Associations between hematologic dynamics during pregnancy and obstetric complications: A retrospective observational study

This is an uncorrected proof. Figures Abstract Background Pregnancy alters hematologic state as measured by complete blood count (CBC), but the longitudinal changes in CBC indices that define healthy pregnancies are not well established. In a large cohort based at an academic health system in the United States, we aimed to define reference intervals and typical longitudinal changes in CBC indices during pregnancy. We then tested for associations between extreme CBC values for gestational age or extreme longitudinal changes in CBC indices and obstetric complications. Methods and findings We studied nine CBC indices in individuals with singleton pregnancies who delivered after 30 weeks’ gestation and presented for prenatal care prior to 20 weeks. The electronic health record (EHR)-based Maternal Health Cohort (Massachusetts General Hospital; 1998–2016) formed our discovery cohort of 45,992 pregnancies, 18% of which had relevant complications. We developed a validation cohort of 48,868, 27% with complications from EHR data in the Mass General Brigham healthcare system from 2016 to 2024. In pregnancies without complications in the discovery cohort, we derived gestational-age-specific reference intervals (2.5th–97.5th percentile) and established typical intra-pregnancy longitudinal changes. In the validation cohort, we then tested CBC values outside of the 26–29 weeks’ gestation reference interval and CBC rare changes (uncommon changes in magnitude and direction) between 7–14 and 26–29 weeks’ gestation for association with a composite outcome (hypertensive disorders of pregnancy, small for gestational age birthweight, preterm birth) and its individual components using generalized estimating equations. Derived reference intervals differed from those in the literature for mean red cell volume, mean red cell hemoglobin, red cell count, and mean red cell hemoglobin concentration; reference intervals for other indices were similar to those previously published. In validation, hematocrit, hemoglobin, and red cell count values above their gestational-age specific reference intervals were associated with increased risk of the composite obstetric outcome: odds ratios (ORs) of 1.4 (95% CI [1.2, 1.5 p 0.67 g/dL) or red cell count (>0.07 106/mm3) between 7–14 weeks’ and 26–29 weeks’ gestation were associated with increased risk for preterm birth, OR for hemoglobin 1.9 (95% CI [1.5, 2.5] p 1.8% in HCT was associated with the composite outcome (1.6 95% CI [1.3, 1.9], as was an increase of >0.67 g/dL in HGB (2.0 95% CI [1.6, 2.5]), an increase >0.07 106/mm3 in RBC (1.9 95% CI [1.6, 2.4]), a decrease of less than 0.05 103/µl or an increase in PLT (1.2 95% CI [1.1, 1.3]), and a decrease in MCV greater than 0.75 fL (1.3 95% CI [1.2, 1.4], Tables 2 and S14). No significant associations were found between rare longitudinal changes and peripartum transfusion (S14 Table). Consistent results were also found in the validation cohort using the thresholds defined in the discovery cohort (Tables 2 and S15). For the composite outcome, an increase of >1.8% in HCT was significantly associated with an OR of 1.5 (95% CI [1.3, 1.7] p 0.67 g/dL in HGB, with an OR of 1.5 (95% CI [1.3, 1.8] p 0.07 106/mm3 in RBC had an OR of 1.7 (95% CI [1.5, 2.0] p 7 weeks for both the discovery cohort and the validation cohort (7.9 weeks for HGB and 7.5 for RBC). We evaluated the robustness of the thresholds identified in the discovery cohort through three sensitivity analyses. These included: (1) expanding the cohort to include pregnancies complicated by stillbirth or blood disorders (S16 Table); (2) additionally adjusting for autoimmune conditions, infections, and smoking status (S17 Table); and (3) restricting the analysis to pregnancies not exposed to medications that could plausibly affect CBC dynamics (S18 Table). Across all analyses, results were consistent with the primary findings (Fig F in S1 Appendix). Because iron supplementation initiated after a diagnosis of anemia at 7–14 weeks’ gestation could potentially influence CBC values, we also assessed the PPV of the identified thresholds separately in pregnancies with and without anemia. PPVs were consistent across these groups, indicating that the predictive performance of the thresholds was not driven by early anemia or its treatment (S19 Table). Discussion Here, we derive gestational age-specific reference intervals for CBC indices and their changes during pregnancy in a large retrospective study of more than 94,000 pregnancies. Using these reference intervals, we found that pregnancies in our study cohort with elevations (either absolute or relative to baseline) in certain CBC indices were at significantly greater risk of obstetric complications. In particular, elevations in red blood cell-related indices were associated with increased risk of subsequent HDP and preterm birth. We also found a doubling in risk of preterm birth for pregnancies with larger increases in HGB or RBC between 7–14 and 26–29 weeks’ gestation. These findings highlight an opportunity to improve risk stratification for pregnant individuals by systematically assessing longitudinal changes in routinely available CBC indices. The reference intervals identified in our cohort were noticeably wider than previously published trimester-specific intervals for RBC, MCV, MCH, and MCHC [3,4,12,18,19,40]. These differences may in part reflect the smaller size of previous study cohorts [3,4,28]. Recent larger studies outside the US found wider intervals for MCV and MCH, consistent with our results [5,41,42]. It is