Formulating a Meaningful and Comprehensive Placental Phenotypic Classification

Abstract

Introduction:

While many placental lesions have been identified and defined, the significance of multiple overlapping lesions has not been addressed. The purpose of our analysis was to evaluate overlapping patterns of placental pathology and determine meaningful phenotypes associated with adverse birth outcomes.

Methods:

Placental pathology reports were obtained from a single hospital between 2009 and 2018. Placental lesions were grouped into four major categories: acute inflammation (AI), chronic inflammation (CI), maternal vascular malperfusion (MVM), and fetal vascular malperfusion (FVM). Within each category, lesions were classified as not present, low grade or high grade. Combinations of pathologies were evaluated in relation to preterm birth (<37 weeks) and small for gestational age (SGA) infant (birthweight <10th percentile).

Results:

During the study period, 19,027 placentas were reviewed by pathologists. Results from interaction models indicate that MVM and MVM in combination with CI and/or FVM are associated with the greatest odds of SGA infant and PTB. When incorporating grade, we identified 21 phenotype groups, each with characteristic associations with the SGA infant and patterns of PTB.

Discussion:

We have developed a comprehensive and meaningful placental phenotype that incorporates severity and multiplicity of placental lesions. We have also developed a web application to facilitate phenotype determination (https://placentaexpression.shinyapps.io/phenotype).

Introduction

The Amsterdam consensus conference of placental pathology terminology¹ and the subsequent Dublin conference² went far to define, grade, and stage the major pathologic categories (acute inflammation (AI), chronic inflammation (CI), maternal vascular malperfusion (MVM), and fetal vascular malperfusion (FVM)), and many minor pathologic lesions in the placenta. The goal of standardized criteria and terminology is to allow research into the significance and clinical implications of these placental lesions. Furthermore, standardized criteria and terminology can help individual clinicians and pathologists to interpret the findings in any one placenta. As a result, over 40 placental lesions have been defined and have become easier to diagnose and report. Furthermore, researchers have identified that placental lesions are associated with preterm birth,^3,4 poor fetal growth,^5,6 and stillbirth,^7,8 as well as long-term outcomes for both the infant^9,10 and the parent.¹¹

However, gaps remain regarding the most ideal reporting and interpretation of placental pathologic findings. First, of the four main placental lesions, not all are graded or staged in terms of severity. There is currently no comprehensive grading system for the constellation of CI, except for chronic villitis, and no accepted grading system for MVM. Second, the interpretation of multiple placental lesions of varying types and duration is not well understood, and how a certain combination of placental lesions compare with one lesion in isolation or a different combination of lesions is also problematic. One such example is the placenta with low-grade chronic villitis, low-stage acute inflammation and high-grade fetal vascular malperfusion. Is the significance of these lesions dependent upon the highest grade lesion? The most acute lesion? The most chronic lesion? Or a combination of all of these factors? Recognizing that lesion multiplicity may be important, some studies have utilized a cumulative score based on number of lesions present and reported significant associations with preeclampsia, intrauterine growth restriction, and neonatal encephalopathy.^12-14 However, this assumes all included lesions contribute similarly to risk and does not capture important differences in severity and timing across lesions. One study that evaluated overlapping lesions reported that severity and timing matter for infant neurologic impairment.¹⁵

Our goal was to first develop grading systems for all major placental lesions, and then define specific placental phenotypes based on a combination of all lesions present by grade. The overall effect would be a comprehensive placental phenotypic classification system that could be applied to any placental examination. Such a system could potentially improve the clinical understanding of placental pathology and its association with adverse pregnancy outcomes, and advance research into outcomes by ensuring comparison of similar placental phenotypes.

Therefore, we sought to 1) define a finite number of placental phenotypes, characterized by the constellation of all low and high-grade lesions in the major placental pathologic categories (acute and chronic inflammation and maternal and fetal vascular malperfusion) using a large, well-characterized cohort and 2) compare the prevalence of each phenotype amongst two well-known adverse outcomes: the small for gestational age (SGA) infant and preterm birth (PTB).

Materials and Methods

Study Sample

We utilized data from singleton livebirths at a single Chicago hospital between January 2009 and March 2018. Demographic characteristics, birth outcomes, and the final diagnosis text field from the placental pathology report were abstracted from medical records. Information from medical records was obtained through the Northwestern Medicine Enterprise Data Warehouse and the study was approved by the Institutional Review Board at Northwestern University.

During the study period, there were 103,688 eligible singleton live births and 19,027 (18.4%) had complete placental pathology reports. Indications for pathologic review included abnormal placental appearance, intrauterine growth restriction, preterm delivery <34 weeks, PPROM <34 weeks, severe preeclampsia, clinical concern for TORCH infection, chorioamnionitis, or abruption, and certain neonatal conditions. Pathology reports were completed by perinatal pathologists. All placentas were grossed in a routine, systematic manner by trained staff and trimmed weight was recorded. Histologic examination included routine review of sections of the extraplacental membranes, umbilical cord, 2 full thickness parenchymal sections, biopsies of the maternal surface, and representative sections of all gross lesions. Microscopic pathology was recorded in the pathology report according to accepted terminology for the time period and detailed the presence of acute inflammatory lesions, chronic inflammatory lesions, maternal vascular pathology, fetal vascular pathology and miscellaneous lesions such as massive perivillous fibrin deposition, chorangiosis, and meconium-associated vascular necrosis.

