CRH as a Predictor of Preterm Birth in Minority Women

Abstract

Objective:

To determine the predictive capability of corticotropin-releasing hormone (CRH) as a biomarker of preterm birth (PTB) in minority women.

Study Design:

Venous blood samples were obtained at 22–24 weeks’ gestation in a prospective, descriptive study of 707 minority women experiencing low-risk pregnancies. CRH was analyzed using a radioimmunoassay and methanol extraction protocol.

Result:

CRH predicted PTB in both African American and Hispanic women. The odds ratio was 1.8 times greater for having a PTB if the CRH level was >24 pg/ml. The median CRH for African American women having a PTB was 46.6 pg/ml and for Hispanic women was 35.03 pg/ml. Using a receiver–operating characteristic curve, the threshold for CRH among the African American women was 30.6 pg/ml and among the Hispanic women was 27.4 pg/ml.

Conclusion:

CRH may be an important biomarker for predicting PTB in minority women, especially when combined with other predictors.

Minority women, particularly African American and Hispanic women (Alexander, Wingate, Bader, & Kogan, 2008), are more at risk for poor infant outcomes, specifically preterm birth (PTB) and the resulting neonatal morbidity and mortality, compared to White women (Bryant, Worjoloh, Caughey, & Washington, 2010; Cabacubgan, Ngui, & McGinley, 2012; Gennaro, 2005). Nationally, the rate of PTB, defined as birth at <37 weeks’ gestation, is 16.2% for African American women, 11.3% for Hispanic women, and 10.17% for White women (Hamilton, Martin, Osterman, & Curtin, 2014).

A potential contributing factor to the ethnic disparity in the PTB rate is the differences in functioning of the hypothalamic–pituitary–adrenal (HPA) axis and placenta. The HPA axis and placental corticotrophin-releasing hormone (CRH) play vital roles in the length of the pregnancy and the size of the fetus (McLean et al., 1995; Wadhwa et al., 2004; Weinstock, 2005). Higher levels of CRH, a master stress hormone produced by both the placenta and the hypothalamus, have been shown to increase the risk for PTB (Holzman, Jetton, Siler-Khodr, Fisher, & Rip, 2001; McLean et al., 1995; Wadhwa, Porto, Garite, Chicz-DeMet, & Sandman, 1998). Progesterone (P4) decreases and glucocorticoids upregulate levels of placental CRH in pregnancy (Petraglia, Imperatore, & Challis, 2010). Recent evidence has linked the combination of a shortened cervix (<25 mm) and increases in CRH (activation of the maternal–fetal HPA axis) with spontaneous PTB at 24–28 weeks’ gestation (Moroz & Simhan, 2014).

In spite of the abundance of evidence supporting a role for CRH in PTB, findings have been contradictory regarding its effectiveness as a predictor of PTB in a low-risk population (Kramer et al., 2013). Methodological issues may have contributed to these differences in findings between studies. Because there is positive feedback between cortisol and CRH and cortisol has a strong diurnal rhythm, the time of day for sample collection needs to be controlled. Cortisol levels rise in the evening and peak in the morning soon after awakening (Aubuchon-Endsley, Bublitz, & Stroud, 2014; Entringer et al., 2010; Lange, Dimitrov, & Born, 2010). Thus, late morning or early to midafternoon are the most effective times for CRH sample collection, as practiced in O’Keane et al. (2011), for example. Also, CRH levels rise exponentially throughout pregnancy. Therefore, all CRH samples in a study should ideally be collected at a gestational age that falls within a 2- to 3-week window (McLean et al., 1995). In some studies, there has not been enough variability in outcomes (PTB vs. term birth) to detect differences (Kramer et al., 2013; Mancuso, Schetter, Rini, Roesch, & Hobel, 2004; Sibai et al., 2005). Finally, because CRH secretion profiles and susceptibility to the effects of stress vary across ethnicities, it is also important to control for ethnicity (Glynn, Schetter, Chicz-DeMet, Hobel, & Sandman, 2007; Tse, Rich-Edwards, Koenen, & Wright, 2012).

