Evidence of placental haemorrhage and preterm delivery

Abstract

Please cite this paper as: Gargano J, Holzman C, Senagore P, Reuss M, Pathak D, Williams M, Fisher R. Evidence of placental haemorrhage and preterm delivery. BJOG 2010;117:445–455.

Objective  To evaluate evidence of placental haemorrhage (PH) obtained through maternal interviews, patient charts and placental pathology examinations as potential indicators of a ‘bleeding pathway’ to preterm delivery (PTD).

Design  Prospective cohort.

Setting  Fifty-two clinics in five communities in Michigan, USA (1998–2004).

Population  A subset (n = 996) of cohort participants with complete placental pathology data.

Methods  First-trimester bleeding and placental abruption were ascertained by mid-trimester interviews and chart review, respectively. Disc-impacting blood clot was defined as a gross placental examination finding of a blood clot impacting adjacent tissue. Microscopic haemorrhage was defined as ‘high’ (top quintile) scores on an aggregate measure of placental pathology findings suggestive of atypical maternal vessel haemorrhage. These four PH indicators were compared with one another and with risk of PTD assessed by logistic regression analyses.

Main outcome measures  Preterm delivery and PTD subtypes (i.e. <35 weeks, 35–36 weeks; spontaneous, medically indicated) compared with term deliveries.

Results  Placental abruption cases had 2.3-fold to 5.5-fold increased odds of the other three PH indicators. Disc-impacting blood clots and microscopic haemorrhage were associated with one another (odds ratio [OR] = 4.6), but not with first-trimester bleeding. In a multivariable model that included all four PH indicators and confounders, risk of PTD < 35 weeks was elevated with first-trimester bleeding (OR = 1.9 [1.0, 3.4]), placental abruption (OR = 5.2 [1.7, 16.2]), disc-impacting blood clots (OR = 2.3 [1.0, 5.0]) and microscopic haemorrhage (OR = 2.4 [1.4, 4.2]).

Conclusions  Multiple clinical and subclinical PH indicators are associated with PTD, particularly early PTD.

Introduction

Several aetiological pathways have been implicated in preterm delivery (PTD), including infection, stress, abnormal uterine distension and uterine bleeding.¹ Various types of direct or indirect evidence of placental haemorrhage (PH) may be manifestations of a uterine bleeding pathway. Vaginal bleeding in early to mid-pregnancy has been associated with PTD in many studies.^2–4 Placental abruption, or premature separation of a normally placed placenta, may originate with bleeding into the decidua from ruptured decidual blood vessels⁵ and can result in major obstetric haemorrhage.⁶ Placental abruption is present in a larger proportion of preterm deliveries than term deliveries.^7,8 Subchorionic or retroplacental bleeding detected by prenatal ultrasonography^9,10 and haemosiderin (a breakdown product of old blood) in placental tissues detected through histological examination¹¹ have also been associated with PTD, although these have not been reported on extensively.

It is unclear whether these variable manifestations of bleeding belong to a single pathway with a common set of maternal risk factors. Vaginal bleeding early in pregnancy has been linked to placental abruption,^2,3,12 suggesting that bleeding and abruption may share one or more common pathways. Many other maternal characteristics and clinical circumstances are shared risk factors for both PTD and placental abruption, including hypertension,^13–15 African American identity,^16,17 smoking,^14,18 cocaine abuse¹⁹ and low body mass index (BMI).^20,21 Less has been published on maternal characteristics associated with other evidence of PH, such as bleeding early in pregnancy²² or placental pathology findings.

While placental abruption is rare, complicating approximately 1% of pregnancies,⁶ placental pathology findings consistent with abruption are more common.⁶ We hypothesised that placental abruption may only represent a portion of the uterine bleeding that affects delivery timing—the ‘tip of the iceberg’—because subclinical gross or microscopic evidence of PH apparent only in placental examinations could also be related to PTD risk. The prospective Pregnancy Outcomes and Community Health (POUCH) Study collected multiple sources of evidence of PH, including mid-pregnancy maternal reports of prior vaginal bleeding episodes, clinical data from medical records and gross and microscopic placental pathology findings. The aim of this study was to evaluate four indicators of PH as potential components of a common ‘bleeding pathway’ by assessing their mutual associations, describing their prevalence according to maternal characteristics, and estimating their odds of PTD and its subtypes.

Methods

Study protocol

The POUCH Study enrolled 3019 pregnant women from five Michigan communities at 15–27 weeks of gestation (1998–2004). English-speaking women aged ≥15 years who had maternal serum alpha-fetoprotein (α-FP) screening at 15–22 weeks of gestation, a singleton pregnancy with no known congenital anomalies or chromosomal abnormalities, and no pre-existing diabetes were eligible. The α-FP was of particular interest in the POUCH Study’s design because of this biomarker’s consistent association with PTD;²³ for this reason, the POUCH cohort oversampled women who had unexplained high α-FP (more than two multiples of the median, 7% of cohort). The POUCH study protocol received institutional review board approval from Michigan State University, the Michigan Department of Community Health, and all nine delivery hospitals. All women provided informed written consent.

