Dietary Polyphenols and Inflammatory Markers in Rural Adult Women in Puno, Peru

Abstract

Background

Dietary polyphenols are bioactive compounds credited with health protection and reducing inflammation. Evidence on their effects is limited in low- and middle-income countries, especially among rural populations.

Objectives

To evaluate the association between dietary polyphenols intake and markers of inflammation and endothelial adhesion over 12 mo of follow-up among rural adult women participating in the Cardiopulmonary Outcomes and Household Air Pollution randomized controlled trial.

Methods

We used data at baseline and 12-mo visit in 100 healthy women aged 25–64 y living in rural communities in Puno, Peru. We assessed dietary intake using 24-h dietary recalls and measured inflammatory cytokines [interleukin (IL)-10, IL-6, IL-1β, C-reactive protein, tumor necrosis factor-α] and endothelial adhesion (vascular cell adhesion protein-1, intercellular adhesion molecule-1) biomarkers in dried blood spots using immunoassay analyses. We quantified polyphenol intake using the Phenol-Explorer database. We used multivariable linear regression to assess the association between changes (12-mo of follow-up minus baseline) in dietary polyphenols intake, and inflammatory and endothelial adhesion markers at 12-mo.

Results

At baseline, mean (±standard deviation) participant age was 49 ± 10 y, mean total polyphenol intake was 335 ± 193 mg/d, and mean phenolic acids intake was 195 ± 80 mg/d. Lower IL-1β concentrations at 12-mo were associated with higher energy-adjusted phenolic acid intake [adjusted mean difference −0.35 pg/mL per mg/d; 95% confidence interval (CI): −0.63, −0.07]. In contrast, higher anti-inflammatory cytokines IL-10 concentrations at 12 mo were associated with higher energy-adjusted intake of stilbenes (0.42 pg/mL per mg/d; 95% CI: 0.18, 0.66) and other polyphenols (0.28 pg/mL per mg/d; 95% CI: 0.10, 0.46).

Conclusions

Higher polyphenol intake, particularly phenolic acids and stilbenes, was associated with an improved anti-inflammatory profile in rural women.

Introduction

Dietary polyphenols are secondary plant metabolites that are recognized for reducing inflammation and preventing oxidative stress, which may in turn reduce the risk of chronic noncommunicable diseases [[1], [2], [3]]. Polyphenols reduce inflammation by suppressing proinflammatory cytokines, including tumor necrosis factor alpha (TNF-⍺), interleukin (IL)-6, and C-reactive protein (CRP), and endothelial adhesion molecules like vascular cell adhesion protein-1 (VCAM-1) and intercellular adhesion molecule-1 (ICAM-1) [4]. Polyphenols also have antioxidant properties, neutralizing reactive oxygen species (ROS) and mitigating oxidative damage [5]. Specific classes of dietary polyphenols, such as flavonoids, phenolic acids, and stilbenes, have demonstrated efficacy attenuating key inflammatory signaling pathways [6].

Polyphenols are abundant in plant-based foods like fruits, vegetables, tea, coffee, spices, nuts, seeds, legumes, and whole grains [[7], [8], [9]]. Dietary patterns that include these foods, such as the Mediterranean diet, are well documented for their anti-inflammatory and antioxidant effects [7,10]. However, evidence on the beneficial health effects of polyphenols is based on studies in high-income countries (HICs), predominantly among European cohorts [7,[10], [11], [12], [13]]. Few studies have reported polyphenol intake in low- and middle-income countries (LMICs) [14,15]. We sought to evaluate the association between dietary polyphenol intake and inflammatory markers in adult women living in rural high-altitude areas of Peru.