noteworthy that this study cohort’s HGB reference intervals include values low enough to be considered anemic according to some current guidelines [30]. This difference may arise because current HGB thresholds for clinical management of anemia were, in some cases, derived from studies in which participants were iron-supplemented, potentially affecting their applicability to all pregnant individuals [43]. The absence of association in our study between decreased red cell indices and adverse outcomes may be surprising and may have been affected by the fact that individuals with low HGB or HCT values in our study were likely supplemented with iron based on current recommendations [44–46]. However, associations between increasing red cell indices and adverse complications are not driven by early anemia or iron supplementation, as indicated by our sensitivity analysis on PPVs of pregnancies with and without anemia. Overall, we found that red cell values above the reference interval were significantly associated with adverse outcomes compared to values below. The findings that elevations in HCT, HGB, and RBC were associated with a higher risk of preeclampsia, HDP, SGA, and preterm birth are consistent with a smaller study, which found that higher first-trimester HGB concentrations were associated with an increased risk of gestational hypertension, preeclampsia, preterm birth, and SGA [47]. Three other smaller studies have linked higher HGB concentrations to stillbirths, HDP, and lower birth weight [8,48,49]. We speculate that high HGB values might be a sign of hemoconcentration or might be associated with abnormal placentation, both of which have been hypothesized to be involved in the pathophysiology of HDP, SGA, and preterm birth [47,50]. We also found that, although not significant after multiple testing, elevated platelet (PLT) count was nominally significantly associated with the composite outcomes and HDP, in line with previous findings on platelets dynamics during pregnancy [17]. Current guidelines on CBC interpretation during pregnancy focus primarily on low values, which in our analysis were not significantly associated with adverse outcomes. Previous studies suggest that the timing of assessment may influence observed associations [17,51]. Our results, derived from a narrower gestational window than most previous studies, indicate that it might be beneficial to consider high CBC values at this specific stage of pregnancy as early indications of subsequent complications. Because CBCs are already routinely collected, incorporating the evaluation of high values may offer a cost-effective approach to detecting high-risk pregnancies. CBC changes during pregnancy were often substantial, for instance with HGB and RBC for most pregnancies decreasing from baseline by more than twice what is expected for normal biological variation outside pregnancy [31]. These changes are consistent with those reported in previous studies, which were smaller or cross-sectional [3,15,17–19,21–23]. By assessing the magnitude and direction of a change in a CBC index, each patient is used as their own baseline, an approach which has been shown to be informative outside pregnancy [7]. The intra-pregnancy CBC changes reported in this study can help identify deviations that are not likely to be detected if the absolute CBC results remain within reference. Indeed, most pregnancies with rare changes and subsequent complications had CBC indices within the 26–29 weeks’ reference interval. We found that the rare increases in RBC or HGB between 7–14 and 26–29 weeks’ gestation were associated with increased risk of the examined pregnancy complications, including preterm birth. The majority of pregnancies without complications had a decrease in RBC and HGB during this gestational period, presumably reflecting modulations of red cell mass and plasma volume typical of healthy pregnancies. The association of increases in red cell dynamics with complications may therefore be related to hemoconcentration, which has been suggested to be associated with obstetric complications in previous studies [47,50,52]. We also found an association between an extreme decrease in MCV and the composite outcome, and this decreased MCV may reflect sub-clinical iron deficiency that may have been undetected and untreated, but further studies assessing iron status are necessary. Our study has several strengths. First, we defined gestational-age-specific intervals in a cohort in the US that is significantly larger than those typically used in prior studies that did not involve meta-analysis [3,4,28]. Second, we systematically analyzed longitudinal intra-pregnancy CBC dynamics and explored their use as prognostic factors for pregnancy complications. Third, our validated definitions of pregnancy complications were derived using rigorous methods and are likely more reliable than those based solely on administrative data [24]. Fourth, our key findings were validated in an out-of-sample cohort, reducing the possibility of type I error. Our study also has limitations. First, it was retrospective, and our discovery cohort was limited to pregnancies with prenatal care starting no later than 20 weeks’ of gestation, and included some pregnancies with anemia or other pathologies. While our sensitivity analyses did not show major changes when anemic individuals were excluded, further investigations are warranted. Second, our findings regarding CBC-associated risk considered only uncomplicated pregnancies or those that developed complications after 29 weeks’ gestation, and therefore these results may not generalize beyond this target population to pregnancies with complications occurring