Objective
The purpose of this study was to determine whether prenatal inflammation (as assessed by clinical chorioamnionitis, maternal temperature >38°C, or histologic chorioamnionitis) is associated with a composite adverse neonatal outcome.
Study Design
We performed a prospective cohort study of women at 22 weeks to 33 weeks 6 days’ gestation with symptoms of labor (April 2009 to March 2012). Relevant maternal and neonatal exposures and outcomes were recorded. Multivariable logistic regression was performed to determine the association between prenatal inflammation and neonatal outcomes that were controlled for potential confounders.
Results
We analyzed 871 mother-infant pairs. The preterm birth rate was 42.0%. When we controlled for infant sex and modified the data by gestational age at delivery, prenatal inflammation remains a significant risk factor for adverse neonatal outcomes, despite advancing gestational age: clinical chorioamnionitis at 32 weeks’ gestation (odds ratio [OR], 3.12; 95% confidence interval [CI], 1.02–9.52], at 36 weeks’ gestation (OR, 8.88; 95% CI, 4.32–18.25), and at 40 weeks’ gestation (OR, 25.30; 95% CI, 9.25–69.19); maternal temperature >38°C at 32 weeks’ gestation (OR, 3.18; 95% CI, 0.66–15.42), at 36 weeks gestation (OR, 8.40; 95% CI, 3.60–19.61), and at 40 weeks gestation (OR, 22.19; 95% CI, 8.15–60.44); histologic chorioamnionitis at 32 weeks gestation (OR, 1.25; 95% CI, 0.64–2.46), at 36 weeks gestation (OR, 2.56; 95% CI, 1.54–4.23), and at 40 weeks gestation (OR, 5.23; 95% CI, 1.95–13.99).
Conclusion
The protective association with advancing gestational age is diminished when prenatal inflammation is present.
Over the last 20 years, evidence has emerged to suggest that prenatal inflammation and/or intrauterine infection leads to the activation of localized inflammatory pathways, plays a critical role in at least 25-40% of spontaneous preterm births, and is a significant contributor to the development of adverse neonatal outcomes. Prenatal inflammation that predisposes to spontaneous preterm birth and adverse neonatal outcomes may occur in the form of either histologic (HCA) and/or acute clinical chorioamnionitis (CCA).
The association between acute CCA or perinatal infection and adverse neonatal outcomes (which includes sepsis, respiratory compromise, necrotizing enterocolitis, intraventricular hemorrhage, and long-term outcomes such as cerebral palsy and neonatal death ) has been well-established. However, most of these studies have been small and retrospective and have focused on the prevalence of adverse neonatal outcomes in cohorts of preterm infants. As a result, many investigators attribute the risk of adversity to gestational age at delivery rather than prenatal inflammation. Furthermore, studies that have examined the risks of inflammation to term neonates have concluded that inflammation poses only minimal risk. As a result, the relative contributions of preterm birth and prenatal inflammation to adverse neonatal outcomes are yet to be determined.
To that end, our primary objective was to determine whether prenatal inflammation that is assessed by acute CCA, maternal temperature >38°C in the 24 hours preceding delivery, or postpartum diagnosis of HCA was associated with a composite adverse neonatal outcome variable (COMP). Secondary objectives were to determine the association between prenatal inflammation and the individual adverse neonatal outcomes sepsis and respiratory compromise. Our hypotheses were that, for all outcomes, prenatal inflammation would be associated with an increased risk for neonatal adversity and that these associations would persist when adjustment was made for gestational age at delivery.
Materials and Methods
We performed a prospective cohort study at a single, urban tertiary care center. The cohort consisted of women with singleton pregnancies at 22 weeks to 33 week 6 days’ gestational age who came to the labor and delivery triage unit with complaints concerning preterm labor. Patients were excluded for multiple-gestation, major fetal anomaly, intrauterine fetal death, severe preeclampsia before enrollment, chronic steroid or immunosuppressive drug use, active immunologic disease, acute systemic febrile illness, and/or pregestational diabetes mellitus. Patients who either were not delivered at our institution or whose infants were transferred to a different hospital for care were also excluded from these analyses.
Patients were enrolled in the study by trained clinical research coordinators who obtained informed consent at the time of enrollment. Once a patient was enrolled in the study, all treatment decisions were made by the treating physician according to the standard of care at our institution. Women were enrolled from April 2009 through March 2012.