Methods

Study design and participants

This study was a longitudinal analysis of dietary polyphenols and inflammatory and endothelial biomarkers collected as part of the Cardiopulmonary Outcomes and Household Air Pollution trial (NCT02994680), a randomized controlled trial designed to test a multicomponent liquefied petroleum gas (LPG) stove intervention for cooking as a strategy to mitigate household air pollution (HAP) exposure and improve cardiopulmonary health outcomes in 180 women aged 25–64 y from rural communities in Puno, Peru who used biomass fuels on a daily basis [16,17]. These rural communities were located along the shores of Lake Titicaca, at an elevation of 3825 m above sea level. Participants were randomly assigned in a 1:1 ratio to the intervention arm, in which they received an LPG stove and free fuel delivery, or the control arm, in which they continued usual cooking practices with biomass fuels. Details from the data collection are described elsewhere [16,17].

A random subset of 100 participants were assigned to complete 24-h dietary recalls on three nonconsecutive days at enrollment (baseline) and at the 12-mo visit (Figure 1). The three nonconsecutive days included two weekdays and one weekend day as previously described [18].

FIGURE 1.

CONSORT flowchart of study participant selection.

All participants provided written informed consent according to a protocol approved by the Ethics Committees of A.B. PRISMA and Universidad Peruana Cayetano Heredia (CE2402.16) in Lima, Peru, and the Institutional Review Board of the Johns Hopkins Bloomberg School of Public Health (7128) in Baltimore, United States [16].

Estimation of nutrient and dietary polyphenol intake

We used information from the three nonconsecutive 24-h recalls to estimate dietary polyphenol intake. Briefly, participants were asked by trained nutritionists in face-to-face interviews to provide accurate information of food items taken the previous day and details about portion size and cooking methods. Yield and raw-cooked conversion factors were used to estimate the net weight of each food item recorded in each 24-h recall per participant per day. The details of the dietary assessment are described elsewhere [18]. Energy and nutrient intake were estimated linking the food items with the Peruvian food composition table [19].

We estimated the total polyphenol content in foods and beverages using the Phenol-Explorer 3.6 database (www.phenolexplorer.eu/) and individual polyphenol content in 5 major classes: 1) flavonoids, 2) phenolic acids, 3) stilbenes, 4) lignans, and 5) other polyphenols, a miscellaneous class consisting of minor polyphenols [8,9]. We used data of phenolic compounds quantified using reversed phase liquid chromatography, except for proanthocyanidins for which normal phase liquid chromatography was used. In the case of lignans and phenolic acids in certain foods, such as root, cereal products, and olives, the quantification of these compounds was estimated by chromatography after hydrolysis because this is essential for the release of phenolic compounds [7,11]. For foods not included in the Phenol-Explorer database, such as quinoa (Chenopodium quinoa Willd) and pepino fruit (Solanum muricatum), estimates were derived from previous reports [20,21]. Polyphenol content in annatto (Bixa orellana L.), airampo (Tunilla corrugate), pacae (Inga feuilleei), purple corn (Zea mays L.), lemon verbena (Aloysia citrodora), coca leaf (Erythroxylum coca), spiked pepper (Piper aduncum), cat's claw (Uncaria tomentosa), maca (Lepidium meyenii), honey, carob (Ceratonia siliqua), and prickly pear (Opuntia spp.) was not estimated because of nonavailable information in the scientific literature of polyphenol profile using high-accuracy method like liquid chromatography couple to mass spectrometry.

We calculated polyphenol intake from each food by multiplying the polyphenol content of each food by the daily consumption of each food (g/d); we determined total polyphenol intake (mg/d) by adding up the intake of individual polyphenol compounds over the day for each study participant and calculated the mean over the three days. Total and polyphenol class intake were adjusted for daily energy intake (kcal/d) using the residual method to attenuate the correlation with energy intake [22]. We calculated changes in dietary polyphenols as the difference between the 12-mo and baseline visits.

Measurement of Inflammatory markers

We collected dried blood spot (DBS) samples at baseline and at the 12-mo visit. The procedure included alcohol finger swabbing, lancet blood collection, and application to Guthrie DBS cards. After labeling, the cards were air dried for ≥10 h, placed in labeled Ziploc bags with desiccant and humidity indicators, and stored at −20°C before analysis [23].

