Recent Choline Intake Is Inversely Associated with Inflammation in Pregnancy – Evidence from the Canadian APrON Cohort

Abstract

Background

Maternal nutrition plays a role in regulating inflammation during pregnancy, which can impact maternal and fetal health.

Objectives

This study explored the association between recent maternal dietary choline intake and high-sensitivity C-reactive protein (hs-CRP) in the third trimester of pregnancy, leveraging data from the Alberta Pregnancy Outcomes and Nutrition (APrON) cohort.

Methods

Dietary choline intake was assessed using a validated 24-h dietary recall and hs-CRP was measured from nonfasting maternal blood samples. Statistical analyses included natural spline regression models to assess the relationship between recent choline intake and hs-CRP, with interaction terms for consumption of other methyl donor nutrients. We also assessed the likelihood of clinically elevated hs-CRP based on choline intake categories.

Results

Our analyses of 1300 pregnant people revealed a significant nonlinear inverse association between maternal choline intake and hs-CRP concentrations. Additionally, participants with higher choline intakes had reduced odds of having had hs-CRP above the clinical cutoff of 5 mg/L when compared with those with lower choline intakes (e.g. intake >700 compared with 200 mg/d, odds ratio = 0.07, 95% confidence interval: 0.02, 0.15).

Conclusions

These findings suggest that higher dietary choline intake may be associated with lower inflammation during pregnancy, highlighting the importance of adequate choline consumption for maternal health.

Introduction

Maternal nutrition is crucial for regulating inflammation during pregnancy and for preventing inflammation-related homeostatic disruptions that can have lasting negative impacts on both mother and fetus. Both acute and chronic inflammation during pregnancy increase the risk for adverse pregnancy outcomes such as preterm birth and preeclampsia and negatively impact offspring outcomes [[1], [2], [3], [4], [5], [6], [7]].

Evidence suggests that dietary intake of choline and other methyl donor nutrients may contribute to the regulation of inflammation [8]. Early evidence for the beneficial effect of choline on inflammation comes from animal models, where choline supplementation has been shown to ameliorate allergic asthma- and hypertension-associated inflammation [9]. In addition, in 2 separate human experimental studies in adults with metabolic syndrome, daily consumption of ≥400 mg/d choline led to reductions in IL-6, TNF-a, complement-reactive protein (CRP), and insulin resistance when compared with those on a lower choline diet [10,11].

Other methyl donor nutrients, folate, vitamin B12, and betaine, have also demonstrated an alleviatory impact on inflammation [12,13]. These nutrients, along with choline, function as part of the folate cycle, where their roles are intertwined [14]. This complexity is important to acknowledge when assessing the role choline may play in regulating inflammation as potential biological interactions may complicate efforts to study any one methyl donor’s impact in isolation.

However, most research on the anti-inflammatory benefits of choline, alone or in conjunction with other methyl donor nutrients, has been conducted in nonpregnant populations. To our knowledge, only 1 study in pregnant women, from our research group, has been conducted thus far [8].

This work aims to investigate the potential anti-inflammatory role of choline during pregnancy, leveraging the large and robust data collected from pregnant women in Alberta, Canada, as part of the Alberta Pregnancy Outcomes and Nutrition (APrON) cohort []. We aim to assess the association between recent maternal self-reported dietary choline intake and high-sensitivity C-reactive protein (hs-CRP) in the third trimester. Furthermore, we aim to assess potential nutrient–nutrient methyl donor interactions, in the hopes of gaining further insights into this complex biological pathway. We hypothesize that women with higher dietary choline intake will have lower inflammation and that this relationship will depend, in part, on their intakes of other methyl donor nutrients.

Methods

Study population

Women from the central and southern regions of Alberta, Canada (N = 2210) were recruited to the longitudinal APrON cohort study from 2009 to 2012. The original aim of the cohort study was to examine how maternal nutrient intake and status before, during, and after gestation influenced maternal mood, birth and obstetric outcomes, and infant neurodevelopment [15]. Details on the cohort recruitment process are provided elsewhere [15].

