Association between Endotype of Prematurity and Mortality: A Systematic Review, Meta-Analysis, and Meta-Regression

Introduction: Preterm birth represents the leading cause of neonatal mortality. Pathophysiological pathways, or endotypes, leading to prematurity can be clustered into infection/inflammation and dysfunctional placentation. We aimed to perform a systematic review and meta-analysis exploring the association between these endotypes and risk of mortality during first hospital admission Methods: PROSPERO ID: CRD42020184843. PubMed and Embase were searched for observational studies examining infants with gestational age (GA) ≤34 weeks. Chorioamnionitis represented the infectious-inflammatory endotype, while dysfunctional placentation proxies were hypertensive disorders of pregnancy (HDP) and small for GA (SGA)/intrauterine growth restriction (IUGR). A random-effects model was used to calculate odds ratios (ORs) and 95% confidence intervals. Heterogeneity was studied using random-effects meta-regression analysis. Results: Of 4,322 potentially relevant studies, 150 (612,580 infants) were included. Meta-analysis showed positive mortality odds for chorioamnionitis (OR: 1.43, 95% confidence interval: 1.25–1.62) and SGA/IUGR (OR: 1.68, 95% confidence interval: 1.38–2.04) but negative mortality odds for HDP (OR 0.74, 95% confidence interval: 0.64–0.86). Chorioamnionitis was associated with a lower GA, while HDP and SGA/IUGR were associated with a higher GA. Meta-regression showed a significant correlation between these differences in GA and mortality odds. Conclusion: Our data suggest that the infectious/inflammatory endotype of prematurity has a greater overall impact on mortality risk as it is the most frequent endotype in the lower GAs. However, when the endotype of placental dysfunction is severe enough to induce growth restriction, it is strongly associated with higher mortality rates even though newborns are more mature.


Introduction
Preterm birth is defined by the World Health Organization as delivery before 37 completed weeks of gestational age (GA) and is further subdivided into extreme (less than 28 weeks), very (28 to less than 32 weeks), moderate (32 to less than 34 weeks), and late (34 to less than 37 weeks) preterm birth. Although survival has increased substantially in the last few decades, preterm birth and its associated complications represent the leading cause of both neonatal and childhood mortality worldwide [1][2][3].
Although GA is the main prognostic factor in prematurity, there is growing recognition that the pathological karger@karger.com www.karger.com/neo conditions triggering preterm birth play a central role in the development of its complications [4][5][6][7][8]. McElrath et al. [9] proposed two main pathways to extreme preterm birth: (1) infection/inflammation and (2) dysfunctional placentation. The first group includes chorioamnionitis, pre-labor premature rupture of membranes, placental abruption, and cervical insufficiency. The second group is characterized by the presence of histological features of placental vascular dysfunction and is associated with hypertensive disorders of pregnancy (HDP) and the entity identified as fetal indication/intrauterine growth restriction (IUGR) [9]. These two etiological pathways possess the characteristics to be considered as the main endotypes of very and extreme preterm birth [4,7].
The term endotype refers to "a subtype of a condition, which is defined by a distinct functional or pathophysiological mechanism" [10]. Endotypes are thus a different form of classification from clinical phenotypes and describe distinct disease entities with a defining etiology and/or a consistent pathophysiological mechanism. In line with the approach of personalized medicine, endotyping will facilitate the design of more targeted therapeutic and prognostic approaches [4,[11][12][13][14].
Despite the growing awareness about the role of the pathophysiological mechanism, or endotype, triggering prematurity in its outcome [4][5][6][7][15][16][17][18][19], this variable is rarely taken into account in the models predicting mortality in the preterm newborn [20][21][22]. Furthermore, to the best of our knowledge, the association between the endotype of prematurity and mortality during primary hospital admission has not been systematically reviewed. The purpose of the present study was to address this gap in the literature. Following the methodology described in our previous metaanalyses [7,23], the infectious-inflammatory endotype was represented by chorioamnionitis, while the dysfunctional placentation endotype was represented by HDP and IUGR. Since many researchers use small for gestational age (SGA) as a proxy for IUGR, and although the two conditions are not necessarily synonymous [24][25][26][27], we also collected data on SGA. From now on, this group will be referred to as SGA/IUGR. In addition, through the use of metaregression, we aimed to unravel the role of variables such as GA, infant sex, or use of antenatal corticosteroids in the association between endotype of prematurity and mortality.