At enrolment, women participated in a structured interview with a study nurse and completed a self-administered questionnaire. The interview elicited information on demographics (including self-reported race/ethnicity), height and pre-pregnancy weight (used to calculate BMI), reproductive history, medical conditions, medication use and events during pregnancy. During the interview, women were asked, ‘Have you had any spotting or bleeding so far during this pregnancy?’ and were prompted to describe the timing (gestational week) and heaviness (spotting, slight, about the same as usual period, or heavier than usual period) for up to seven episodes. The questionnaire was designed to collect data on potentially sensitive questions including substance use and physical abuse. Women were asked how often in the last 6 months they had been ‘shoved, hit, or physically abused by [their] parents or a partner or husband’ (never, once or twice, several times, often or very often).

Gestational age was estimated by last menstrual period, corroborated by an ultrasound scan conducted before 25 weeks (available for 97% of women). If the estimates differed by more than 2 weeks (17%) or if last menstrual period was unavailable (3.6%), the ultrasound date was used. Preterm delivery was defined as delivery before 37–0/7 weeks of gestation.

Subcohort sample

A subcohort (n = 1371) was selected for detailed study of biological samples and medical chart-level data. The subcohort included all PTD, all women with unexplained high α-FP and a stratified random sample of women with normal α-FP and term deliveries, with oversampling of African Americans. Subcohort analyses incorporated sampling weights to reconstitute the cohort distributions and further correct for over-representation of high α-FP in the cohort, such that weighted proportions based on the subcohort can be interpreted as prevalence or risk.

Placenta protocol

After subcohort deliveries, placentas were fixed in formalin at the delivery hospital before transport to the study’s pathology laboratory. For the gross examination, parallel slices were made through the placental disc 1 cm apart. The pathologist noted clots in the cut surface, along with indicators of adjacent tissue involvement, that is dissecting haemorrhage, and tissue infarction, compression, or red/brown discolouration. ‘Disc-impacting blood clot’ was defined as a gross examination finding of a retroplacental or intraplacental blood clot impacting adjacent tissue.

The microscopic placental examination was conducted without the pathologist’s knowledge of gross examination findings and clinical outcome. Nine tissue samples per placenta were selected: two from the membrane roll, two from the cord and five from the disc.²⁴ The disc samples were chosen as follows: one at the cord insertion, one in grossly normal central tissue, one from marginal tissue and two additional samples from central tissue, the latter representing grossly abnormal tissue if present.

Microscopic vascular-related findings that fell within five constructs adapted from a diagnostic coding tool were recorded. Items in the ‘Maternal Vascular—Disturbance of Integrity’ (MV-I) construct, that is microscopic evidence of retroplacental blood with adjacent disc disruption/compression, decidual haemorrhage in the basal plate and decidual haemosiderin-like pigment in the membranes or basal plate, were used to calculate an MV-I score for each woman. ‘Microscopic haemorrhage’ was defined as the top quintile of MV-I scores based on the distribution of scores among term deliveries with normal α-FP. This serves as a possible indicator of atypical maternal vessel haemorrhage. This distributional cut-point along a continuum of findings was previously shown to correlate with PTD risk in the POUCH Study.²⁵

Medical record abstraction

Study nurses abstracted subcohort medical records in detail. For all PTD, an additional brief abstraction was completed by a physician with expertise in obstetrics. A pool of possible placental abruption cases, identified by bleeding near delivery or any mention of placental abruption in the abstracted data, was later reviewed by three clinicians with labour and delivery experience who were unaware of the placental pathology findings recorded by the POUCH Study pathologist (P.K.S.). Placental abruption was defined as (1) documented signs and symptoms consistent with abruption (e.g. vaginal bleeding, pain, increased uterine tone, fetal distress); or (2) retroplacental haematoma visualised on a prenatal ultrasound scan. Disagreements among the reviewers were resolved by discussion with the principal investigator (C.B.H.) and study pathologist (P.K.S.) until a consensus was reached.

Other relevant information abstracted from patient charts included trauma or injuries during pregnancy, episodes of vaginal bleeding during pregnancy and date of occurrence, blood pressure and proteinuria values, and details of the delivery process including timing of membrane rupture, cervical dilatation, and interventions. Hypertension was defined as documented diastolic blood pressure ≥90 mmHg and/or systolic blood pressure ≥140 mmHg on ≥2 days or a documented diagnosis and/or history of hypertension, or antihypertensive medication before 20 weeks. Pre-eclampsia, gestational hypertension and chronic hypertension were not considered separately because numbers of each were small. Preterm delivery subtypes, including medically indicated, spontaneous labour and preterm prelabour rupture of membranes (PPROM) were determined based on chart-level data as previously described.²⁶ Spontaneous PTD included spontaneous preterm labour and PPROM.