At our institution, maternal fever is diagnosed as a temperature >38°C. Maternal temperature is recorded on laboring patients every 4 hours while membranes are intact and hourly after membranes are ruptured. A diagnosis of temperature >38°C was obtained from review of the electronic nursing records in the 24 hours preceding each patient’s delivery. Acute CCA is diagnosed in the setting of maternal temperature >38°C and at least 1 of the following occurrences: maternal tachycardia (≥100 beats per minute), fetal tachycardia (>160 beats per minute), and/or fundal tenderness. Patients who receive a diagnosis of acute CCA who are not in spontaneous labor are induced at the time this diagnosis is made. HCA is a diagnosis that is made by the pathologist after microscopic examination of the placenta and is defined as the presence of neutrophils in the chorion or amnion. Pathologists were unaware of whether patients were enrolled in the study and/or whether their neonates had any adverse outcomes. If the placenta was not sent to pathology, then HCA was assumed to be not present.
Data collection
After enrollment, each patient was tracked for the remainder of her pregnancy. Relevant maternal information that included pertinent medical, surgical, and gynecologic histories was recorded; relevant delivery information was obtained through chart review. Whether antenatal corticosteroids were administered, whether HCA or CCA was diagnosed, and the highest maternal temperature within 24 hours before delivery were also noted.
Relevant neonatal information that included gestational age at delivery, mode of delivery, infant sex, birthweight, and nursery admission were recorded. Neonatal chart abstraction was performed, and the presence of key diagnoses was determined by review of the documentation of the attending neonatologist or pediatrician who assessed each infant. Such key diagnoses included presumed (diagnosis by neonatology attending based on clinical presentation for which ≥7 days of antibiotics were administered) or proved (culture positive) sepsis, respiratory compromise (need for continuous positive airway pressure or mechanical ventilation beyond the delivery room), seizures, necrotizing enterocolitis (diagnosed both clinically and radiographically), intraventricular hemorrhage (grades III-IV), and/or death before discharge. Neonatal information was limited to those events that occurred between delivery and the infant’s discharge home from the hospital.
The use of a surrogate composite variable for cases in which individual outcomes were of low prevalence has been performed in previous neonatal studies. Therefore, a COMP was created to summarize neonatal outcomes for the primary outcome of this study. A neonate was classified as having the primary composite outcome if at least one of the adverse outcomes listed earlier (sepsis, respiratory compromise, seizures, necrotizing enterocolitis, intraventricular hemorrhage, and/or death) was present. A neonate was classified as having one of the secondary outcomes if either presumed or proven sepsis or respiratory compromise were present, as defined earlier. Preterm birth was defined as delivery at <37 weeks’ gestational age.
Data analyses
Associations between categoric variables were examined with chi-squared analyses. Normality for continuous variables was assessed with the Shapiro-Wilk test for normality. Where a normal distribution could be ruled out, the Kruskal Wallis equality-of-populations rank test was used to test differences.
Although our choice of 38°C as a cut point for the variable to assess elevated maternal temperature was made (38°C represents a fever at our institution) before any analyses were performed, we verified that this cut-point was the most associated with COMP among candidate cut-points of 37, 37.5, 38, and 38.5°C in both raw models of COMP that were associated with a “yes/no” higher temperature and full models that included interaction of temperature with gestational age at delivery and adjustment by male sex (data not shown).
Multivariable logistic regression was performed to determine the association between each form of prenatal inflammation (CCA, temperature >38°C [T38], and HCA) and both composite and individual adverse neonatal outcomes (neonatal sepsis and respiratory compromise). To control for potential confounding of the association of composite neonatal outcome and prenatal inflammation, we selected potential confounders for testing based on the association with adverse neonatal outcome found in the literature, which included gestational age at presentation with preterm labor, antenatal corticosteroid treatment, maternal age, and maternal race. Modeling strategy was change-in-estimate of the inflammation variable over a multivariable baseline model of adverse outcome and inflammation that included gestational age at delivery in weeks and infant sex, which were selected because their clinical relevance. Potential confounders were examined individually and, if associated with a ≥5% change in the regression coefficient of the indicator of prenatal inflammation (ie, the “change in 5% rule”), was retained in the model. Because we believed that gestational age would modify the association between inflammation and adverse neonatal outcomes, we tested the interaction of gestational age at delivery in weeks with inflammation in association with adverse neonatal outcomes. Goodness of fit for the multivariable models was assessed with the use of the Hosmer-Lemeshow χ 2 test, where a large probability value indicates that the model is well-fit. We used a threshold of a probability value of > .05. Areas under the receiver operating characteristic curve (AUC) for each multivariable model were calculated and compared with the use of a statistical receiver-operator curve area comparison test. Multivariable models with each prenatal inflammation variable were compared with each other variable to determine whether 1 particular prenatal inflammation variable had superior discriminatory ability to the others and to verify that each multivariable model that included a prenatal inflammation variable was significantly more discriminatory than the model that included only gestational age at delivery and infant sex.