Petraglia, Imperatore, and Challis (2010) contend that CRH should be used as a biomarker to indicate women at risk for PTB among populations who have a higher prevalence of PTB, such as African American or Hispanic women. The purpose of the present study was to examine the predictive capability of a 1-time measurement of CRH as a biomarker of PTB among African American and Hispanic women using a radioimmunoassay technique in which binding protein is extracted prior to running the assay.

Method

Participants

The participants in this study were recruited as part of two larger National Institutes of Health/National Institute of Nursing Research (NIH/NINR)-funded studies on the psychoneuroimmunology underlying PTB in Hispanic (2008–2012; Ruiz et al., 2015) and African American women (2009–2014). In one arm of the study, 192 women, aged ≥ 18 years, who were experiencing normal pregnancies and who self-reported as African American (African American, Caribbean American, or African American Hispanic) provided informed consent during routine prenatal clinic visits at one of the two medical centers in the Bronx, NY, at 22–24 weeks’ gestation. We were not able to use samples from 12 of these women because their CRH levels were below the assay’s detectable limit. In the other arm of the study, 515 women, aged 14–43 years, who were experiencing normal pregnancies, self-identified as Hispanic and spoke English or Spanish provided informed consent and were entered into the study at 22–24 weeks’ gestation. These women were recruited from private physicians’ offices. As we had planned for a sample of 500 women, we analyzed samples from only 503 of these women for measurement of CRH. In both arms of the study, the length of gestation was confirmed by accurate last normal menstrual period and ultrasound. All women were pregnant with a singleton intrauterine pregnancy.

Formal exclusion criteria for women across the study were (a) inability to read English or Spanish (at the Texas site); (b) known uterine or cervical abnormalities; (c) kidney disease, pyelonephritis in the current pregnancy, or chronic hypertension; (d) heart disease, coronary artery disease, and history of peripartum cardiomyopathy; (e) autoimmune disorders (i.e., lupus and antiphospholipid syndrome); (f) Type 1 or Type 2 diabetes or insulin-dependent gestational diabetes; (h) asthma requiring use of steroid inhaler; (i) preeclampsia at the time of data collection; (j) oral steroid use within 1 month prior to the time of enrollment; (k) congenital anomalies as determined by fetal ultrasound, especially those leading to hydramnios, trisomies, or major structural anomalies such as neural tube defects, ventral wall defects, or congenital heart disease; (l) blood group isoimmunization; or (m) active cervicovaginal bleeding or placenta previa. These exclusion criteria were necessary to control for differences in the causation of PTB. In particular, they included chronic diseases that are associated with vascular problems but not with spontaneous preterm labor (the kind most associated with early preterm delivery).

The medical centers in the Bronx and the University of Texas Medical branch and the University of Texas at Austin provided institutional review board approval. We obtained parental consent and child assent for any participants who were 14–18 years of age.

Procedures

After women provided informed consent, we collected demographic data, including age and educational level. We obtained CRH specimens from 2 to 3 cc of venous blood drawn into silicone-covered EDTA-treated vacutainers for all participants at 22–24 weeks’ gestation. We collected these samples between 10 a.m. and 5 p.m. to control for diurnal variation. The protocol for CRH sample preparation and assay, which was previously published (Latendresse & Ruiz, 2008), included methanol extraction of free CRH. We obtained neonatal gestational age at delivery from chart review (based on maternal ultrasound at ≤20 weeks’ gestation). We used ≤37 weeks’ gestation as the cutoff for the definition of PTB.

In New York, we transported CRH specimens on ice to the research lab in the hospital where women were receiving care and stored them at −80°C. In Texas, samples were kept on ice for 1–2 hr at the physicians’ offices until they could be transported and stored in the Biobehavioral Laboratory at the University of Texas at Austin in a −80°C freezer for pending analysis. In the laboratory, we separated serum into aliquots of plasma prior to freezing, for a total of one aliquot for each CRH assay to be completed. Specimens from New York were shipped on dry ice to the Clinical Research Laboratory at Ohio state where they were analyzed using exactly the same protocol as was used at the Biobehavioral Laboratory at the School of Nursing at the University of Texas at Austin. The primary author, who originated the CRH-analysis protocol, confirmed that the protocol was followed meticulously at both locations, so the possibility of systematic error was minimized. Laboratory managers at both locations checked for positive controls for all assays and all results fell within expected ranges as noted in the assay kit protocol. We determined an expected range of 35.7–48.7 pg/ml by extending the mean of eight assays by 2 standard deviations (SDs) on the normal curve. We confirmed all results for accuracy (i.e., coefficients of variation between duplicates run on the same sample were not greater than 15%). Laboratory technicians were blinded as to sample characteristics and infant outcomes.