Evidence of placental haemorrhage

The four principal PH indicators compared in this study were (1) placental abruption, (2) disc-impacting blood clots, (3) microscopic haemorrhage and (4) first-trimester bleeding (any versus none, ascertained during enrolment interview). In selected analyses, we replaced first-trimester bleeding (any versus none) with variables that captured more detailed information: (1) first-trimester bleeding from the enrolment interview categorised as heavier than spotting, spotting only or none, and (2) first-trimester and second-trimester vaginal bleeding (excluding bleeding in the week before delivery) recorded in prenatal charts categorised as first-trimester only, second-trimester only, both trimesters, or none.

Study sample

The derivation of the study sample is shown in Figure 1. From the subcohort, we sequentially excluded 158 women whose placentas were not saved, 127 women whose placental examinations were not yet complete, six women with placenta praevia at delivery and 84 women who identified themselves as any race/ethnic category other than ‘White or Caucasian’ (referred to as ‘white’) or ‘Black or African American’ (referred to as ‘African American’). After these exclusions, a sample of 996 women remained.

Figure 1.  Flow diagram showing derivation of study sample from the Pregnancy Outcomes and Community Health (POUCH) Study cohort. *Sampling scheme was stratified by race, alpha-fetoprotein (α-FP) and delivery timing (term/preterm), see text for details. **‘Other’ race includes Asian, Hispanic, Native American and others.

Descriptive statistics

As a first step, we calculated the prevalence of the four PH indicators and maternal characteristics overall and stratified by delivery timing (term/preterm). Next, we calculated the prevalence of each PH indicator among women with the other three indicators, and calculated bivariate odds ratios (OR) and 95% confidence intervals (CI) for each pair of PH indicators to examine their co-occurrence. Finally, we estimated the prevalence of each indicator according to selected maternal characteristics, and tested for differences in proportions using modified Rao–Scott chi-square tests for complex survey designs.²⁷

Modelling strategy

We used weighted logistic regression models to assess relations between the four PH indicators and PTD. We calculated ORs for each PH indicator individually, then entered all four variables into the same model (i.e., a ‘mutually adjusted’ model). Because the PH indicators were not highly concordant, it was possible to include all four in the same model without introducing multicollinearity problems. Next, we added maternal characteristics to the model, and retained those that changed ORs for any of the PH indicators by ≥5% from the mutually adjusted estimates. Potential confounders considered were those associated with PTD or any of the PH indicators in univariate analyses but not directly in the pathway between PH indicators and PTD. Finally, we added hypertension separately to examine its effect on other variables in the model. Hypertension may have a direct, proximate effect on placental abruption, placental pathology findings and delivery timing.

Weighted polytomous logistic regression models were used to assess the relations among PH indicators and PTD subtypes defined by timing (35–36 weeks, <35 weeks, term [referent]) or clinical circumstances (spontaneous, medically indicated, term [referent]; and spontaneous labour, PPROM, medically indicated, term [referent]). All statistical analyses were conducted using SAS 9.1.3 (Statistical Analysis Software, Cary, NC, USA).

Results

Descriptive statistics

Characteristics of the subcohort sample are listed in Table 1, overall and stratified by delivery timing. In weighted analyses, 10.7% of pregnancies ended in PTD. In the total sample, prevalence of the four PH indicators ranged from 2.0% for placental abruption and 5.6% for disc-impacting blood clots to about 20% for microscopic haemorrhage and first-trimester bleeding. All four PH indicators were more prevalent among PTD than term deliveries. Medicaid insurance, lack of high-school education, spending >1 hour per week in a smoky room, hypertension and parity/prior PTD were also significantly associated with PTD in these univariate analyses. Although the relation between PTD and BMI categories was not statistically significant overall (P = 0.13), preterm deliveries involved larger proportions of both underweight (6.1%) and obese (30.1%) women than term deliveries (3.3% and 25.8%, respectively).

