Association of Periconceptional Dietary Inflammatory Index and Hypertensive Disorders of Pregnancy and Its Subtype
Funding: This research was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (RTI International grant U10 HD063036, HD063072, U10 HD063047, U10 HD063037, U10HD063041, U10 HD063020, U10 HD063046, U10 HD063048, U10 HD063053). In addition, support was provided by respective Clinical and Translational Science Institutes to Indiana University (grant UL1TR001108) and University of California, Irvine (grant UL1TR000153). Dr. Venkatesh was supported by the Care Innovation and Community Improvement Program.
ABSTRACT
Background
The association between diet and hypertensive disorders of pregnancy (HDP) is inconclusive due incomplete assessment of dietary measures and HDP subtypes.
Objective
We determined the association between the energy-adjusted Dietary Inflammatory Index (E-DII) and the risk of HDP subtypes in a large, diverse cohort of pregnant individuals.
Methods
We completed a secondary analysis of data from the nuMoM2b study. Participants (n = 7309) were recruited from eight sites across the United States in the first trimester and followed through delivery. Participants completed the Block Food Frequency Questionnaire during the first trimester, which assessed the periconceptional period. Calculated macro and micronutrients from the questionnaire were used to calculate E-DII scores. HDP subtypes included gestational hypertension, preeclampsia without severe features, and preeclampsia with severe features. Multivariable ordinal logistic regression models were used to assess the association between E-DII and HDP subtypes, adjusting for baseline age, body mass index, education, race, and ethnicity as a social construct, and insurance status.
Results
Among 7309 assessed individuals, the mean E-DII score was −0.61, and the frequency of gestational hypertension was 15.5%, preeclampsia without severe features was 4.3%, and preeclampsia with severe features was 4.1%. A higher E-DII score was associated with greater odds of developing an HDP, with worsening E-DII scores associated with greater odds of experiencing a more severe HDP subtype (adjusted OR = 1.15, 95% CI: 1.08, 1.22).
Conclusions
In a multisite prospective US cohort of nulliparous individuals, higher periconceptional E-DII scores indicative of a pro-inflammatory diet was associated with a slightly increased risk of HDP. And the severity of HDP progressively increased with worsening dietary inflammation.
1 Introduction
Hypertensive disorders of pregnancy (HDP) affect approximately 15% of pregnant individuals in the United States [1], and is a leading cause of maternal morbidity and mortality [1, 2]. HDP increases the risk of adverse maternal and child outcomes, including birth complications, chronic hypertension, premature cardiovascular disease, and related mortality [3, 4]. In clinical practice, HDP can be classified into multiple subtypes, from isolated hypertension to organ failure or death, which dictate its management from least to most severe, and current clinical classifications are gestational hypertension (GHTN), preeclampsia (PE) without severe features, and PE with severe features [5, 6].
The pathogenesis of HDP is multi-factorial, and its exact mechanism is poorly understood [4, 6-8]. Common themes in proposed disease pathways focus on systemic inflammation, abnormal placentation, and oxidative stress in pregnancy [4, 6, 8]. Recent studies have demonstrated a relationship between derangements in systemic inflammatory factors (specifically upregulation of proinflammatory cytokines and downregulation of anti-inflammatory cytokines) and the risk of developing HDP [9-11], and that derangements in systemic inflammation can been seen prior to pregnancy in women who go on to develop HDP [12, 13]. It is hypothesized a heightened inflammatory state prior to pregnancy contributes to a maladaptive immune response to pregnancy that contributes to HDP [11, 14]. While adaptations to immune function are a critical physiologic event in pregnancy, overresponses or pathologic changes in the immune response such as a Th1/TH2 shift and M1/M2 imbalance are thought to be on the causative pathway of HDP [14, 15]. Furthermore, other pathophysiologic pathways contributing to HDP such as hypoxia from poor placentation, and oxidative stress also create a positive feedback cycle that further prompts alterations in systemic inflammation [10]. Exploring factors thought to impact systemic inflammation may be relevant to understanding risk for HDP.