Plasma CRH was measured by a radioimmunoassay provided by Phoenix Pharmaceuticals, Inc. (Burlingame, CA). The assay is based upon competition between radioactive iodine (I-¹²⁵) and CRH antigens for binding sites with CRH antibodies. There are limited numbers of antibody-binding sites specific to CRH. This method is very sensitive and quantifies the amount of actual antigen (CRH) in the sample. We extracted samples prior to assay using a methanol-based procedure to remove interfering binding protein. The lowest detection limit of the assay, as reported by the manufacturer, is 26.51 pg/ml. Interassay coefficient of variance is 14.2%, and intraassay coefficient of variance is 3.2%.

Statistical Analysis

We present normal continuous variables using mean and SD. We describe nonnormal data (CRH) using median and interquartile range. We used a Fisher’s exact test to compare the preterm rate between ethnic groups. To induce normality in CRH levels, we used a log transformation. We compared continuous variables using unpaired t-tests and categorical variables using Fisher’s exact tests. We evaluated diagnostic performance for CRH (sensitivity, specificity, positive predictive value, negative predictive value, and odds ratio) at the median threshold and determined the predictive performance of CRH for PTB using receiver–operating characteristic (ROC) analysis. We used ROC area under the curve (AUC) to summarize the overall predictive ability of CRH. Finally, we determined the optimum threshold for CRH to differentiate PTB from term birth with maximum sensitivity and maximum specificity separately and adjusted it for the two minority populations.

Results

Among the 57 Hispanic women in the Texas arm of the study who experienced PTBs, 11 (19.3%) had high blood pressure and 5 (8.8%) had premature rupture of membranes. Spontaneous labor occurred in 36 (63.2%) of these PTBs. There were missing data regarding the cause of PTB in only 1 of these 57 cases. Of the 10 infants in this arm of the study that were small for gestational age, commonly defined as a weight below the 10th percentile for gestational age (Pilliod, Cheng, Snowden, Doss, & Caughey, 2012), only 4 were preterm. All women who completed the study in Texas had samples for CRH and delivery data to determine gestational length, especially the PTBs.

Among the 13 PTBs in the New York sample of African American women, 1 case had missing data because the participant delivered at another institution. Of the remaining 12 women, 2 (16.7%) had high blood pressure, 3 (25%) experienced preeclampsia, and 3 (25%) developed gestational diabetes. A total of 8 (66.7%) of these women experienced spontaneous preterm labor, and 5 (41.7%) had premature rupture of membranes.

Table 1 lists participant characteristics by ethnic group. Median CRH, prevalence of infection, and the rate of PTB did not differ significantly between the groups. African American women were significantly older and more educated compared to Hispanic women. Table 2 lists participant characteristics according to preterm status within each ethnic group. The median CRH level was significantly higher in women who had a PTB as compared to women who had a term birth among both the Hispanic (p = .037) and the African American women (p = .007). The median level of CRH (46.6 pg/ml) was slightly higher in the African American women who had a PTB than in the Hispanic women who had a PTB (35.03 pg/ml), but the difference was not significant. Age, presence of infection, and education did not differ significantly between Hispanic women who had a PTB versus a term birth. While infection and years of education also did not differ significantly by birth status among African American women, the average age of African American women who had a PTB was significantly higher than those who had a term birth.

Table 1.

Characteristics of Participants by Ethnic Group.

Variables	African American	Hispanic	p Value
	n = 180	n = 503
Age (years), mean (SD)	26.03 (6.14)	24.61 (5.81)	.005
Education (years), mean (SD)	12.76 (1.81)	11.82 (2.42)	.000
CRH pg/ml, median (IQR)	24.08 (15.35, 37.37)	24.04 (13.52, 44.19)	.875
Infection, n (%)	72 (40)	174 (35.58)	.421
Preterm birth, n (%)	13 (7.2)	57 (11.33)	.577

Note. CRH = corticotropin-releasing hormone; SD = standard deviation; IQR = interquartile range.