Table 1.   Indicators of placental haemorrhage, maternal characteristics and risk factors, overall and by preterm delivery status

	Total		Term		Preterm		P**
	n	%*	n	%*	n	%*
*Percentages have been weighted (inverse of sampling probability) to reflect distribution in cohort. **P-value for preterm versus term comparison by Rao–Scott chi-square test (SAS PROC SURVEYFREQ). ***P-value cannot be calculated in univariate analyses because exposure and outcome are sampling strata.
Total	996		758	89.3	238	10.7
Indicators of placental haemorrhage (not mutually exclusive)
Placental abruption	31	2.0	12	1.4	17	7.3	<0.0001
Cut surface clot impacting adjacent tissue	62	5.6	36	5.0	26	10.4	0.005
Microscopic haemorrhage	215	20.4	143	19.3	72	29.4	0.002
First-trimester vaginal bleeding	220	19.2	154	18.2	66	27.8	0.005
Maternal characteristics
African American	416	24.6	340	23.6	76	32.9	***
Medicaid-insured	536	45.6	411	44.8	125	52.7	0.03
Education: did not complete high school	203	16.3	153	15.7	50	21.6	0.03
Smoked during pregnancy	282	26.8	209	26.3	73	30.6	0.23
Smoked prior to pregnancy	343	33.1	250	32.4	93	38.5	0.10
Smoky room >1 hour/week	479	44.9	349	43.6	130	55.6	0.002
Cocaine use (ever)	82	8.5	25	10.7	57	8.3	0.29
Trauma/injury noted in medical record	116	10.7	89	10.6	27	11.8	0.62
Physical abuse (previous 6 months)	105	9.1	83	9.1	22	8.8	0.89
Hypertensive disorder (includes pre-eclampsia, pregnancy-induced hypertension and chronic hypertension)	99	8.6	61	7.7	38	16.3	<0.01
Caesarean deliveryc	267	26.6	194	26.2	73	30.0	0.28
Age
<20 years	160	13.3	122	12.9	38	16.3	0.36
20–34 years	765	78.6	583	78.8	182	77.0
≥35 years	71	8.1	53	8.3	18	6.7
Parity/prior preterm birth
Nulliparous	407	40.8	304	40.5	103	43.6	<0.0001
Parous/no prior preterm birth	449	49.2	380	51.7	69	28.6
Prior preterm birth	139	9.9	74	7.8	65	27.7
Prepregnancy body mass index
<18.5	44	3.6	30	3.3	14	6.1	0.13
18.5–24.9	438	45.4	336	45.8	102	42.5
25–29.9	228	24.7	179	25.1	49	21.3
≥30	286	26.3	213	25.8	73	30.1

Table 2 shows pairwise associations among the four PH indicators. Placental abruption was associated with all three other PH indicators (versus disc-impacting blood clot, OR = 5.5 [1.7, 17.3]; versus microscopic haemorrhage, OR = 2.3 [1.0, 5.5]; versus first-trimester bleeding, OR = 3.4 [1.4, 8.5]). However, for each pairwise comparison, <50% of placental abruption cases had one of the other PH indicators (23.0%, 37.0% and 43.8%, respectively). Disc-impacting blood clot and microscopic haemorrhage were strongly associated with one another (OR = 4.6 [2.3, 8.9]), but first-trimester bleeding was not associated with disc-impacting blood clot or microscopic haemorrhage (ORs 1.1 and 0.9).

Table 2.   Prevalence of four indicators of placental haemorrhage overall and among women with other indicators of placental haemorrhage (weighted row %), and odds ratios for bivariate comparisons

	Placental abruption		Disc-impacting blood clots		Microscopic haemorrhage		First-trimester bleeding
	%	OR (95% CI)	%	OR (95% CI)	%	OR (95% CI)	%	OR (95% CI)
Among all women	2.0		5.6		20.4		19.2
Among women with placental abruption (n = 31)	–		23.0	5.5 (1.7, 17.3)	37.0	2.3 (1.0, 5.5)	43.8	3.4 (1.4, 8.5)
Among women with disc-impacting blood clots (n = 62)	8.4	5.5 (1.7, 17.3)	–		51.1	4.6 (2.3, 8.9)	20.7	1.1 (0.5, 2.4)
Among women with microscopic haemorrhage (n = 215)	3.7	2.3 (1.0, 5.5)	13.9	4.6 (2.3, 8.9)	–		18.5	0.9 (0.6, 1.5)
Among women with first-trimester bleeding (n = 220)	4.6	3.4 (1.4, 8.5)	6.0	1.1 (0.5, 2.4)	19.6	0.9 (0.6, 1.5)	–

Figure 2 shows the number of women with all combinations of the four PH indicators. A total of 413 women (38.9% weighted) had at least one PH indicator, and 321 women (31.9% weighted) had only one PH indicator. Nine of 31 placental abruption cases (35.2% weighted) had no other PH indicators.

Figure 2.  Venn diagram showing 15 possible combinations of the four indicators of placental haemorrhage. Oval with double outline, disc-impacting blood clot; rectangle with double outline, microscopic haemorrhage; rectangle with single outline, first-trimester bleeding; oval with single outline, placental abruption. Note: 583 women with none of the four types of evidence are not represented in the figure.