We assessed inflammatory cytokines (IL-10, IL-6, IL-1β, TNF-α, CRP) and endothelial adhesion proteins (ICAM-1 and VCAM-1) in duplicate from DBS samples as previously described [24]. Laboratory analyses were conducted at Emory University, where immunoassay techniques were performed using MSD Multiplex Proinflammatory 96-well plate (Meso Scale Discovery Rockville) for IL-10, IL-6, IL-1β, and TNF-α; and MSD Multiplex Vascular Injury 96-well plate (MSD, Meso Scale Discovery Rockville) for ICAM, VCAM, and CRP. Concentrations were determined using external standard calibration plots, assuming a DBS volume of 70 μL with 100% extraction efficiency from the card. Quality control measures included duplicate analyses of all samples, with relative standard deviations (RSDs) required to be within 10% of RSD for validity. Additionally, quality control samples were required to be within 20% of expected values, with reanalysis performed if this threshold was exceeded. The limit of detection (LOD) specified in the immunoassay kits was 13.8 pg/mL for CRP, 0.089 pg/mL for IL-10, 0.181 pg/mL for IL-6, 0.135 pg/mL for IL-1β, 0.081 for TNF-α pg/mL, 0.017 ng/mL for ICAM-1, and 0.091 ng/mL for VCAM-1. Concentrations below the LOD were replaced by LOD/$\sqrt{2}$.

Anthropometry and sociodemographic variables

Trained personnel assessed height and body weight. We measured height using a local stadiometer with an accuracy of 0.1 cm and weight using the SECA 220089 scale (SECA GMBH & Co) with an accuracy of 0.1 kg. We calculated BMI (in kg/m²) as body weight in kg divided by square height in m [17]. Sociodemographic and lifestyle information, including age, education level, main occupation, and income, was collected through a structured questionnaire. Wealth index was estimated following a standardized methodology and categorized into poorest, poor, and middle [18,25].

Biostatistical methods

Descriptive analyses were conducted for baseline characteristics and presented as mean and SD, or frequencies and percentages for continuous and categorical variables, respectively. Normal distribution of continuous variables was assessed using the Shapiro-Wilk test. T-tests were used to compare each inflammatory marker and polyphenol class intake between baseline and 12-mo visit.

Independent associations between changes in dietary polyphenol intake (12-mo - baseline), and inflammation and endothelial adhesion markers at 12-mo visit were examined using multivariable linear regression models with robust variances. Two models of increasing complexity were used: an unadjusted model, and a model adjusted for baseline values of each inflammation and endothelial marker, age, BMI, wealth index. An interaction term between polyphenol intake and arm was also included in the adjusted model to evaluate for effect modification.

Changes in dietary polyphenols values and concentration of inflammatory markers were normalized and scaled by multiples of 1-SD (one unit of SD) using Blom’s inverse normalization to produce more normally distributed values to help meet linear regression assumptions [26]. No missing imputation was applied in our analysis.

Statistical analyses were conducted using Stata 16.0 (StataCorp) and RStudio version 4.4.2 [27].

Results

Participant characteristics

We analyzed data of 100 participants. We summarized participant characteristics in Table 1. Mean (±SD) age at baseline was 49 ± 10 y, with a mean BMI of 27 ± 4. Most participants were categorized into the poorest national wealth quintile (55%), followed by the poor (37%) and the middle (8%) quintiles.

TABLE 1.

Participant characteristics at baseline

Variables	Total (n = 100)
Age (y), mean ± SD	49 ± 10
BMI (kg/m2), mean ± SD	27 ± 4
Wealth index quintiles, n (%)
Poorest	55 (55)
Poor	37 (37)
Middle	8 (8)
Harvest season, n (%)
Preharvest	33 (33)
Harvest	25 (25)
Postharvest	42 (42)

Values are expressed as mean ± SD or n (%).

Abbreviation: BMI, body mass index.

In Table 2, we summarize DBS concentrations of inflammatory and endothelial markers at baseline and at 12 mo. Concentrations of ICAM-1, IL-1β, and TNF-α were lower at 12-mo than at baseline, whereas CRP increased from a median of 0.2 pg/mL (0.1–0.5 pg/mL) at baseline to 1.6 pg/mL (0.8–4.9 pg/mL) (P = 0.001) at 12-mo visit.

