4 Discussion
4.1 Main Findings
In a cohort of well-characterized nulliparous pregnant people, a GRSBW, derived from a set of previously was modestly associated with infant BW. However, its association with BW was not statistically significant among participants who self-identified as Black, Asian, or had an unknown race, or among those with AFR, EAS, SAS, or UNK genetically predicted ancestry. Our findings suggest that the GRSBW does not fully generalize to racially or genetically diverse groups.
4.2 Interpretation
Our findings are concordant with other studies assessing the relationship between race and fetal growth. Across a variety of contexts, studies have found that race is associated with differences in fetal growth among both unselected and low risk groups.(16-18) In our study, both self-identified race and genetically predicted ancestry were associated with gestational age-adjusted BW, even after controlling for sex and GRSBW. Our finding that the GRSBW was not consistently associated with BW across non-European ancestry groups is also consistent with existing studies of other conditions. Polygenic risk scores derived in primarily European cohorts perform significantly less well in participants of non-European descent for multiple conditions, including venous thromboembolism, coronary artery disease, heart disease, hypertension, chronic kidney disease, and cancer.(19-25) The non-generalizability of genetic findings to diverse populations is a critical gap with the potential to exacerbate existing disparities.(26, 27) Our findings add to this important body of work by extending it to fetal growth, which holds considerable clinical relevance in perinatal medicine.
Our findings have several implications for future efforts in this area. First, our results demonstrating that GRSBW is not associated with BW in many ancestry groups, and that genetically predicted ancestry remains independently associated with BW after controlling for GRSBW, suggests that additional work is needed to achieve equity in the performance of genetic risk scores for BW prediction. Methods to support multi-ancestry polygenic risk score derivation are now available and are promising in their ability to equitably leverage genotypes for trait prediction. However, but such methods still depend on the availability of discovery cohorts that themselves are diverse, if not globally representative.(28-32) As precision medicine advances its ability to improve recognition of diseases such as fetal growth restriction and thereby allow for earlier surveillance or treatment, genetic risk scores that perform better in some populations than others have the potential to exacerbate inequities in adverse pregnancy outcomes. Second, two results suggest that there are additional unaccounted-for factors linking race to fetal growth: the lack of association between GRSBW and BW in multiple self-identified racial groups, and that self-identified race remains associated with BW after controlling for GRSBW. As noted, the GRSBW is likely insufficiently capturing the genetic components of this association. However, as race is a social construct, the persistent association between race and BW can also be linked to systematic differences in environmental and social exposures that are known to contribute to racial health disparities. It is also plausible that there are epigenetic influences reflecting the transgenerational impact of racism and other forms of deprivation, oppression, and hardship imposed on minoritized populations. These factors and their complex relationships to the genetics of fetal growth remain to be clarified and warrant further investigation.
4.3 Strengths and Limitations
Strengths of our study included the use of a large, multicenter U.S. obstetric cohort with geographical and racial diversity. The nuMoM2b protocol provides both standardized specimen collection and validated outcomes ascertainment, and we used externally derived BW-associated SNPs for GRSBW assessment, adding to the rigor and validity of our analysis. Additionally, our assessment of GRSBW using two distinct approaches (both self-identified race and genetically predicted continental ancestry groups) demonstrates that the lack of generalizability is a robust finding.
Our study also had limitations. The need to map SNPs derived from reference build GRCh37 to GRCh38, ultimately leading to the use of 73 rather than 86 SNPs may have reduced the strength of the overall association between the GRSBW and BW. Also, it is possible that the lack of association between the GRSBWand BW is due to the sample sizes of each group, especially for the smallest groups, such as Native Hawaiian/Pacific Islander or American Indian/Alaska Native. However, sample size limitations are unlikely to fully explain the lack of association, as the GRSBW was associated with BW in the multiracial group (n=508) and was very nearly significant among those of SAS predicted ancestry (n=274), both of which had smaller sample sizes than the largest groups in which GRSBW was not associated with BW.