Individual diagnostic lines of the pathology report were searched for words used to describe specific types of lesions using R version 4.0 (stringr package). A subset of diagnostic lines flagged for a lesion were reviewed to ensure a correct identification and lines not flagged, but containing some (but not all) of the diagnostic terms were reviewed. Manual changes were made to address some errors, such as typographical errors in the pathology reports. A complete list of lesions and the search terms used to identify them can be found in Table S1. To validate the automated process for identifying lesions, 20 reports from each year of the study were randomly selected (200 total) and manually reviewed. Manual text review involved reading pathology reports and recording lesion presence and did not involve reviewing the slides themselves. The lesions identified manually were compared to those identified through the automated text search process. Kappa statistics were calculated to assess interrater reliability between the automated and manual text reviews of the pathology diagnoses.¹⁶

Placental Lesion Definitions

Placental gross and histologic lesions of interest were grouped into four major categories: acute inflammation (AI), chronic inflammation (CI), fetal vascular malperfusion (FVM), and maternal vascular malperfusion (MVM). Within each category, lesions were classified as not present, low grade, and high grade (Table 1). ‘Grade’ of AI was determined based on the accepted Amsterdam staging system.¹ A placenta was classified as having high-grade AI if either high-stage maternal or high-stage fetal acute inflammation was present. Chronic inflammatory lesions were summarized by compartment¹⁷ (chorionic plate/membranes, basal plate, villi, intervillous space, and fetal vasculature) to capture the extent of chronic inflammation. Linear trends were identified between number of compartments with CI and adverse outcomes; as the number of compartments affected by CI increased, the prevalence of SGA and preterm birth increased (p < 0.01). Based on the distribution of the scores, the presence of CI in two or more compartments was considered high grade (Figure S1). FVM was identified based on presence of a thrombus or intramural fibrin deposition in at least one fetal vessel (umbilical, chorionic, velamentous, stem villous) and/or avascular villi/villous stromal-vascular karyorrhexis. FVM was considered high grade based on Amsterdam criteria (more than one focus of avascular villi (≥45 villi) or two or more occlusive or nonocclusive thrombi in the chorionic plate or stem villi, or multiple nonocclusive thrombi), documented using the phrase, “fetal thrombotic vasculopathy” or “high-grade fetal vascular malperfusion” (see Table S1). MVM was scored based on the number of lesions present, with one point for each of the following lesions: fibrinoid necrosis/acute atherosis, muscularization of basal plate arterioles, mural hypertrophy of membrane arterioles, basal decidual vascular thrombus, single infarct, increased syncytial knots, villous agglutination, increased perivillous fibrin deposition, distal villous hypoplasia/small terminal villous diameters, retroplacental blood/hematoma, and two points for 3 specific lesions: multiple infarcts, retroplacental blood/hematoma with hemosiderin or infarct, and an SGA placenta. An SGA placenta alone was not considered indicative of MVM; however, in the presence of ≥1 MVM lesion, two points were added to the score. Based on distributions, MVM scores of 0 and 1 were considered not to be MVM, scores of 2–3 were considered low grade and scores of ≥4 were considered high grade (Figure S2). We also identified a few other significant placental conditions not classified as CI, FVM or MVM, but felt to be of potential importance for adverse outcomes, including chorioamnionic hemosiderosis, massive perivillous fibrin deposition, fetal vessel myonecrosis, amnion nodosum, delayed villous maturation, chorangiosis, and isolated SGA placenta. We assigned placentas with lack of lesions in any of the above categories as “histologically normal”.

Table 1.

Key Variable Definitions.

Variable	Lesions Includeda		Definition
Acute inflammation	Maternal: acute subchorionitis, acute chorionitis, acute amnionitis, acute necrotizing chorioamnionitis Fetal: acute chorionic vasculitis, phlebitis, arteritis/panvasculitis, funisitis	High-grade	High stage maternal and/or fetal, based on Amsterdam Criteria
		Low-grade	Low-stage maternal and/or fetal, based on Amsterdam Criteria
		None	No acute inflammatory lesions
Chronic inflammation	Membranes/chorionic plate: chronic chorionitis, chronic amnionitis Basal plate: chronic deciduitis with plasma cells, chronic decidual perivasculitis Villi: villitis, intravillous plasma cells Intervillous space: chronic intervillositis Fetal: chronic eosin/t-cell vasculitis, chronic fetal inflammation- not specified	High-grade	≥2 compartments with chronic inflammation
		Low-grade	1 compartment with chronic inflammation
		None	0 compartments with chronic inflammation (no chronic inflammatory lesions)
Fetal vascular malperfusion	Thrombi or intramural fibrin deposition in chorionic vessel, velamentous vessel, stem villous vessel, and/or umbilical vessel; avascular villi/villous stromal vascular karyorrhexis	High-grade	High grade FVM or fetal thrombotic vasculopathy, defined as more than one focus of avascular villi (≥45 villi) or two or more occlusive or nonocclusive thrombi in the chorionic plate or stem villi, or multiple nonocclusive thrombi
		Low-grade	Any fetal vascular lesions, not indicated as high grade
		None	No fetal vascular lesions
Maternal vascular malperfusion	1 point: fibrinoid necrosis/acute atherosis, muscularization of basal plate arterioles, mural hypertrophy of membrane arterioles, basal decidual vascular thrombus, single infarct, increased syncytial knots, villous agglutination, increased perivillous fibrin deposition, distal villous hypoplasia, retroplacental blood/hematoma 2 points: multiple infarcts, retroplacental hematoma with hemosiderin or infarct, placental hypoplasiab	High-grade	Score of ≥4
		Low-grade	Score of 2–3
		None	Score of 0–1
Other significant pathology	Hemosiderosis, massive perivillous fibrin deposition, myonecrosis, amnion nodosum, delayed villous maturation, chorangiosis, isolated SGA placenta (placental hypoplasia)	Not graded	Presence of ≥1 lesion