prior to 29 weeks. Third, we evaluated typical longitudinal changes in pregnancy at three time points, but more frequent time sampling under controlled conditions could be more accurate. In particular, because the discovery cohort was comprised of individuals who received obstetric care in our health system, not all participants had pre-pregnancy CBC data, which limited our sample size for examining typical longitudinal changes in CBC indices that occur between pre-pregnancy and the first trimester. Thus, our analysis was limited to the analysis of out-of-range and rare behavior of CBCs at prenatal visits with higher CBC coverage (7–14 and 26–29 weeks’ gestation). Fourth, discovery and validation cohorts spanned different time periods with evolving practices and diagnostic coding, leading to differences in prevalences and possible misalignment [38]. However, the consistent replication of associations across cohorts suggests that the identified biological signal is robust. Fifth, we evaluated univariate associations between CBC and obstetric complications, and it is possible that a multivariate approach could lead to stronger findings. Finally, the study population and chosen time windows might not accurately reflect obstetric practices outside of our health system. In this study, we derive gestational-age-specific reference intervals for CBC indices and establish typical longitudinal changes in CBC indices across gestation in a large US-based cohort. We find that elevations and rare longitudinal increases in red cell indices are associated with complications of pregnancy. Future work is needed to determine whether these findings can help shed light on the physiologic processes that contribute to obstetric complications. Future prospective studies should determine whether assessment of longitudinal changes in CBC indices can be integrated into clinical practice to improve risk stratification. Supporting information S1 Appendix. Supplementary Methods. Text A. List of hematology-related problems used for exclusion. Text B. Detail on instruments used to measure Complete Blood Counts. Text C. Detail on inclusion criteria for routine pre-pregnancy CBCs. Text D. Estimation of body mass index (BMI) in discovery and validation cohorts. Text E. Mathematical formulation of mixed-effect model for longitudinal dynamics. Text F. Calculation of biological variation. Text G. Definition of hypertensive disorders of pregnancy. Text H. Definition of small for gestational age. Text I. Mathematical formulation for association analysis with adverse outcomes. Text J. Rare dynamic definition. Text K. List of phecodes used to detect pre-existing and novel autoimmune conditions and perinatal infections. Text L. List of medications not likely to affect complete blood counts used to define subset for sensitivity analysis. Fig A. 95% intervals of complete blood count (CBC) indices of interest on the Siemens Advia 2120, Sysmex XE-5000 instruments, and Sysmex XN-9000 instruments. Discrepancies between machines were so great for MPV that it was subsequently excluded from all further analysis. Abbreviations: HCT,hematocrit; HGB, hemoglobin; WBC, white cell count; RBC, red cell count; PLT, platelet count; MCV, mean corpuscular volume; MCH, mean corpuscular hemoglobin; MCHC, mean corpuscular hemoglobin concentration; MPV, mean platelet volume. Fig B. Frequency of available CBCs by gestational week in discovery cohort pregnancies. Shaded red areas indicate chosen windows for which we considered CBCs in gestation (7–14 weeks and 26–29 weeks). The pre-delivery time point is not shaded as it is individualized and fell between 30 and 41 weeks gestation. Fig C. Sensitivity analysis of the effects of including patients with a diagnosis of anemia on reference interval determination. The figure compares gestational-age-specific intervals for all term pregnancies without complications (blue error bars), pregnancies with no diagnosis of anemia (<11 g/dL HGB at PP, 7–14 and PD, and <10.5 g/dL at 26–29 weeks, orange error bars), pregnancies considered anemic (green error bars), and literature trimester-specific intervals (red error bars), all in the discovery cohort. Up to 15% of pregnancies met criteria for anemia at each time point. Abbreviations: HCT, hematocrit, HGB, hemoglobin, WBC, white cell count; RBC, red cell count; PLT, platelet count; MCV, mean corpuscular volume; MCH, mean corpuscular hemoglobin; MCHC, mean corpuscular hemoglobin concentration; MPV, mean platelet volume. Fig D. Sensitivity analysis of the effects of including patients with recorded iron supplementation on reference interval determination. The figure compares gestational-age-specific intervals for all term pregnancies without complications (blue error bars), pregnancies with no record of either oral or IV iron supplementation in their medical record (orange bars), and pregnancies with recorded oral or IV iron supplementation (green error bars, all in the discovery cohort. N = 1,503 of 37,709 term pregnancies without complications had recorded IV or oral iron in their medical record. Abbreviations: HCT, hematocrit; HGB, hemoglobin; WBC, white cell count; RBC, red cell count; PLT, platelet count; MCV, mean corpuscular volume; MCH, mean corpuscular hemoglobin; MCHC, mean corpuscular hemoglobin concentration; MPV, mean platelet volume. Fig E. Sensitivity analysis of the effects of including patients with chronic or pregnancy-related conditions on reference intervals. Comparison of gestational-age-specific intervals for term pregnancies without complications in discovery, pregnancies in individuals without chronic or pregnancy-related conditions (N = 1,460) in discovery and literature