Cardiovascular health in early pregnancy and circulating biomarkers of placental health and function

Sample

A subset of participants from the Nulliparous Pregnancy Outcomes Study: Monitoring Mothers to Be (nuMoM2b) prospective cohort comprised the sample for this study. Full methods for the cohort have been published previously.²⁸ Briefly, nulliparous pregnant people with a singleton gestation were recruited in early pregnancy (6⁰-13⁶ weeks gestation) and assessed three times throughout pregnancy (baseline, 16⁰-21⁶ weeks gestation and 22⁰-29⁶ weeks gestation) and at delivery. Assessments included questionnaires, clinical measurements, and blood draws. Ethical approval was obtained at each clinical site and participants provided informed consent prior to data collection. Biomarker assays were conducted on 1^st and 2^nd trimester samples using a case-subcohort design drawing from individuals who delivered after 20 weeks of gestation and provided sufficient serum.²⁹ All cases of APO (preterm birth, preeclampsia, stillbirth, and/or small-for-gestational-age birth), as well as a random sample of 12% of the remaining individuals with uncomplicated, term deliveries, were included. The sampling proportion of 12% was determined by the funds available for assays.

Cardiovascular Health Score

Each component of the CVH score was assessed in the 1^st trimester at the baseline visit and assigned a score on a scale of 0-100 according to the Life’s Essential 8 scoring algorithm, in which higher scores indicate more favorable health.¹² The full scoring algorithm can be found in Table S1.

Physical activity was assessed via a self-report questionnaire, in which participants identified frequency, duration, and type of activities over the past 4 weeks. Average minutes per week of moderate (3.0-5.9 metabolic equivalents of task [METs]) and vigorous (>6.0 METs) leisure time physical activity were calculated based on corresponding MET values from the Compendium of Physical Activities³⁰, and weekly duration of MVPA as it pertains to U.S. federal guidelines was calculated as minutes of moderate plus 2 x minutes of vigorous activity.³¹ CVH scores ranged from 0 (no weekly MVPA) to 100 (>150 minutes/week MVPA) with intermediate scores for values between 0-150 minutes/week. Physical activity was additionally categorized into none (no weekly MVPA), insufficient (1-149 minutes/week MVPA), sufficient (150-299 minutes/week MVPA), and high (300+ minutes/week MVPA).

The Healthy Eating Index (HEI) score was used to assess diet quality and was determined using a modified Block Food Frequency Questionnaire^32,33 with reference to the 3- month period immediately preceding conception. CVH scores were based on percentiles of HEI scores in the full nuMoM2b sample; for example, 0 points corresponded to HEI <25^th percentile, while 100 points corresponded to HEI >95^th percentile.

Self-reported weekday and weekend hours of nightly sleep were used to determine sleep duration. Weighted averages were calculated to account for differences between weekday and weekend sleep duration, and CVH scores were assigned based on the U-shaped association between sleep duration and cardiovascular health (i.e., both low and high sleep durations are adverse).

Participants self-reported combustible tobacco use and exposure to secondhand smoke in the home. CVH scoring for smoking was adapted from the standard scoring algorithm (see Table S1). A score of 0 indicated current smoking during pregnancy, while a score of 100 indicated never smoking tobacco; intermediate scores were assigned for former smokers. 20 points were subtracted in cases of exposure to secondhand smoke in the home for all except current smokers.

Blood pressure, height, and weight were measured in person by research personnel using standardized techniques. Body mass index (BMI) was calculated from height and weight measures in kg/m². The standard scoring algorithms for blood pressure and BMI were applied to calculate component scores. Blood draws were obtained in a non-fasted state, so the lipid and glucose components of the composite CVH score were excluded from this analysis.

The composite CVH score was derived by averaging the scores of the 6 available components: physical activity, diet, sleep, smoking, blood pressure, and BMI.