Diet is one factor associated with inflammation, and as such, several studies have examined the relationship between dietary patterns and pregnancy outcomes [16-18]. Specifically, studies have shown that a high-quality diet, which includes more fruits and vegetables, whole grains, healthy fats, and lean proteins in conjunction with fewer processed/fried foods, sugary drinks/snacks, and saturated fats, is associated with a lower risk for adverse pregnancy outcomes [16-20]. The Dietary Inflammatory Index (DII) and energy-adjusted Dietary Inflammatory Index (E-DII) were developed to objectively quantify the inflammatory potential of a given diet [21-23]. Higher E-DII has been associated with higher plasma levels of pro-inflammatory factors [22-24]. The E-DII has been validated across cultural and regional differences in dietary patterns [21, 22]. Prior data outside of pregnancy show that higher E-DII scores are associated with higher odds of cardiovascular disease, hypertension, diabetes, and obesity [25-29]. Less is known about how the E-DII is associated with pregnancy outcomes. Some studies have found a relationship between increased E-DII and increased maternal inflammation, oxidative stress, and HDP [20, 24, 30]; however, other studies have found no relationship [31-33]. Such differences in findings may be due to small sample sizes and lack of assessment of HDP subtype (GHTN, PE without severe features, PE with severe features). The objective of the current study was to examine the associations between E-DII and HDP and its subtypes in a large, diverse cohort of nulliparous individuals.
2 Methods
2.1 Study Setting and Participants
We conducted a secondary analysis of data from participants in the nuMoM2b study (n = 10 038). The primary aim of nuMoM2b was to investigate factors associated with adverse pregnancy outcomes in a nulliparous population [34]. Study procedures have been previously described [34]. Briefly, participants were enrolled in the first trimester and followed through delivery. Participants completed questionnaires at study visits, and their medical records were reviewed for relevant data about pregnancy outcomes using data abstracted by trained study personnel. Participants were excluded from the present analysis if they had preexisting diabetes or hypertension, were missing dietary or pregnancy outcome data, had a non-live birth, or were missing covariate data. All procedures received Institutional Review Board approval at each institution, and participants provided written, informed consent.
2.2 Exposure
Participants completed the Block Food Frequency Questionnaire at the first study visit at 6–13 weeks’ gestation to capture periconceptional diet. NutritionQuest (Berkley, California) analyzed the Food Frequency Questionnaire data to convert participant responses into average daily intake of macro and micronutrients, which were then used to calculate the E-DII. Methods to calculate the E-DII have been previously described [21, 22, 35]. Briefly, participant's intake was first converted to energy-density metrics per 1000 kcals consumed. Next, macro and micronutrients were compared to the global average intake of each macro and micronutrient (also converted to be per 1000 kcals consumed) to create a z-score. Z scores were multiplied by an “article effect score” for each parameter. Article effect scores reflect the association between food parameters and circulating inflammatory marker data synthesized from over 2000 articles. Food parameter scores were then summed to create the overall E-DII score. Lower, more negative scores reflected a more anti-inflammatory diet, while higher and more positive scores reflected a more pro-inflammatory diet. Dietary data were available for 7981 participants. There are known sociodemographic differences in the nuMoM2b cohort between dietary responders and non-responders, with responders typically being older, more educated, less likely to have government insurance, and more likely to be non-Hispanic White [36]. However, there are not differences in the rate of HDP.
2.3 Outcome
The primary outcome was defined as a four-level outcome to capture HDP severity and included in an ordinal or ordered fashion GHTN, PE without severe features, and PE with severe features compared with no HDP as the reference. The outcome was defined using data from pregnancy through the delivery admission, and participants were classified per their most severe HDP diagnosis. GHTN and PE classifications in nuMoM2b were defined according to the American College of Obstetrics and Gynecology criteria [3, 37]. GHTN was defined as new-onset hypertension (≥140/90 mmHg) after 20 weeks of gestation with no proteinuria or other signs of end-organ involvement. PE without severe features was defined as new-onset hypertension (≥140/90 mmHg) after 20 weeks of gestation in the setting of proteinuria. PE with severe features was defined as new-onset hypertension with either SBP ≥ 160 or DBP ≥ 110 or other evidence of end-organ injury.
Descriptive statistics were used to report the characteristics and outcomes of the population. Participants with missing covariate data (n = 139) were excluded from the analysis. We first used a univariate model with E-DII as the predictor and HDP as the outcome to assess that the Brant test of cumulative probability (p = 0.22) was met. Because we had a non-binary outcome that was categorized by underlying disease severity, we conducted ordinal logistic regressions to determine whether there was an association between higher E-DII and increased severity of HDP (i.e., GHTH vs. PE without severe features vs. PE with severe features). We assessed unadjusted and adjusted models. We a priori decided to adjust for age, government-issued insurance (yes/no), college level education (yes/no), self-reported minoritized race and ethnicity (non-Hispanic White yes/no), and body mass index (BMI) based on conceptional associations between these factors and the study variables. Self-reported race was understood as a social construct that has been associated with HDP due to social drivers of health. Categorical variables were dummy coded into dichotomous variables for ease of statistical interpretation. Finally, because of the complex relationship between diet and BMI, we assessed interaction effects by BMI, and a priori decided we would stratify by BMI if interaction analyses were significant (p < 0.05). All analyses were conducted using R 4.4.1 [38]. Ordinal regressions, assumption tests, and pseudo R2 were conducted with the ordinal, companion, MASS, and brant packages [39-42].