trimester-specific intervals. Abbreviations: HCT, hematocrit; HGB, hemoglobin; WBC, white cell count; RBC, red cell count; PLT, platelet count; MCV, mean corpuscular volume; MCH, mean corpuscular hemoglobin; MCHC, mean corpuscular hemoglobin concentration; MPV, mean platelet volume. Fig F. Odds ratio comparison between main analysis and sensitivity analyses for out of range associations (panel A) and rare dynamics (panel B). Circled in black those associations that were significant in both main and sensitivity analysis. Depicted odds ratios are limited to those previously found significant in the main analysis. The hemoglobin (HGB) outlier can be attributed to a larger confidence interval [1.49, 26.49] that no longer meets Bonferroni correction, but is still nominally significant. Abbreviations: HCT, hematocrit, HGB, hemoglobin, WBC, white cell count; RBC, red cell count; PLT, platelet count; MCV, mean corpuscular volume; MCH, mean corpuscular hemoglobin; MCHC, mean corpuscular hemoglobin concentration; MPV, mean platelet volume; HDP, hypertensive disorder of pregnancy; SGA, small for gestational age; O.R., Odds ratio. Fig G. Histogram of 50 most frequent PheCodes for visits corresponding to the pre-pregnancy CBCs considered. Considered pre-pregnancy CBCs are described in Text C. https://doi.org/10.1371/journal.pmed.1004747.s002 (DOCX) S1 Table. The number of total pregnancies and the number of pregnancies with various complications in both the discovery and validation cohorts. Pregnancies with a first elevated blood pressure before 29 weeks’ gestation are excluded from the group with a CBC at 26–29 weeks’ gestation and the group with CBCs at both 7–14 and 26–29 weeks’ gestation. Abbreviations: CBC, complete blood count; HDP, hypertensive disorders of pregnancy. https://doi.org/10.1371/journal.pmed.1004747.s003 (XLSX) S2 Table. Comparison of demographics included/excluded pregnancies during the development of discovery and validation cohorts. Abbreviations: BMI, body mass index. https://doi.org/10.1371/journal.pmed.1004747.s004 (XLSX) S3 Table. Tabular version of reference intervals derived from discovery cohort and literature intervals. https://doi.org/10.1371/journal.pmed.1004747.s005 (XLSX) S4 Table. Beta coefficients of CBC dynamics in pregnancy. Beta coefficients measure the effect of gestational age on CBC values and were estimated on the subset of pregnancies with all four time points available (N = 2,791). Random effects were added for the individual pregnancy, the individual, parity, OB-GYN site of care, and for year of delivery. P-values are calculated from the t-distribution using t-statistics generated from calculated beta cofficients divided by the calculated standard error for each coefficient. The p-value threshold with Bonferroni correction for multiple testing was p = 0.002. https://doi.org/10.1371/journal.pmed.1004747.s006 (XLSX) S5 Table. Percentile changes of CBC indices in pregnancy. PP is pre-pregnancy, PD is pre-delivery. https://doi.org/10.1371/journal.pmed.1004747.s007 (XLSX) S6 Table. Rare CBC changes (<5%) during pregnancy occur for red blood cell count, hematocrit, hemoglobin, platelet count, and mean corpuscular volume. Percentages of pregnancies with decreasing, stable, or increasing CBC indices between subsequent timepoints for analysis of common and uncommon behavior in blood dynamics during pregnancy are shown. We considered a difference between two timepoints to be stable if the absolute value of the change was less than the biological variation for that index, increasing if it was positive and greater than biological variation, and decreasing if it was negative and greater in magnitude than biological variation (see S6 Text for further details). Additionally, percentage of pregnancies with stable (within biological variation) indices throughout gestation are reported in the last column. The CBC indices for which fewer than 5% of pregnancies remain stable are bolded. https://doi.org/10.1371/journal.pmed.1004747.s008 (XLSX) S7 Table. Beta coefficients are consistent when inferred on all available pregnancies with CBCs for at least two consecutive time points. Beta coefficients were inferred on the subset of pregnancies without complications for which at least one consecutive pair of time points was available (N = 28,450) and compared to the beta coefficients in the main analysis. P-values are calculated from the t-distribution using t-statistics generated from calculated beta cofficients divided by the calculated standard error for each coefficient. P-values were corrected for multiple testing with Bonferroni correction so that p < 0.002 was considered significant. All absolute differences in beta coefficients between sensitivity and main analysis fall within 2× biological variation. https://doi.org/10.1371/journal.pmed.1004747.s009 (XLSX) S8 Table. Elevated CBC index values are associated with adverse pregnancy outcomes. Pregnancies with a CBC at 26–29 weeks’ gestation were considered (N = 34,159), and those with abnormal tests for HDP or preeclampsia diagnosis before 29 weeks’ gestation were excluded (N = 254). The odds ratios of developing examined complications (composite of HDP, SGA, and preterm delivery; each of those adverse events individually; and need for transfusion at or after delivery) are shown in table for a CBC index value above, outside, or below the study intervals at 26–29 weeks’ gestation. Significant odds ratios are marked with a 1 in the corresponding column. Prevalence and positive predictive value (PPV) are also reported. P-values are generated from Wald tests. Significance was evaluated with a Bonferroni-corrected p-value of 0.0003. Abbreviations: PPV, positive predictive value; HDP, hypertensive disorder of pregnancy; SGA, small for gestational age. https://doi.org/10.1371/journal.pmed.1004747.s010 (XLSX) S9 Table. Parity, and to a lesser extent maternal age, BMI, and public insurance status, were significantly associated with predicting the composite outcome in out-of-range analysis of discovery cohort. Pregnancies with a CBC at 26–29 weeks’ gestation were considered (N = 34,159), and those with abnormal tests for HDP or preeclampsia diagnosis before 29 weeks’ gestation were excluded (N = 254). OR and p-values for all covariates used in our logisitic regression for each CBC index are reported in the table. P-values are generated from Wald tests. Abbreviations: BMI, Body Mass Index; OR, Odds ratio. https://doi.org/10.1371/journal.pmed.1004747.s011 (XLSX) S10 Table. Odds ratios and PPVs for extreme CBC index values are consistent in an out-of-sample validation cohort. Pregnancies with a CBC at 26–29 weeks’ gestation were considered (N = 48,490), and those with abnormal tests for HDP or preeclampsia diagnosis before 29 weeks’ gestation were excluded (N = 387). We analyzed associations between adverse outcomes and the presence of a CBC index above, outside, or below the study reference intervals derived in discovery at 26–29 weeks’ gestation. Only combinations of CBC index/adverse outcomes that were at least nominally significant in the discovery cohort were considered for this analysis. In the table, we report ORs, prevalence, positive predictive value (PPV), and negative predictive value (NPV). P-values are generated from Wald tests. Significance was evaluated with a Bonferroni corrected p-value of 0.004 and significant ORs in validation are marked with a 1 in the corresponding column. https://doi.org/10.1371/journal.pmed.1004747.s012 (XLSX) S11 Table. Sensitivity analysis of associations on an expanded discovery cohort including pregnancies ending in stillbirth and blood disorders previously excluded yielded near identical OR in out-of-range analysis. In this sensivitiy analysis, the discovery cohort was re-derived to include pregnancies ending in stillbirth and blood disorders (N = 41,153). Pregnancies ending in stillbirth were labelled as having the composite outcome, and marked as preterm or HDP when applicable. The presence of a blood-related disorder was added as a binary covariate to our logistic regression. Odds ratios, positive predictive values and negative predictive values for each CBC index, outcome pair are listed in the table, as is the prevalence of each outcome. P-values are generated from Wald tests. Significance is denoted if p < 0.0003. Abbreviations: PPV, positive predictive value; NPV, negative predictive value; HDP, hypertensive disorder of pregnancy; SGA, small for gestational age. https://doi.org/10.1371/journal.pmed.1004747.s013 (XLSX) S12 Table. Sensitivity analysis of associations in the discovery cohort (N = 34,159) including the presence/absence of smoking, auto-immune disorders, and infections during pregnancy yielded near identical OR in out-of-range analysis. In this sensitivity analysis, information on infections during pregnancy, pre-existing auto-immune disorders and smoking status was acquired from the electronic health record, and binary covariates were developed reflecting their presence or absence (see S11 Text for details). Odds ratios, positive predictive values, and negative predictive values for each CBC index, outcome pair are listed in the table, as is the prevalence of each outcome. P-values are generated from Wald tests. Significance is denoted if p < 0.0003. Abbreviations: PPV, Positive Predictive value; HDP, Hypertensive disorder of pregnancy; SGA, Small for gestational age. https://doi.org/10.1371/journal.pmed.1004747.s014 (XLSX) S13 Table. Sensitivity analysis of associations in a subset of the discovery cohort (N = 6,709) with no medications that may cause hematological change yielded similar OR in out-of-range analysis. In this sensitivity analysis, information on medications prescribed during pregnancy (conception to 30 weeks’ gestation) was acquired from the electronic health record, and a subset of the discovery cohort was developed of pregnancies without medications that could plausibly cause changes to blood cells (see S12 Text for details on medications). Odds ratios, positive predictive values, and negative predictive values for each CBC index, outcome pair are listed in the table, as is the prevalence of each outcome. P-values are generated from Wald tests. Significance is denoted if p < 0.0003. Abbreviations: PPV, positive predictive value; HDP, hypertensive disorder of pregnancy; SGA, small for gestational age. https://doi.org/10.1371/journal.pmed.1004747.s015 (XLSX) S14 Table. Pregnancies with rare longitudinal changes between 7−14 and 26−29 weeks’ gestation are at higher risk of complications. Assocations were tested in N = 29,416 after which those with preterm birth at <30 weeks’ gestation or evidence of HDP or preeclampsia before 29 weeks were excluded (N = 297). Rare change thresholds are reported in the following units: HGB (g/dL), HCT (%), RBC (106/mm3), PLT (103/µL), MCV (fL), WBC (103/µL), RDW (%), MCH (pg), MCHC (g/dL). In the table, we report the odds ratio, the number of pregnancies with the rare behavior (Total N), the number of complicated pregnancies with a rare behavior (Complicated N), the number of complicated pregnancies with a rare behavior that fall within the reference interval at 26−29 weeks’ gestation (Complicated N in range), and their percentage and associated confidence interval. We also report the prevalence of the condition in the dataset, the positive