^aSee Table S1 for specific lesion definitions.

^bPlacental hypoplasia only considered in the presence of ≥1 maternal vascular lesion (isolated placental hypoplasia not scored).

Outcome Definitions

Gestational age at delivery and birthweight were abstracted from medical records. Birthweight percentiles were determined based on Fenton growth charts.¹⁸ An SGA infant was identified based on a birthweight <10th percentile for gestational age and infant sex. Fetal-placental weight ratio was determined based on birthweight and placental weight. Preterm birth was defined as gestational age <37 weeks. We also evaluated early preterm birth (<34 weeks) and late preterm birth (34-36 weeks).

Statistical Analysis

We used logistic models to evaluate interactions between the four placental lesion categories simplified into dichotomous variables (absent vs. present (low or high grade)) to assess the impact of lesion multiplicity on SGA infant and PTB. Additionally, to incorporate the differences between low and high grade placental lesions, we calculated the prevalence of SGA infant and PTB among 81 combinations of the three-level AI, CI, FVM, and MVM variables (3⁴) and 3 combinations of AI (none, low, high) with other significant pathology for 84 total combinations (referred to as phenotypes). In the tables, we generally use 4-letter sequences to reflect the phenotypes, with A indicating acute inflammation, C indicating chronic inflammation, F indicating fetal vascular malperfusion, and M indicating maternal vascular malperfusion. Capitalized letters indicate high-grade pathology and lower case letters indicate low-grade pathology. Lastly, we grouped phenotypes together with similar characteristics and prevalence of SGA infant and PTB, and referred to these as phenotype groups.

We used logistic regression to calculate odds ratios for SGA infant by comparing each phenotype group to the histologically normal group. As a sensitivity analysis, we investigated associations with SGA defined as <5th percentile. We also used linear regression to estimate associations between phenotype groups and fetal-placental weight ratio. In addition, we compared distributions of early and late preterm birth by phenotype group. To account for selection bias due to indication for pathologic review, we replicated findings using stabilized inverse probability weights.¹⁹ To calculate the analysis weights, we used the full sample to model the probability of being selected into the pathology sample based on maternal and pregnancy characteristics. After taking the inverse of the probabilities, those who are overrepresented in the pathology sample relative to the full sample (high probability of selection) have a smaller weight and those who are underrepresented in the pathology sample relative to the full sample (low probability of selection) have a larger weight. Weighting the sample by the inverse probability of selection ensures that the distribution of maternal and pregnancy characteristics in the pathology sample is similar to that of the full sample. SAS version 9.4 (SAS Institute INC., Cary, North Carolina) was used for statistical analysis and an alpha level of 0.05 was used to determine statistical significance. Bonferonni corrections were applied to account for multiple testing where appropriate.

Results

Placental pathology reports were completed for 19,027 of the 103,688 eligible patients (18.4%). The average age of patients with completed placental pathology reports was 32.0 years (standard deviation 5.5), and 53.6% were identified as white race in the EMR (Table 2). One quarter of deliveries were preterm and 15.9% were SGA. Patients with completed pathology reports were more likely to be nulliparous, deliver via cesarean section, deliver preterm, and deliver an SGA infant than patients without pathology reports. In a review of 200 randomly selected pathology reports, kappa statistics for key variables (grade of AI, CI, FVM, MVM, and other significant pathology) were all >0.90, indicating that the automated text review was able to accurately identify lesions listed in the pathology reports.¹⁶

Table 2.