TABLE 2.

Inflammatory markers at baseline and 12-mo visit

Inflammatory markers	Baseline					12 mo					P
	N	Mean ± SD	Median	25th	75th	N	Mean ± SD	Median	25th	75th
Endothelial adhesion proteins
VCAM-1 (ng/mL)	99	256.9 ± 188.9	176.1	122.8	364.5	99	203.9 ± 364.0	94.3	72.8	132.1	0.189
ICAM-1 (ng/mL)	99	704.8 ± 1333.2	255.2	150.8	541.5	99	269.9 ± 510.9	123.5	95.9	178.1	0.003
Inflammatory cytokines
CRP (pg/mL)	100	0.7 ± 1.6	0.2	0.1	0.5	100	10.1 ± 28.1	1.6	0.8	4.9	0.001
IL-10 (pg/mL)	100	0.2 ± 0.2	0.2	0.1	0.3	49	0.2 ± 0.1	0.3	0.1	0.3	0.222
IL-1 β (pg/mL)	100	5.9 ± 6.9	3	1.7	7.7	49	2.5 ± 3.3	1.1	0.4	3.2	0.001
IL-6 (pg/mL)	100	0.4 ± 0.2	0.4	0.4	0.4	49	0.6 ± 0.9	0.4	0.4	0.4	0.104
TNF-α (pg/mL)	100	0.9 ± 0.4	0.8	0.7	1.2	49	0.6 ± 0.4	0.4	0.3	0.7	<0.001

Statistical analyses were carried out using the t-test.

Abbreviations: CRP, C-reactive protein; ICAM, intercellular adhesion molecule; VCAM, vascular cell adhesion protein.

P values refer to differences between baseline and 12-mo of follow-up.

P < 0.05 indicates statistical significance.

We summarized dietary and polyphenol intake in Table 3. Regarding food groups, consumption of grains and roots, fruits, vegetables, pulses, coffee and cocoa products, dairy products, and eggs remained stable, while meat, poultry, and fish, increased from 52 ± 46 g/d to 62 ± 47 g/d (P = 0.048). Polyphenol intake was generally lower at 12-mo visit than at baseline, however, none of these differences were significant at the 0.05 level.

TABLE 3.

Dietary intake at baseline and 12-mo visit

Dietary intake	Baseline (N = 100)	12 mo (N = 100)	P
Energy and nutrient intake
Energy (kcal/d)	1927 ± 605	1954 ± 505	0.657
Fat (% energy)	19 ± 7	19 ± 8	0.981
Carbohydrate (% energy)	77 ± 8	76 ± 9	0.749
Protein (% energy)	12 ± 3	12 ± 3	0.209
Fiber (g/d)	28 ± 13	29 ± 12	0.550
Food groups
Grain and roots (g/d)	716 ± 279	713 ± 250	0.917
Fruits (g/d)	76 ± 68	72 ± 66	0.595
Vegetables (g/d)	85 ± 47	90 ± 54	0.372
Pulses (g/d)	35 ± 37	31 ± 35	0.448
Coffee and cocoa products (g/d)	0.1 ± 0.4	0.4 ± 1.9	0.238
Dairy (g/d)	65 ± 105	86 ± 171	0.282
Meat, poultry, and fish (g/d)	52 ± 46	62 ± 47	0.048
Eggs (g/d)	13 ± 18	14 ± 22	0.664
Alcoholic beverages (g/d)	6 ± 26	0 ± 0	0.033
Sweet beverages (g/d)	135 ± 193	79 ± 150	0.008
Herbs (g/d)	0.6 ± 3.5	0.4 ± 1.8	0.569
Infusions (g/d)	1.1 ± 1.8	1.3 ± 1.6	0.254
Spices (g/d)	0.1 ± 0.2	0.1 ± 0.5	0.254
Polyphenol intake
Total polyphenols (mg/d)	335 ± 193	303 ± 161	0.174
Flavonoids (mg/d)	130 ± 151	106 ± 122	0.174
Phenolic acids (mg/d)	195 ± 80	188 ± 75	0.515
Lignans (mg/d)	0.5 ± 0.4	0.4 ± 0.3	0.590
Stilbenes (mg/d)	0.01 ± 0.1	0 ± 0.0	0.462
Other polyphenols (mg/d)	9.4 ± 27.2	8.1 ± 15.3	0.692

Values are expressed as mean ± SD. Statistical analyses were carried out using the t-test.