3 Results
Of the nuMoM2b study sample (n = 10 038), 2057 were removed for lack of dietary data, 354 were removed for lack of pregnancy outcome data, 179 had preexisting hypertension, and 139 had missing covariate data, namely BMI (n = 109), yielding a final sample size of 7309. Characteristics of the analytical sample stratified by HDP subtypes, are reported in Table 1. Most of the sample was non-Hispanic White (65%), had a college level education (56%), and had private insurance (76%). Participants’ mean age at first visit was 27.4 years (standard deviation [SD] ± 5.5 years) and mean BMI was 26.0 (SD ± 5.93). Compared to individuals who remained normotensive, individuals with HDP had higher, more inflammatory E-DII scores, but overall, average E-DII scores (−0.61) reflect a neutral inflammatory diet. Individuals with PE with or without severe features were more likely to be non-White and have less than a college education and government insurance than individuals who remained normotensive or developed GHTN.
| Total (N = 7309) | No HDP (N = 5607) | GHTN (N = 1088) | PE without severe features (N = 312) | PE with severe features (N = 302) | |
|---|---|---|---|---|---|
| Insurance | |||||
| Private | 5538 (76%) | 4268 (76%) | 839 (77%) | 213 (68%) | 218 (72%) |
| Government | 1771 (24%) | 1339 (24%) | 249 (23%) | 99 (32%) | 84 (28%) |
| Race | |||||
| NH White | 4743 (64.8%) | 3634 (64.8%) | 733 (67.3%) | 194 (62.2%) | 182 (60.2%) |
| NH Black | 753 (10.3%) | 499 (8.9%) | 166 (15.3%) | 41 (13%) | 47 (15.7%) |
| Hispanic | 1170 (16.0%) | 965 (17.2%) | 104 (9.6%) | 54(17.5%) | 47 15.4% |
| Asian | 307 (4.2%) | 251 (4.5%) | 40 (3.7%) | 6 (1.9%) | 10 (3.3%) |
| Othera | 336 (4.6%) | 258 (4.6%) | 45 (4.1%) | 17 (5.4%) | 16 (5.4%) |
| Education | |||||
| No college | 3235 (44%) | 2465 (44%) | 446 (41%) | 161 (52%) | 163 (54%) |
| College | 4074 (56%) | 3142 (56%) | 642 (59%) | 151 (48%) | 139 (46%) |
| Fetal sex | |||||
| Male | 3736 (51%) | 2879 (51%) | 534 (49%) | 173 (55%) | 150 (50%) |
| Female | 3553 (49%) | 2709 (48%) | 554 (51%) | 139 (45%) | 151 (50%) |
| Ambiguous | 2 (0%) | 1 (0%) | 0 (0%) | 0 (0%) | 1 (0%) |
| Missing | 18 (0%) | 18 (0%) | 0 (0%) | 0 (0%) | 0 (0%) |
| Body mass index | |||||
| Mean (SD) | 26.0 (5.93) | 25.3 (5.48) | 27.8 (6.53) | 28.5 (7.28) | 28.6 (7.32) |
| Med. [min, max] | 24.4 [14.4, 59.8] | 23.9 [14.4, 59.7] | 26.2 [15.9, 53.5] | 26.3 [17.1, 58.2] | 26.6 [16.0, 59.8] |
| Gestational age at delivery (weeks) | |||||
| Mean (SD) | 38.9 (1.95) | 39.0 (1.86) | 39.0 (1.71) | 38.4 (1.87) | 36.9 (3.04) |
| Med [min, max] | 39 [20, 43] | 39.0 [20, 43] | 39.0 [26, 42] | 39 [28, 42] | 38 [24, 42] |
| Missing | 4 (0%) | 4 (0%) | 0 (0%) | 0 (0%) | 0 (0%) |
| Age (years) | |||||
| Mean (SD) | 27.4 (5.5) | 27.3 (5.4) | 27.9 (5.6) | 26.8 (5.7) | 27.4 (5.8) |
| Med [min, max] | 28 [13, 45] | 28 [13, 45] | 28 [14, 44] | 27 [16, 41] | 28 [16, 42] |
| Dietary Inflammatory Iandex | |||||
| Mean (SD) | −0.61 (2.07) | −0.68 (2.06) | −0.48 (2.10) | −0.21 (2.04) | −0.26 (2.07) |
| Med. [min, max] | −0.66 [−5.7, 4.7] | −0.74 [−5.7, 4.7] | −0.57 [−5.4, 4.4] | −0.09 [−4.6, 4.2] | −0.16 [−5, 4.4] |
- Abbreviations: HDP: hypertensive disorders of pregnancy; GHTN: gestational hypertension; max: maximum; Med.: median; min: minimum; NH: non-Hispanic; PE: preeclampsia; SD: standard deviation.