For this study of maternal choline intake and inflammation, we included participants from the APrON cohort who had dietary intake data and hs-CRP measurement at the third trimester visit, and who met our eligibility criteria (n = 1300). We excluded those women who had a pre-existing diagnosis of type-1 (21 women) or type-2 diabetes mellitus (2 women), documented alcohol or cigarette use in pregnancy (269 women), and/or antibiotic use in the second trimester of pregnancy (16 women) and have detailed this in the supplemental participant flowchart (Supplemental Figure 1). Exclusion criteria were selected as they are known confounders (e.g., antibiotics may cause gastrointestinal distress and prompt changes in dietary intake and directly impact inflammation) [16,17]. Excluded women were similar to included women (Supplemental Table 1), but fewer of the women missing hs-CRP identified as White or had incomes >69k/y.

Dietary assessment

At the third trimester pregnancy visit, conducted at mean (SD) 32.5 (1.3) wk of pregnancy, women were asked to detail the quantity and type of all foods and supplements consumed in the previous 24 h, midnight to midnight [15]. For the first 600 participants, trained research assistants conducted interviews using a multiple-pass 24-h dietary recall method (AMPM) [18]. Food models helped participants estimate portion sizes, and additional details, such as cooking methods, were recorded. The AMPM was evaluated for validity against the 3-d diet record for use during pregnancy and was deemed to be sufficiently accurate for use in the study [18]. Starting in August 2010, APrON transitioned to a 24-h food recall completed using the online Food Behaviour Questionnaire developed at the University of Waterloo, which employed a similar multipass structure to the AMPM [15,18]. The choline content of the women’s diet was estimated using a comprehensive choline database from the University of Alberta, which was developed specifically for use with the 24-h dietary intake recall data to estimate the choline content of foods consumed by the APrON participants [19]. The database contains information [19] on total choline content in foods, as well as on the 5 most common dietary forms of choline (free choline, glycerophosphocholine, phosphocholine, phosphatidylcholine, and sphingomyelin) and betaine. This database includes the choline content of roughly 2707 foods, and roughly 4 times more choline containing foods than the USDA database, making it the most detailed tool at the time for estimating dietary choline intake [19].

Blood collection and CRP measurement

Phlebotomists collected nonfasting maternal blood samples at the third-trimester pregnancy visit, concurrently with routine prenatal phlebotomy and with the 24-h dietary recall. The blood sample was then used to quantify hs-CRP, which is a more sensitive version of the historical CRP assay [20,21]. Of note, although both CRP and hs-CRP are latex particle-enhanced immunoturbidimetric assays that detect CRP in the blood, hs-CRP is more suitable than CRP for detecting low-grade inflammation [20]. Although hs-CRP is not known to fluctuate with recent intake, at the time of blood collection participants were asked about their food intake over the preceding three hours [22]. Additionally, as acute illness could impact CRP concentrations, participants were rescheduled if they reported any symptoms of illness. Samples were processed immediately, and serum, plasma (with EDTA), buffy coat, and red blood cells were aliquoted and stored at −80°C for future analyses at the Li Ka Shing Centre for Health Research Innovation in Edmonton, Alberta, Canada. Once thawed, hs-CRP was assessed via a commercially available sandwich ELISA kit (R&D Systems) [15]. This quantification occurred between 2017 and 2021.

Covariates

At enrollment in the study, participants reported their self-identified race and ethnicity, date of birth, household income, educational attainment, and parity. Researchers collected demographic and health covariates, including maternal prepregnancy height and weight, by reviewing medical records provided through participant consent [15]. Gestational age at each visit was quantified from the date of the last menstrual period and confirmed via ultrasound. We obtained information on diet-related covariates from the third-trimester 24-h recall (betaine, vitamin B12, omega-3 fatty acids, SFAs, total energy intake) or from blood samples (red blood cell folate) [15].