P values refer to differences between baseline and 12-mo of follow-up.

P <0.05 indicates statistical significance.

In Supplemental Table 1, we display the mean intake of total polyphenol and polyphenol classes at baseline by sociodemographic variables, reporting statistically significant differences across wealth quantiles (P = 0.025). In Supplemental Table 2, we report the contribution of polyphenol classes to total polyphenol intake and their primary food sources at baseline. Phenolic acids contributed most of total polyphenol intake (58.2%), whereas flavonoids contributed 38.8%. Major contributors included black grape, apple, and wheat for flavonoids, and potato and peppermint for phenolic acids. No significant changes in polyphenol intake were observed between control and intervention arms (Supplemental Table 3).

Association between inflammatory/endothelial markers and polyphenol intake

In Figure 2 and Supplemental Table 4, we summarize the association between changes in energy-adjusted polyphenol intake (12-mo - baseline) and inflammatory/endothelial markers at 12-mo. In unadjusted analysis, a 1-SD increase in total polyphenol intake was associated with lower values of CRP at 12-mo visit [adjusted mean difference of −0.19 pg/mL per mg/d; 95% confidence interval (CI): −0.37, −0.002; P = 0.047] (Supplemental Table 4). However, in adjustment analysis we did not find an association between total polyphenol intake and changes in either inflammatory or endothelial markers (Figure 2 and Supplemental Table 4).

FIGURE 2.

Changes in energy-adjusted polyphenol intake from baseline to 12-mo visit, and inflammatory and endothelial markers at 12-mo follow-up. Statistical analyses were carried out using multiple linear regression, adjusting by baseline levels of each inflammatory marker; age, BMI, and wealth index. An interaction term between the change in polyphenol intake and arm was also included in the adjusted model. P values <0.05 were considered statistically significant. β, nonstandardized coefficient; CRP, C-reactive protein; ICAM, intercellular adhesion molecule; VCAM, vascular cell adhesion protein.

Because of the heterogeneity of energy-adjusted dietary polyphenol intake, the primary classes were analyzed separately, including total energy-adjusted flavonoids, phenolic acids, lignans, stilbenes, and other polyphenols, as shown in Figure 2. We found that significant associations were observed between changes in energy-adjusted phenolic acids, stilbenes, and other polyphenols from baseline to 12-mo visit, with specific inflammatory markers at 12-mo after full adjustment using multiple linear regression. Phenolic acid intake showed a significant negative association with IL-1β concentrations (adjusted mean difference of −0.35 pg/mL per mg/d; 95% CI: −0.63, −0.07; P = 0.017). Furthermore, a 1-SD increase in the intake of stilbenes and other polyphenols was associated with a higher level of IL-10 at 12-mo visit, with an adjusted mean difference of 0.42 per mg/d (95% CI: 0.18, 0.66; P = 0.002) for stilbene intake; and an adjusted mean difference of 0.28 pg/mL per mg/d (95% CI: 0.10, 0.46; P = 0.004) for other polyphenols intake. Predicted values were calculated from the fully adjusted regression model and are presented in Supplemental Figure 1. Although higher resveratrol intake was significantly associated with higher IL-10 in the regression model, the predicted line Supplemental Figure 1may appear slightly negative in some regions because of covariate adjustment, interaction terms, and the distribution of observed intake values.