Table 2.

Participant Characteristics by Birth Status and Ethnic Group.

	African American			Hispanic
Characteristic	Term Birth	Preterm Birth	p Value	Term Birth	Preterm Birth	p Value
Age (years), mean (SD)	25.80 (6.17)	29.06 (6.08)	.039	24.65 (5.81)	25.11 (5.81)	.575
Education (years), mean (SD)	12.77 (1.83)	12.40 (1.55)	.446	11.84 (2.39)	11.77 (2.76)	.839
Infection, n (%)	64 (39.75)	6 (35.29)	.799	144 (33.88)	23 (43.40)	.173
CRH, median (IQR)	23.40 (13.31, 35.01)	46.58 (19.71, 57.53)	.007	23.51 (13.45, 42.81)	35.03 (15.72, 58.11)	.037

Note. SD = standard deviation; IQR = interquartile range; CRH = corticotropin-releasing hormone.

The median overall (both term and PTB across both ethnic groups) CRH level was 24 pg/ml. Using this median level as a threshold for categorizing CRH levels for predicting PTB, we found that the odds of PTB were 1.8 times greater for women with CRH levels ≥24 pg/ml as compared to women with CRH levels <24 pg/ml. The overall sensitivity was 63%, specificity was 52%, positive predictive value was 14%, and negative predictive value was 92%. The African American women had a slightly higher sensitivity of 71% for CRH as compared to the Hispanic women, with a sensitivity of 60%. Specificity and predictive values were similar for the two ethnic groups (Table 3).

Table 3.

Diagnostic Accuracy of Corticotropin-Releasing Hormone (CRH) at the Median Threshold for Differentiating Between Preterm and Term Birth.

Diagnostic Measure	Overall (Estimate [95% CI])	African American (Estimate [95% CI])	Hispanic (Estimate [95% CI])
Sensitivity	0.63 [0.50, 0.74]	0.71 [0.44, 0.90]	0.60 [0.46, 0.73]
Specificity	0.52 [0.48, 0.56]	0.53 [0.45, 0.60]	0.51 [0.47, 0.56]
Positive predictive value	0.14 [0.10, 0.18]	0.14 [0.07, 0.22]	0.14 [0.10, 0.19]
Negative predictive value	0.92 [0.88, 0.95]	0.94 [0.88, 0.98]	0.91 [0.86, 0.94]
Odds ratio	1.79 [1.08, 2.94]	2.65 [0.93, 7.54]	1.59 [0.90, 2.80]

Note. CI = confidence interval.

ROC analysis showed the CRH-level threshold for discriminating between PTB and term birth was 30.6 pg/ml within the African American cohort, with more than 60% sensitivity and specificity. In comparison, the CRH-level threshold for discriminating between PTB and term birth was 27.4 pg/ml within the Hispanic cohort. The AUC for predicting PTB using CRH level was 70% (95% CI [54%, 85%]) and 59% (95% CI [51%, 68%]) for the African American and Hispanic cohorts, respectively. The effectiveness of CRH level in predicting PTB was slightly larger for the African American cohort as compared to the Hispanic cohort, though the difference did not reach statistical significance (p = .24; Figure 1). After controlling for age and education, the AUC for predicting PTB using CRH across the two ethnic groups was 62%.

Figure 1.

Receiver–operating characteristic curves for corticotropin-releasing hormone in discriminating between preterm and term birth according to ethnic group.

Discussion

At 22–24 weeks’ gestation, the predictive capability of CRH level for PTB was relatively high for both African American and Hispanic women in the present study. It was best for the African American cohort, which is important due to the higher rate of PTB among those women (Hamilton et al., 2014). The CRH-level threshold that served to discriminate between PTB and term birth differed only slightly between the two ethnic groups. The overall odds ratios for the risk of PTB were greater than 1.5 for both ethnic groups using their respective median CRH levels as a threshold.

There were significant differences between the mean ages of the Hispanic and African American women and between the African American women with PTB and those with term birth. However, the mean ages for all groups and subgroups were within the range of normal childbearing age. The mean educational level also differed significantly between groups, but there was actually only 1 year of difference in educational level between them, which likely has little clinical significance. After we adjusted for the effects of age and education level, the predictive capability of CRH remained high and significant across the two ethnic groups (AUC = 0.62).