In all analyses, a probability value of < .05 was considered statistically significant, and probability value of < .10 was the threshold for display of an interaction. Stata software (version 10.1; StataCorp, College Station, TX) was used for all analyses.
Sample size
A priori sample size calculations were performed to determine how many mother-infant pairs would be needed to detect a 2.5-fold increase in the prevalence of COMP between those infants who were exposed to prenatal inflammation compared with those who were not exposed. Based on previously published data from our institution, we estimated that the prevalence of COMP among infants who were not exposed to prenatal inflammation was 6.0%. Assuming 90% power, a type I error of 5%, and an enrollment ratio of 1:4 between those with and without prenatal inflammation exposure, we determined we would require 850 mother-infant pairs.
This study was approved by the institutional review board at the University of Pennsylvania (Philadelphia, PA).
Results
Delivery information was available to analyze for 871 women and their infants. The rate of preterm birth in this high-risk cohort was 42.0%. The cohort prevalence of CCA was 6.5% (n = 57), of T38 was 5.6% (n = 49), and of HCA was 25.6% (n = 223). Most patients with CCA or T38 had other inflammatory exposures. Specifically, 87.7% of the women with CCA also had HCA, and 64.9% of the women with CCA had T38; 75.5% of the women with T38 also had CCA, and 79.6% of the women with T38 had HCA. However, only 22.4% of the women with HCA had CCA, and 17.5% of the women with HCA had T38.
We compared demographic variables among women with and without exposure to prenatal inflammation. The demographic variables that were associated significantly with HCA exposure ( Table 1 ) were preterm birth, gestational age at delivery, antenatal corticosteroids, and gestational age at enrollment. For CCA, associations were limited to antenatal corticosteroids and gestational age at enrollment, and the associations for T38 were limited further to gestational age at delivery. Preterm birth, gestational age at delivery, antenatal corticosteroids, and gestational age at enrollment were also associated significantly with the primary outcome COMP ( Table 2 ).
| Demographic characteristics | Exposure | P value a | |
|---|---|---|---|
| Present | Not present | ||
| Clinical chorioamnionitis b | |||
| Preterm birth <37 wk, n (%) | 28 (49) | 338 (42) | .26 |
| Gestational age at delivery, wk c | 34.93 ± 5.48 | 36.60 ± 3.38 | .34 |
| Race, n (%) | .29 | ||
| Black | 47 (82) | 668 (82) | |
| White | 4 (7) | 94 (12) | |
| Asian | 4 (7) | 24 (3) | |
| Other | 2 (4) | 28 (3) | |
| Infant male sex, n (%) | 33 (58) | 436 (54) | .53 |
| Corticosteroids, n (%) | 27 (47) | 269 (33) | .03 |
| Maternal age, y c | 24.97 ± 6.08 | 25.62 ± 6.08 | .48 |
| Gestational age at enrollment, wk c | 29.35 ± 3.24 | 30.36 ± 3.66 | .03 |
| Temperature >38°C d | |||
| Preterm birth <37 wk, n (%) | 15 (31) | 351 (43) | .10 |
| Gestational age at delivery, wk c | 37.41 ± 3.89 | 36.44 ± 3.55 | .003 |
| Race, n (%) | .60 | ||
| Black | 40 (82) | 675 (82) | |
| White | 4 (8) | 94 (11) | |
| Asian | 3 (6) | 25 (3) | |
| Other | 2 (4) | 28 (3) | |
| Infant male sex, n (%) | 32 (65) | 437 (53) | .10 |
| Corticosteroids, n (%) | 12 (24) | 284 (35) | .15 |
| Maternal age, y c | 24.37 ± 5.84 | 25.65 ± 6.09 | .12 |
| Gestational age at enrollment, wk c | 30.12 ± 3.19 | 30.31 ± 3.67 | .50 |
| Histologic chorioamnionitis e | |||
| Preterm birth <37 wk, n (%) | 160 (72) | 206 (32) | < .001 |
| Gestational age at delivery, wk c | 33.75 ± 4.72 | 37.44 ± 2.47 | < .001 |
| Race, n (%) | .16 | ||
| Black | 192 (86) | 523 (81) | |
| White | 16 (7) | 82 (13) | |
| Asian | 8 (4) | 20 (3) | |
| Other | 7 (3) | 23 (4) | |
| Infant male sex, n (%) | 121 (54) | 348 (54) | .89 |
| Corticosteroids, n (%) | 125 (56) | 171 (26) | < .001 |
| Maternal age, y c | 25.62 ± 6.18 | 25.57 ± 6.05 | .93 |
| Gestational age at enrollment, wk c | 29.86 ± 3.77 | 30.45 ± 3.59 | .04 |