Maternal characteristics had few statistically significant associations with evidence of PH (Table 3). Women with <12 years of education had a lower prevalence of first-trimester bleeding than women with more education. Smokers had a lower prevalence of microscopic haemorrhage than nonsmokers. Hypertensive women had a higher prevalence of disc-impacting blood clots than nonhypertensive women. Nulliparous women had a lower prevalence of all four PH indicators than parous women with and without prior PTD, although only the association with placental abruption was statistically significant. Seven percent of underweight women had placental abruption, but this was not statistically significantly higher than the proportions in women in other BMI categories (range 1.5 to 2.1%).

Table 3.   Prevalence (weighted row %) of four indicators of placental haemorrhage according to selected maternal characteristics

	Placental abruption	Disc-impacting blood clot	Microscopic haemorrhage	First-trimester bleeding
*P < 0.05 from Rao–Scott modified chi-square test.
Overall	2.0	5.6	20.4	19.2
Race/ethnicity
African American	3.2	5.9	18.9	18.2
White	1.7	5.4	20.9	19.5
Medicaid-insured
Yes	2.2	5.7	17.7	16.9
No	1.9	5.4	22.7	21.1
Education
<2 years	2.9	6.0	16.6	13.4*
≥12 years	1.9	5.5	21.1	20.3
Smoked during pregnancy
Yes	1.1	4.6	15.0*	15.8
No	2.4	5.9	22.4	20.5
Smoked 6 months before pregnancy
Yes	1.3	4.5	16.7	15.5
No	2.4	6.1	22.3	21.0
Time spent in smoky room
≥1 hour/week	2.3	5.5	19.2	19.2
<1 hour/week	1.8	5.6	21.4	19.2
Ever used cocaine
Yes	2.0	4.7	13.4	16.1
No	2.0	5.6	21.1	19.5
Trauma/injury noted in medial record
Yes	5.4	3.6	17.5	17.0
No	1.6	5.8	20.8	19.5
Physical abuse (previous 6 months)
Yes	1.5	8.1	23.0	21.7
No	2.1	5.3	20.1	18.9
Hypertensive disease
Yes	4.0	15.3*	24.3	19.3
No	1.9	4.6	20.0	19.2
Age
<20 years	2.1	5.4	20.7	14.5
20–34 years	2.1	5.4	20.3	19.8
≥35 years	1.4	7.0	20.7	21.1
Parity/prior preterm birth
Nulliparous	0.8*	4.4	18.5	17.6
Parous/no prior preterm birth	2.7	6.1	20.6	19.3
Prior preterm birth	3.9	7.3	27.3	25.6
Prepregnancy body mass index
<18.5	7.1	7.7	16.3	20.1
18.5–24.9	2.1	4.5	21.7	21.3
25–29.9	1.7	6.6	20.7	20.6
≥30	1.5	6.0	18.5	14.1

Model results

For PTD < 37 weeks, there were elevated odds for each PH indicator in unadjusted analyses (Table 4). In a mutually adjusted model (i.e. a model that included all four PH indicators), all four ORs were attenuated, and disc-impacting blood clot lost statistical significance. Only minor changes occurred after adding maternal characteristics (maternal race, marital status and BMI) to the model. Adding hypertension to the model slightly attenuated the ORs for placental abruption and disc-impacting blood clot but had no impact on the other two PH indicators. When we considered two alternative specifications of vaginal bleeding during pregnancy in these models, we found that heavier bleeding had a stronger association with PTD than spotting, and bleeding in both first and second trimesters (from patient charts) had a stronger association with PTD than bleeding confined to either the first or second trimester in unadjusted but not in adjusted analyses (not shown). However, neither of these alternative specifications changed the interpretations for the other three PH indicators, hence all further analyses use first-trimester bleeding (any versus none) reported from the maternal interviews.

Table 4.   Odds ratios and 95% confidence intervals for the association between four indicators of placental haemorrhage and preterm delivery, overall and stratified by delivery timing (n = 996)