Descriptive Characteristics of the Study Sample (n = 19,027) and the Sample Without Pathology (n = 84,666).

n (%) or Mean (SD)	Included Pathology Sample (n = 19,027)	Excluded Sample Without Pathology (n = 84,666)
Maternal age	32.0 (5.5)	31.8 (5.2)
Maternal race/ethnicity*
White	9,540 (53.6)	47,678 (59.9)
Black	2,649 (14.9)	7,571 (9.5)
Hispanic	3,113 (17.5)	14,559 (18.3)
Asian	1,487 (8.4)	5,842 (7.3)
Other	995 (5.6)	3,930 (4.9)
Nulliparous	12,146 (64.2)	40,782 (48.5)
Cesarean section	7,143 (37.6)	19,875 (23.5)
Sex, male	9,895 (52.0)	43,045 (50.8)
Preterm birth, <37 weeks	4,900 (25.8)	3,200 (3.8)
Preterm birth, <34 weeks	2,004 (10.5)	1,045 (1.2)
Small for gestational age infant	3,020 (15.9)	4,391 (5.2)

^*Missing for 1,243 of the sample with pathology and 5,086 of the sample without pathology.

Interaction

In the interaction model for SGA infant, FVM and MVM alone were associated with increased odds of SGA infant, while AI and CI alone were not associated with increased odds of SGA infant (Figure 1). MVM in combination with either or both CI and FVM was associated with greater odds of SGA infant than MVM alone (odds ratio range: 3.8 to 5.0). Similarly, CI and FVM together were associated with greater odds of SGA infant than either alone. AI in combination with FVM was less likely to result in SGA infant than FVM alone.

Figure 1.

Associations between combinations of placental pathology and small for gestational age (SGA) infant. Each type of pathology (AI, CI, FVM, MVM) is represented by a color (orange for AI, blue for CI, green for FVM and red for MVM) and associations (odds ratios and confidence intervals) between individual types of pathology and SGA infant are represented by solid color rectangles and color-corresponding dashed lines. Combinations of pathology are represented by multi-colored rectangles. To determine if a combination of pathology has increased or decreased odds of SGA infant as compared to an individual type of pathology, compare one of the multi-colored rectangles to either of the color-corresponding dashed lines. If the confidence interval does not intersect the dashed line, then the combination of pathologies is statistically significantly different from the individual pathology alone. Both MVM and FVM are associated with statistically significantly increased odds of an SGA infant while CI and AI are not. MVM alone has the highest odds of SGA infant when compared with any other individual pathologic categories. Combinations of FVM and CI or both FMV and CI with MVM statistically increase the odds of SGA infant over MVM alone while MVM paired with AI does not. The combination of FVM and CI has statically significantly increased odd of SGA infant over either pathology alone. Abbreviations: AI – acute inflammation; CI – chronic inflammation; FVM – fetal vascular malperfusion; MVM – maternal vascular malperfusion.

In the interaction model for PTB, CI and MVM alone were associated with increased odds of PTB and MVM and CI in combination were associated with even greater odds of PTB (odds ratio range: 2.3 to 3.3) (Figure 2). AI and FVM alone were not associated with increased odds of PTB.

Figure 2.

Associations between combinations of placental pathology and preterm birth (PTB, <37 weeks). Each type of pathology (AI, CI, FVM, MVM) is represented by a color (orange for AI, blue for CI, green for FVM and red for MVM) and associations (odds ratios and confidence intervals) between individual types of pathology and PTB are represented by solid color rectangles and color-corresponding dashed lines. Combinations of pathology are represented by multi-colored rectangles. To determine if a combination of pathology has increased or decreased odds of PTB as compared to an individual type of pathology, compare one of the multi-colored rectangles to either of the color-corresponding dashed lines. If the confidence interval does not intersect the dashed line, then the combination of pathologies is statistically significantly different from the individual pathology alone. Both MVM and CI are associated with statistically significantly increased odds of PTB while FVM and AI are not associated with increased odds of PTB. MVM alone has the highest odds of PTB when compared with any other individual pathologic categories. The combination of MVM and CI is associated with even greater odds of PTB than either MVM or CI alone. Abbreviations: AI – acute inflammation; CI – chronic inflammation; FVM – fetal vascular malperfusion; MVM – maternal vascular malperfusion.

Phenotypes

Among the 84 phenotypes that incorporate grading, the most common were AI alone (low-grade: 8.6%, high-grade: 5.7%) and the histologically normal phenotype (7.4%) (Table S2). From the 84 phenotypes, ten main phenotype groups were derived, six of which required subcategories resulting in 21 total groups (group membership is indicated using numbers) (Tables 3 and 4). A decision tree for identifying the 21 groups is presented in Figure 3 and relationships with SGA infant and PTB are discussed below. Additional details on how the 84 phenotypes are grouped into 21 groups can be found in Table S2.

Table 3.

Prevalence and Odds Ratios of Small for Gestational Age (SGA) Infant and Fetal-Placental Weight Ratio for Each Placental Pathology Group (n = 19,027).