In spite of our findings of a significant relationship between CRH levels and PTB among participants in the present study, a number of previous studies have had contradictory findings. One potential reason for these differences in results is that the methods used to measure CRH have differed across studies. Enzyme-linked immunoassay methodology is not as sensitive as the radioimmune assay (RIA) technique, which may explain why the findings of some studies showed no relationship between CRH level and PTB (Himes & Simhan, 2011; Sibai et al., 2005). In another study, in which researchers found no relationships among stress, CRH, and PTB, authors used an RIA but did not disclose whether they extracted the hormone’s binding protein prior to running the assay (Kramer et al., 2013). This step is essential in measuring the level of free CRH and was included in the present study.

The results presented here are consistent with the findings of numerous other studies that have evaluated CRH as a predictive biomarker for PTB. In a German study of 79 women in preterm labor, Makrigiannakis and colleagues (2007) extracted the bound CRH, used an RIA methodology to measure CRH levels, and found that the levels were significantly higher at all measured time points in the PTB group than in the women who delivered at term. Specifically, at 24–29 weeks’ gestation, women in the preterm delivery group had CRH levels of 14 pg/ml, while women who experienced term births had mean CRH levels of 10.2 pg/ml at the same time point. The authors concluded that CRH may be used as a biomarker of PTB. In another study that used the same methodology that we used in the present study (Holzman et al., 2001), scientists found that increased levels of CRH as early as 15–19 weeks’ gestation were associated with an increased risk of PTB and that CRH levels varied by race and ethnicity. The median CRH level at 15–19 weeks for African American women who experienced PTB at <35 weeks’ gestation was 73.6 pg/ml versus 50.7 pg/ml for those women who delivered at term (p < .001). In our study, the median level for CRH was 46.58 pg/ml at 22–24 weeks’ gestation for women who delivered preterm at <37 weeks’ gestation and were African American. The sensitivity of CRH level in the present study was much better than that in the Holzman et al. study, at .67 versus .41. In a recent study examining cervical length via ultrasound and its association with CRH level, Moroz and Simhan (2014) found that women who had a short cervical length (<25 mm) at 24 weeks’ gestation also had higher levels of CRH (r ² = .34, p = .014), indicating activation of the maternal–fetal HPA. The temporal relationship between cervical shortening and changes in CRH level is unknown, however.

A number of studies have explored relationships between environmental stressors and the physiological stress response that might help to explain the differences we found by ethnic group in the present study. The placenta is the primary source for CRH in the maternal blood during pregnancy. It may be the master controller of the maternal–fetal response to physiological or psychological stress (Himes & Simhan, 2011). Cortisol forms a positive feedback loop with CRH (Kramer et al., 2013) and thus may play a role in driving the CRH response as well. Some investigators (Tse et al., 2012) have found that the effects of stress on prenatal CRH levels may be mediated by factors that differ between ethnic groups. Specifically, Tse et al. found that racial discrimination, community violence, and cumulative stress were related to CRH levels in African Americans but not Hispanics. Ruiz et al. (2015) found that CRH levels in Hispanics were related to acculturation, depression, and coping. Thus, acculturation may play a larger role in activating the physiological stress response in pregnant Hispanic women than in pregnant African American women.

Results of the present study support the use of CRH as an important biomarker related to PTB. Although CRH level was predictive of PTB in both ethnic groups studied, its predictive capability was significantly greater among African American women, a notable finding given that African American women experience a higher rate of PTB. Consideration of the relationship of ethnicity to the predictive threshold of CRH level is important, as we found that the threshold differed between African American and Hispanic women. Further research on CRH and PTB is indicated related to variation across subgroups of Hispanics, such as Puerto Ricans and Cubans. Nevertheless, the thresholds we determined in the present study should serve as a step toward the clinical use of CRH for prediction of PTB among African American and Hispanic women. CRH may be especially useful as a negative predictor to screen for lower risk of PTB. Based on our findings as well as the findings of numerous previous studies, we conclude that CRH, in combination with other screening tools such as cervical length on ultrasound, holds promise as tool for PTB risk assessment.