We conducted stratified analyses to examine the associations between changes in energy-adjusted polyphenol intake, from baseline to 12-mo visit, and inflammatory and endothelial markers at 12-mo by arm (Supplemental Figure 2). In the control arm (traditional cooking fuel, higher HAP exposure), higher intake of flavonoids and phenolic acids was associated with lower IL-1β concentrations (flavonoids: −0.33 pg/mL per mg/d; 95% CI: −0.62, −0.03; P = 0.047; phenolic acids: adjusted mean difference −0.43 pg/mL per mg/d; 95% CI: −0.67, −0.18; P = 0.004), whereas higher intake of stilbenes and other polyphenols was associated with higher IL-10 (stilbenes: 0.42 pg/mL per mg/d; 95% CI: 0.10, 0.74; P = 0.023; others: 0.40 pg/mL per mg/d; 95% CI: 0.08, 0.72; P = 0.028). In the intervention arm (LPG cooking, lower HAP exposure ), higher intake of total polyphenols and phenolic acids was associated with lower CRP (total polyphenols: −0.24 pg/mL per mg/d; 95% CI: −0.46, −0.01; P = 0.045; phenolic acids: −0.39 pg/mL per mg/d; 95% CI: −0.63, −0.15; P = 0.004) and lower TNF-α (total polyphenols: −0.38 pg/mL per mg/d; 95% CI: −0.69, −0.06; P = 0.027). Higher intake of other polyphenols was associated with higher VCAM-1 (0.24 pg/mL per mg/d; 95% CI: 0.05, 0.43; P = 0.015).

Discussion

This study assessed the association between polyphenol intake and inflammatory biomarkers in rural women over a period of 12 mo. We found that a higher energy-adjusted intake of total polyphenol, phenolic acid, stilbene, and other polyphenols was associated with lower proinflammatory and higher anti-inflammatory cytokines. To our knowledge, this is the first study to examine changes in dietary polyphenol intake and their association with inflammatory markers in rural women from LMICs, where environmental exposure to HAP is relevant.

We observed that higher total polyphenol intake was associated with lower CRP concentrations, suggesting that it might help reduce systemic inflammation. Observational studies in European populations, including Spain and Italy, have linked adherence to the Mediterranean diet with lower levels of inflammatory markers such as CRP, an association partly attributed to its high content of polyphenols [7,10,12]. Additionally, the association between polyphenols and inflammation markers could be influenced by the health condition of the participants. For instance, in a cross-sectional study of 391 Iranian women with overweight or obesity attending health centers in Tehran, higher total polyphenol intake was inversely associated with CRP [28]. In contrast, studies by Taguchi et al. [29] and Sut et al. [30] found no differences between polyphenol intake and CRP concentrations in healthy Japanese and Polish men with chronic coronary syndrome. Overall, the anti-inflammatory properties of polyphenols, attributed to their antioxidant activity, play a role in their protective effect against inflammation. Mechanisms proposed include antioxidant activity and modification of signaling pathways that impact oxidative stress and immune cell function [5,31].

We had an opportunity to carefully characterize the association between polyphenol classes and inflammation. Phenolic acid intake, primarily from root and cereal consumption, was linked to lower IL-1β concentrations. Although previous research supports the anti-inflammatory properties of phenolic acids, our study did not find associations with the same cytokines. A study of Iranian women revealed that phenolic acid intake was associated with lower CRP concentrations [28]. Bumrungpert et al. [32] found that daily supplementation of 1000 mg ferulic acid for 6 wk led to a 33% decrease in high-sensitivity CRP and a 13% reduction in TNF-α in individuals with hyperlipidemia, potentially through nuclear factor kappa B (NF-kB) signaling inhibition. Another trial showed that overweight/obesity subjects had reduced TNF-α concentrations after 4 wk and an increase in IL-10 after 8 wk on a diet rich in phenolic acids [33]. The antioxidant properties of phenolic acids on inflammation have been described in several studies [13,[31], [32], [33], [34], [35], [36], [37], [38]]. Phenolic acids are known for their antioxidative capacity because of the hydroxyl groups that possess effective free radical scavenging abilities capable of donating an electron or hydrogen atom to reactive oxygen, nitrogen, and chlorine species [5,35,39,40].