Group Name and Description	Group Number	Included Phenotypes	N	Percent	SGAa			Fetal-Placental Weight Ratiob
					Percent SGA	OR	95% CI	Mean	β (95% CI)
Multiple high-grade pathologies (C, F, M)	1		771	4.1%	39.0%	14.08	10.49, 18.90	7.77	0.86 (0.75, 0.96)
All three high-grade pathologies (triple threat)	1a	CFM, aCFM, ACFM	34		38.2%
High-grade fetal AND maternal vascular pathology	1b	FM, cFM, aFM, acFM, AFM, AcFM	113		48.7%
High-grade maternal vascular pathology AND high-grade chronic inflammation	1c	CM, CfM, aCM, aCfM, ACM, ACfM	536		38.2%
High-grade fetal vascular pathology AND high-grade chronic inflammation	1d	CF, CFm, aCF, aCFm	88		31.8%
High-grade acute inflammation AND high-grade fetal vascular pathology (with or without chronic inflammation and/or low-grade maternal vascular pathology)	2	AF, AFm, AcF, AcFm, ACF, ACFm	122	0.6%	11.5%	2.85	1.54, 5.26	7.05	0.01 (−0.22, 0.23)
High-grade fetal vascular pathology
With or without low-grade acute inflammation	3a	F, aF	160	0.8%	14.4%	3.69	2.21, 6.15	6.97	−0.10 (−0.30, 0.10)
WITH low-grade chronic inflammation AND/OR low-grade maternal vascular pathology (with or without low-grade acute inflammation)	3b	cF, Fm, cFm, acF, aFm, acFm	171	0.9%	29.8%	9.34	6.16, 14.17	7.22	0.24 (0.05, 0.44)
High-grade chronic inflammation
With or without low-grade fetal vascular pathology	4a	C, Cf	654	3.4%	13.5%	3.42	2.43, 4.81	7.22	0.33 (0.22, 0.45)
WITH any acute inflammation (with or without low-grade fetal vascular pathology)	4b	aC, aCf, AC, ACf	719	3.8%	11.0%	2.71	1.92, 3.84	7.28	0.26 (0.15, 0.37)
WITH low-grade maternal vascular pathology (with or without low-grade fetal vascular pathology)	4c	Cm, Cfm	263	1.4%	27.0%	8.13	5.59, 11.82	7.50	0.54 (0.38, 0.70)
WITH low-grade maternal vascular pathology AND any acute inflammation	4d	aCm, aCfm, ACm, ACfm	276	1.5%	19.9%	5.47	3.70, 8.09	7.64	0.59 (0.43, 0.75)
High-grade maternal vascular pathology (with or without any acute inflammation and/or other low-grade pathology)	5	M, cM, fM, cfM, aM, acM, afM, acfM, AM, AcM, AfM, AcfM	2673	14.0%	30.1%	9.46	7.22, 12.38	8.03	0.95 (0.87, 1.03)
Low-grade fetal vascular pathology OR chronic inflammation
With or without low-grade acute inflammation	6a	c, f, ac, af	2574	13.5%	9.4%	2.29	1.72, 3.06	7.16	0.21 (0.13, 0.29)
WITH high-grade acute inflammation	6b	Ac, Af	917	4.8%	7.8%	1.87	1.32, 2.66	7.04	−0.01 (−0.11, 0.09)
Low-grade maternal vascular pathology OR any combination of low-grade chronic inflammation and fetal vascular malperfusion
Without acute inflammation	7a	m, cm, cf, fm, cfm	1680	8.8%	18.5%	4.97	3.74, 6.61	7.47	0.44 (0.36, 0.53)
WITH low-grade acute inflammation	7b	am, acm, acf, afm, acfm	1554	8.2%	18.7%	5.07	3.80, 6.75	7.52	0.50 (0.41, 0.58)
WITH high-grade acute inflammation	7c	Am, Acm, Acf, Afm, Acfm	924	4.9%	16.0%	4.19	3.07, 5.72	7.48	0.37 (0.27, 0.47)
Other significant pathology Acute inflammation alone	8		1455	7.6%	19.4%	5.29	3.96, 7.05	7.89	0.88 (0.79, 0.97)
High-grade acute inflammation	9a	A	1082	5.7%	4.3%	1.00	0.68, 1.47	6.91	−0.17 (−0.27, −0.08)
Low-grade acute inflammation	9b	a	1630	8.6%	4.9%	1.13	0.81, 1.60	6.95	−0.04 (−0.13, 0.04)
Histologically normal	10		1402	7.4%	4.3%	Reference		6.87	Reference

Abbreviations: SGA – small for gestational age, OR – odds ratio, CI – confidence interval, A – high-grade acute inflammation, a – low-grade acute inflammation, C – high-grade chronic inflammation, c – low-grade chronic inflammation, F – high-grade fetal vascular malperfusion, f – low-grade fetal vascular malperfusion, M – high-grade maternal vascular malperfusion, m – low-grade maternal vascular malperfusion.

^aOdds ratios from logistic regression (null value is 1).

^bParameter estimates (β) from linear regression (null value is 0); model adjusted for gestational age and infant sex (excludes 454 fragmented placentas).

Table 4.

Prevalence of Pathology Groups by Gestational Age Category (n = 19,027).

Group Name and Description	Group Number	Included Phenotypes	Percent Term ≥37 Weeks n = 14,127)	Percent Preterm <37 Weeks (n = 4,900)	Percent Preterm 34–36 Weeks (n = 2,896)	Percent Preterm <34 Weeks (n = 2,004)	Adjusted* p-Value, 34–36 Weeks vs. <34 Weeks
Multiple high-grade pathologies (C, F, M)	1		3.0%	7.1%	7.0%	7.4%	0.99
All three high-grade chronic pathologies (triple threat)	1a	CFM, aCFM, ACFM
High-grade fetal AND maternal vascular pathology	1b	FM, cFM, aFM, acFM, AFM, AcFM
High-grade maternal vascular pathology AND high-grade chronic inflammation	1c	CM, CfM, aCM, aCfM, ACM, ACfM
High-grade fetal vascular pathology AND high-grade chronic inflammation	1d	FC, CFm, aFC, aCFm
High-grade acute inflammation AND high-grade fetal vascular pathology (with or without chronic inflammation and/or low-grade maternal vascular pathology)	2	AF, AFm, AcF, AcFm, ACF, ACFm	0.8%	0.2%	0.0%	0.4%	0.06
High-grade fetal vascular pathology
With or without low-grade acute inflammation	3a	F, aF	0.9%	0.5%	0.7%	0.3%	0.99
WITH low-grade chronic inflammation AND/OR low-grade maternal vascular pathology (with or without low-grade acute inflammation)	3b	cF, Fm, cFm, acF, aFm, acFm	0.9%	0.8%	1.0%	0.5%	0.93
High-grade chronic inflammation
With or without low-grade fetal vascular pathology	4a	C, Cf	3.0%	4.7%	5.9%	3.0%	<0.01
WITH any acute inflammation (with or without low-grade fetal vascular pathology)	4b	aC, aCf, AC, ACf	4.4%	1.9%	2.2%	1.5%	0.99
WITH low-grade maternal vascular pathology (with or without low-grade fetal vascular pathology)	4c	Cm, Cfm	1.1%	2.2%	2.6%	1.6%	0.29
WITH low-grade maternal vascular pathology AND any acute inflammation	4d	aCm, aCfm, ACm, ACfm	1.6%	1.0%	1.1%	1.0%	0.99
High-grade maternal vascular pathology (with or without any acute inflammation and/or other low-grade pathology)	5	M, cM, fM, cfM, aM, acM, afM, acfM, AM, AcM, AfM, AcfM	10.2%	25.1%	19.0%	34.0%	<0.01
Low-grade fetal vascular pathology OR chronic inflammation
With or without low-grade acute inflammation	6a	c, f, ac, af	14.5%	10.8%	14.2%	5.7%	<0.01
WITH high-grade acute inflammation	6b	Ac, Af	5.5%	2.7%	0.8%	5.4%	<0.01
Low-grade maternal vascular pathology OR any combination of low-grade chronic inflammation and fetal vascular malperfusion
Without acute inflammation	7a	m, cm, cf, fm, cfm	7.3%	13.4%	14.7%	11.4%	0.01
WITH low-grade acute inflammation	7b	am, acm, acf, afm, acfm	9.5%	4.2%	3.8%	4.8%	0.99
WITH high-grade acute inflammation	7c	Am, Acm, Acf, Afm, Acfm	5.5%	3.1%	1.4%	5.7%	<0.01
Other significant pathology	8		8.9%	4.2%	4.9%	3.0%	0.02
Acute inflammation alone
High-grade acute inflammation	9a	A	6.4%	3.6%	1.9%	6.0%	<0.01
Low-grade acute inflammation	9b	a	10.2%	4.0%	4.9%	2.7%	<0.01
Histologically normal	10		6.3%	10.5%	13.9%	5.6%	<0.01

Abbreviations: SGA – small for gestational age, OR – odds ratio, CI – confidence interval, A – high-grade acute inflammation, a – low-grade acute inflammation, C – high-grade chronic inflammation, c – low-grade chronic inflammation, F – high-grade fetal vascular malperfusion, f – low-grade fetal vascular malperfusion, M – high-grade maternal vascular malperfusion, m – low-grade maternal vascular malperfusion.

^*P-values adjusted using Bonferonni correction.

Figure 3.

Decision tree for determining phenotype group membership based on acute inflammation (A/a), chronic inflammation (C/c), fetal vascular malperfusion (F/f), and maternal vascular malperfusion (M/m). Capital letters (‘big’) indicate high-grade pathology and lower case letters (‘little’) indicate low-grade pathology.

Overall, the most common groups among SGA infants were groups 5 (high-grade MVM, 26.6%), 7a (low-grade MVM or any combination of low-grade CI and FVM, 10.3%), 1 (multiple high-grade pathologies (C, F, M), 10.0%), and 7b (low-grade MVM or any combination of low-grade CI and FVM with low-grade AI, 9.6%) (Figure 4). Similarly, the most common groups among PTBs were groups 5 (25.1%) and 7a (13.4%). Groups 6a (low-grade FVM or CI with or without low-grade AI, 10.8%) and 10 (histologically normal, 10.5%) were also common among PTB.

Figure 4.

Distribution of pathology phenotype groups among small for gestational age (SGA) infants and preterm births.