Ethics approval

The APrON study was approved by the University of Calgary Health Research Ethics Board (REB14-1702) and University of Alberta Health Research Ethics Biomedical Panel (Pro00002954). All participants gave written informed consent, and all procedures followed recognized standards for ethical research, including those outlined in the Declaration of Helsinki. This secondary data analysis was reviewed by Cornell University Institutional Review Board and determined to be exempt as the data was deidentified.

Data analysis

Explanatory and response variables were first visualized through histograms to ascertain their distribution and identify potential skewness. Our response variable, plasma hs-CRP was right skewed, with several high values in the right tail, >20 mg/L, perhaps because of unidentified current illness [[23], [24], [25]] (Supplemental Figure 2). To account for skewness, we natural-log transformed our response variable. The transformed variable is used in the reporting of each analysis, unless otherwise specified.

We first used locally estimated scatterplot smoothing (LOESS) to visually assess the shape of the relationship between choline intake and log-transformed CRP. The LOESS curve suggested a nonlinear association, with changes in slope at lower and higher intakes. On the basis of this pattern, we fit a natural spline model with knots at 450 mg and 900 mg—intakes corresponding to doses commonly tested in choline supplementation trials [[26], [27], [28]]. This approach captures the nonlinear relationship by allowing for local changes in the slope, which are represented by the changing tangent lines along the curve and allowed us to assign intervention-relevant thresholds to our spline structure. As a sensitivity analysis, we also fit a natural spline model with 3 degrees of freedom to ensure that our findings were not dependent on the choice of knot placement. The results were consistent across both approaches. The choice of spline type was guided, in part, by the Akaike Information Criterion and the Bayesian Information Criterion, while also selecting for splines that would minimize instability at the boundaries.

For choline subtypes, we also first visualized with a LOESS curve. Each subtype also demonstrated a nonlinear relationship, and thus, here too, we employed a linear regression with natural splines. For these subtypes, knots were placed at the first and third quartile locations.

In our simplest model, we estimated a linear regression model with natural splines to examine associations of choline intake (total choline and by choline subtype) and our biomarker of interest, hs-CRP. We also conducted multivariable analyses that included dietary covariates known to influence inflammation (omega-3 fatty acids, SFAs, betaine, vitamin B12, total energy intake from the 24-h recall, and red blood cell folate from blood analysis), as well as sociodemographic and perinatal confounders (prepregnancy BMI, gestational age at the time of blood draw, household income, maternal age at enrollment, and parity) [[32], [33], [34]]. An alpha level of 0.05 indicated a significant difference.

To assess for interactions between intake of choline and intake of other methyl donor nutrients, we implemented separate interaction models for each methyl donor of interest (betaine, vitamin B12, folate). In each analysis, choline intake was modeled using a natural spline, and an interaction term was included between the spline-transformed choline variable and each respective methyl donor nutrient. We used an alpha level of 0.05 to indicate a significant interaction term.

To evaluate the influence of potentially acute inflammatory responses on the relationship between maternal dietary choline intake and hs-CRP, we conducted sensitivity analyses using truncated hs-CRP values. Specifically, we restricted the analytic sample to participants with hs-CRP values below 4 thresholds: <1 mg/L, <5 mg/L, <10 mg/L, and <20 mg/L. For each subset, we assessed the association between self-reported choline intake and hs-CRP and compared model fit using residual SE values.

Additionally, we performed an analysis dichotomizing hs-CRP at >5 compared with ≤5 mg/L, and hs-CRP at >1 compared with ≤1 mg/L as hs-CRP >5 mg/L indicates clinically elevated inflammation and hs-CRP >1 mg/L indicates elevated subacute inflammation [29]. To assess whether the proportion of participants with elevated hs-CRP differed by choline intake, we grouped participants into 6 intake categories: <300 mg/d, 300–399 mg/d, 400–499 mg/d, 500–599 mg/d, 600–700 mg/d, and >700 mg/d. We first applied a chi-squared test with Bonferroni correction to test for overall differences across intake groups. We then estimated odds ratios for elevated hs-CRP within each intake category and evaluated statistical significance using Fisher’s exact test.

We evaluated each nutrient as a continuous variable and evaluated hs-CRP along a continuous scale. We also assessed for potential collinearity between methyl-donor nutrients using a Variance Inflation Factor. We performed all statistical analyses using R-studio Statistical Software (v4.1.2; R Core Team 2022).

Results

Demographic and health characteristics of the mothers at their first study visit are detailed in Table 1. The 1300 participants included in our analysis were predominately White/Caucasian, highly educated, and of high socioeconomic status. The mean (SD) age of the women at enrollment was 31.8 (4.3) and the average prepregnancy BMI was in the normal range (Table 1).

TABLE 1.

Characteristics of maternal participants.

Age1 (y)	31.8 ± 4.3
Education (% college graduate or greater)	71.7%
Race and ethnicity (% Caucasian)	81.8%
Household income (% >69k/y)	79.7%
Prepregnancy BMI1 (kg/m2)	23.8 ± 4.2
Nulliparous (% Yes)	51.9%
Choline intake1,2 (mg)	365.9 ± 159
Red blood cell folate1,3 (nmol/L)	1509.6 ± 381
Vitamin B122 (μg)	4.7 ± 3.5
Betaine2 (mg)	236.1 ± 195.4

¹Mean (SD).

²From dietary recall assessments.

³From laboratory values.

Dietary intakes of choline were modestly right skewed, with an estimated mean (SD) intake from the 24-h recall of 365.9 (159), which is below the recommendations for pregnancy of 450 mg/d [30]. Before log transformation, hs-CRP was right skewed, as is expected in a healthy population where most CRP values should fall below the reference range of 5 mg/L [[23], [24], [25]]. Average untransformed hs-CRP was 3.2 mg/L; however, 249 participants had values above the reference concentration (>5 mg/L) and 38 participants had markedly elevated CRP values of >10 mg/L. Log transformation normalized the skew of the values.

We observed a significant, nonlinear relationship between maternal third-trimester choline intake and hs-CRP. Table 2 reports the CRP estimated predictions for specific choline intakes, obtained from the weighted sum of the 3 splines or $\hat{y}$ = β0 + β₁B₁(x) + β₂B₂(x) + β₃B₃(x). For example, when hs-CRP was back transformed from the log scale, it reached predicted concentrations of 2.2 mg/L [95% confidence interval (CI): 2.0, 2.4] at 200 mg/d choline and 1.7 mg/L (95% CI: 1.2, 2.2) at 900 mg/d choline intake. These patterns persisted in the partially and fully adjusted models, but the CIs were substantially wider in the fully adjusted model and should be interpreted cautiously.

TABLE 2.

CRP predicted estimates at specific choline intakes.

Choline intake	Unadjusted hs-CRP1 estimate (95% CI) (mg/L)	Partially adjusted2 hs-CRP1 estimate (95% CI) (mg/L)	Fully adjusted3 hs-CRP1 estimate (95% CI) (mg/L)
200 mg/d	2.17 (1.98, 2.39)	2.39 (1.95, 2.93)	2.17 (0.84, 5.56)
300 mg/d	1.98 (1.85, 2.11)	1.96 (1.72, 2.23)	1.82 (0.72, 4.62)
400 mg/d	1.91 (1.85, 2.11)	1.85 (1.59, 2.17)	1.75 (0.68, 4.49)
500 mg/d	1.99 (1.83, 2.17)	2.13 (1.78, 2.56)	2.03 (0.78, 5.26)
600 mg/d	2.19 (1.93, 2.48)	2.68 (1.99, 3.61)	2.56 (0.96, 6.89)
700 mg/d	2.31 (1.92, 2.78)	2.85 (1.92, 4.21)	2.76 (0.99, 7.71)
800 mg/d	2.17 (1.73, 2.72)	1.94 (1.03, 3.65)	1.95(0.62, 6.11)
900 mg/d	1.65 (1.21, 2.23)	0.66 (0.14, 3.06)	0.70 (0.12, 4.18)

Abbreviation: hs-CRP, high-sensitivity C-reactive protein.

¹hs-CRP back-transformed from the log scale.

²Unadjusted + betaine, B12, red blood cell folate, omega-3 fatty acids, saturated fat.

³Unadjusted + maternal prepregnancy BMI, household income, gestational age at measurement, parity, maternal age at enrollment, total energy intake.

We further investigated relationships between maternal self-reported dietary choline intake and hs-CRP using subsets of CRP values to assess the potential influence of high hs-CRP values, which may be indicative of acute illness. In this analysis, we assessed the relationship at truncated hs-CRP values of 1 mg/L, 5 mg/L, 10 mg/L, and 20 mg/L and found the best fit when hs-CRP was <1 mg/L and <5 mg/L, as evidenced by the lowest residual SEs (Supplemental Table 2). This may indicate a clinical relevance of the anti-inflammatory impact of choline particularly during chronic low-grade inflammation and was supported by a significant inverse linear relationship between choline intake and log transformed hs-CRP in a linear regression in the <1 mg/L truncated data (β = −0.001, 95% CI: = −0.001, −0.0002) (Figure 1).

FIGURE 1.

Relationship between choline intake and truncated <1 mg/L log-transformed high-sensitivity C-reactive protein.

Analysis of the association between the recent intake of various choline derivatives and hs-CRP found no significant relationships between any derivative and maternal inflammation (Table 3). Furthermore, there were no significant interactions between choline and other methyl donor nutrients in their association with maternal hs-CRP (Figure 2).

TABLE 3.

Estimated hs-CRP concentrations at the 5th–95th percentiles of choline metabolite intake.

	GPC	PCH	PC	SM	Free choline
5th percentile	1.94 (1.70, 2.21)	2.14 (1.87, 2.44)	2.30 (2.00, 2.64)	2.04 (1.78, 2.33)	2.18 (1.42, 3.33)
25th percentile	1.99 (1.82, 2.16)	2.03 (1.87, 2.19)	2.00 (1.87, 1.15)	1.93 (1.79, 2.09	2.16 (1.51, 3.10)
50th percentile	1.97 (1.83, 2.11)	1.95 (1.81, 2.10)	1.89 (1.75, 2.04)	1.94 (1.80, 2.09	2.14 (1.59, 2.87)
75th percentile	1.96 (1.81, 2.12)	1.91 (1.76, 2.07)	1.90 (1.76, 2.05)	2.03 (1.88, 2.20	2.11 (1.71, 2.62
95th percentile	2.11 (1.87, 2.38)	1.98 (1.74, 2.24)	2.04 (1.79. 2.33)	2.08 (1.84, 2.35)	2.07 (1.87, 2.29)

Abbreviations: GPC, glycerophosphorylcholine; hs-CRP, high-sensitivity C-reactive protein; PC, phophatidylcholine; PCH, phosphocholine; SM, sphingomyelin.

FIGURE 2.

Interaction effects of choline intake and methyl donor nutrients on predicted high-sensitivity C-reactive protein concentrations.

However, we did find significant differences in the proportion of participants with elevated hs-CRP by choline intake categories, Supplemental Figure 3 demonstrates the differences in elevated hs-CRP (above the reference range of 5 mg/dL) by choline intake groups. Here, >6% of participants had hs-CRP above the reference range at the lowest intake (< 300 mg/d), whereas <1% had elevated CRP from intakes of 600 mg/d and higher. Chi-squared tests with Bonferroni corrections revealed significant differences (P < 0.001) between intake categories that were two apart. For example, participants consuming <300 mg/d choline showed a statistically higher proportion of elevated hs-CRP compared with those consuming 400–500 mg/d choline, and those consuming 400–500 mg/d choline had a significantly higher proportion of elevated hs-CRP compared with those consuming 600–700 mg/d choline. Odds ratios were also revealed to be significantly different via Fisher’s exact test, wherein each category above the <300 mg/d category had significantly lower odds of elevated CRP when compared with the <300 mg/d category (Figure 3). The effect was even more pronounced when we looked at low-grade subclinical inflammation—identifying the odds of having hs-CRP >1 mg/L (Figure 4).

FIGURE 3.

Unadjusted odds of high-sensitivity C-reactive protein >5 mg/L by choline intake category.

FIGURE 4.

Unadjusted odds of high-sensitivity C-reactive protein >1 mg/L by choline intake category.

Discussion

In this study of the Canadian APrON pregnancy cohort, we explored third trimester associations of recent maternal prenatal choline intake and a biomarker of maternal inflammation, as measured by hs-CRP. In this population, we found a significant, nonlinear, inverse association between recent maternal choline intake and hs-CRP [30]. However, higher hs-CRP at ∼600 mg/d choline intake warrants further investigation. The lower hs-CRP observed from 200 to 900 mg/d choline intakes was roughly 0.5 mg/L across that range. Additionally, at higher choline intakes, participants were less likely to have hs-CRP above the reference range; a finding may have clinical implications for choline recommendations during in pregnancy. We did not observe any choline derivative associations, indicating that the totality of choline intake may be more important than any one derivative. We also did not observe any interactions between choline and other methyl donor nutrients. These findings partially support the hypothesis that participants with higher dietary choline intake will have lower concentrations of inflammation markers, although the lack of effect from choline and methyl donor interactions deviates from our expectations.

The inverse association between choline intake and risk of elevated hs-CRP in this pregnant population confirms previous findings in nonpregnant populations [9,10,31,35]. There are several mechanisms to support these associations, including through the conversion of homocysteine in the methionine cycle, as an antioxidant, and through methylation-induced epigenetic changes [[36], [37], [38], [39], [40]]. In the former, choline is oxidized to betaine in a 2-step process, facilitated by the enzymes, choline dehydrogenase, and betaine aldehyde dehydrogenase. The choline metabolite, betaine, is then able to convert homocysteine to methionine with the assistance of betaine–homocysteine methyltransferase. This action and the resultant decrease in plasma homocysteine are physiologically advantageous as homocysteine is a known risk factor for atherosclerosis, and it plays a role in the orchestration of inflammatory response. With the latter mechanism, again driven by the choline metabolite betaine, changes in the amount of 5-methylcytosine in DNA can drive hyper (increased) or hypo (decreased) DNA methylation [41]. This mechanism is responsible for the expression or suppression of various genes, including those involved in regulating inflammation.

Although our findings align with previous research in nonpregnant populations, some of our findings were inconsistent with our previous work in the Project Viva pregnancy cohort, a maternal–infant dyad cohort in Boston, MA, United States [8]. In the Project Viva work, a longitudinal research study of women and their children, with an initial aim to examine relationships between maternal diet and other factors on offspring health, we analyzed cross-sectional associations between midpregnancy choline intake and CRP, IL-6, and TNF-a in healthy pregnant women. Here we found that the inverse relationship between choline intake and IL-6 relied on the intake of other methyl donor nutrients such as betaine, vitamin B12, and folate. Notably, we did not see any interactions between choline and other methyl donor nutrients as we did with Project Viva, and we saw significant results in the association with CRP, whereas our work in Project Viva only found a significant relationship between choline and IL-6. For the latter, the most plausible explanation is that during the data collection period for Project Viva in the early 2000s, hs-CRP was not yet available. Despite having similar minimum concentrations in both cohorts (0.006 mg/L in APrON and 0.005 mg/L in Project Viva), the high end of the Project Viva range was ∼5 mg/L, whereas the APrON cohort had plasma concentrations >20 mg/L. Additionally, Project Viva had more clustering or repeat values around the limit of detection, which could limit the variability of the data and increase statistical dependence. Consequently, the less sensitive test used in Project Viva was less likely to detect a relationship between choline and CRP even if there was one [42]. We also did not find an association between choline derivatives and CRP. This may be because the combined impact, as seen in overall choline intake, is greater than the impact of any one subtype or derivative.