predictive value (PPV), and the negative predictive value (NPV). P-values are generated from Wald tests. Threshold for statistical significance of the odds ratio was p < 9*10−6 with Bonferroni correction. Abbreviations: HDP, hypertensive disorders of pregnancy; SGA, small for gestational Age; PPV, positive predictive value; NPV, negative predictive value; OR, odds ratio; CI, confidence interval; https://doi.org/10.1371/journal.pmed.1004747.s016 (XLSX) S15 Table. Odds ratios and PPVs for rare behaviors are consistent in an out-of-sample validation cohort. Out-of-sample validation of rare dynamics for hemoglobin HGB (g/dL), red blood cell count RBC (106/mm3), hematocrit HCT (%), platelet count PLT (103/µL), and mean red cell volume MCV (fL) shows odds ratios, PPV, and NPV consistent with the discovery cohort. Associations were tested in N = 48,490 pregnancies (with N = 378 preterm births at <30 weeks’ gestation or pregnancies with evidence of HDP or preeclampsia before 29 week excluded). In the table, we report the odds ratio, the number of pregnancies with the rare behavior (Total N), the number of complicated pregnancies with a rare behavior (Complicated N), the number of complicated pregnancies with a rare behavior that fall within the reference interval at 26–29 weeks’ gestation (Complicated N in range), and their percentage and associated confidence interval. We also report the prevalence of the condition in the dataset, the positive predictive value (PPV), and the negative predictive value (NPV). We also show the prevalence of each adverse outcome and the associated confidence interval obtained via bootstrapping. P-values are generated from Wald tests. Threshold for statistical significance of the odds ratio was p < 0.003 with Bonferroni correction. https://doi.org/10.1371/journal.pmed.1004747.s017 (XLSX) S16 Table. Sensitivity analysis of associations on an expanded discovery cohort including pregnancies ending in stillbirth and blood disorders previously excluded yielded near identical OR in rare behavior analysis. In this sensivitiy analysis, the discovery cohort was re-derived to include pregnancies ending in stillbirth and blood disorders (N = 29,775). Pregnancies ending in stillbirth were labelled as having the composite outcome, and marked as preterm or HDP when applicable. The presence of a blood-related disorder was added as a binary covariate to our logistic regression. In the table, we report the odds ratio, the number of pregnancies with the rare behavior (Total N), the number of complicated pregnancies with a rare behavior (Complicated N), the number of complicated pregnancies with a rare behavior that fall within the reference interval at 26–29 weeks’ gestation (Complicated N in range), and their percentage and associated confidence interval. We also report the prevalence of the condition in the dataset, the positive predictive value (PPV), and the negative predictive value (NPV). We also show the prevalence of each adverse outcome and the associated confidence interval obtained via bootstrapping. Threshold for statistical significance of the odds ratio was p < 0.0009 with Bonferroni correction. Abbreviations: PPV, positive predictive value; HDP, hypertensive disorder of pregnancy; SGA, small for gestational age. https://doi.org/10.1371/journal.pmed.1004747.s018 (XLSX) S17 Table. Sensitivity analysis of associations in the discovery cohort (N = 34,159) including the presence/absence of smoking, auto-immune disorders, and infections during pregnancy yielded near identical OR in rare behavior analysis. In this sensitivity analysis, information on infections during pregnancy, pre-existing auto-immune disorders, and smoking status was acquired from the electronic health record, and binary covariates were developed reflecting their presence or absence (see S11 Text for details. In the table, we report the odds ratio, the number of pregnancies with the rare behavior (Total N), the number of complicated pregnancies with a rare behavior (Complicated N), the number of complicated pregnancies with a rare behavior that fall within the reference interval at 26–29 weeks’ gestation (Complicated N in range), and their percentage and associated confidence interval. We also report the prevalence of the condition in the dataset, the positive predictive value (PPV), and the negative predictive value (NPV). We also show the prevalence of each adverse outcome and the associated confidence interval obtained via bootstrapping. P-values are generated from Wald tests. Threshold for statistical significance of the odds ratio was p < 0.0009 with Bonferroni correction. Abbreviations: PPV, positive predictive value; HDP, hypertensive disorder of pregnancy; SGA, small for gestational age. https://doi.org/10.1371/journal.pmed.1004747.s019 (XLSX) S18 Table. Sensitivity analysis of associations in a subset of the discovery cohort (N = 5,764) with no medications that may cause hematological change yielded similar OR in rare behavior analysis. In this sensitivity analysis, information on medications prescribed during pregnancy (conception to 30 weeks’ gestation) was acquired from the electronic health record, and a subset of the discovery cohort was developed of pregnancies without medications that could plausibly cause changes to blood cells (see S12 Text for details on medications). In the table, we report the odds ratio, the number of pregnancies with the rare behavior (Total N), the number of complicated pregnancies with a rare behavior (Complicated N), the number of complicated pregnancies with a rare behavior that fall within the reference interval at 26–29 weeks’ gestation (Complicated N in range), and their percentage