In our study, higher IL-10 concentrations were linked to increased stilbene intake, from resveratrol-rich grapes and red wine. Clinical trials show that resveratrol reduces inflammatory markers, particularly CRP and TNF-α, in conditions like type 2 diabetes mellitus, metabolic syndrome, and nonalcoholic fatty liver disease (NAFLD) [[41], [42], [43], [44]]. Observational studies showed the association between stilbene intake with higher anti-inflammatory cytokines, such as IL-4 and IL-5 and CRP concentrations, although no association with IL-10 was found [13,28]. In animal models, resveratrol activates the silent information regulator T1 (SIRT1), which suppresses inflammation by inhibiting ROS production, stimulating adenosine monophosphate-activated protein kinase, inhibiting NF-kB activation, and decreasing glutathione S-transferase expression [45,46]. Resveratrol may also modulate IL-10 expression [47].

Increased intake of other polyphenols, primarily from grains (rich in alkylphenols) and olives (rich in tyrosols), was associated with higher IL-10 concentrations at 12 mo. Mosca et al. [48] demonstrated that supplementation with hydroxytyrosol and vitamin E increased IL-10 concentrations in children with NAFLD. Conversely, in patients with breast cancer, supplementation with hydroxytyrosol, omega-3 fatty acids, and curcumin maintained stable IL-10 concentrations, with only a decrease in CRP concentrations observed [49]. This association may be attributed to the action of tyrosol to upregulate SIRT1, thereby suppressing inflammation [50].

Flavonoids, although the second-largest contributor to total polyphenol intake, were not associated with inflammatory markers overall; however, stratified analysis showed that women in the control arm with higher flavonoid intake had lower IL-1β, suggesting context-dependent effects. Preclinical studies documented in 2 systematic reviews showed the anti-inflammatory role of flavonoids [51,52]. Flavonoids can modulate macrophage activity, inhibiting NF-kB, reducing proinflammatory cytokines and increasing antioxidant signaling pathway [52].

Beyond intrinsic biological mechanisms, contextual factors in rural LMIC populations shape the association between polyphenol intake and inflammation. Dietary pattern, lifestyle, and socioeconomic condition differ substantially from those in HICs and may alter the biological effect of phenolic compounds [53,54]. For example, in HIC, major sources of polyphenols include fruits, vegetables, tea, coffee, and red wine as part of Mediterranean diet [7,10,11]. By contrast, in rural LMICs, roots, grains, and locally available products are the predominant sources [14,15]. These differences in food sources and living condition likely contribute to population-specific responses. In our study, phenolic acids, particularly hydroxycinnamic acids, were the main contributors to the total polyphenol intake, which is consistent with findings of a study in Brazil [14,15,55]. In our study, polyphenol intake varied across wealth index quintiles; with women in the poorest quintile reporting the lowest intake (296.4 ± 157.6 mg/d). Similarly, nationally surveys in Brazil found that polyphenol intake differed by income levels [14]. Individuals in the lowest income group consumed 113.9 mg/d of phenolic acid and 47 mg/d of flavonoids, compared with 348.9 mg/d of phenolic acids and 1558.5 mg/d of flavonoids in the highest income group [14].

Equally critical is the role of environmental exposures such as HAP. In this cohort of rural women, we found that the associations between polyphenol intake and inflammatory marker differed by exposure to HAP. Among women continuing to use traditional cooking fuels (control arm), higher intakes of flavonoids and phenolic acids were associated with lower IL-1β, and higher intakes of stilbenes and other polyphenols were linked to higher IL-10. In contrast, among women who transitioned to LPG cooking (intervention arm) with reduced HAP exposure, polyphenol intake was associated with lower CRP and TNF-α. Prolonged HAP exposure promotes ROS production and chronic immune activation, contributing to long-term cardiopulmonary complications [56]. Consistent with this, in the same cohort, higher black carbon levels in kitchen areas were positively associated with TNF-α and inversely associated with IL-10 [24]. Together, these findings suggest that the biological activity of polyphenols is context dependent, shaped not only by diet and socioeconomic conditions but also by environmental exposures, an aspect poorly explored up to date and critical for advancing our understanding of how dietary bioactive compounds influence inflammation in LMICs.