a Determined by χ 2 test (categoric data) and Kruskal-Wallis rank test (continuous data)
b Present, 57; not present, 814
d Present, 49; not present, 822
| Demographic characteristic | Composite adverse neonatal outcome variable (n = 168) | No composite adverse neonatal outcome variable (n = 703) | P value a |
|---|---|---|---|
| Preterm birth <37 wk, n (%) | 143 (85) | 223 (32) | < .001 |
| Gestational age at delivery, wk b | 32.01 ± 4.59 | 37.57 ± 2.21 | < .001 |
| Race, n (%) | .73 | ||
| Black | 140 (83) | 575 (82) | |
| White | 16 (10) | 82 (12) | |
| Asian | 7 (4) | 21 (3) | |
| Other | 5 (3) | 25 (4) | |
| Infant male sex, n (%) | 98 (58) | 371 (53) | .19 |
| Corticosteroids, n (%) | 112 (67) | 184 (26) | < .001 |
| Maternal age, y b | 26.07 ± 6.12 | 25.46 ± 6.07 | .15 |
| Gestational age at enrollment, wk b | 29.13 ± 3.77 | 30.58 ± 3.56 | < .001 |
a Determined by χ 2 test (categoric data) and Kruskal-Wallis rank test (continuous data)
To address our primary objective, we determined whether prenatal inflammation (measured by CCA, T38, and HCA) was associated with COMP. In bivariate analyses, all measures of prenatal inflammation were associated significantly with increased risk. Furthermore, gestational age at delivery significantly modified the association of both CCA ( P < .01) and HCA ( P < .05) on COMP and trended towards significance of T38 on COMP ( P = .061; Table 3 ). The inclusion of the potential confounders antenatal corticosteroid use, maternal age, or maternal race 1 at a time in the multivariable main-effects models of COMP resulted in changes of <5% of the regression coefficient of any of the indicators of prenatal inflammation and therefore were not included in the final model.
| Prenatal Inflammation | Composite adverse neonatal outcome variable, n (%) | Odds ratio (95% CI) | P value | |||
|---|---|---|---|---|---|---|
| Present a | Not present b | Crude | Adjusted c | Crude/adjusted odds ratio | Interaction | |
| Clinical chorioamnionitis, wk | 32 (19.0) | 25 (3.6) | 6.38 (3.66–11.11) | < .001 | ||
| 32 | 3.12 (1.02–9.52) | < .05 | ||||
| 36 | 8.88 (4.32–18.25) | < .001 | < .01 | |||
| 40 | 25.30 (9.25–69.19) | < .001 | ||||
| Temperature >38°C, wk | 20 (11.9) | 29 (4.1) | 3.14 (1.73–5.70) | < .001 | ||
| 32 | 3.18 (0.66–15.42) | .15 | ||||
| 36 | 8.40 (3.60–19.61) | < .001 | .061 | |||
| 40 | 22.19 (8.15–60.44) | < .001 | ||||
| Histologic chorioamnionitis, wk | 101 (60.1) | 122 (17.4) | 7.18 (4.98–10.34) | < .001 | ||
| 32 | 1.25 (0.64–2.46) | .51 | ||||
| 36 | 2.56 (1.54–4.23) | < .001 | < .05 | |||
| 40 | 5.23 (1.95–13.99) | < .001 | ||||
To better understand the interaction of gestational age at delivery and prenatal inflammation, we graphed the odds ratio (OR) of prenatal inflammation in association with COMP using CCA ( Figure 1 ), T38 ( Figure 2 ), and HCA ( Figure 3 ) as a function of gestational age at delivery from 25-41 weeks. All models were modified by gestational age at delivery and controlled for infant sex. The OR of each model is significantly different from 1.0, which is the OR of the group without inflammation exposure, starting at week 32 for CCA, 33 for T38, and 34 for HCA and continues to increase with increasing age. In infants at 32-34 weeks’ gestational age that were not exposed to prenatal inflammation, there was a sharp decrease in COMP. However, in those infants who were exposed, the magnitude of this decrease was not as great, which suggests that the protective association with advancing gestational age is diminished when prenatal inflammation is present (data not shown). To attach numbers to the ORs and 95% confidence intervals (CIs) found in the Figures, we selected 3 gestational ages (32, 36, and 40 weeks; Table 3 ). The inclusion of the potential confounders antenatal corticosteroid use, maternal age, or maternal race 1 at a time in the multivariable main-effects models of COMP resulted in changes of <5% of the regression coefficient of any of the indicators of prenatal inflammation and therefore were not included in the final model.