- a Other race includes American Indian, Native Hawaiian, and Multiracial.
The frequency of HDP subtypes was 15.5% GHTN, 4.3% PE without severe features, and 4.1% PE with severe features. The average E-DII score for individuals with GHTN was −0.48; PE without severe features was −0.21, and PE with severe features was −0.26.
In multivariable analyses (Table 2), the odds of experiencing a more severe HDP subtype were slightly higher for each 1 SD increase in pro-inflammatory diet [aOR 1.15 (95% CI: 1.08–1.22). For example, someone with an E-DII score at the mean had a 14% probability of developing GHTN, while someone with an E-DII two SDs above the mean had a 17% probability of developing GHTN. There were no significant interaction effects for BMI (p = 0.11).
| Model 1 | Coefficients | β | OR | OR 95% CI | p | |
|---|---|---|---|---|---|---|
| EDII | 0.14 | 1.16 | 1.09 | 1.22 | <0.001 | |
| Threshold coefficients | β | SE | p | |||
| Normal | GHTN | 1.20 | 0.03 | <0.001 | |||
| GHTN | PE | 2.40 | 0.04 | <0.001 | |||
| PE | PE with severe features | 3.15 | 0.06 | <0.001 | |||
| Pseudo R2 | ||||||
| Cox and Snell | 0.004 | |||||
| Nagelkerke | 0.005 | |||||
| Model 2 | Coefficients | β | aOR | aOR 95% CI | p | |
|---|---|---|---|---|---|---|
| EDII | 0.14 | 1.15 | 1.08 | 1.22 | <0.001 | |
| Government insurance | 0.10 | 1.10 | 0.94 | 1.29 | 0.23 | |
| College | 0.07 | 1.07 | 0.92 | 1.24 | 0.37 | |
| Non-White | −0.10 | 0.90 | 0.80 | 1.02 | 0.11 | |
| Age | 0.02 | 1.02 | 1.01 | 1.03 | 0.01 | |
| Body mass index | 0.07 | 1.07 | 1.06 | 1.08 | <.001 | |
| Threshold coefficients | β | SE | p | |||
| Normal | GHTN | 3.54 | 0.21 | <0.001 | |||
| GHTN | PE | 4.78 | 0.21 | <0.001 | |||
| PE | PE with severe features | 5.54 | 0.22 | <0.001 | |||
| Pseudo R2 | ||||||
| Cox and Snell | 0.04 | |||||
| Nagelkerke | 0.05 | |||||
- Note: Model 1 reflects the unadjusted model while Model 2 reflects the adjusted model.
- Abbreviations: E-DII: energy adjusted Dietary Inflammatory Index; GHTN: gestational hypertension; PE: preeclampsia.
4 Discussion
In a multisite prospective US cohort of nulliparous individuals, higher periconceptional E-DII scores indicative of a pro-inflammatory diet was associated with a slightly increased odds of HDP, and the severity of HDP progressively increased with worsening dietary inflammation.
Numerous studies have reported elevated systemic inflammatory markers in early pregnancy in individuals who consequently develop HDP [11, 43, 44]. The exact cause of the elevated systemic inflammatory markers, whether it is developing HDP pathology, or a reflection of the pre-pregnancy state is uncertain [6, 15]. Growing evidence does support that heightened systemic inflammation or chronic inflammation may be present prior to pregnancy in women who develop HDP [11-13, 45]. In the current study, we used the E-DII score to measure inflammation exposure through diet as it is based on empirical data between dietary patterns and inflammatory markers, such as IL-6, IL-10, and TNF-α [21, 23]. Here, we hypothesize that a more pro-inflammatory diet in the periconceptional period contributed to elevated systemic inflammation contributing to an altered immune response to pregnancy during the time of placenta development and implantation, and thus greater odds of developing HDP. The increasing odds for more severe types of HDP as dietary inflammation increases may reflect differing mechanisms of HDP subtypes [8, 46]. It is possible that mechanisms leading to preeclampsia with severe features are more susceptible to pre-pregnancy systemic inflammation (here induced by inflammatory diet). Potentially, improving diet prior to pregnancy by consuming a low inflammation diet may reduce the likelihood of developing severe forms of HDP. However, while our findings are statistically significant, the effect size is small, and further intervention work would be required to determine if pre-pregnancy diet can influence risk for different types of HDP, as well as the mechanisms by which risk reduction may occur.