Methods
The study was conducted according to the PRISMA and MOOSE guidelines. Detailed information on methods is provided as online supplementary Material (for all online suppl. material, see www.karger.com/doi/10.1159/000530127). Review protocol was registered in PROSPERO database (ID = CRD42020184843). The Population, Exposure, Comparison, and Outcome question was: Do preterm infants (P) exposed to chorioamnionitis, HDP, or growth restriction during pregnancy (E) have a higher risk of mortality (O) than preterm infants with no history of exposure (C)?

Sources and Search Strategy
A comprehensive literature search was undertaken using the PubMed and EMBASE from their inception up to March 2022. Search strategy is detailed in the online supplementary material.

Study Selection
Studies were included if they examined extreme to moderate preterm infants (GA ≤34 weeks and/or birth weight ≤1,500 g) and reported primary data that could be used to measure the association between exposure to CA, HDP, or SGA/IUGR and mortality before hospital discharge.

Data Extraction, Definitions, and Quality Assessment
Two investigators (T.H., E.V.-M.) independently extracted data from relevant studies using a predetermined data extraction form. Any definition of chorioamnionitis, HDP, or SGA/IUGR was accepted, but we performed subgroup analysis based on the different definitions. Methodological quality was assessed using the Newcastle-Ottawa Scale (NOS) for cohort or case-control studies [28].

Statistical Analysis
Studies were combined and analyzed using comprehensive meta-analysis V3.0 software (Biostat Inc., Englewood, NJ, USA). Summary statistics were calculated with a random-effects model and subgroups were combined with a mixed-effects model [29]. For dichotomous outcomes, the odds ratio (OR) with 95% confidence interval was calculated. For continuous outcomes, the mean difference with 95% confidence interval was calculated. Statistical heterogeneity was assessed by Cochran's Q statistic and by the I 2 statistic. Potential sources of heterogeneity were assessed through subgroup analysis and/or random effects (method of moments), univariate meta-regression analysis [30]. We used the Egger's regression test and funnel plots to assess publication bias. A probability value of less than 0.05 (0.10 for heterogeneity) was considered statistically significant.

Results
The PRISMA flow diagram of the search process is shown in online supplementary eFigure 1 in the Supplement. Of 4,322 potentially relevant studies, 150 met the inclusion criteria. These studies included 612,580 infants. Characteristics of the studies and quality assessment are summarized in online supplementary eTable 1 in the Supplement. All studies received an NOS score of at least six points, indicating a low to moderate risk of bias. 408 Main Meta-Analyses Meta-analysis showed a significant association between chorioamnionitis and increased odds of mortality ( Fig. 1, online suppl. eFig. 2 in the Supplement). Heterogeneity was moderate (Fig. 1). When subdividing by chorioamnionitis definition, the association with mortality remained significant for both clinical and histological chorioamnionitis. Meta-regression did not show significant differences depending on chorioamnionitis definition (p = 0.375). When the meta-analysis was limited to the subgroup of infants with GA ≤28 weeks, no significant association between chorioamnionitis and mortality was observed (Fig. 2, online suppl. eFig. 3 in the Supplement). Meta-regression showed significant differences (p = 0.021) between studies that only included infants with GA ≤28 weeks and those that also included older infants (online suppl. eTable 2 in the Supplement).
Meta-analysis showed a significant association between HDP and decreased odds of mortality ( Fig. 1, online suppl. eFig. 4 in the Supplement). Heterogeneity was moderate (Fig. 1). When subdividing by HDP definition, the association with mortality remained significant for any HDP and preeclampsia but not for preeclampsia/HELLP. Meta-regression showed a significant difference (p = 0.013) between the ORs for any HDP and preeclampsia. When the meta-analysis was limited to the subgroup of infants with GA ≤28 weeks, no significant association between HDP and mortality was observed (Fig. 2, online suppl. eFig. 5 in the Supplement). Metaregression showed significant differences (p < 0.001) between studies that only included infants with GA ≤28 weeks and those that also included older infants (online suppl. eTable 2 in the Supplement).
Meta-analysis showed a significant association between SGA/IUGR and increased odds of mortality ( Fig. 1, online suppl. eFig. 6 in the Supplement). Heterogeneity was high (Fig. 1). When subdividing by definition of SGA/IUGR, the association with mortality remained significant for the three thresholds of birth weight (BW < P10, <P5, and <P3) but not for IUGR (Fig. 1). Metaregression analysis did not show significant differences in the association between SGA/IUGR and mortality depending on the criteria used to define SGA/IUGR (p = 0.783). The association between SGA/IUGR and increased odds of mortality was also observed when the meta-analysis was limited to the subgroup of infants with GA ≤28 weeks (Fig. 2  In contrast, a slight publication bias was observed in the HDP meta-analysis (p = 0.016).