These current findings from the APrON cohort mark an important translation of previous studies of choline and inflammation from rodents and nonpregnant adults with chronic disease to an apparently healthy, pregnant, population. As inflammation presents a significant threat to the health of both mother and fetus, avoiding inflammatory states during pregnancy is a crucial step in maintaining healthy pregnancies [[4], [5], [6], [7],39,[43], [44], [45], [46], [47]]. We found that elevated hs-CRP above the reference range is not uncommon in healthy pregnancies and that higher choline intake is associated with a lower likelihood of its occurrence. In addition to this, the combined results from the APrON and Project Viva cohorts provide initial support for future experimental work that could examine the therapeutic anti-inflammatory role of choline in pregnancy. Moreover, the results of this study raise concerns about the current low choline intakes in women of reproductive age. In North America, the average intake of choline is ∼325 mg/d—far below the adequate intake of 450 mg/d for pregnancy [19,30,[48], [49], [50], [51]]. The results of this study generally suggest a beneficial association of higher choline intakes for lower inflammation during pregnancy. These results suggest that further research on defining the optimal choline intake for pregnancy may be necessary and that perinatal care practitioners may need to include information about choline as part of patient education. It is possible that hs-CRP could be considered as a clinical target for generating evidence regarding the optimum choline intakes during pregnancy.

This study had many strengths. First, this work builds upon our previous work in the Project Viva cohort, which collectively represent the first work on the association between estimated dietary choline intake and clinical biomarkers of inflammation during pregnancy. In addition, this work addresses some of the limitations of our work with Project Viva, by using a larger sample size and more sensitive analysis methods for CRP. This work was also the first to assess not only total choline intake and inflammation, but also to assess associations between choline derivatives and inflammation. Lastly, the study used a multipass 24-h recall, obtaining more robust estimates of recent choline intake and sensitive measures of hs-CRP [42].

However, there were still limitations of this study. The cross-sectional design of the analysis limits the ability to infer temporality and causality between dietary intake and inflammation. Although the food-based Food Frequency Questionnaire employed by the Project Viva cohort may have limitations regarding its ability to estimate micronutrient intakes and has a high respondent burden, the 24-h dietary recall employed in the APrON cohort also has limitations [52]. Reliance on single 24-h dietary recall to estimate choline intake may not accurately reflect habitual dietary patterns, although third-trimester intake of choline was similar to that of first-trimester intakes [19,52]. Relatedly, lack of an available, valid, and reliable, biomarker for choline intake and status limits exposure measurement to self-reported intake methods, which are always subject to recall bias [52,53]. Lastly, this work featured a predominately White and well-educated population, potentially limiting generalizability.

In conclusion, overall, this cohort study demonstrated a significant, nonlinear, inverse association between higher maternal dietary choline intake and inflammation in the third trimester of pregnancy and found fewer participants with relatively high choline intake having hs-CRP above the reference range. Because of the long- and short-term damage caused by inflammation, including alterations to fetal development, shifts in metabolic programming, and an increased risk of pregnancy loss, this finding of lower inflammation at higher choline intakes may be clinically relevant. Although this is the first time these results have been shown in pregnant populations, they are supported by evidence in animal models and experimental studies in nonpregnant humans. Further examination of these initial findings in pregnancy could be an important area for future research as we consider updated recommendations for choline.

Author contributions

The authors’ responsibilities were as follows – EAL: designed and conducted the data analysis and generated the first draft of the manuscript; WP, LES, CJF, AF, AL, AS: contributed to revision of the manuscript and interpretation of the analysis results; and all authors: approved of the final version of the manuscript.

Funding

Research reported was supported by the NIH under award 2 T32 HD087137, an interdisciplinary team grant from Alberta Innovates–Health Solutions (grant/award number N/A) and by the Canadian Institute for Health Research (PS 156069).

Conflict of interest

The authors report no conflicts of interest.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

figs1.