Associations With SGA Infant

The placental group with the highest prevalence of SGA infants was group 1 (multiple high-grade pathologies (C, F, M) – includes 1a, 1b, 1c, 1d) (Table 3). On average, 39.0% of infants with placentas in group 1 were SGA, with a range of 31.8% (group 1d, high-grade FVM and high-grade CI) to 48.7% (group 1b, high-grade FVM and high-grade MVM). Placentas in group 1 were 14.1 times as likely to result in an SGA infant as compared to histologically normal placentas (95% confidence interval (CI): 10.5, 18.9). A high prevalence of SGA infants (>25%) was also observed for groups 5 (high-grade MVM), 3b (high-grade FVM with other low-grade pathology, without high-grade AI), and 4c (high-grade CI with low-grade MVM). Odds ratios for these groups ranged from 8.1 to 9.5.

A moderately high prevalence of SGA infants (15-25%) was observed for five groups. This group included high-grade CI with low-grade MVM (group 4d, 19.9%), all three groups characterized by low-grade MVM, with or without acute inflammation or other low-grade pathology (groups 7a, 7b, and 7c, range: 16.0% to 18.5%), and other significant pathology (group 8, 19.4%). Odds ratios for these groups ranged from 4.2 to 5.5.

The prevalence of SGA infants was mildly elevated (<15%) for six groups, including groups 3a (high-grade FVM with or without low-grade AI), 4a (high-grade CI with or without low-grade FVM), 2 (high-grade AI and high-grade FVM), and 4b (high-grade CI and any AI with or without low-grade FVM). A mild prevalence of SGA infants was also observed for both groups characterized by low-grade FVM or CI with or without AI (groups 6a and 6b). Odds ratios for groups with a mildly elevated prevalence of SGA infants ranged from 1.9 to 3.7.

Among placentas in groups 9a (high-grade AI alone) and 9b (low-grade AI alone), 4.3% and 4.9% of infants were SGA, respectively. These groups were not associated with odds of SGA infant as compared to histologically normal placentas (odds ratios range: 1.0 to 1.1). Among histologically normal placentas, 4.3% of infants were SGA.

Estimates of associations between placental groups and SGA infant were consistent when stabilized inverse probability weights were used to account for selection into the pathology sample (Table S3) and when SGA infant was defined as birthweight <5th percentile (Table S4).

Associations With Fetal-Placental Weight Ratio

Among histologically normal placentas, the mean fetal-placental weight ratio was 6.9, indicating that fetal weight was nearly 7 times greater than placental weight (Table 3). In a model adjusted for gestational age and infant sex, twelve of the phenotypes were associated with an increase in fetal-placental weight ratio as compared to the histologically normal group. The largest increases were observed for group 5 (high-grade MVM), group 8 (other significant pathology), and group 1 (multiple high-grade pathologies (C, F, M)). Group 5 was associated with a 0.95-point increase in fetal-placental weight ratio (95% CI: 0.87, 1.03), indicating that, compared to histologically normal placentas, fetal size was increased relative to placental size for placentas with high-grade MVM.

Patterns Among PTB

When comparing early PTB (<34 weeks) to late PTB (34-36 weeks), six groups were more common in late PTB (all adjusted p-values <0.05) (Table 4). Group 4a (high-grade CI with or without low-grade FVM) was present in 5.9% of late PTB and 3.0% of early PTB, group 6a (low-grade FVM or CI with or without low-grade AI) was present in 14.2% of late PTB and 5.7% of early PTB, and group 7a (low-grade MVM or any combination of low-grade CI and FVM) was present in 14.7% of late PTB and 11.4% of early PTB. Groups 8 (other significant pathology), 9b (low-grade AI alone) and 10 (histologically normal) were also more common in late PTB.

Four groups were more common in early PTB as compared to late PTB (Table 4). Group 5 (high-grade MVM) was the most common group among all PTB (25.1%) and was present in 34.0% of early PTB as compared to 19.0% of late PTB. The three other groups were all characterized in part by the presence of high-grade AI and included groups 6b (high-grade AI and low-grade FVM or CI), 7c (high-grade AI and low-grade MVM or any combination of low-grade CI and FVM), and 9a (high-grade AI alone).

Discussion

Our results indicate that severity and multiplicity of placental lesions are important when evaluating risk of SGA infant and PTB. Based on our findings, we have developed a novel classification system that considers both severity and multiplicity of placental lesions to identify placental phenotypes that have similar associations with two common and meaningful outcomes, the SGA infant and PTB. To facilitate clinical interpretation, we have generated a decision tree and web application (Figure 1; https://placentaexpression.shinyapps.io/phenotype). Pathologists can use the decision tree and/or web application to easily determine a placenta’s phenotype and phenotype group and generate an evidence-based pathology report with comments about the prevalence of the phenotype in PTB and the risk for SGA infant. We believe that these tools will help to acknowledge the importance of lesion severity and multiplicity for both clinical and research use.

