In line with existing reviews and international studies, we found a higher risk of recurrence of PTB and SGA overall. We found that effect sizes for recurrent PTB generally ranged from two to five times higher than incident PTB. Similarly, a four-fold risk was observed for recurrent SGA in two studies. We did not find any U.S. studies on the recurrence of any other outcomes. With respect to race and ethnicity comparisons, the results were not consistent in showing a higher risk of recurrence across Black race and Hispanic ethnicity, and they were inconclusive for Asian and Pacific Islander mothers. Using the Woolf test for homogeneity of race and ethnicity-stratified risk ratios, we found evidence of effect modification by race for the risk of a recurrent PTB among Black and White mothers in two of four studies only, and the evidence regarding race as a modifier was weak for SGA in the same. While comparisons of Black and White births suggested a higher recurrent PTB risk among Black mothers, race-stratified results also indicated that recurrent PTB could be an equally severe issue for White mothers. Stratified estimates for three of the four studies showed that Black mothers’ risk of a recurrent PTB was relatively lower than that of White mothers, i.e., Black mothers with a prior PTB had a smaller risk ratio of a second PTB than White mothers, pointing to the potential effect of higher risk in the first PTB, which, in turn, affects their likelihood of experiencing another preterm birth. One reason for observing a lower recurrence risk among Black women might also be fetal loss due to their heightened risk for adverse birth outcomes [24].
Only one study produced estimates for Hispanic and Asian mothers, finding that Hispanic women experience a higher risk if they had a prior PTB compared to a prior term birth. However, this risk is similar to that of White mothers. For Asian women, the risk was not higher for a subsequent PTB given a prior PTB. The stratified effect sizes varied across the groups and produced inconsistent evidence of Black race as a factor in recurrent PTB. For SGA, the stratified analysis showed that risk was higher for all races and Hispanic ethnicity, but one study found these risks to be similar to those of White women, i.e., pooled estimates would be recommended, but another showed statistically different estimates for Black and White women. Without data for covariates, these comparisons were largely unadjusted. None of the international studies made any comparisons by socioeconomic status or other socially salient categories such as race and ethnicity in the U.S. Data of PTB by race and ethnicity in the U.S. has established clear evidence of higher risk among Black women but evidence regarding higher risk among Asian or Hispanic women is less consistent [24]. Our research similarly found that Asian women’s risk of either PTB or SGA was not different from that of White women, but for Hispanic women, the stratified risk ratios were statistically different from White women.
Research also suggests a genetic link between repeat adverse outcomes, especially across more than two pregnancies, given the likelihood of the same mother delivering a preterm baby at the same gestation over consecutive pregnancies [34]. The authors argue that the likelihood of similar social and environmental exposures occurring at the same time in consecutive pregnancies is slim, suggesting an underlying genetic pathology. Similarly, genetic risk may differentially predispose Black and White women to heightened inflammatory response to infection during pregnancy [26]. However, to our knowledge, no study has provided a direct genetic analysis of recurrent preterm birth, even in the absence of other risk factors. Furthermore, given the legacy of anti-Black racism in the U.S., and lower rates of preterm birth among foreign-born Black women, relying on genetic risk to explain the Black-White disparity in the recurrence of adverse birth outcomes is not recommended [27].
Our findings were limited in the range of outcomes from similar studies using national registries and hospital data from Europe (Sweden, Finland, Malta, Scotland, Norway) [28,29,30,31], Australia [32], India [33], Indonesia [34], Brazil [35], Tanzania [36], and several other developing countries [37], which have shown that having a stillbirth, PTB, LBW, or neonatal mortality increases the risk of having another similar outcome. Further, we excluded recurrent cross-matched outcomes due to poor reporting of race and ethnicity measures. However, other published research has also shown higher risks for adverse cross-outcomes across pregnancies. A 2017 review and meta-analysis by Malcova and colleagues (17 studies) found an increased risk of stillbirth, PTB, and SGA if a previous birth had resulted in any of the outcomes, with higher risk by the degree of severity of the prior outcome [2]. Meta-analyses should only be conducted if the statistical combination of individual study results is meaningful. Metelli and Chaimani (2020) describe the challenges of running a meta-analysis with observational study data, including increased risk of reporting and publication bias, high heterogeneity, and confounding [38]. Our study reported too a high level of heterogeneity to draw conclusions from the meta analysis.
We extend some caution that the studies in the U.S. are based on a limited number of states, have overlapping cohorts, and vary significantly in study design and population coverage. For example, the two studies disaggregating data for Hispanic and Asian mothers had a relatively small sample of Black women (∼ 100 births each) from the NICHD cohort for the same period. Moreover, all comparisons of effect sizes should be made with caution due to the variation in covariates, sample size for underlying risk adjustments, and health system capacity across studies. Any conclusions about the risk of recurrence and repeat PTB and SGA events in the U.S. should be avoided unless estimates can be made across adequately powered samples of racially and ethnically diverse populations and stratified analysis is produced.
A strength of our study is that it is the only review in the U.S. or internationally that reports findings for socially salient categories such as race and ethnicity, which begins to consider the relevance and importance of maternal health disparities across these groups. For example, while we found evidence of recurrence risk for PTB across all races and Hispanic ethnicity, these differences were also conditional on the risk of an incident PTB, which is known to be higher among Black women in the U.S. compared to all other racial or ethnic categories. Therefore, recurrence risk is not indicative of the overall risk for an adverse outcome such as PTB. Similar disparities will likely exist in other countries, whether by race, ethnicity, or socioeconomic status, which is important for global maternal health equity. Second, the outcome of each pregnancy in these studies was established through direct measurement, i.e., from a record of each birth itself through either vital or hospital records. As Adams 2001 found, relying on data from the second pregnancy’s birth certificate for the outcome of the first pregnancy will grossly undercount adverse outcomes in the first pregnancy, and direct measurement at the time of birth is recommended [39].
A limitation of this systematic review was high heterogeneity influenced by factors such as differences in underlying study populations, study inclusion and exclusion criteria, adjusted variables, and analytical methods. For this reason, we chose not to present the results of our meta-analysis. Possible reasons for high heterogeneity could be the wide variation in sample sizes and underlying populations (especially by race and ethnicity), adjustments for underlying medical conditions and comorbidities, setting, period and duration of data collection, and varying exposure and outcome definitions (e.g., spontaneous versus indicated PTB). This heterogeneity can be addressed with more information on the characteristics of the included populations in each study. Overall, we questioned the validity of conducting a meta-analysis given persistent heterogeneity despite attempts to adjust for study differences and confounding variables. This limitation reflects the quality of included studies and the nature of combining observational studies versus RCTs rather than an inherent flaw in our methods. Our conclusions are also limited by the lack of adjusted estimates that fit our research question. While the chosen studies aligned with our selection criteria, they provided results that limited comparisons by race and ethnicity. For example, we decided to produce stratified unadjusted estimates from study tables to make comparisons across the studies since these were either not provided or were presented with varying levels of disaggregation for maternal age, type of PTB (early or late), and periods. As noted earlier, we were disappointed to have to drop one study with a large sample size due to unresolvable errors in the published table.
Add Comment