Secondary objectives were to determine the association between prenatal inflammation and individual adverse outcomes of neonatal sepsis and respiratory compromise.
As with COMP, gestational age at delivery significantly modified the association with both CCA ( P < .01) and T38 ( P < .05) on neonatal sepsis but was not significant between HCA and neonatal sepsis ( P = .59). Through selection of 3 gestational ages at delivery (32, 36, and 40 weeks), we found that the associations between prenatal inflammation and neonatal sepsis continued to be significant when modified by gestational age at delivery and controlling for infant sex, again with relatively larger ORs for neonatal sepsis associated with more advanced gestational ages (CCA at 32 weeks [OR, 2.80; 95% CI, 1.36–5.77], at 36 weeks [OR, 7.23; 95% CI, 3.25–16.10], and at 40 weeks [OR, 18.70; 95% CI, 5.79–60.39]; T38 at 32 weeks [OR, 3.03; 95% CI, 1.05–8.74], at 36 weeks [OR, 7.09; 95% CI, 3.07–16.35], and at 40 weeks [OR, 16.58; 95% CI, 5.17–53.23]). Because the interaction of HCA and age was not significant, the association of HCA with neonatal sepsis was estimated in a main effect model when controlled for gestational age at delivery and infant sex. It was significant ( P < .01) with an OR of 3.18 (95% CI, 1.65–6.13).
Similarly, gestational age at delivery modified the association with CCA ( P < .05) and T38 ( P = .052, with the probability value of a <.10 threshold for reporting results of an interaction) on respiratory compromise but did not modify the effect between respiratory compromise and HCA ( P = .18). Through sampling the same 3 gestational ages, we found that the associations between prenatal inflammation and respiratory compromise continued to be significant at older gestational ages when modified by gestational age at delivery and controlling for infant sex (CCA at 32 weeks [OR, 2.24; 95% CI, 0.79–6.33], at 36 weeks [OR, 5.74; 95% CI, 2.60–12.63], and at 40 weeks [OR, 14.69; 95% CI, 4.24–50.91]; T38 at 32 weeks [OR, 1.94; 95% CI, 0.47–7.94], at 36 weeks [OR, 5.27; 95% CI, 2.25–12.35], and at 40 weeks [OR, 14.32; 95% CI, 4.21–48.65]) The association of HCA with respiratory compromise, which was estimated in a main effects model that controlled for gestational age at delivery and infant sex, was not significant (OR, 1.48; 95% CI, 0.87–2.52) in contrast with neonatal sepsis.
AUC for each multivariable logistic model to discriminate between those patients who did vs did not experience COMP were calculated and compared. Accounting for effect modification by gestational age at delivery and controlling for infant sex, the discriminatory ability of each prenatal inflammation variable was very strong (CCA AUC, 0.89; T38 AUC, 0.89; HCA AUC, 0.88). Although the multivariable model that included only gestational age at delivery in weeks and infant sex had similar clinical ability to discriminate for COMP (AUC, 0.86), the AUC was significantly smaller than a model that also included either CCA ( P < .05), T38 ( P < .05), or HCA ( P < .05).
Finally, we performed analyses parallel to those that are reported in Table 3 on subsets of patients without preeclampsia ( Table 4 ) and without induction of labor ( Table 5 ). In both restricted samples, similar results were found to those in the complete cohort. Probability values of the interaction between prenatal inflammation and gestational age at delivery were within 0.10, which was our threshold for display. All indices of prenatal inflammation showed increasing OR in association with COMP with increasing gestational age at delivery (data not shown).