and associated confidence interval. We also report the prevalence of the condition in the dataset, the positive predictive value (PPV), and the negative predictive value (NPV). We also show the prevalence of each adverse outcome and the associated confidence interval obtained via bootstrapping. P-values are generated from Wald tests. Threshold for statistical significance of the odds ratio p < 0.0009 with Bonferroni correction. Abbreviations: PPV, positive predictive value; HDP, hypertensive disorder of pregnancy; SGA, small for gestational age. https://doi.org/10.1371/journal.pmed.1004747.s020 (XLSX) S19 Table. Sensitivity analysis of anemia in rare dynamics. We considered the clinical definition of anemia of hemoglobin <11 mg/dL. N(%) denotes the number of individuals with or without anemia and have rare dynamics. The percentage is calculated on the total of individuals with anemia or in those without anemia, to reflect an increased prevalence of rare dynamics in pregnancies with clinical anemia. For each group, we then evaluated the capability of detecting complications and reported the number of pregnancies with the Composite outcome, the sensitivity, and the positive predictive value (PPV). https://doi.org/10.1371/journal.pmed.1004747.s021 (XLSX) References - 1. Chandra S, Tripathi AK, Mishra S, Amzarul M, Vaish AK. Physiological changes in hematological parameters during pregnancy. Indian J Hematol Blood Transfus. 2012;28(3):144–6. pmid:23997449 - 2. Soma P, Pillay P, Nelson C, Piercy C, Tolppanen H, Mebazaa A. Physiological changes in pregnancy: review articles. Cardiovasc J Afr. 2016;27:89–94. - 3. Lockitch G. Handbook of diagnostic biochemistry and hematology in normal pregnancy. Boca Raton: CRC Press; 1993. - 4. Abbassi-Ghanavati M, Greer LG, Cunningham FG. Pregnancy and laboratory studies: a reference table for clinicians. Obstet Gynecol. 2009;114(6):1326–31. pmid:19935037 - 5. Barakauskas VE, Bohn MK, Branch E, Boutin A, Albert A, Luke S, et al. Mining the gap: deriving pregnancy reference intervals for hematology parameters using clinical datasets. Clin Chem. 2023;69(12):1374–84. pmid:37947280 - 6. Bohn MK, Adeli K. Physiological and metabolic adaptations in pregnancy: importance of trimester-specific reference intervals to investigate maternal health and complications. Crit Rev Clin Lab Sci. 2022;59(2):76–92. pmid:34587857 - 7. Foy BH, Petherbridge R, Roth MT, Zhang C, De Souza DC, Mow C, et al. Haematological setpoints are a stable and patient-specific deep phenotype. Nature. 2025;637(8045):430–8. pmid:39663453 - 8. Nasiri M, Faghihzadeh S, Alavi Majd H, Zayeri F, Kariman N, Safavi Ardebili N. Longitudinal discriminant analysis of hemoglobin level for predicting preeclampsia. Iran Red Crescent Med J. 2015;17(3):e19489. pmid:26019901 - 9. Shi H, Chen L, Wang Y, Sun M, Guo Y, Ma S, et al. Severity of anemia during pregnancy and adverse maternal and fetal outcomes. JAMA Netw Open. 2022;5(2):e2147046. pmid:35113162 - 10. Ukah UV, Payne B, Karjalainen H, Kortelainen E, Seed PT, Conti-Ramsden FI, et al. Temporal and external validation of the fullPIERS model for the prediction of adverse maternal outcomes in women with pre-eclampsia. Pregnancy Hypertens. 2019;15:42–50. pmid:30825926 - 11. Martin J, Osterman M. Changes in prenatal care utilization: United States, 2019-2023. Natl Vital Stat Rep. 2023;72. - 12. Balloch AJ, Cauchi MN. Reference ranges for haematology parameters in pregnancy derived from patient populations. Clin Lab Haematol. 1993;15(1):7–14. pmid:8472501 - 13. Beguin Y, Lipscei G, Thoumsin H, Fillet G. Blunted erythropoietin production and decreased erythropoiesis in early pregnancy. Blood. 1991;78(1):89–93. pmid:2070061 - 14. Choi JW, Pai SH. Tissue plasminogen activator levels change with plasma fibrinogen concentrations during pregnancy. Ann Hematol. 2002;81(11):611–5. pmid:12454697 - 15. Dundar O, Yoruk P, Tutuncu L, Erikci AA, Muhcu M, Ergur AR, et al. Longitudinal study of platelet size changes in gestation and predictive power of elevated MPV in development of pre-eclampsia. Prenat Diagn. 2008;28(11):1052–6. pmid:18973157 - 16. Karim SA, Khurshid M, Rizvi JH, Jafarey SN, Rizwana I. Platelets and leucocyte counts in pregnancy. J Pak Med Assoc. 1992;42(4):86–7. pmid:1597923 - 17. Reese JA, Peck JD, Deschamps DR, McIntosh JJ, Knudtson EJ, Terrell DR. Platelet counts during pregnancy. N Engl J Med. 2018;379:32–43. - 18. Milman N, Bergholt T, Byg K-E, Eriksen L, Hvas A-M. Reference intervals for haematological variables during normal pregnancy and postpartum in 434 healthy Danish women. Eur J Haematol. 2007;79(1):39–46. pmid:17598837 - 19. Qvist I, Abdulla M, Jägerstad M, Svensson S. Iron, zinc and folate status during pregnancy and two months after delivery. Acta Obstet Gynecol Scand. 1986;65(1):15–22. pmid:3716775 - 20. Horowitz K, Ingardia C, Borgida A. Anemia in pregnancy – ClinicalKey. Clin Lab Med. 2013;33:281–91. - 21. Milman N, Graudal N, Nielsen OJ, Agger AO. Serum erythropoietin during normal pregnancy: relationship to hemoglobin and iron status markers and impact of iron supplementation in a longitudinal, placebo-controlled study on 118 women. Int J Hematol. 1997;66(2):159–68. pmid:9277046 - 22. Romslo I, Haram K, Sagen N, Augensen K. Iron requirement in normal pregnancy as assessed by serum ferritin, serum transferrin saturation and erythrocyte protoporphyrin determinations. Br J Obstet Gynaecol. 1983;90(2):101–7. pmid:6824608 - 23. Bar A, Mendelsohn-Cohen N, Kohanim YK, Mayo A, Toledano Y, Alon U. Pregnancy and postpartum dynamics revealed by an atlas of millions of lab tests. bioRxiv. 