The findings of the current study add to and extend to the growing evidence that periconceptional dietary patterns, here inflammatory dietary patterns, may impact pregnancy outcomes, including HDP [47-49]. Previous clinical trials focused on diet modification in pregnancy have not significantly decreased HDP [50-52]. However, many of these trials initiated dietary intervention after 12 weeks of gestation. It is possible that modification of diet later in pregnancy may have limited impact to alter the placentally driven pathophysiology of HDP, which likely occurs in early pregnancy. Dietary modification before pregnancy, and dietary modification focused on mechanisms associated with HDP such as an anti-inflammatory diet, may more effectively reduce risk for HDP by promoting placental development and function [48]. In the current study, dietary data reflected the periconceptional period (i.e., 3 months around conception), offering further support that periconceptional or preconception dietary intervention may still be a fruitful target to potentially mitigate HDP [53].
While the preconception period is difficult to target with a significant portion of pregnancies unplanned or unintended, growing evidence supports the feasibility of preconception interventions [54-58]. Of the successful preconception dietary interventions, interventions have focused on micronutrient supplementation and engagement of the woman and her partner in the intervention [56-58]. In a systematic review of preconception interventions, health behavior education delivered by primary care providers during routine primary care was effective at improving overall health prior to conception and ultimately decreasing adverse outcomes such as miscarriage, preterm birth, and low birthweight [54]. Anti-inflammatory diet patient education, such as the Mediterranean Diet for cardiovascular disease prevention [59], already exists and could also be delivered in primary care to reproductive aged individuals with the goal of improving systemic inflammation prior to conception to improve HDP outcomes. Evidence suggests there is a general lack of patient knowledge of the importance of preconception health for all [60-63]. While more direct intervention is likely necessary, ongoing conversations between patients and care providers during the reproductive years at routine care appointments can help bridge the patient knowledge gap of the importance of preconception health [60, 61].
The current analysis has some limitations. This is an observational study, and we cannot assess causality between dietary inflammation and the risk of HDP. Second, while the Block Food Frequency questionnaire is a well-validated instrument for assessing individual dietary patterns, including in pregnancy, the instrument relies on participant recall which may not be accurate. Also, the current study included the Block Food Frequency questionnaire at a single time point, and did not have data on potential changes across trimester of pregnancy, which may also impact HDP risk. Dietary data were not available for all participants, and participants without dietary data were more likely to be less educated, have government insurance, be non-Hispanic or non-White, be younger, and have a higher BMI. This study did not account for other health behaviors and lifestyle factors that may impact both inflammation and HDP risk, resulting in residual confounding. This analysis was restricted to a prospective cohort of nulliparous individuals, however, the association between dietary inflammation and HDP risk would not be expected to vary by parity. Finally, we categorized HDP by current ACOG classifications, however, other HDP subtypes exist such as early onset versus late onset and immunologic versus angiogenic [6, 64]. The association between diet and other subtypes may differ than our findings reported here, and is an avenue for continued investigation.
The current analysis also has strengths. Data were collected prospectively and longitudinally from a diverse cohort of pregnant individuals across the United States. Importantly, sufficient frequencies of HDP subtypes allowed analysis of diet with these subtypes, which is an approach largely missing from previous analyses. Finally, the E-DII metric is an exposure that allows for re-assessment of these results across populations and geographical locations, and can also be used in clinical practice and future research studies.
5 Conclusion
In summary, a more pro-inflammatory diet in the periconceptional period was significantly associated with a slightly increased risk of HDP as well as a progressively increased risk of more severe HDP, albeit the overall effect of diet remains small. Preconception diet interventions may prove to be fruitful and effective interventions to improve HDP outcomes, which will require further study. Future research should also examine the relative role of inflammation as a mechanism linking dietary patterns to pregnancy outcomes.
Ethics Statement
All study procedures received Institutional Review Board Approval, and all participants provided written informed consent.
Conflicts of Interest
The authors declare no conflicts of interest.
Open Research
Data Availability Statement
The data that support the findings of this study are openly available in DASH data repository.