Additional Meta-Analyses, Sensitivity Analysis, and Meta-Regression
To investigate the potential sources of heterogeneity, we conducted additional meta-analyses exploring the differences in baseline characteristics (GA, sex, exposure to antenatal corticosteroids, rate of cesarean section, and outborn) between the groups exposed and non-exposed to chorioamnionitis, HDP, or SGA/ IUGR. We also performed meta-regression and subgroup analysis.
Infants exposed to chorioamnionitis had a significantly lower GA than infants not exposed to the condition, while infants exposed to HDP or belonging to the SGA/IUGR group had a higher GA than their respective controls (Table 1). Meta-regression showed that these differences in GA significantly correlated with the effect size of the association between mortality and chorioamnionitis, HDP, or SGA/IUGR (Fig. 3, online suppl. eTable 3 in the Supplement).
To further assess the effect of GA on the meta-analyses, we carried out a subgroup analysis of studies where the difference in GA between infants exposed to the chorioamnionitis, HDP, or SGA/IUGR and their respective controls was not statistically significant (p > 0.05). In these subgroups of studies, neither chorioamnionitis nor HDP were significantly associated with mortality odds, but the association between SGA/IUGR and higher odds of mortality was maintained (online suppl. eTable 4 in the Supplement).
Chorioamnionitis, HDP and SGA/IUGR were all associated with higher odds of exposure to antenatal corticosteroids compared to their respective control groups (online suppl. eTable 6 in the Supplement). Metaregression showed that the higher odds of exposure to antenatal corticosteroids were significantly correlated with the effect size of the association between mortality and SGA/IUGR but not for chorioamnionitis or HDP (online suppl. eTable 6 in the Supplement).
Chorioamnionitis was negatively associated with cesarean section, while HDP and SGA/IUGR were positively associated with cesarean section (Table 1). In addition, HDP was negatively associated with male sex and being outborn ( Table 1). As expected, exposure to chorioamnionitis was negatively associated with exposure to HDP or being SGA/IUGR and a strong positive association was observed between HDP and SGA (Table 1).  We also analyzed the possible effect of study design, median year in which the study was conducted, and geographical location (continent) on the results of the meta-analysis. Meta-regression showed no significant differences when studies were grouped based on whether they were prospective or retrospective (online suppl. eTable 2 in the Supplement). When cohort and casecontrol studies were compared, a significant difference was observed for the HDP meta-analysis (p = 0.030) but not for the meta-analyses on chorioamnionitis or SGA/ IUGR (online suppl. eTable 2 in the Supplement). Metaregression did not show a significant correlation between the median year in which the studies were conducted and the effect size of the association between mortality and chorioamnionitis, HDP, or SGA/IUGR (online suppl. eTable 3 in the Supplement). In addition, meta-regression also showed no significant differences when studies were grouped according to whether the inclusion of infants took place before or after the year 2000 (eTable 2 in the Supplement). Finally, metaregression showed no significant differences when the studies were grouped by continent (online suppl. eTable 5 in the Supplement).