Placental blood analytes

Full methodology of the biomarker assays have been published previously²⁹; in brief, blood serum samples collected at the study visits in the 1^st and 2^nd trimester were stored at -70° C, then assayed with enzyme-linked immunosorbent assay (sEng), electrochemiluminescence assay (VEGF and sFlt-1), and lanthanide-based time-resolved fluorometry (PlGF and PAPP-A). Concentrations below the limits of detection were assigned values equal to half the lower limit, while those above the upper limit were assigned a value equal to the upper limit.

Covariates

Demographic information including age, race/ethnicity, educational attainment, income, and health insurance status were self-reported via questionnaire at the baseline visit. At the second visit, participants completed the Krieger Discrimination Scale³⁴ to assess experiences with racial discrimination and the Connor-Davidson Resilience Scale³⁵ to assess resilience.

Statistical Analyses

Latent (unobserved) classes of adherence to combinations of the four health behaviors were constructed empirically using the “gsem” command in Stata, which can compute the models using continuous values. Models were built using 1^st trimester CVH scores on a scale of 0-100 for each of the four health behaviors (physical activity, diet, sleep duration, and smoking) based on the full nuMoM2b sample (N=10,038), and the best-fitting model was selected based on values of Akaike/Bayesian Information Criterion, sample size in each class, and model interpretability; posterior probabilities were also assessed.

Distributions of placental blood biomarkers were examined for normality and log- transformed if appropriate. Descriptive characteristics were compared among behavioral latent classes using Chi-square tests for categorical variables and analysis of variance for continuous variables.

Associations between composite CVH scores, behavioral latent classes, and individual health behavior scores with each placental blood biomarker were examined using linear regression. Two sets of analyses were conducted. First, a cross-sectional analysis examined the association between 1^st trimester exposures (composite CVH scores, behavioral latent classes, and each individual health behavior) and 1^st trimester serum biomarker concentrations. Next, a longitudinal analysis examining the association between 1^st trimester CVH exposures and 2^nd trimester biomarker concentrations was performed. Scatterplot visualizations of the associations between physical activity scores and the biomarkers suggested a nonlinear association, so physical activity was analyzed categorically (none, insufficient, sufficient, and high); diet, sleep duration, and smoking were analyzed as continuous CVH component scores.

Sampling weights were added to all analyses to account for the overrepresentation of APO in the placental analyte subsample, wherein each observation was weighted based on the inverse of the sampling proportion (i.e., a weight of 1 was applied to cases of APO, while a weight of 8.48 was applied to non-APO observations). All linear regression models were adjusted for gestational age at the time of the blood draw, and longitudinal analyses were additionally adjusted for 1^st trimester biomarker concentrations. These minimally adjusted models were further adjusted for demographic characteristics, BMI (latent class and individual health behavior analyses only), perceived racial discrimination, and resilience in a backwards stepwise approach in which covariate associations with a p-value <0.2 were retained in the fully adjusted model. All analyses were conducted in Stata SE version 18 (StataCorp, College Station, TX) and significant associations were accepted at p<0.01 to account for multiple comparisons.

Participants

Of the original 10,038 nuMoM2b participants, 2,188 were included in this analysis (Figure 1). Mean baseline age was 26.9±5.8 years and mean BMI was 26.9±6.8 kg/m². Prior to applying sampling weights, those included in this analysis had significantly higher BMI, were more likely to report non-Hispanic Black race, less education, lower perceived racial discrimination, and, by design, more likely to experience an APO than the full cohort. After sample weights were applied, participants included in this analysis had similar rates of APO but had significantly lower BMI and were more likely to report non-Hispanic White race, lower perceived racial discrimination, and use of commercial insurance than the full cohort (Table S2).

• Download figure
• Open in new tab

Figure 1.

CONSORT diagram of included and excluded participants. nuMoM2b: Nulliparous Pregnancy Outcomes Study: Monitoring Mothers-to-be; GA: gestational age; SGA: small for gestational age; PTB: preterm (<37 weeks GA) birth.