BMI, body mass index; PTD, preterm delivery. *Model includes all four types of evidence of placental haemorrhage. **Model includes all four types of evidence of placental haemorrhage, maternal race, marital status, and BMI (categorical). ***Model includes all four types of evidence of placental haemorrhage, maternal race, marital status, BMI (categorical), and hypertension. ****OR and 95% CI calculated using weighted logistic regression models. *****OR and 95% CI for PTD in both time intervals calculated simultaneously using weighted polytomous logistic regression models. Boldface denotes P < 0.05.
	PTD (n)	Unadjusted		Mutually adjusted*		Adjusted for maternal characteristics**		Adjusted for maternal characteristics and hypertension***
		OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
Preterm delivery overall
<37 weeks****	238
Placental abruption	19	5.5	2.4, 12.8	4.2	1.8, 9.8	4.0	1.7, 9.7	3.8	1.5, 9.5
Disc-impacting blood clot	26	2.2	1.3, 4.0	1.7	0.9, 3.1	1.8	1.0, 3.2	1.5	0.8, 2.9
Microscopic haemorrhage (top quintile)	72	1.7	1.2, 2.5	1.6	1.1, 2.3	1.6	1.1, 2.4	1.6	1.1, 2.3
First-trimester bleeding	66	1.7	1.2, 2.5	1.6	1.1, 2.4	1.7	1.2, 2.5	1.7	1.2, 2.5
Preterm delivery stratified by timing
<35 weeks*****	77
Placental abruption	11	10.2	3.9, 26.5	6.4	2.4, 17.4	5.5	1.7, 17.2	5.3	1.7, 17
Disc-impacting blood clot	14	4.2	2.0, 8.5	2.6	1.2, 5.5	2.8	1.3, 6.0	2.4	1.1, 5.3
Microscopic haemorrhage (top quintile)	33	3.0	1.8, 4.9	2.4	1.4, 4.1	2.5	1.4, 4.3	2.5	1.4, 4.3
First-trimester bleeding	25	2.1	1.2, 3.5	1.8	1.0, 3.2	1.9	1.1, 3.5	1.9	1.0, 3.5
35–36 weeks*****	161
Placental abruption	8	3.5	1.3, 9.6	3.0	1.1, 8.4	3.1	1.1, 8.8	2.9	1.0, 8.5
Disc-impacting blood clot	12	1.4	0.7, 2.9	1.2	0.6, 2.7	1.3	0.6, 2.7	1.1	0.5, 2.5
Microscopic haemorrhage (top quintile)	39	1.3	0.8, 2.0	1.2	0.8, 1.9	1.3	0.8, 2.0	1.3	0.8, 2.0
First-trimester bleeding	41	1.6	1.0, 2.4	1.5	1.0, 2.4	1.6	1.1, 2.5	1.6	1.1, 2.5

When we stratified PTD by gestational week, all four PH indicators had stronger associations with PTD at <35 weeks of gestation than with PTD at 35–36 weeks in all models (Table 4). All four PH indicators were more strongly associated with PTD < 35 weeks than PTD at 35–36 weeks.

When we stratified PTD by delivery circumstances, ORs did not differ meaningfully between spontaneous and medically indicated PTD in unadjusted, mutually adjusted, or maternal characteristics-adjusted models (Table 5). After adding hypertension, the OR of 2.2 for disc-impacting blood clot and medically indicated PTD observed in the maternal characteristics-adjusted model was reduced to 1.1. This finding may be attributable to the expected strong association between hypertension and medically indicated PTD (OR = 9.9, unweighted) in conjunction with a more modest association between hypertension and disc-impacting blood clot among term (OR = 3.0 unweighted) and indicated preterm (OR = 2.6 unweighted) deliveries. However, small numbers in some cells (i.e. only nine women with indicated PTD had disc-impacting blood clots, and six of these had hypertension) warrant caution in interpreting this finding.

Table 5.   Odds ratios and 95% confidence intervals for the association between four indicators of placental haemorrhage and spontaneous or medically indicated preterm delivery

	PTD (n)	Unadjusted		Mutually adjusted*		Adjusted for maternal characteristics**		Adjusted for maternal characteristics and hypertension***
		OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
PTD, preterm delivery. *Model includes all four types of evidence of placental haemorrhage. **Model includes all four types of evidence of placental haemorrhage, maternal race, marital status and BMI (categorical). ***Model includes all four types of evidence of placental haemorrhage, maternal race, marital status, BMI (categorical) and hypertension. ****OR and 95% CI for both PTD subtypes calculated simultaneously using weighted polytomous logistic regression models. Boldface denotes P < 0.05.
Preterm delivery, spontaneous****	169
Placental abruption	13	5.4	2.2, 13.2	4.3	1.7, 10.6	4.0	1.6, 10.1	3.9	1.5, 10.0
Disc-impacting blood clot	17	2.0	1.0, 3.8	1.5	0.8, 3.0	1.6	0.8, 3.1	1.7	0.9, 3.4
Microscopic haemorrhage, top quintile	51	1.7	1.1, 2.5	1.6	1.0, 2.4	1.6	1.1, 2.5	1.6	1.1, 2.5
First-trimester bleeding	44	1.6	1.1, 2.4	1.5	1.0, 2.3	1.6	1.0, 2.5	1.6	1.1, 2.5
Preterm delivery, indicated****	69
Placental abruption	6	5.9	2.0, 18.1	4.1	1.3, 13.1	4.0	1.2, 13.9	3.0	0.6, 14.6
Disc-impacting blood clot	9	2.9	1.3, 6.6	2.2	0.9, 5.2	2.2	0.9, 5.3	1.1	0.4, 3.4
Microscopic haemorrhage, top quintile	21	1.8	1.0, 3.2	1.6	0.9, 2.8	1.6	0.9, 3.0	1.7	0.8, 3.2
First-trimester bleeding	22	2.0	1.2, 3.6	1.9	1.1, 3.4	2.1	1.2, 3.9	2.2	1.1, 4.3

Although numbers were very small in some cells, we further subdivided spontaneous PTD into spontaneous preterm labour and PPROM. Only microscopic haemorrhage was associated with PPROM (adjusted OR = 2.0, 95% CI 1.1, 3.5); the other three measures had stronger associations with spontaneous preterm labour (not shown).

Finally, we considered the possibility that mode of delivery may impact placental pathology findings so we repeated the analyses in Tables 4 and 5 after excluding 73 preterm and 194 term caesarean deliveries. Results for disc-impacting blood clot and microscopic haemorrhage were not meaningfully different, and the associations between placental abruption and PTD subtypes were of a greater magnitude. We also repeated the analyses after excluding women who reported taking aspirin (n = 45), having an amniocentesis (n = 21), or having chorionic villus sampling (n = 6) during the index pregnancy, and results were not meaningfully different.

Discussion

A key finding of this study was that evidence of PH identified through objective gross and microscopic placental pathology examinations (i.e. disc-impacting blood clot and microscopic haemorrhage) were associated with PTD at <35 weeks even after accounting for clinically evident bleeding (i.e. placental abruption and first-trimester bleeding). Previously, Salafia et al. found that haemosiderin in the decidua or extraplacental membranes was more common in very preterm (<32 weeks of gestation) deliveries than in a sample of uncomplicated, healthy term deliveries in an unadjusted analysis.¹¹ Associations between early pregnancy vaginal bleeding and subsequent PTD^2–4,28 or placental abruption^3,8,12 have previously been identified. Our study builds on these earlier findings by considering these clinically evident manifestations of PH in conjunction with subclinical placental pathology findings.

Previous studies have shown that pathology findings consistent with placental abruption are more common than clinical diagnoses.^5,29 Some have suggested that placental pathology-based abruption-related findings may be inconsequential in the absence of clinical suspicion, given that most pregnancies with such findings have unremarkable outcomes.⁶ In the POUCH Study, two types of evidence of PH from pathology examinations—disc-impacting blood clot (prevalence 5.6%) and microscopic haemorrhage (prevalence 20.4%, based on a distributional cut-point)—were clearly more common than placental abruption cases (prevalence 2.0%), and the majority of women with disc-impacting blood clots and high microscopic haemorrhage scores delivered at term without placental abruption. However, after accounting for placental abruption cases in multivariable models, we identified excess PTD risk associated with both of these placental pathology findings, particularly for PTD occurring at <35 weeks. Few studies have empirically demonstrated an association between subclinical evidence of haemorrhage in the delivered placenta and PTD by comparing preterm and term placentas. In previous work from the POUCH Study,²⁵ the microscopic haemorrhage construct was found to be associated with both spontaneous and medically indicated PTD occurring at <35 weeks. However, information on previous vaginal bleeding episodes and placental abruption was not considered in that study, so it was unknown whether the observed results were primarily attributable to clinically evident haemorrhage.

Retroplacental clots are sometimes employed in placental abruption clinical diagnoses, and have been required to confirm placental abruption cases in some epidemiological studies,^30–32 so the limited association between placental abruption and disc-impacting blood clots is noteworthy. It is important to distinguish the disc-impacting blood clot construct captured by the POUCH Study pathologist from adherent retroplacental clots visualised on a freshly delivered placenta by an attending clinician. Data on the latter are not uniformly recorded in patient charts so we relied on a gross placental pathology protocol to infer the presence of a clot before delivery. Collected blood not resulting from haemorrhage can become sequestered in the space between the membranes and disc and become firm with formalin fixation, resembling a true clot from haemorrhage. For this reason, the pathologist identified clots associated with disc tissue changes that would support a significant retroplacental clot on gross examination alone. Disc-impacting blood clots undiagnosed as placental abruption may signal concealed (possibly intraplacental or dissecting) haemorrhage, a less severe clinical picture, or a case with a low index of suspicion for placental abruption. Placental abruption diagnoses in the absence of disc-impacting blood clots could occur in very acute abruptions in which clots do not have time to organise and cause tissue reactions before delivery.³³

We found that microscopic haemorrhage was modestly associated with PTD < 37 weeks (mutually adjusted OR = 1.6) and PTD < 35 weeks (mutually adjusted OR = 2.4), and these associations remained after maternal demographic characteristics and hypertension were added to the models. Although it is possible that findings in this construct represent intermediate steps on a pathway that sometimes leads to overt placental abruption, most cases of placental abruption (63%, see Table 2) did not have high scores on this construct. We hypothesised that gestational vaginal bleeding remote from delivery reported by mothers, which has been linked to placental abruption and PTD in this and other studies,^{2–4,8,12,34} might signal chronic, slow haemorrhage, and would therefore be associated with high microscopic haemorrhage scores, but this was not the case. Microscopic haemorrhage was associated with PPROM, while the three other manifestations of PH were not. Possible mechanisms linking specific items included in our microscopic haemorrhage construct to spontaneous PTD and PPROM have previously been suggested: thrombin generated in response to decidual haemorrhage may cause cervical ripening, contractions and membrane degradation resulting in preterm labour or PPROM;³⁵ and tissue haemosiderin, an iron compound produced during the breakdown of blood, may exert tissue irritant and proinflammatory effects that contribute to preterm labour.¹¹ Additional research is needed to discover the antecedents of high microscopic haemorrhage scores.

The strengths and limitations of this study primarily derive from the prospective cohort design and the placental pathology protocol. Statistical power to detect significant associations differed among the four PH indicators given their widely varying prevalence, which ranged from 2.0 to 20.4%. Sample size limitations precluded assessing risk of very early PTD (i.e. <32 weeks). However, by performing the pathology examination on a large sample of term and preterm deliveries we were able to empirically evaluate the relations among PTD and four PH indicators. Like most studies that have investigated gestational vaginal bleeding in relation to pregnancy outcome, we cannot discern the actual origin of the blood—some may be from the placenta, whereas other reported bleeding may have originated in the cervix or vaginal tract. Data on first-trimester bleeding and substance use were ascertained prospectively; therefore this information was not subject to differential reporting based on mother’s or clinician’s knowledge of pregnancy outcome.

The study population was drawn from a well-characterised cohort, which was very similar to community women giving birth in the study years based on race-specific birth certificate comparisons.³⁶ The incidence of placental abruption observed in this study (2.0%) is higher than recent population-based estimates using administrative data and vital records in the United States.^7,37 Birth certificates and hospital discharge data have been shown to have limited sensitivity for ascertainment of maternal morbidity compared with medical chart review and application of standardised definitions,³⁸ the method used in the POUCH Study. Furthermore, the POUCH Study included a relatively high proportion of African American women, and this demographic group is known to have a higher rate of abruption diagnoses compared with whites.³⁷ The higher than expected incidence of placental abruption was not otherwise attributable to the sampling scheme, because appropriate weighted analyses were used to account for oversampling of PTD and women with high α-FP.

Each of the four manifestations of PH considered in this study contributed information regarding PTD risk. Some of the PH indicators were associated with one another but others were not. Many placental abruption cases did not show strong evidence of PH in their placental pathology examinations. The PH indicators diverged somewhat with respect to their relations with maternal characteristics and PTD subtypes. Taken together, these findings suggest potential heterogeneity in the hypothesised PH ‘iceberg’. In the future, it may be helpful to consider both clinical and subclinical manifestations of PH in relation to biomarker data (e.g. gene polymorphisms,³⁹ antiangiogenic factors) to gain insight into broader pathways to PTD that may involve disrupted uteroplacental vascular integrity.

Disclosure of interests

None.

Contribution to authorship

J.G. contributed to the concept, design and drafting of manuscript and to data analysis and interpretation. C.H. contributed to the concept, design of manuscript, data acquisition, analysis and interpretation and to the critical revision for intellectual content. P.S. contributed to the acquisition of data, performance of placental pathology examinations and critical revision for intellectual content. M.R. and M.W. contributed to data interpretation and to critical revision for intellectual content. D.P. contributed to analysis and interpretation of data and to the critical revision for intellectual content. R.F. contributed to data interpretation and manuscript review. All authors gave their final approval of the version to be published.

Details of ethics approval

The POUCH study has maintained IRB approval from all participating institutions since the Study’s inception.

Funding

This work was supported by a Perinatal Epidemiological Research Initiative Program Grant from the March of Dimes Foundation (Grants 20FY01-38 and 20-FY04-37), the National Institute of Child Health and Human Development and the National Institute of Nursing Research (Grant R01 HD34543), the Thrasher Research Foundation (Grant 02816-7) and the Centers for Disease Control and Prevention (Grant U01 DP000143-01).

Acknowledgements

The authors would like to acknowledge the work contributed by the histotechnicians in the MSU Investigative Histopathology Laboratory. The authors would also like to thank Dr Bertha Bullen, project director, the Prematurity Study Group, and Drs Joseph Marshall and Judith Suess, physician abstractors, and Drs Nazish Siddiqi and Judith Suess and Ms Lynn Thelen for reviewing suspected cases of placental abruption.