Discussion
This is the first systematic review and meta-analysis that comprehensively addresses the association between endotype of prematurity and mortality during first hospital admission. Meta-analysis showed an association between the inflammatory/infectious endotype, represented by chorioamnionitis, and increased odds of mortality. In contrast, meta-analysis of the two proxies of the placental dysfunction endotype showed diametrically opposite results; HDP were associated with decreased odds, while SGA/IUGR was associated with positive odds of mortality. Interestingly, when analysis was restricted to extremely preterm infants, only the association between SGA/IUGR and mortality remained significant. In addition, our results confirm that conditions triggering prematurity differ in important characteristics that may play a relevant role in mortality risk, particularly the GA, which was lower in the chorioamnionitis group than in the HDP and SGA/IUGR groups. Meta-regression demonstrated a significant correlation between these differences in GA and the effect size of the association between mortality and endotype. The mode of delivery CA, chorioamnionitis; CI, confidence interval; GA, gestational age; HDP, hypertensive disorders of pregnancy; IUGR, intrauterine growth retardation defined on the basis of fetal growth assessment; K, number of studies; MD, mean difference (exposed minus unexposed); OR, odds ratio; Outborn, birth outside tertiary specialized perinatal centers and transport to tertiary center after birth; SGA, small for gestational age as defined by birth weight percentile. also differed between the two endotypes. As expected, the rate of caesarean section was higher in the placental dysfunction endotype. Finally, meta-regression also showed that the association between endotype of prematurity and mortality is relatively homogeneous across continents and has not changed significantly over the years.
Since very and extremely preterm birth is by definition a pathological condition, there is no healthy control group with which to compare infants exposed to a given prenatal insult [4,5,9,31]. Although some placentas may combine lesions of infection/inflammation and vascular dysfunction [32], fetuses exposed to chorioamnionitis will be less likely exposed to HDP and vice versa [4,5,9,31]. The main originality of our meta-analysis lies in the comparison of the three pathological conditions most frequently associated with preterm birth. However, both our identification of the two endotypes and the validity of chorioamnionitis, HDP, and SGA/IUGR as their proxies may be questionable. Moreover, there is great heterogeneity in the potential clinical applicability of these three proxies. For example, histological chorioamnionitis may be a good indicator of the presence of the infectious endotype, but information about its presence may be not available at birth [8].
Regarding the endotype of placental dysfunction, the criteria for the diagnosis of HDP have been changing over the years and even those on which there is greater consensus may be far from being implemented worldwide [33]. Finally, the most questionable assumption of our study may be the use of SGA as proxy for placental dysfunction. SGA is a statistical definition based on BW, with the 10th percentile as the most commonly used threshold [26,34,35]. Therefore, SGA definition also encompasses constitutionally small infants who do not suffer growth restriction. Conversely, growth-restricted infants who have a BW above the 10th percentile may be misclassified as normally grown [26]. It has been suggested that a cutoff value below the 5th or 3rd percentile may reflect a degree of smallness that is more likely to be pathological rather than constitutional [26,36]. However, even this pathology may be of fetal origin as there are conditions in which restricted growth is not caused primarily by placental insufficiency but indirectly leads to it [37].
Regardless of its specificity as a marker of placental dysfunction, our data confirmed that SGA preterm infants are at increased risk of death. The association between mortality and SGA was observed even for the most conservative threshold (10th percentile). Preterm SGA a b c Fig. 3. Meta-regression plot showing the correlation between the effect size of the association endotype-mortality and the mean difference (MD) in gestational age (GA) between exposed and non-exposed groups. a Univariate regression model correlating the difference in GA between chorioamnionitis-exposed and chorioamnionitis-unexposed infants. A total of 53 studies were included (coefficient −0.178; 95% CI: −0.298 to −0.068; p = 0.002; R2 analog 0.37). Each week that chorioamnionitis-exposed infants were born earlier than control infants resulted in an increase in mortality log OR of 0.178 (the equivalent of going from an OR of 1.00 to an OR of 1.19). b Univariate regression model correlating the difference in GA between hypertensive disorders of pregnancy (HDP)-exposed and HDP-unexposed infants. A total of 32 studies were included (coefficient −0.249; 95% CI: −0.403 to −0.094; p = 0.002; R2 analog 0.17). c Univariate regression model correlating the difference in GA between small for GA (SGA)/intrauterine growth restriction (IUGR)-exposed and SGA/IUGR-unexposed infants. A total of 34 studies were included (coefficient −0.282; 95% CI: −0.437 to −0.129; p < 0.001; R2 analog 0.38).