Composite CVH Scores and Health Behavior Latent Classes

The mean composite CVH score in this sample was 71.1 ± 15.4 points. Thirty percent of participants achieved a “high” CVH score (>80 points), 56% had “moderate” CVH (score of 50-79 points), and 14% had “low” CVH (<50 points) as defined by the American Heart Association.¹²

For the behavioral latent class analysis, the model fit criteria (Table S3) indicated a 4- class solution was the most appropriate: Healthy Behaviors (n=756 [34.6%]), Active with Poor Diet (n=449 [20.5%]), Inactive with Poor Diet (n=717 [32.8%], and Less Healthy Behaviors (n=265 [12.1%]) (Figure 2).

• Download figure
• Open in new tab

Figure 2.

Weighted mean (SD) cardiovascular health (CVH) scores for each health behavior component in each latent class. Higher score indicates more ideal CVH; scores <50 are “poor” CVH; 50-79 are “moderate” CVH, and scores >80 are “high” CVH as indicated by markers on the y-axis.

Participants who were older, with a lower BMI, non-Hispanic White, and who had more education, higher income, or commercial insurance were significantly more likely to belong to the Healthy Behaviors class compared to the other classes. Furthermore, the composite CVH score was significantly higher and the rate of APO was significantly lower in the Healthy Behaviors class compared to the other classes in this sample (Table 1). Absolute concentrations of each placental biomarker by latent class in each trimester can be found in Table S4.

Placental growth and development: PAPP-A

A higher (more favorable) 1^st trimester composite CVH score was associated with higher (more favorable) 1^st trimester PAPP-A concentration (Table 2, p<0.001); this association was not detected in longitudinal analysis evaluating the association of 1^st trimester composite CVH score with 2^nd trimester PAPP-A concentrations. No significant associations were found between the behavioral latent classes and PAPP-A concentrations in either cross-sectional or longitudinal analyses (Tables S5 & S6).

Angiogenic balance: VEGF, PlGF, sFlt-1, sEng

A higher (healthier) 1^st trimester composite CVH score was associated with a more anti- angiogenic 1^st trimester biomarker profile, as reflected by a significantly lower concentration of VEGF and a significantly higher concentration of sEng in cross-sectional analyses (Table 2, p<0.001 for each).

Within the behavioral latent class analyses, a significant association was found between latent class and 1^st trimester sFlt-1 concentration (Figure 3), in which those in the Inactive with Poor Diet class had a more anti-angiogenic profile, as characterized by higher sFlt-1 concentrations than those in the Healthy Behaviors class (p=0.008).

• Download figure
• Open in new tab

Figure 3.

Healthy behaviors latent class demonstrated significantly lower 1^st trimester sFlt-1 concentration than the inactive with poor diet latent class, *p=0.008. sFlt-1: soluble FMS-like tyrosine kinase 1.

No longitudinal associations were detected between either composite CVH or the latent classes and 2^nd trimester angiogenic biomarker concentrations (Tables S5 & S6).

Individual Health Behaviors

Placental growth and development

In the longitudinal analysis, 1^st trimester physical activity was associated with lower 2^nd trimester concentration of PAPP-A, though a significantly lower proportion of those reporting sufficient weekly MVPA had an APO compared to those reporting no weekly MVPA (Figure 4a, p<0.01). Participants reporting no weekly MVPA had significantly higher PAPP-A than those reporting insufficient, sufficient, or high MVPA (Figure 4b, p<0.001 for all). No significant associations were found in 1^st trimester cross-sectional analysis for any of the individual health behaviors and 1^st trimester PAPP-A concentration, and no significant longitudinal associations were found between component scores for the other health behaviors (diet, sleep, or smoking) and PAPP-A concentrations in the 2^nd trimester (Tables S7 & S8).

• Download figure
• Open in new tab

Figure 4.

Differences between physical activity categories, defined as: None = 0 minutes/week; Insufficient = 1-149 minutes per week; Sufficient = 150-299 minutes/week; High = >300 minutes/week. *Indicates p<0.01 (a) Significantly lower weighted rate of APO in those with sufficient weekly physical activity compared to none; (b) Any 1^st trimester physical activity associated with lower 2^nd trimester PAPP-A than none; (c) No weekly 1^st trimester physical activity associated with lower 1^st trimester VEGF than those with insufficient or high weekly physical activity (p<0.001), while those with sufficient weekly physical activity were non-significantly lower (p=0.02); (d) High weekly 1^st trimester physical activity associated with lower 1^st trimester sFlt-1 than no weekly physical activity. PAPP-A: pregnancy-associated plasma protein A; sFlt-1: soluble FMS-like tyrosine kinase 1; VEGF: vascular endothelial growth factor. Data presented as weighted percentages (a) or weighted means with 95% confidence intervals (b-d).

Angiogenic balance

First trimester physical activity was associated with a pro-angiogenic 1^st trimester profile: more weekly physical activity was associated with a higher VEGF concentration (Figure 4c, p<0.001 for insufficient and high MVPA compared to none), and high MVPA was associated with lower first trimester sFlt-1 concentration versus no weekly MVPA (Figure 4d, p=0.003). Longitudinal associations between 1^st trimester physical activity categories and 2^nd trimester concentrations of angiogenic biomarkers were nonsignificant, and no significant cross-sectional or longitudinal associations were found between component CVH scores for diet, sleep, or smoking and biomarkers of angiogenic balance (Tables S7 & S8).

Placental growth and development

PAPP-A is a component of the insulin-like growth factor pathway.³⁶ PAPP-A concentrations increase across gestation in normal pregnancy^37,38 and display dysregulated trajectories in cases of APOs.³⁹ In preeclampsia, 1^st trimester PAPP-A is lower than normal, then climbs dramatically in the 3^rd trimester to levels above normal.⁴⁰

In line with our hypothesis, higher (more favorable) 1^st trimester CVH scores were associated with higher 1^st trimester circulating PAPP-A. As higher concentrations of this biomarker in early gestation are associated with better pregnancy outcomes, this finding suggests that higher early-pregnancy CVH may be related to more robust placentation and concomitant higher concentrations of PAPP-A. Whether early intervention to improve parameters of CVH can improve placentation through PAPP-A concentrations is worth future study.

In contrast, the association between greater 1^st trimester physical activity and lower 2^nd trimester concentrations of PAPP-A suggests that physical activity in early pregnancy is related to less robust placental growth and development by mid-pregnancy. This finding is in the opposite direction from our hypothesis, yet it is important to note that it did not appear to be pathological – participants who self-reported at least 150 minutes/week of 1^st trimester MVPA had a significantly lower rate of APOs than those who self-reported no weekly MVPA. We speculate this finding may be due to the timing of the PAPP-A measurement. In the 2^nd trimester, PAPP-A concentrations are similar between individuals who ultimately develop healthy and adverse pregnancy outcomes; low PAPP-A is adverse in the 1^st trimester, but high PAPP-A is adverse in the 3^rd trimester.^36,40 It is possible that the lower 2^nd trimester PAPP-A concentrations we found in those engaging in physical activity may be indicative of physical activity eliciting a protective effect against high late-pregnancy concentrations. While very little is known regarding the regulation and control of PAPP-A during pregnancy, outside of pregnancy it is produced by several tissue types and is upregulated by inflammatory factors.³⁶ Given that one benefit of physical activity is reduced inflammation⁴¹, it is possible that the lower concentrations of 2^nd trimester PAPP-A seen in those participants reporting higher physical activity is due to this mechanism. Additional research examining the mechanisms of this association is warranted.

Angiogenic balance

All significant associations examining 1^st trimester CVH were with 1^st trimester biomarkers of angiogenic balance, suggesting that while early-pregnancy CVH may play a role in angiogenesis in the first several weeks of gestation, it may not be directly related to mid- pregnancy angiogenic balance. However, prior work has demonstrated that components of CVH, themselves, are subject to change across pregnancy.⁴² Therefore, as we did not examine changes in CVH across pregnancy, mid-pregnancy associations between health behaviors or composite CVH with biomarkers of angiogenic balance cannot be ruled out.

Interestingly, we found that healthier 1^st trimester composite CVH scores were associated with a more anti-angiogenic 1^st trimester profile, manifesting as lower concentrations of VEGF and higher concentrations of sEng. This was in the opposite direction as we hypothesized. VEGF is an angiogenic protein that is active in the early stages of placental development, and is also known to regulate vascular permeability.⁴³ Correlates of early pregnancy VEGF have not been extensively described; this finding will need to be replicated in future studies to confirm its validity. Further, higher concentrations of sEng in early- and mid- pregnancy are associated with higher risk for preeclampsia.^44–46 However, limited data suggest that certain components of CVH, namely BMI and smoking, are inversely related to sEng with higher concentrations in those with lower BMI and less smoking.⁴⁷ Prior work in this cohort found natriuretic peptide concentrations that would be considered adverse outside of pregnancy were, in fact, correlated with lower risk of APO and future maternal hyptertension;⁴⁸ we speculate that sEng concentrations may be elevated in a similarly physiologically appropriate fashion in the first trimester, but this result must be confirmed in future studies.

In line with our hypothesis, we found that more self-reported 1^st trimester physical activity was associated with a pro-angiogenic profile via higher VEGF and lower sFlt-1 concentrations in the 1^st trimester. Higher VEGF concentrations have been detected in term placenta tissue in active versus inactive pregnant people;⁴⁹ our finding suggests this difference may be present as early as the 1^st trimester. Additionally, we confirmed the findings of Whorton, et al.²⁰, who found significantly lower sFlt-1 concentrations in those above versus below 150 min/week MVPA. Our results expand on this finding, as we demonstrated that the lowest concentration of sFlt-1 was seen in those reporting >300 min/week MVPA and that those reporting 150-299 min/week were not significantly different from those reporting no weekly MVPA. This suggests that physical activity may indeed be a potential mechanism to reduce the antiangiogenic sFlt-1 concentration in early pregnancy²⁰, but it may require high volumes of MVPA to reach this target.

A similar sFlt-1 result was found in the latent class analysis, suggesting that physical activity was likely driving the differences found between latent classes for this biomarker. Only the Inactive with Poor Diet class demonstrated significantly higher sFlt-1 concentration than the Healthy Behaviors class, despite the Less Healthy Behaviors class displaying the lowest mean composite CVH score. This indicates that low physical activity, the unique feature of the Inactive with Poor Diet class, may have been the predominant contributor to this finding.

Strengths and Limitations

This study is bolstered by several strengths. We used a diverse national cohort of nulliparous individuals, in which robust data were collected across pregnancy. This allowed us to examine both cross-sectional and longitudinal associations between CVH and placental biomarkers in the 1^st and 2^nd trimesters, which is of particular importance given the dynamic nature of placental development throughout pregnancy. This is also the first examination to our knowledge of the ways in which pregnant individuals engage in health behaviors in combination, which provides novel data that can be translated into behavioral interventions.

Certain limitations are acknowledged. The sample for this analysis was comprised of all cases of APO and a random sample of remaining participants with uncomplicated full-term pregnancies from the parent study. While we used sampling weights to account for this, significant differences remained between our weighted sample and the full cohort. Multiple comparisons may have led to a Type I error and paradoxical results suggest these findings should be interpreted with caution, though we used an a priori cutoff of p<0.01 to mitigate this risk. By only analyzing those whose pregnancies continued beyond 20 weeks of gestation our results may be impacted by survivor’s bias. Health behaviors were all self-reported which may have introduced recall bias, although participants were not aware of biomarker concentrations at the time of behavior recall so non-random bias should be minimal. The diet assessment referenced the 3 months prior to conception and may not have been representative of 1^st trimester dietary intake. Finally, we did not adjust for medications; while it is possible that aspirin therapy could have impacted our results, prior analysis in this cohort found less than 2% of participants reported aspirin use during their pregnancy.⁵⁰