2023;:2023.05.11.540359. - 24. Hsu S, Selen DJ, James K, Li S, Camargo CA Jr, Kaimal A, et al. Assessment of the validity of administrative data for gestational diabetes ascertainment. Am J Obstet Gynecol MFM. 2023;5(2):100814. pmid:36396038 - 25. Maya J, Selen DJ, Thaweethai T, Hsu S, Godbole D, Schulte CCM, et al. Gestational glucose intolerance and birth weight-related complications. Obstet Gynecol. 2023;142(3):594–602. pmid:37539973 - 26. Selen DJ, Thaweethai T, Schulte CCM, Hsu S, He W, James K, et al. Gestational glucose intolerance and risk of future diabetes. Diabetes Care. 2023;46: 83–91. - 27. Fraser C. Biological variation: from principles to practice. Washington DC: AACC Press; 2001. - 28. Cunningham FG. Normal reference ranges for laboratory values in pregnancy – UpToDate. In: UpToDate [Internet]. Jan 2023 [cited 9 Feb 2023]. Available from: https://www-uptodate-com.treadwell.idm.oclc.org/contents/normal-reference-ranges-for-laboratory-values-in-pregnancy?search=reference%20ranges%20in%20pregnancy&source=search_result&selectedTitle=1~150&usage_type=default&display_rank=1 - 29. Normal reference ranges and laboratory values in pregnancy. In: perinatology.com [Internet]. 22 Dec 2023 [cited 22 Dec 2023]. Available from: https://www.perinatology.com/Reference/Reference%20Ranges/Reference%20for%20Serum.htm - 30. Anemia in pregnancy: ACOG practice bulletin, number 233. Obstet Gynecol. 2021;138:e55. - 31. Aarsand A, Fernandez-Calle P, Webster C, Coskun A, Gonzales-Lao E, Diaz-Garzon J, et al. The EFLM biological variation database. [cited 26 Jan 2023]. Available from: https://biologicalvariation.eu/ - 32. Halekoh U, Højsgaard S, Yan J. TheRPackagegeepackfor generalized estimating equations. J Stat Soft. 2006;15(2). - 33. Yan J. Geepack: Yet another package for generalized estimating equations. R News. 2002;2:12–4. - 34. Yan J, Fine J. Estimating equations for association structures. Stat Med. 2004;23(6):859–74; discussion 875-7,879-80. pmid:15027075 - 35. Gestational hypertension and preeclampsia: ACOG practice bulletin, number 222. Obstet Gynecol. 2020;135(6):e237–60. pmid:32443079 - 36. Leonard SA, Kennedy CJ, Carmichael SL, Lyell DJ, Main EK. An expanded obstetric comorbidity scoring system for predicting severe maternal morbidity. Obstet Gynecol. 2020;136(3):440–9. pmid:32769656 - 37. Oken E, Kleinman KP, Rich-Edwards J, Gillman MW. A nearly continuous measure of birth weight for gestational age using a United States national reference. BMC Pediatr. 2003;3:6. pmid:12848901 - 38. Freaney PM, Harrington K, Molsberry R, Perak AM, Wang MC, Grobman W, et al. Temporal trends in adverse pregnancy outcomes in birthing individuals aged 15 to 44 years in the United States, 2007 to 2019. J Am Heart Assoc. 2022;11: e025050. - 39. Reveiz L, Gyte GM, Cuervo LG, Casasbuenas A. Treatments for iron-deficiency anaemia in pregnancy. Cochrane Database Syst Rev. 2011;2011(10):CD003094. pmid:21975735 - 40. Ahmed Y, van Iddekinge B, Paul C, Sullivan HF, Elder MG. Retrospective analysis of platelet numbers and volumes in normal pregnancy and in pre-eclampsia. Br J Obstet Gynaecol. 1993;100(3):216–20. pmid:8476825 - 41. Bakrim S, Motiaa Y, Ouarour A, Masrar A. Hematological parameters of the blood count in a healthy population of pregnant women in the Northwest of Morocco (Tetouan-M’diq-Fnideq provinces). Pan Afr Med J. 2018;29:205. pmid:30100959 - 42. Li A, Yang S, Zhang J, Qiao R. Establishment of reference intervals for complete blood count parameters during normal pregnancy in Beijing. J Clin Lab Anal. 2017;31(6):e22150. pmid:28105762 - 43. O’Brien KO, Ru Y. Iron status of North American pregnant women: an update on longitudinal data and gaps in knowledge from the United States and Canada. Am J Clin Nutr. 2017;106(Suppl 6):1647S-1654S. pmid:29070557 - 44. Gomes F, King SE, Dallmann D, Golan J, da Silva ACF, Hurley KM, et al. Interventions to increase adherence to micronutrient supplementation during pregnancy: a systematic review. Ann N Y Acad Sci. 2021;1493(1):41–58. pmid:33400303 - 45. Sununtnasuk C, D’Agostino A, Fiedler JL. Iron+folic acid distribution and consumption through antenatal care: identifying barriers across countries. Public Health Nutr. 2016;19(4):732–42. pmid:26022914 - 46. Achebe MM, Gafter-Gvili A. How I treat anemia in pregnancy: iron, cobalamin, and folate. Blood. 2017;129(8):940–9. pmid:28034892 - 47. Gaillard R, Eilers PHC, Yassine S, Hofman A, Steegers EAP, Jaddoe VWV. Risk factors and consequences of maternal anaemia and elevated haemoglobin levels during pregnancy: a population-based prospective cohort study. Paediatr Perinat Epidemiol. 2014;28(3):213–26. pmid:24506330 - 48. Amburgey OA, Ing E, Badger GJ, Bernstein IM. Maternal hemoglobin concentration and its association with birth weight in newborns of mothers with preeclampsia. J Matern Fetal Neonatal Med. 2009;22(9):740–4. pmid:19557662 - 49. Stephansson O, Dickman PW, Johansson A, Cnattingius S. Maternal hemoglobin concentration during pregnancy and risk of stillbirth. JAMA. 2000;284(20):2611–7. pmid:11086368 - 50. Chaiworapongsa T, Chaemsaithong P, Yeo L, Romero R. Pre-eclampsia part 1: current understanding of its pathophysiology. Nat Rev Nephrol. 2014;10(8):466–80. pmid:25003615 - 51. Young MF, Oaks BM, Rogers HP, Tandon S, Martorell R, Dewey KG, et al. Maternal low and high hemoglobin concentrations and associations with adverse maternal and infant health outcomes: an updated global systematic review and meta-analysis. BMC Pregnancy Childbirth. 2023;23(1):264. pmid:37076797 - 52. de Haas S, Ghossein-Doha C, van Kuijk SMJ, van Drongelen J, Spaanderman MEA. Physiological adaptation of maternal plasma volume during pregnancy: a systematic review and meta-analysis. Ultrasound Obstet Gynecol. 2017;49(2):177–87. pmid:28169502