412
Neonatology 2023;120:407-416 DOI: 10.1159/000530127 infants add to their immaturity the technical challenges of their small size that complicate procedures such as endotracheal intubation or vascular access [38,39]. Of note, the meta-analysis could not demonstrate a significant association between mortality and growth restriction based on fetal assessment. However, this analysis was based on very few studies that showed a high degree of heterogeneity. Damodaram et al. [19] conducted a metaanalysis in which cohorts of growth-restricted preterm infants were compared with cohorts of infants with normal growth and found a marked increase in mortality across all GAs.
A common conundrum in perinatal medicine is the extent to which complications of pregnancy harm preterm infants through triggering prematurity or through disturbances in fetal homeostasis and development [4,40]. In the last few years, we conducted meta-analyses on the association between chorioamnionitis and complications of prematurity such as bronchopulmonary dysplasia [23], retinopathy of prematurity [41], intraventricular hemorrhage (IVH) [42], sepsis [43], or patent ductus arteriosus [44]. Similar to the present findings, meta-analyses showed an association between chorioamnionitis and all these conditions [23,[41][42][43][44]. In addition, meta-regression analysis showed a correlation between the earlier GA of the chorioamnionitis-exposed infants and the risk of developing the abovementioned complications [23,[41][42][43][44]. These data suggest that a significant component of the pathological effect of chorioamnionitis is related to the lower GA of the infants rather than to the inflammatory stress that generates in the fetus. This has been confirmed in a recent meta-analysis of our group in which we investigated the effects of funisitis, the histological equivalent of the fetal inflammatory response syndrome [45]. Funisitis was associated with increased mortality only when the control group consisted of infants without funisitis or chorioamnionitis. In contrast, when the control group consisted of infants without funisitis but with chorioamnionitis, there was no significant effect of funisitis on mortality [45]. Therefore, the fetal inflammatory response (i.e., funisitis) was not an additional pathogenic factor in the very preterm infant when maternal inflammation (i.e., chorioamnionitis) was already present [45].
In contrast to what we observed for chorioamnionitis and for SGA/IUGR, exposure to HDP was associated with lower odds of mortality in the newborn. HDP and fetal growth restriction are closely related because they often share a common underlying etiology of placental insufficiency. Placental dysfunction is a multifactorial condition, with fetuses becoming hypoxemic, undernourished, and hypercortisolemic, each of these variables being capable of independently affecting fetal homeostasis and development [46]. In addition, not all HDP are associated with growth retardation, and not all cases of growth retardation are related to placental pathology. Notwithstanding this consideration, our data suggest that placental dysfunction is associated with increased mortality when it results in a decrease in infant birth weight. When this does not occur, i.e., in cases of HDP with a normal birth weight, preterm infants may be in more favorable conditions to cope with extrauterine life. Nevertheless, it should be noted that, as discussed above, the "control" group would consist mainly of preterm infants who were born as a result of the infectious/inflammatory endotype. Moreover, HDP-exposed infants had a higher GA and were more often girls, which may have contributed to the lower mortality [47].
Even among infants of comparable GA at delivery, the endotype of prematurity may be a determining factor in the time and cause of death. Most deaths in extreme preterm infants occur in the first days of life as a consequence of pulmonary immaturity [6,48]. It is noteworthy that the intrauterine stresses associated with both the infectious/inflammatory endotype and the placental dysfunction endotype have been considered as inducers of pulmonary maturity, leading to a lower rate of respiratory distress at birth [49,50]. Although very limited by the clinical and statistical heterogeneity, the results of our previous meta-analyses suggest that the lower incidence of respiratory distress is observed only in the SGA/ IUGR group [7,23]. Nevertheless, other conditions such as pulmonary hemorrhage or pulmonary hypertension may contribute to increased early respiratory mortality in SGA infants [6,48].
Severe IVH is another major cause of early mortality among extremely preterm infants [6,48,51]. As mentioned above, in a previous meta-analysis, we observed an association between chorioamnionitis and the risk of developing IVH [42]. Interestingly, this association was independent on the effect of chorioamnionitis on GA [42]. It is therefore likely that a part of the early mortality induced by chorioamnionitis is related to clinical instability that may increase the chances of developing severe IVH. Conversely, and although to our knowledge this has not been systematically reviewed, several cohort studies suggest that exposure to HDP or being SGA may be associated with a lower risk of developing severe IVH [6,48,52,53].
An important limitation of our study is the moderate to high heterogeneity that we found in most of the metaanalyses. Through subgroup analysis and meta-regression, we aimed to investigate possible sources of heterogeneity. Neither the association between chorioamnionitis and high odds of mortality nor the association between HDP and reduced odds of mortality were observed in the subgroup of studies that exclusively included extremely preterm infants (GA ≤28 weeks). In contrast, the association between SGA/IUGR and high odds of mortality was maintained in the subgroup of extremely preterm infants. Together, these data suggest that at lower GAs, mortality is relatively independent of the endotype of prematurity. However, infants who combine extreme prematurity and being SGA would have the highest mortality risk.
Another limitation of our meta-analyses is that they are based on a "births-based" instead of "fetuses-atrisk" formulation [54,55]. Fetuses who would be viable but die in utero are not considered when assessing mortality from a births-based perspective [54,55]. Therefore, if we intend to analyze the overall impact of an endotype of prematurity on perinatal health, we should also consider how many children are not born prematurely because they die as a result of the severely adverse intrauterine environment that is induced by their particular endotype.
Advances in perinatology over the past 2 decades, such as wide implementation of antenatal steroids and surfactant, have changed care for extremely preterm infants and have been associated with a decrease in mortality [56]. However, neither meta-regression nor division of studies into two subgroups (median year of the cohort before or after 2000) could demonstrate an effect of study date on the association between endotype of prematurity and mortality. Finally, it is interesting to note our finding that in the three groups analyzed (chorioamnionitis, HDP, and SGA/IUGR), the odds of receiving antenatal corticosteroids was higher than that of the respective controls. We speculate that this finding may be related to the more heterogeneous nature of the control groups. Nevertheless, our group is conducting a meta-analysis on the association between antenatal corticosteroids and endotype of prematurity.
In summary, the present data suggest that the infectious/inflammatory endotype of prematurity has a greater overall impact on mortality risk as it is the most frequent endotype in the lower and more vulnerable GAs. However, when the endotype of placental dysfunction is severe enough to induce growth restriction, it is strongly associated with higher mortality rates even though newborns are more mature. Nevertheless, neither of the two endotypes is desirable and extremely preterm birth should be prevented regardless of the etiopathogenic process that caused it. Moreover, interventions aimed at prolonging gestation to increase fetal maturity imply, in many cases, a prolongation of fetal exposure to an unfavorable intrauterine environment. Therefore, delay in delivery may not benefit some infants, and might even increase risk, if prematurity is triggered by a severe pathology and postponement of delivery is not adequately addressing the underlying condition [15]. Finally, since some placental diagnoses may be critical for personalized clinical care of extremely preterm newborns, efforts should be made to have information on placental pathology available in the first days of life [8,32].

Statement of Ethics
As this systematic review and meta-analysis did not involve animal subjects or personally identifiable information on human subjects, ethics review board approval and patient consent were not required.

Conflict of Interest Statement
The authors declare no conflict of interest. The views expressed in this paper are those of the authors and do not necessarily reflect the policies of Statistics Netherlands.

Funding Sources
This research did not receive any specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data Availability Statement
All data relevant to the study are included in the article or uploaded as supplementary information. Additional data are available upon reasonable request.