The main strength of our study is the longitudinal design, which facilitated the assessment of polyphenol intake and inflammatory biomarkers over time. The application of multiple linear regression models allowed for control of potential confounders, increasing the robustness of our results. Additionally, the use of DBSs provided a rapid, simple, and cost-effective technique for evaluating inflammatory markers in epidemiological studies. Our study also has some potential shortcomings. First, dietary intake was evaluated using 24-h dietary recalls, which can introduce recall bias. Adults with obesity tend to underreport energy intake because of weight stigma or personal judgment regarding food and weight [57]. Therefore, we included BMI as a covariate in our regression analyses. Future studies should use more objective measures, such as biomarkers, to more accurately assess polyphenol intake [58]. Second, although polyphenol intake was estimated using an updated comprehensive database (Phenol-Explorer), information on the phenolic content of some local foods is lacking, which could affect the accuracy of the estimates. However, we minimized this limitation by incorporating some regional food composition data from scientific literature [20,21]. Nonetheless, the lack of data for many plant-based foods common in LMICs represents both a limitation for our study and a broader gap in global food composition research. Future studies should prioritize the characterization of phenolic profiles in these foods using high-accuracy methods like liquid chromatography coupled to mass spectrometry. Furthermore, the phenolic content in specific foods might vary based on agri-environmental, storage, transport, and processing conditions [59,60]. Third, residual confounding remains a limitation, despite efforts to adjust for covariates. Fourth, the exact volume of blood in DBS samples could not be precisely determined, which may introduce error in our marker measurement. Finally, the generalizability of our results may be limited to the study sample, consisting of women from a specific rural area with dietary patterns and cultural influences that may differ from those of men and urban populations. Thus, our findings may not be broadly applicable to the entire Peruvian population.

In conclusion, we found that dietary polyphenol intake was associated with an anti-inflammatory profile among rural women. Phenolic acid and stilbene were the main phenolic compounds that contributed to a lower inflammatory profile. Differences in polyphenol intake by socioeconomic status underscore the importance of strategies to promote polyphenol-rich diets among vulnerable populations. Further studies are necessary to understand the mechanisms of these associations and investigate the clinical effects of polyphenol-rich diets as a nutritional intervention in inflammatory conditions.

Author contributions

The authors’ responsibilities were as follows – WC, CT-M: conceived the overall study. EPL-S, CT-M: conducted statistical analyses and drafted the first version of the manuscript; WC, RML-R: provided scientific support over the course of this work; and all authors: revised the manuscript critically for intellectual content and approved the published version.

Funding

The CHAP trial received funding from the United States National Institutes of Health through various Institutes and Centers, including the Fogarty International Center, National Institute of Environmental Health Sciences, National Cancer Institute, and Centers for Disease Control, under award numbers U01TW010107 and U2RTW010114 (MPIs: Checkley, Gonzales, Naeher, Steenland). Furthermore, the trial received partial support from the Clean Cooking Alliance of the United Nations Foundation (UNF-16-810, PI: Checkley). EPL-S was supported by the “Juan de la Cierva-Formación” grant (JDC2022-049842) funded by the Ministerio de Ciencia e Innovación (MICIN) (MCIN/AEI/10.13039/501100011033) and, by European Union NextGeneration EU/PRTR. INSA-UB is a María de Maeztu Center of Excellence (grant CEX2021-001234-M both funded by MICIN/AEI/FEDER, UE). ISGlobal was supported from the grant CEX2023-0001290-S funded by MCIN/AEI/ 10.13039/501100011033, and from the Generalitat de Catalunya through the CERCA Program. CT-M was supported by the D43-funded Fogarty Research training in Chronic, Non-Communicable Respiratory Diseases in Peru (PulmPERU) training grant (D43TW011502).

Conflict of interest

RML-R reports personal fees from Cerveceros de España, personal fees, and others from Adventia, UNIDECO SA, Wine in Moderation, Ecoveritas S.A., outside the submitted work. All other authors declare no conflict of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article: