Iranian adults, demonstrating robust correlations between intakes estimated by a similar FFQ and multiple 24-hour dietary recalls (24). For each item, participants reported consumption frequency over the prior year using daily, weekly, monthly, or yearly response options. Standard portion sizes were taken from U.S. Department of Agriculture (USDA) references when available; otherwise, household measures were recorded and converted to grams and servings for consistency. Given the limitations of the Iranian Food Composition Table (FCT), nutrient estimates were primarily sourced from the USDA FCT; when traditional or national foods were not listed in the USDA database, values were taken from the Iranian FCT.

The DI-GM was derived from 12 components available in the dataset, adapted from the original 14-item index proposed by Kase et al. [16]. Eight components were classified as supportive of microbial diversity—fermented dairy products, chickpeas, soybeans, whole grains, dietary fiber, cranberries, broccoli, and coffee—whereas four were designated as detrimental: refined grains, red meat, processed meats, and a high-fat diet (≥ 40% of total energy from fat). For each beneficial component, a value of 1 was assigned if intake exceeded the sex-specific median (higher intake considered favorable for gut microbiota). For each detrimental component, a value of 1 was assigned if intake was below the corresponding median (lower intake considered favorable). Component scores were summed to yield a DI-GM ranging from 0 to 12, with higher scores indicating dietary patterns more conducive to gut microbiota diversity. The original DI-GM additionally includes avocado and green tea as beneficial components; however, these items were not explicitly captured by the MMRT FFQ and therefore could not be quantified directly in this cohort. Accordingly, our primary analyses used the 12-component DI-GM, whereas a 14-component version incorporating avocado and green tea was generated for sensitivity analyses through an imputation procedure described in the Statistical analysis section.

Non-dietary assessment

Sociodemographic characteristics, health behaviors, and medical history were obtained using a standardized, interviewer-administered questionnaire. Socioeconomic status was operationalized via a wealth score index (WSI) constructed with Multiple Correspondence Analysis (MCA) from asset indicators—ownership of household appliances, digital devices, and transportation means—along with television type and history of international travel [17]. Physical activity was quantified with the Modifiable Activity Questionnaire (MAQ) to estimate weekly metabolic equivalent task (MET) hours; the MAQ has been validated for use in Iranian populations [18].

Anthropometric and biochemical measurements

Measurements were obtained by trained personnel using standardized protocols. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m²). Waist circumference (WC) was measured in the standing position at the midpoint between the lower margin of the last palpable rib and the top of the iliac crest, at the end of a normal expiration, using a non-stretchable tape held horizontally and without compressing the skin.

Blood pressure was measured with an automated sphygmomanometer after participants had been seated at rest for 5 min; the mean of two readings was used for analysis.

Fasting venous blood was collected following an 8–12 h overnight fast. Serum fasting blood sugar (FBS), triglycerides (TG), total cholesterol (TC), and high-density lipoprotein cholesterol (HDL-C) were quantified by enzymatic colorimetric methods on an automated analyzer using the AutoAnalyzer system (Selectra E, Vitalab, Holliston, the Netherlands) with Pars Azmoon kits. Low-density lipoprotein cholesterol (LDL-C) was estimated using the Friedewald equation [19]. Insulin concentrations were measured by electrochemiluminescence immunoassay, and insulin resistance was calculated as HOMA-IR = [FBS (mg/dl) × fasting insulin (µU/mL)]/405. Liver enzymes, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), and gamma-glutamyl transferase (GGT), were determined using standard enzymatic assays. C-reactive protein (CRP) was measured using a high-sensitivity immunoturbidimetric assay on the AutoAnalyzer system (Selectra E, Vitalab, Holliston, the Netherlands).

Hepatic steatosis was evaluated by transient elastography (FibroScan™, Echosens, Paris, France). For participants with overweight, the XL probe was used. Hepatic fat content was quantified using the Controlled Attenuation Parameter (CAP), expressed in decibels per meter (dB/m).

Disease definitions and outcome

Non-alcoholic fatty liver disease (NAFLD) was defined as hepatic steatosis on transient elastography with a Controlled Attenuation Parameter (CAP) threshold of ≥ 238 dB/m [20]. Type 2 diabetes mellitus (T2DM) was defined in accordance with American Diabetes Association criteria as either a self-reported physician diagnosis or current treatment with glucose-lowering agents, or by meeting any of the following biochemical thresholds: fasting plasma glucose ≥ 126 mg/dl, glycated hemoglobin (HbA1c) ≥ 6.5%, or 2-hour postprandial plasma glucose (2hPP) ≥ 200 mg/dl [15]. Cardiovascular disease status was ascertained by participant self-report and, when available, corroborated through review of medical records.

Hypertension was defined in accordance with JNC-7 as SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg, based on the mean of two seated measurements obtained after 5 min of rest with an automated sphygmomanometer, or the use of antihypertensive medication [21]. Use of antihypertensive medications was self-reported by participants and subsequently confirmed by a study physician, based on review of prescription labels when available.

Statistical analysis

Baseline characteristics were summarized across quartiles of the DI-GM. Continuous variables are reported as mean ± SD and compared using linear regression; categorical variables are presented as counts (percentages) and compared using χ² tests. P values for linear trend across DI-GM quartiles were obtained by modeling the quartile variable as an ordinal term.

Associations between DI-GM and the incident hypertension were estimated using multivariable logistic regression to obtain odds ratios (ORs) and 95% confidence intervals (CIs). DI-GM was modeled both categorically (quartiles) and continuously (per one-unit increment). For tests of trend across quartiles, the median value of each category was entered as a single continuous term. Candidate covariates were identified a priori from prior literature [16, 22, 23]. Variables demonstrating a univariate association with the outcome at p < 0.20 were subsequently included in multivariable models. The final adjusted model therefore accounted for established and data-supported confounders, including sociodemographic characteristics, lifestyle factors, baseline clinical measures, and total energy intake.

Potential non-linear associations between DI-GM and incidence of hypertension were examined using restricted cubic spline regression in the multivariable-adjusted model, with four knots placed at the 5th, 35th, 65th, and 95th percentiles of the DI-GM distribution. Overall association and non-linearity were evaluated with Wald tests of the spline terms.

Prespecified subgroup analyses evaluated effect modification by age, sex, BMI, smoking status, physical activity, and wealth score index. Within each stratum, models used the same adjustment set as the primary analysis, omitting the continuous form of the stratifying variable to avoid redundancy. Statistical interaction was tested by adding a cross-product term between DI-GM and the stratifying variable; heterogeneity was assessed with Wald tests for interaction.

Sensitivity analyses probed robustness by (i) excluding participants with T2DM at baseline; (ii) excluding those with NAFLD at baseline; (iii) excluding participants who experienced > 10% weight gain during follow-up; and (iv) adding further adjustments, one at a time, for potential intermediates or correlates, including HOMA-IR, triglyceride-to-HDL-cholesterol ratio, CRP, dietary sodium-to-potassium ratio, saturated fat intake, and dietary sugar. To evaluate the influence of the two DI-GM components that were not captured by our FFQ (avocado and green tea), we additionally constructed an extended 14-component DI-GM and repeated all regression models using this index. Because the questionnaire did not include specific items for avocado or green tea, their intakes were unobserved. In the sensitivity analysis, we therefore generated binary component scores for avocado and green tea using a simple probabilistic imputation approach: for each participant, scores were drawn from Bernoulli distributions with quartile-specific probabilities conditional on the 12-component DI-GM, with higher probabilities assigned to higher DI-GM quartiles. The probabilities were specified a priori on the basis of published data on avocado and green tea consumption, which indicate low overall use of these foods in adult populations and particularly limited consumption of green tea in Iran, and were chosen such that the imputed prevalence of a “high” intake classification was approximately 14% for avocado and 6% for green tea. Each imputed component contributed one point to the DI-GM, yielding a score ranging from 0 to 14, and this 14-component index was analysed with the same modelling strategy and covariate adjustment as used for the primary 12-component DI-GM [24–26].

Mediation analysis was conducted to determine whether insulin resistance mediated the association between DI-GM and incident hypertension. Analyses were implemented in R (version 4.5.0) using the mediation package, with mediator and outcome models specified using the same covariates as in the primary multivariable analysis. Nonparametric bootstrapping with 10,000 resamples was used to estimate indirect (mediated), direct, and total effects, along with 95% CIs. The proportion mediated by HOMA-IR was subsequently calculated to quantify mediation via insulin resistance.

All analyses were two-sided with α = 0.05 and were conducted on complete cases (no missing data for the variables analyzed). Statistical analyses were performed using IBM SPSS Statistics version 26.0 (IBM Corp., Armonk, NY, USA) and R version 4.5.0 (R Foundation for Statistical Computing, Vienna, Austria).

Results

In this prospective cohort study, 5,185 adults with normal blood pressure at baseline participated. Participants’ mean age ± SD was 42.3 ± 0.14 years, and their BMI was 27.8 ± 0.06 kg/m². Women comprised 50.5% of the cohort. The baseline means SBP and DBP were 118.1 ± 0.10 mmHg and 72.1 ± 0.12 mmHg, respectively. Table 1 represents the baseline characteristics across quartiles of the DI-GM. With increasing DI-GM, participants were younger and exhibited higher WSI and greater physical activity level (P-value for trend < 0.001). Adiposity indices were lower with higher DI-GM levels (P-value for trend < 0.001). SBP and DBP were modestly lower at higher DI-GM levels (P-value for trend < 0.001). Indices of glucose homeostasis (FBS, insulin, and HOMA-IR) were progressively lower, the lipid profile was more favorable (TG, TC, and LDL-C decreased while HDL-C increased), and liver enzymes (ALT, AST, GGT), hepatic fat burden (CAP), and systemic inflammation (CRP) were reduced (P-value for trend < 0.001). The baseline prevalence of T2DM and NAFLD was correspondingly lower in the higher DI-GM quartiles, whereas the distributions of sex and smoking status varied minimally across quartiles (P-value < 0.001).

Table 1.

Baseline characteristics of participants across quartiles of dietary index for gut microbiota.a

Characteristics	Total population (N = 5185)	Quartile 1 (N = 1546)	Quartile 2 (N = 873)	Quartile 3 (N = 1778)	Quartile 4 (N = 988)	P-valueb
Age	42.3 ± 9.9	44.3 ± 9.9	42.7 ± 10.0	41.3 ± 9.8	40.6 ± 9.6	< 0.001
Men, n (%)	2568 (49.5)	774 (50.1)	409 (46.8)	899 (50.6)	486 (49.2)	0.32
Smoker, n (%)	577 (11.1)	180 (11.6)	94 (10.8)	204 (11.5)	99 (10.0)	0.69
Education, n (%)
Low	943 (18.2)	257 (16.6)	170 (19.5)	347 (19.5)	169 (17.1)	0.14
Middle	3065 (59.1)	931 (60.2)	490 (56.1)	1050 (59.1)	594 (60.1)
High	1177 (22.7)	358 (23.2)	213 (24.4)	381 (21.4)	225 (22.8)
Family history of hypertension, n (%)	2661 (51.4)	796 (51.5)	452 (51.9)	900 (50.6)	513 (52.0)	0.88
WSI	70.5 ± 14.0	67.6 ± 13.5	70.4 ± 14.4	71.9 ± 14.0	72.7 ± 13.6	< 0.001
Physical activity (MET/H/week)	23.7 ± 2.7	22.8 ± 2.2	23.4 ± 2.6	24.2 ± 2.9	24.5 ± 2.8	< 0.001
Weight (kg)	86.8 ± 13.9	92.8 ± 13.3	88.6 ± 14.1	83.9 ± 13.0	81.0 ± 12.1	< 0.001
Height (cm)	166.5 ± 6.1	166.6 ± 6.2	166.2 ± 6.1	166.6 ± 6.0	166.2 ± 6.0	0.29
Waist circumference (cm)	87.1 ± 15.6	93.6 ± 17.7	89.1 ± 16.3	83.9 ± 14.1	81.1 ± 11.2	< 0.001
Body mass index (Kg/m2)	27.8 ± 4.5	29.3 ± 4.2	28.3 ± 4.5	26.9 ± 4.3	26.3 ± 4.3	< 0.001
SBP (mmHg)	118.1 ± 7.3	118.3 ± 7.4	117.8 ± 7.2	117.8 ± 7.2	117.0 ± 7.1	< 0.001
DBP (mmHg)	72.1 ± 8.5	73.2 ± 8.5	72.3 ± 8.5	71.5 ± 8.6	71.1 ± 8.4	< 0.001
Fasting blood sugar (mg/dl)	98.7 ± 14.8	101.3 ± 14.5	100.4 ± 14.5	97.2 ± 14.7	96.1 ± 14.8	< 0.001
Fasting Insulin (µU/mL)	8.9 ± 3.0	10.7 ± 3.0	9.3 ± 3.0	8.0 ± 2.6	7.2 ± 2.0	< 0.001
HOMA-IR	1.4 ± 0.6	1.8 ± 0.5	1.5 ± 0.6	1.2 ± 0.5	1.0 ± 0.3	< 0.001
Triglyceride (mg/dl)	139.7 ± 17.2	144.6 ± 18.2	142.7 ± 18.5	136.6 ± 15.8	134.7 ± 13.9	< 0.001
Total cholesterol (mg/dl)	172.3 ± 27.5	179.2 ± 30.1	173.1 ± 28.3	169.3 ± 25.4	166.3 ± 23.5	< 0.001
LDL-C (mg/dl)	95.6 ± 15.3	98.8 ± 15.3	97.0 ± 15.2	93.9 ± 15.3	92.4 ± 14.5	< 0.001
HDL-C (mg/dl)	43.9 ± 6.9	41.1 ± 5.9	42.9 ± 6.5	45.4 ± 7.0	46.5 ± 7.1	< 0.001
ALT (U/L)	27.2 ± 4.7	29.0 ± 4.6	27.6 ± 4.6	26.4 ± 4.5	25.5 ± 4.3	< 0.001
AST (U/L)	24.8 ± 5.0	26.8 ± 4.8	25.2 ± 4.7	24.0 ± 4.8	22.8 ± 4.5	< 0.001
GGT (U/L)	26.7 ± 9.6	33.4 ± 7.5	28.7 ± 9.1	23.5 ± 8.2	20.0 ± 6.2	< 0.001
CAP score (dB/m)	188 ± 46	220 ± 38	197 ± 46	172 ± 41	156 ± 30	< 0.001
CRP (mg/dl)	2.4 ± 2.1	3.8 ± 2.3	2.8 ± 2.3	1.7 ± 1.7	1.2 ± 0.9	< 0.001
Type 2 diabetes Mellitus, n (%)	457 (8.8)	222 (14.4)	108 (12.4)	90 (5.1)	37 (3.7)	< 0.001
Non-alcoholic fatty liver disease	982 (18.9)	586 (37.9)	224 (25.7)	159 (8.9)	13 (1.3)	< 0.001

Quartile 1 (Q1: DI-GM score 1–5), Quartile 2 (Q2: DI-GM score =6), Quartile 3 (Q3: DI-GM score 7–8), and Quartile 4 (Q4: DI-GM score 9–12).^a Data are presented as mean ± standard deviation for continuous variables and number (percent) for non-continuous variables. ^b P-value was obtained from linear regression analysis for continuous variable, and χ2 test for categorical variables, respectively, according to the category of dietary index for gut microbiota score.Abbreviations: WSI: Wealth Score Index; SBP, systolic blood pressure; DBP, diastolic blood pressure; HOMA-IR, Homeostatic Model Assessment for Insulin Resistance; LDL, Low Density Lipoprotein cholesterol; HDL, High Density Lipoprotein cholesterol; ALT, Alanine Amino transferase; AST, Aspartate Amino transferase; GGT, Gamma-glutamyl transferase, CRP: C-reactive protein

Table 2 represents the dietary intake of participants across quartiles of the DI-GM. With increasing DI-GM, total energy intake was lower and the macronutrient distribution shifted toward a smaller proportion of energy from fat, with modestly higher contributions from carbohydrate and protein (P-value for trend < 0.001). Food group patterns indicated higher quality diets, with greater consumption of whole grains, legumes, fruits, vegetables, dairy, nuts, and white meat, and lower consumption of refined grains, red and processed meats, and added sugars (P-value for trend < 0.001). Micronutrient density improved across quartiles, including higher intakes of potassium, calcium, magnesium, iron, and total dietary fiber, lower sodium intake, and a reduced sodium to potassium ratio (P-value for trend < 0.001). Fat quality was also more favorable at higher DI-GM, as the percentages of energy from saturated and trans-fatty acids were progressively lower (P-value for trend < 0.001).

Table 2.

Dietary intake of participants across quartiles of dietary index for gut microbiota.a

	Quartile 1	Quartile 2	Quartile 3	Quartile 4	P-valueb
Total energy intake (Kcal)	2331 ± 375	2235 ± 346	2173 ± 322	2124 ± 298	< 0.001
Carbohydrate intake (% of energy)	60.4 ± 3.8	60.4 ± 3.6	60.8 ± 3.6	61.1 ± 3.5	< 0.001
Fat intake (% of energy)	28.9 ± 4.2	28.6 ± 4.0	27.7 ± 3.9	27.3 ± 3.7	< 0.001
Protein intake (% of energy)	12.4 ± 1.2	12.9 ± 1.4	13.6 ± 1.3	13.8 ± 1.1	< 0.001
Whole grain (gram/day)	14.3 ± 8.6	17.6 ± 9.6	21.9 ± 10.	25.6 ± 9.6	< 0.001
Refined grain (gram/day)	416.5 ± 77.5	385.8 ± 72.9	366.8 ± 66.3	347.7 ± 58.3	< 0.001
Legumes (gram/day)	47.4 ± 14.7	58.0 ± 18.6	68.0 ± 18.1	75.7 ± 16.3	< 0.001
Red meat (gram/day)	20.5 ± 8.7	17.5 ± 8.7	15.7 ± 8.4	13.4 ± 7.9	< 0.001
White meat (gram/day)	37.2 ± 13.8	40.5 ± 14.9	44.7 ± 15.3	45.1 ± 14.8	< 0.001
Dairy (gram/day)	159.7 ± 31.8	160.5 ± 31.1	162.5 ± 33.3	165.9 ± 34.6	< 0.001
Fruits (gram/day)	317.2 ± 57.3	337.7 ± 65.8	361.7 ± 61.2	378.1 ± 56.7	< 0.001
Vegetables (gram/day)	440.7 ± 87.6	482.0 ± 102.8	514.9 ± 97.6	553.2 ± 94.1	< 0.001
Nuts (gram/day)	4.8 ± 3.9	4.9 ± 3.9	5.3 ± 4.2	5.3 ± 4.1	< 0.001
Sugar (gram/day)	53.3 ± 24.3	46.0 ± 23.4	40.3 ± 22.1	36.8 ± 20.5	< 0.001
Sodium (mg/day)	3430 ± 901	3154 ± 884	2917 ± 794	2768 ± 716	< 0.001
Potassium (mg/day)	3443 ± 450	3525 ± 478	3622 ± 456	3720 ± 431	< 0.001
Sodium to potassium ratio	1.00 ± 0.26	0.90 ± 0.26	0.81 ± 0.22	0.74 ± 0.19	< 0.001
Calcium (mg/day)	836 ± 105	843 ± 110	861 ± 109	873 ± 104	< 0.001
Iron (mg/day)	14.6 ± 2.0	14.6 ± 2.1	15.0 ± 2.0	15.1 ± 1.9	< 0.001
Magnesium (mg/day)	305 ± 38	308 ± 40	316 ± 37	321 ± 35	< 0.001
Total fiber (gram/day)	23.5 ± 2.8	25.1 ± 3.4	26.7 ± 3.2	28.0 ± 2.7	< 0.001
Trans-fatty acid (% of energy)	0.30 ± 0.13	0.24 ± 0.12	0.20 ± 0.11	0.18 ± 0.10	< 0.001
Saturated fatty acid (% of energy)	10.5 ± 1.9	10.6 ± 1.8	10.2 ± 1.7	10.1 ± 1.6	< 0.001

Quartile 1 (Q1: DI-GM score 1–5), Quartile 2 (Q2: DI-GM score =6), Quartile 3 (Q3: DI-GM score 7–8), and Quartile 4 (Q4: DI-GM score 9–12).^a Data are presented as mean ± standard deviation.^b P-value was obtained from linear regression analysis according to the category of dietary index for gut microbiota score

In this cohort, over a five-year follow-up, 2,150 participants experienced incident hypertension, corresponding to an incidence of 41.5%. Table 3 represents the association between DI-GM and the odds of incident hypertension. In the crude model, participants in the highest DI-GM quartile had 37% lower odds of incident hypertension than those in the lowest quartile (OR = 0.63, 95% CI 0.53–0.74; P-trend < 0.001). With additional adjustment for age, sex, sociodemographic, lifestyle, and baseline clinical factors, the inverse association persisted, with 23% lower odds in the highest versus lowest quartile (OR = 0.77, 95% CI 0.64–0.92; p for trend < 0.01). In the fully adjusted model, participants in the highest quartile had 21% lower odds of incident hypertension compared to those in the lowest quartile (OR = 0.79, 95% CI 0.66–0.95; p for trend = 0.01). The inverse association between DI-GM and the odds of incident hypertension was also evident when DI-GM was modeled continuously; in the fully adjusted analysis, each one-unit increase in DI-GM was associated with a 4% reduction in the odds of incident hypertension (OR = 0.96; 95% CI, 0.93–0.99).

Table 3.

Odds ratios (95% confidence intervals) for hypertension incidence according to the proportion of dietary index for gut microbiota score

	Q1	Q2	Q3	Q4	P-trend	Continues (Per one unit increment	P-value
Median Score	4.0	6.0	7.0	9.0
Cases/populations	738/1546	353/873	697/1778	362/988		2150/5185
Crude model	1.00 (Ref)	0.74 (0.62–0.87)	0.70 (0.61–0.81)	0.63 (0.53–0.74)	< 0.001	0.91 (0.89–0.94)	< 0.001
Model 1a	1.00 (Ref)	0.74 (0.63–0.82)	0.70 (0.61–0.81)	0.63 (0.53–0.75)	< 0.001	0.92 (0.89–0.94)	< 0.001
Model 2b	1.00 (Ref)	0.78 (0.66–0.93)	0.82 (0.71–0.95)	0.77 (0.64–0.92)	< 0.01	0.95 (0.92–0.98)	< 0.001
Model 3c	1.00 (Ref)	0.79 (0.67–0.95)	0.84 (0.72–0.97)	0.79 (0.66–0.95)	0.01	0.96 (0.93–0.99)	0.01

Significant p-values are highlighted in bold.

Quartile 1 (Q1: DI-GM score 1–5), Quartile 2 (Q2: DI-GM score =6), Quartile 3 (Q3: DI-GM score 7–8), and Quartile 4 (Q4: DI-GM score 9–12).

^a Model 1 adjusted for age and sex.

^b Model 2 additionally adjusted for Model 1 and physical activity, and wealth score index, education level, smoking, body mass index, baseline blood pressure, type 2 diabetes mellitus, non-alcoholic fatty liver disease, and family history of hypertension.

^c Model 3 additionally adjusted for Model 2 and dietary energy intake

Figure 2 demonstrates multivariable-adjusted dose–response association between adherence to DI-GM and the odds of incident hypertension. The spline curve shows progressively lower odds with higher DI-GM, and the overall association was statistically significant (P-value < 0.01), with no evidence of non-linearity (P-value = 0.59).

Fig. 2.

Multivariable adjusted spline curve for the association between dietary index for gut microbiota score and the odds of hypertension

Table 4 represents the subgroup analyses of the association between DI-GM and the odds of incident hypertension. In comparisons of the highest versus lowest DI-GM quartiles, the inverse association remained significant among older participants (OR = 0.72; 95% CI, 0.53–0.99), men (OR = 0.68; 95% CI, 0.53–0.8), individuals with overweight or obesity (BMI ≥ 25 kg/m²; OR = 0.77; 95% CI, 0.62–0.96), and nonsmokers (OR = 0.81; 95% CI, 0.67–0.99). Tests for interaction by age, sex, BMI, and smoking were not significant (all p for interaction > 0.05). With respect to socioeconomic position, the association was stronger among participants with a higher WSI, where the highest versus lowest DI-GM quartiles were associated with lower odds of incident hypertension (OR = 0.67; 95% CI, 0.51–0.88). No significant trend was observed in the lower wealth stratum, and the interaction by wealth category was statistically significant (p for interaction = 0.02). Across physical activity strata, trends were generally inverse but did not reach statistical significance, and there was no evidence of interaction by activity level (p for interaction = 0.40).

Table 4.

Subgroup analysis of the association between dietary index for gut microbiota score and the odds of hypertension.a

	Q1	Q2	Q3	Q4	P-trend	P for interaction
Age
< 45 years old	1 (ref)	0.85 (0.67–1.06)	0.92 (0.76–1.12)	0.84 (0.67–1.05)	0.20	0.17
> 45years old	1 (ref)	0.73 (0.56–0.95)	0.71 (0.55–0.91)	0.72 (0.53–0.99)	0.01
Sex
Men	1 (ref)	0.78 (0.61–1.00)	0.73 (0.59–0.91)	0.68 (0.53–0.88)	0.01	0.58
Women	1 (ref)	0.80 (0.63–1.02)	0.96 (0.77–1.19)	0.91 (0.70–1.18)	0.65
Body mass index
< 25	1 (ref)	0.73 (0.49–1.08)	0.91 (0.66–1.26)	0.82 (0.58–1.16)	0.52	0.98
> 25	1 (ref)	0.82 (0.67–0.99)	0.81 (0.68–0.97)	0.77 (0.62–0.96)	0.01
Smoking
No	1 (ref)	0.81 (0.68–0.98)	0.84 (0.72–0.99)	0.81 (0.67–0.99)	0.04	0.82
Yes	1 (ref)	0.60 (0.35–1.02)	0.78 (0.49–1.24)	0.57 (0.32–1.01)	0.09
Wealth score index
Low	1 (ref)	0.74 (0.59–0.93)	0.86 (0.70–1.06)	0.92 (0.71–1.19)	0.41	0.02
High	1 (ref)	0.86 (0.66–1.11)	0.63 (0.50–1.01)	0.67 (0.51–0.88)	< 0.01
Physical activity
Low	1 (ref)	0.81 (0.66–1.01)	0.85 (0.70–1.04)	0.78 (0.60–1.02)	0.05	0.40
High	1 (ref)	0.77 (0.57–1.02)	0.83 (0.65–1.06)	0.79 (0.61–1.04)	0.17

^aMultivariable logistic regression was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs). Models were adjusted for age, sex, physical activity, wealth score index, education level, smoking status, body mass index, baseline blood pressure, type 2 diabetes mellitus, non-alcoholic fatty liver disease, family history of hypertension, and total energy intake. For subgroup analyses, the stratifying variable was included as a main effect with a product (interaction) term, and the corresponding continuous covariate was omitted from the adjustment set for that subgroup to avoid overadjustment (e.g., age was not included when stratifying by age)

Quartile 1 (Q1: DI-GM score 1–5), Quartile 2 (Q2: DI-GM score = 6), Quartile 3 (Q3: DI-GM score 7–8), and Quartile 4 (Q4: DI-GM score 9–12)

Table 5 presents the results of sensitivity analyses evaluating the association between the DI-GM and the odds of incident hypertension. After excluding participants with > 10% weight gain from baseline, the OR for the highest versus lowest quartile was 0.71 (95% CI: 0.56–0.90), with a continuous OR of 0.93 (95% CI: 0.89–0.96). Excluding individuals with T2DM yielded an OR of 0.77 (95% CI: 0.64–0.94), and a continuous OR of 0.96 (95% CI: 0.92–0.99). After excluding participants with NAFLD, the corresponding ORs were 0.77 (95% CI: 0.63–0.94) and 0.95 (95% CI: 0.92–0.99), respectively. With additional adjustment for HOMA-IR, the association between the DI-GM and the incidence of hypertension attenuated and no longer statistically significant, with an OR of 0.87 (95% CI: 0.74–1.01) for the highest versus lowest quartile and 0.97 (95% CI: 0.95–1.00) in the continuous model. After adjusting for baseline CRP, the inverse association also weakened, showing an OR of 0.83 (95% CI: 0.69–1.00) for the highest quartile compared to the lowest and 0.97 (95% CI: 0.94–0.99) per unit increase. Further adjustments for dietary sodium-to-potassium ratio, saturated fatty acid, and dietary sugar intake yielded ORs ranging from 0.79 to 0.81 for the highest versus lowest quartile, with corresponding continuous estimates between 0.96 and 0.96, all within statistically significant confidence intervals. In an additional sensitivity analysis, we examined whether omission of avocado and green tea from the DI-GM influenced the findings by repeating the models with a 14-component DI-GM that incorporated imputed scores for these two foods (Supplementary Table 1). The results were essentially unchanged. In the fully adjusted model, participants in the highest quartile of the 14-component DI-GM had 24% lower odds of incident hypertension compared with those in the lowest quartile (OR = 0.76; 95% CI: 0.64–0.91; P for trend = 0.01). When the 14-component DI-GM was modeled continuously, each 1-unit increase in the score was associated with a 4% reduction in the odds of hypertension incidence (OR = 0.96; 95% CI: 0.93–0.98; P = 0.01).

Table 5.

Sensitivity analysis for incident hypertension according to the proportion of dietary index for gut microbiota score.a

	Q1	Q2	Q3	Q4	P-trend	Continues (Per one unit increment	P-value
Excluding participants with weight gain of > 10% from baseline	1.00 (Ref)	0.85 (0.68–1.05)	0.75 (0.62–0.91)	0.71 (0.56–0.90)	< 0.01	0.93 (0.89–0.96)	< 0.001
Excluding participants with T2DM at baseline	1.00 (Ref)	0.76 (0.63–0.92)	0.82 (0.70–0.97)	0.77 (0.64–0.94)	0.01	0.96 (0.92–0.99)	0.01
Excluding participants with NAFLD at baseline	1.00 (Ref)	0.77 (0.63–0.95)	0.81 (0.68–0.97)	0.77 (0.63–0.94)	0.02	0.95 (0.92–0.99)	< 0.001
Additional adjusted for baseline HOMA-IR	1.00 (Ref)	0.85 (0.74–0.98)	0.87 (0.76–0.98)	0.87 (0.74–1.01)	0.05	0.97 (0.95–1.00)	0.05
Additional adjusted for baseline TG to HDL ratio	1.00 (Ref)	0.76 (0.63–0.92)	0.83 (0.71–0.98)	0.78 (0.65–0.95)	0.02	0.96 (0.93–0.99)	0.01
Additional adjusted for baseline C-reactive protein	1.00 (Ref)	0.81 (0.68–0.97)	0.87 (0.75–1.02)	0.83 (0.69–1.00)	0.06	0.97 (0.94–0.99)	0.04
Additional adjusted for dietary sodium to potassium ratio	1.00 (Ref)	0.81 (0.68–0.96)	0.86 (0.74–1.00)	0.81 (0.68–0.98)	0.04	0.96 (0.93–0.99)	0.02
Additional adjusted for dietary saturated fatty acid	1.00 (Ref)	0.80 (0.67–0.95)	0.84 (0.72–0.98)	0.79 (0.66–0.95)	0.01	0.96 (0.93–0.98)	0.01
Additional adjusted for dietary sugar	1.00 (Ref)	0.83 (0.72–0.95)	0.82 (0.72–0.93)	0.80 (0.69–0.93)	< 0.01	0.96 (0.93–0.98)	0.01

Significant p-values are highlighted in bold.Quartile 1 (Q1: DI-GM score 1–5), Quartile 2 (Q2: DI-GM score =6), Quartile 3 (Q3: DI-GM score 7–8), and Quartile 4 (Q4: DI-GM score 9–12).^a Multivariable logistic regression was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs). Models were adjusted for age, sex, physical activity, wealth score index, education level, smoking status, body mass index, baseline blood pressure, type 2 diabetes mellitus, non-alcoholic fatty liver disease, family history of hypertension, and total energy intake

Figure 3 illustrates the results of the mediation analysis examining the extent to which the association between the DI-GM and hypertension is mediated by insulin resistance, as assessed by the HOMA-IR. The total effect of the DI-GM on hypertension was − 0.0101 (95% CI: −0.0183 to − 0.00029; P = 0.01). The indirect effect via HOMA-IR was statistically significant (− 0.0036; 95% CI: −0.0061 to − 0.0012; P < 0.01), corresponding to a proportion mediated of 34.5% (P = 0.01). The direct effect of the DI-GM on hypertension, independent of HOMA-IR, was not statistically significant (− 0.0068; 95% CI: −0.0148 to 0.0010; P = 0.09).

Fig. 3.

The relationship of DI-GM and the odds of hypertension mediated by Homeostatic Model Assessment for Insulin Resistance

Discussion

In this five-year, community-based cohort of 5,185 Iranian adults with normotension at baseline, higher adherence to a microbiota-supportive dietary pattern, quantified by the DI-GM, was prospectively associated with a lower risk of incident hypertension. The observed five-year incidence of hypertension in our cohort (41.5%) is higher than that reported in many Western cohorts of middle-aged adults [27–29], but consistent with findings from other Middle Eastern and Asian populations undergoing epidemiological transition [30–33]. Several factors may contribute to this elevated incidence: higher baseline cardiometabolic risk among participants (mean BMI 27.8 kg/m², substantial prevalence of insulin resistance, T2DM, and NAFLD), rapid nutritional and lifestyle transitions in Iran, and inclusion of antihypertensive medication use in the outcome definition, which may capture cases not identified by blood pressure measurements alone. While absolute rates differ across populations, the high event rate provides strong statistical power and enhances the clinical relevance of our findings for individuals at elevated cardiometabolic risk. In multivariable models, individuals in the highest DI-GM quartile had markedly reduced odds of incident hypertension compared with those in the lowest quartile, and the dose–response analysis indicated an approximately linear inverse association, with each one-unit increase in DI-GM corresponding to about a 4% reduction in hypertension odds. Mediation analysis further showed that insulin resistance, assessed by HOMA-IR, accounted for roughly one-third of this relationship, indicating that improvements in insulin sensitivity represent a key pathway through which a gut-supportive dietary pattern may contribute to blood pressure control.

Several large cross-sectional analyses based on US NHANES data have recently evaluated DI-GM in relation to hypertension. In a weighted analysis of more than 41,000 adults, each one-unit higher DI-GM was associated with approximately 4% lower odds of hypertension (OR 0.96, 95% CI 0.94–0.98), driven primarily by the beneficial, rather than the unfavorable, component score [10]. Another NHANES study including 20,804 participants reported that individuals in the highest DI-GM category had 13% lower odds of prevalent hypertension compared with those in the lowest category (OR 0.87, 95% CI 0.76–0.99), with a broadly linear inverse trend across DI-GM levels [12]. More recently, Li et al. showed that higher DI-GM was similarly associated with lower hypertension risk (OR per one-unit increase 0.92, 95% CI 0.90–0.94), and that systemic inflammatory markers—including white blood cell count, neutrophils, and the systemic immune-inflammation index—mediated a modest proportion (approximately 4–9%) of this association [11]. Consistent with these prior reports, higher DI-GM scores in our cohort were also associated with a modest but potentially meaningful reduction in the risk of incident hypertension over follow-up. The direction and magnitude of the associations were broadly comparable to those observed in NHANES-based analyses, reinforcing the plausibility and potential generalizability of a protective role for microbiota-supportive dietary patterns in hypertension prevention. In addition, by using a prospective design in a Middle Eastern population free of hypertension at baseline and by quantifying mediation through insulin resistance—rather than inflammatory markers—our study extends this literature from cross-sectional prevalence to incident disease and provides complementary mechanistic evidence focused on glucose–insulin homeostasis.

In prespecified subgroup analyses, the inverse association between DI-GM and incident hypertension was broadly consistent across categories of age, sex, adiposity, smoking status, physical activity, and WSI, with no statistically significant interactions for age, sex, BMI, smoking, or physical activity. Nonetheless, the magnitude of the association appeared greater among men, older adults, individuals with overweight or obesity, and non-smokers. These patterns are compatible with known sex differences in blood pressure regulation, including the vasoprotective effects of endogenous estrogens in premenopausal women [34], and with evidence that ageing is accompanied by a decline in gut microbiota diversity and stability, potentially increasing responsiveness to microbiota-supportive dietary interventions [35]. The more pronounced inverse association observed among participants with overweight or obesity is biologically plausible, because these individuals typically have a higher burden of insulin resistance and may therefore derive greater benefit from improvements in insulin sensitivity [36]. Our mediation analysis showed that insulin resistance, as indexed by HOMA-IR, accounted for about one-third of the association between DI-GM and incident hypertension, supporting insulin resistance as a key pathway. Additionally, smokers exhibited higher risk of incident hypertension, consistent with their typically less healthy dietary habits [35]. It should be noted that blood pressure is not controlled by a simple prescription, but a complex combination of lifestyle modification, adherence to treatment, and a deep understanding of the body’s mechanisms for its control is required [37]. Notably, a statistically significant interaction emerged only for socioeconomic status. Higher DI-GM scores were clearly protective among participants in the higher WSI category, whereas no discernible dose–response gradient was evident among those with lower WSI. This pattern likely reflects structural and economic constraints that shape dietary choices. Microbiota-supportive foods—such as fresh fruits and vegetables, whole grains, nuts, and fermented dairy products—are often relatively more expensive, less accessible, and less consistently available in lower-income neighborhoods, making sustained adherence to such patterns difficult for individuals with limited financial resources. In addition, lower WSI is frequently accompanied by reduced nutrition literacy and competing priorities related to employment and household responsibilities, which can hinder engagement with and implementation of dietary recommendations [38]. From a public health perspective, these findings suggest that the potential benefits of microbiota-supportive dietary patterns for hypertension prevention may be inequitably distributed, and that policies aimed at improving affordability, availability, and education around healthy foods are likely to be necessary to enable lower-SES groups to realize similar cardiometabolic gains.

The beneficial effects of a gut-supportive diet may be partly explained by improved insulin sensitivity, potentially mediated by microbiota-derived metabolites such as short-chain fatty acids (SCFAs), contributing to lower blood pressure [39–41].

Our findings revealed significant stability after conducting a thorough set of sensitivity analyses and adjusting for various potential confounding factors. This statistical stability was even more pronounced when participants with metabolic diseases, such as type 2 diabetes and nonalcoholic fatty liver disease, were excluded from the analysis. Additionally, including important dietary variables such as the sodium-to potassium ratio, saturated fat, and added sugars in the statistical models did not significantly affect the direction or strength of this association. Although the current study accounted for a wide range of demographic, lifestyle, and clinical variables, we cannot completely rule out the possibility of residual confounding factors. These may include chronic psychosocial stress [42], and the use of certain medications [43], all of which can independently influence gut microbiota composition and blood pressure regulation. Furthermore, early-life factors, such as the mode of delivery and infant nutrition, which help shape the early development of the gut microbiome, could also serve as potential confounders [44]. Despite these considerations, the stability and consistency of the findings across various analytical scenarios, combined with the application of advanced statistical methods, enhance the validity of the observed association and suggest that it is not merely a chance finding.

Strengths and limitations

This investigation offers several methodological strengths. First, the prospective design with five years of follow-up and a large sample of 5,185 Iranian adults initially free of hypertension enables robust temporal inference regarding the diet-hypertension relationship. Second, the comprehensive assessment protocol incorporated standardized clinical measurements, biochemical markers, and validated lifestyle questionnaires, thereby minimizing information bias. Third, the application of mediation analysis with bootstrapped confidence intervals allowed us to quantify the extent to which insulin resistance accounts for the observed association, providing mechanistic insight into potential biological pathways. Fourth, the consistency of findings across multiple sensitivity analyses and diverse subgroups strengthens confidence in the robustness of our results. Finally, this study addresses a critical evidence gap in Middle Eastern populations, where dietary patterns and cardiometabolic risk profiles differ substantially from Western cohorts that have dominated prior research.

Several limitations warrant consideration when interpreting these findings. The observational nature of the study precludes definitive causal inference, and residual confounding by unmeasured or imperfectly measured factors—such as medication adherence, stress levels, or sleep quality—cannot be entirely excluded despite extensive covariate adjustment. Dietary intake was assessed using a semi-quantitative food frequency questionnaire, which, although validated in this population, relies on participant recall and may introduce measurement error that could attenuate observed associations toward the null. Critically, we employed a dietary index as a proxy for gut microbiome composition rather than direct microbiological or metabolomic profiling; while DI-GM has demonstrated validity in predicting microbial diversity in prior studies, the absence of direct microbiome data limits our ability to confirm the hypothesized biological mechanism. Antihypertensive medication use was self-reported and physician-confirmed when prescription labels were available, though misclassification remains possible. Certain DI-GM components (avocado, green tea) were infrequently consumed and required imputation, although sensitivity analyses suggested minimal bias. Importantly, DI-GM was originally validated in NHANES data from the United States rather than Iranian populations, necessitating future validation studies in diverse cultural contexts. Finally, recruitment of a convenience sample from urban Iran may limit generalizability to rural or non-Iranian populations with differing dietary patterns and genetic backgrounds.

Notwithstanding these limitations, DI-GM offers a practical tool for assessing dietary quality in relation to gut microbiome health, incorporating elements of established cardioprotective patterns. Our findings suggest that modest shifts toward microbiota-supportive diets may reduce hypertension risk in at-risk populations. Future studies integrating direct microbiome sequencing, metabolomics, and intervention designs are needed to clarify causal mechanisms.

Conclusion

In conclusion, in this prospective cohort of Iranian adults free of hypertension at baseline, higher adherence to a microbiota-supportive dietary pattern, as reflected by a higher DI-GM score, was associated with a lower five-year risk of incident hypertension. Part of this association was mediated by insulin resistance, suggesting that improvements in glucose–insulin homeostasis may be an important pathway linking gut-supportive dietary patterns to blood pressure control. These findings support the potential role of microbiota-supportive diets in the primary prevention of hypertension and indicate that promoting such dietary patterns could help reduce the burden of hypertension at the population level. Further prospective studies that integrate dietary assessment with microbiome and metabolomic profiling, as well as randomized dietary intervention trials, are needed to confirm these results and clarify the underlying mechanisms.

Author contributions

Overall, RH and JMR, supervised the project and approved the final version of the manuscript to be submitted. AN designed the research; AN analyzed and interpreted the data; AN, MM, LS, and EE drafted the initial manuscript. GJ, and NZ, critically revised the manuscript. All authors approved the final version of the manuscript submitted for publication.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

The datasets analyzed in the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The National Nutrition and Food Technology Research Institute (NNFTRI) ethics committee approved the study protocol (Ethics code: IR.SBMU.NNFTRI.REC.1404.003). All participants provided written informed consent and were informed about the study. All procedures performed in studies involving human participants adhered to the ethical standards of the institutional and/or national research committee and to the 1964 Helsinki Declaration and its later amendments, or to comparable ethical standards.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.