Association of Vegetable and Animal Flesh Intake with Inflammation in Pregnant Women from India

Abstract

In pregnant women, studies are lacking on the relationship of vegetable and animal flesh (poultry, red meat and seafood) intake with inflammation, especially in low- and middle-income countries. We conducted a cohort study of pregnant women receiving antenatal care at BJ Medical College in Pune, India. The dietary intake of pregnant women was queried in the third trimester using a validated food frequency questionnaire. Twelve inflammatory markers were measured in plasma samples using immunoassays. Only 12% of the study population were vegetarians, although animal flesh intake levels were lower compared to Western populations. In multivariable models, higher intakes of total vegetables were associated with lower levels of the T-helper (Th) 17 cytokine interleukin (IL)-17a (p = 0.03) and the monocyte/macrophage activation marker soluble CD163 (sCD163) (p = 0.02). Additionally, higher intakes of poultry were negatively associated with intestinal fatty-acid binding protein (I-FABP) levels (p = 0.01), a marker of intestinal barrier dysfunction and Th2 cytokine IL-13 (p = 0.03), and higher seafood was associated with lower IL-13 (p = 0.005). Our data from pregnant women in India suggest that a higher quality diet emphasizing vegetables and with some animal flesh is associated with lower inflammation. Future studies should confirm these findings and test if modulating vegetables and animal flesh intake could impact specific aspects of immunity and perinatal health.

Keywords: gut barrier; inflammation; meat intake; monocyte activation; pregnancy; vegetable intake.

Figure 1

Association of vegetable and fruit intake with markers of inflammation. Using linear regression, the mean difference in log₂ concentration of each inflammation markers and 95% confidence intervals (95% CI) per unit (log_e grams/day) change in total vegetable and fruit intake is shown in Figure 1. We further assessed the relationship with inflammation markers by unit change (log_e grams/day) of (a) total vegetables and (b) fruits, along with sub-categories of vegetables: (i) dry vegetables, (ii) green vegetables and (iii) raw/salad vegetables. Only significant results (p < 0.05) in univariable or multivariable analyses are shown in the forest plot (Figure 1A), with complete data shown in Supplementary Tables S2–S4 and S6–S8. The network graph (Figure 1B) shows statistically significant mean differences of inflammatory markers (gray nodes) for each food group (green nodes). Positive or negative mean differences are represented by yellow or blue edges, respectively. The size of the edges is proportional to log₂ mean differences. Only significant (p < 0.05) multivariable models are shown in this chart. Univariable models are adjusted for average daily energy intake and multivariable models are adjusted for daily energy intake, HIV status, age, education, gestational age at sampling and current smoking status.

Figure 2

Association of animal flesh intake with markers of inflammation.Using linear regression, Table 2. concentration of each inflammation markers and 95% CI per unit (log_e grams/day) change in total animal flesh (poultry + red meat + seafood) intake is shown in Figure 2. We further assessed the relationship of inflammation markers by unit change (log_e grams/day) of sub-categories of animal flesh: (i) poultry, (ii) red meat and (iii) seafood. Only significant results (p < 0.05) in univariable or multivariable analyses are shown in the forest plot (Figure 2A), with complete data shown in Supplementary Tables S5 and S9–S11. The network graph (Figure 2B) shows statistically significant mean differences of inflammatory markers (gray nodes) for each food group (green nodes). Positive or negative mean differences are represented by yellow or blue edges, respectively. The size of the edges is proportional to log₂ mean differences. Only significant (p < 0.05) multivariable models are shown in this chart. Univariable models are adjusted for average daily energy intake and multivariable models are adjusted for daily energy intake, HIV status, age, education, gestational age at sampling and current smoking status.