Results from both the interaction models and the group analysis indicate that MVM (alone or in combination with CI and/or FVM) is associated with the greatest odds of SGA infant. This is consistent with other studies that report associations between MVM and measures of fetal growth, including SGA infant.^6,20-23 Our results also build on the literature by investigating lesion multiplicity; in both analyses, we observed even stronger associations with SGA infant among placentas with multiple high-grade lesions. This suggests that severity of MVM, as well as co-occurrence of other pathology is important for fetal growth. One exception to high-grade lesion multiplicity was group 2 (thromboinflammatory pathology), where high-grade AI is seen in combination with high-grade FVM, with or without high-grade CI. In this case, FVM may be characterized by recent thrombi resulting from endothelial injury induced by high-grade AI. Therefore, FVM may be a consequence of a more acute process that may not have the same impact on fetal growth as FVM lesions that develop independent of acute inflammation.²⁴ This finding was also consistent with the interaction model, where the combination of AI and FVM was associated with lower odds of SGA infant than FVM alone and in the analysis of fetal size relative to placental size, where there was no association between group 2 and fetal-placental weight ratio. Findings of attenuated associations (group 2) or null associations (groups 9a and 9b) between acute inflammation and SGA infant also indicate the importance of lesion duration for SGA infant, as acute inflammation may reflect a more recent pathologic change with less potential to impact fetal growth.

We also observed that high-grade MVM (alone or in combination with CI and/or FVM) was associated with an increase in fetal-placental weight ratio. Our results are consistent with another study that found maternal vascular lesions are associated with a higher fetal-placental weight ratio.²⁵ A larger fetal-placental weight ratio may reflect increased placental efficiency to compensate for MVM; however, higher fetal-placental weight ratios are also associated with adverse outcomes, indicating that compensation may be inadequate.^26,27 Further, a higher fetal-placental weight ratio may also indicate reduced reserve capacity, which may increase susceptibility to additional stressors.²⁶

In the group analysis, we observed that several groups characterized by AI (groups 6b, 7c, 9a) were more common among early PTB. Mechanistically, this may be a result of preterm birth secondary to premature rupture of membranes/clinical chorioamnionitis. Similarly, groups characterized by MVM (groups 5 and 7c) were also more common among early PTB, which may reflect hypertension-related MVM among indicated early PTB. Unfortunately, we were unable to separate indicated and spontaneous PTB to determine whether associations differ for these two groups. In contrast, several groups with CI (groups 4a, 6a, and 7a) were more common among late PTB, which is consistent with proposed etiologies regarding CI and late PTB.²⁸

Other studies have proposed updates to placental pathology reports to facilitate clinical interpretation. One recent study of 315 placentas from intrauterine deaths and 31 controls sought to create a simplified classification system based on the predominant pathologic lesion, with other lesions recorded in the comments.²⁹ Another recent study called for the use of synoptic reporting rather than narrative reporting in pathology reports.³⁰ In the proposed synoptic approach, a standardized form is use to report each lesion as present or absent, with information on grade or stage. These findings are then used to generate a report with clinical implications. Similarly, studies have proposed classifications systems for evaluating risk of SGA infant or fetal growth restriction based on placental histology, with ischemic lesions identified as the strongest indicator of poor fetal growth.^22,31 Thorne et al. (2014) reported that 10% of placentas with uteroplacental ischemia and 22% of all placentas had co-existing pathology, but did not examine whether risk of SGA infant differed based on co-existing pathology.²² Our results confirm that ischemic lesions (MVM) are associated with increased odds of SGA infant. However, our results also demonstrate that multiplicity of lesions is important (presence of two or more of CI, FVM, and MVM is associated with greater odds of SGA infant than MVM alone) and that severity matters (high-grade MVM is associated with higher odds of SGA infant than low-grade MVM). Our proposed classification system builds on these studies by addressing multiplicity and severity of placental lesions to summarize placental findings.

One important limitation of this work is the use of a pathologic sample. While our findings may be informative for a pathologic sample, they are not generalizable to all deliveries. Though results were consistent in a sensitivity analysis utilizing inverse probability weighting to address selection bias, selection bias remains a concern and the reported associations should be interpreted with caution. Replication of these findings in a sample representative of all deliveries is necessary to rule out selection bias as a potential explanation. Further, even though preterm birth and SGA infant are indications for pathologic review, not all placentas from preterm and/or SGA infants are sent to pathology. We were also unable to separate spontaneous and indicated preterm birth, and there are likely differences in the underlying mechanisms and placental phenotypes associated with these outcomes. While our results indicate that placental phenotypes may be important for SGA and preterm birth, the proposed phenotypes do not capture non-placental factors that contribute to risk of SGA and PTB. Finally, decisions about splitting and combining groups were also made based on the distributions in our dataset, which may not be generalizable to other samples.

Strengths of this analysis include the use of a large dataset of singleton livebirths from one hospital, where the majority of placentas sent to pathology were reviewed by trained perinatal pathologists. Additionally, a standardized pathology report was used during the duration of the study period, facilitating the text search algorithm to identify specific lesions of interest. While identifying placental lesions through text-searching is imperfect, in a validation study, kappa statistics were all >0.90 for key lesion categories of interest (AI, CI, FVM, MVM, other significant pathology).

We propose 21 placental phenotype groups that consider severity and multiplicity of lesions. The classifications and decision tree/web application improve clinical utility of pathology reports and offer more informative groupings for use in research studies. Future research should investigate whether results are consistent in a sample representative of all deliveries and whether the proposed classifications are associated with other adverse outcomes, such as intrauterine fetal demise and perinatal brain injury.