In high-income and middle-income countries, women from underserved and under-represented racial and ethnic groups are at increased risk of adverse perinatal outcomes. Black women are consistently at higher risk of all complications such as neonatal death, stillbirth, preterm birth, and small-for-gestational-age babies than White women. The effect varied for other racial and ethnic groups. Adverse outcomes such as preterm birth and small-for-gestational-age babies were higher in Black and south Asian women than in White women irrespective of the geographical region, and over time. Our work highlights the magnitude of disparities facing pregnant women from underserved racial and ethnic backgrounds irrespective of geographical region, emphasising the need for a broad global outlook to tackle these problems. To the best of our knowledge, our IPD meta-analysis is the largest and most comprehensive assessment to date of the magnitude of the association between race and ethnicity and adverse perinatal outcomes across high-income and upper-middle-income countries. Our work was based on a prospectively registered protocol with predefined aims and objectives. The harmonised IPPIC IPD data from multi-country cohorts provided us with a large sample size, facilitating high precision in the findings and increasing the generalisability of these results. We used multiple imputations to deal with missing variables, thereby avoiding the loss of useful information. 
We considered race and ethnicity to be social constructs without biological meaning and reported these terms in line with current recommendations to minimise bias.19 We used the terms race and ethnicity as a lens through which to study the disparities in pregnancy outcomes in women from underserved and under-represented groups because of differential treatment and access to health care. We considered the effects of race and ethnicity to include the effects of a woman’s appearance (phenotype including skin colour) that influences how she is perceived by others, and also the understanding of her appearance affecting her identity and behaviour, her parents’ appearances, and the cultural context.11 Given the different ways in which women of African origin might self-identify their origin, and the varied reporting of these women, we categorised the grouping as Black for the purpose of our analysis, as recommended by current guidance.19 Our subgroup analysis assessed the variations in outcome disparities in Black women by geographical region. The term Asian is broad and includes numerous countries of origin (eg, Bangladesh, China, India, Indonesia, Japan, and Pakistan). Given the significant differences between the various Asian ethnic groups in terms of rates of diabetes, hypertension, and other adverse outcomes, instead of pooling in one category, we reported them separately as south Asians and east Asians.85 Because of the small sample size of east Asians in the IPPIC dataset, we included them in the “other” group for the purpose of analysis. We classified all women of Hispanic identity under the underserved and under-represented race and ethnic category, including those who might identify as White Hispanic. We did so on the basis of how women might be perceived by others, which can affect their experiences and expose them to inequalities in care. In a survey, Hispanic adults said that they are described by most people as Hispanic rather than White.20 In our study, we considered White women to be the reference group through the lens of societal context, irrespective of their majority or minority status,18 where White experience is one of privilege and power across regions and settings86 and White women are expected to have optimal outcomes compared with other groups. 
Confounding variables adjusted for in our analysis were identified a priori by use of a directed acyclic graph, and unlike previous studies in this area87–90 we refrained from the unnecessary adjustment of gestational age and birthweight due to their collider status.30,91 By adjusting for the highest educational level attained as a measure of socioeconomic status, we avoided overadjusting for other factors along the pathway.92 Since the highest educational attainment achieved by an individual is usually reached in early adulthood and is the main marker for upward mobility,93 we consider it to be a key marker of social status such as income, employment, and living environment.93–96 Studies show that the association between education and health is driven by increases in human capital, with people who have lower levels of education experiencing a faster health decline than those with higher levels of education.97 
Our study had some limitations. We only included cohorts of pregnant women shared and harmonised as part of the IPPIC project, and data from studies not in the IPPIC data repository were not considered in the analysis. There were high levels of missing data in variables in some of the cohorts used for the IPD meta-analysis. However, our sensitivity analysis on complete cases resulted in similar results to our imputed dataset. Some of the cohorts included pregnant women over many decades, and the risk of adverse perinatal outcomes could have changed over time. Stillbirth was also variably defined within individual cohorts, which might have affected estimates in our analysis. Our analysis did not consider unmeasured factors that could confound the association between race and ethnicity and perinatal outcomes. The definitions of race and ethnicity differed between studies according to the databases used, the geographical regions, and time of data collection within the included IPPIC cohorts. We were only able to assess for variations in disparities due to race and ethnicity in perinatal outcomes between geographical regions, and not by health systems (private sector, government funded, or mixed models) because of the paucity of reported data. 
We acknowledge that the experiences and challenges faced by women from ethnic groups such as south Asians might vary between regions due to differences in historical immigration patterns (eg, migration to escape civil war, for economic reasons, or to join family members) and policies.98 But overall, we did not observe significant differences between the subgroups categorised by region for increased risk of small-for-gestational-age babies in women of south Asian ethnicity. It is likely that the effects of race and ethnicity on perinatal outcomes might be different within subgroups such as African and African Caribbean women born in a high-income country (eg, the UK, USA, or Canada) compared to first-generation migrants to that country, and also between various south Asian groups (eg, those of Bangladeshi, Indian, and Pakistani origin) who have been reported to have varied health outcomes, such as the highest levels of infant mortality rates in babies born to women of Pakistani origin.99–101 However, we were limited by the paucity of relevant data in the primary studies and were not able to undertake this analysis. We were able to adjust for only one measure of a woman’s socioeconomic status, maternal educational attainment, and not for other measures such as income and occupation, because of the availability of sparse and heterogeneous data in the IPPIC repository. Since the studies involved in our meta-analyses were not specifically done to assess the effects of race and ethnicity on perinatal outcomes, it is difficult to interpret the likelihood of the publication of a study included in our IPD meta-analysis with the magnitude of the association we estimated or the precision of these estimates. Therefore, we refrained from assessing the risk of publication bias.11 
Since the 1980s, neonatal mortality rates have been on the decline in most countries, but this overall trend hides underlying differences within individual racial and ethnic groups.102 For example, in the UK, a 12% fall in stillbirths among White women between 2013 and 2018 contrasts starkly with a contemporaneous 5% rise in stillbirths among Black women.1 The effect of race and ethnicity has often been shown to be associated with adverse perinatal outcomes, but this has mostly been presented in the light of it being modified by socioeconomic status.1,9 Studies such as the UK National Maternity and Perinatal Audit9 and those from the USA103,104 report higher rates of adverse perinatal outcomes in Black and Asian women, as well as women from other underserved groups, than in White women even after adjusting for socioeconomic deprivation,9 implying the contribution of other factors.105 
Our study shows that after controlling for the effect of maternal characteristics, including a woman’s educational attainment, the association between race and ethnicity and adverse perinatal outcomes persists. Complex multifactorial characteristics influence these outcomes in women from underserved racial and ethnic groups. The unique set of challenges posed by pregnancy is further worsened in individuals who are disadvantaged by their sex, race, and ethnicity.106 Racial discrimination is known to be associated with chronic stress that can influence pregnancy outcomes.4 Furthermore, women from underserved racial and ethnic groups encounter discrimination at various levels, contributing to adverse pregnancy outcomes: at the institutional level, leading to differential access to antenatal care; at the interpersonal level in their interactions with health-care professionals who do not acknowledge their concerns; and through internalised racism, where women from marginalised groups accept their perceived incompetence that limits them from seeking timely care.4 These problems are compounded by racial discrimination across generations and gaps in health literacy,4 which are in turn affected by the environment, social relationships, and employment opportunities.5 Previous studies have incorrectly adjusted for birthweight and gestational age at delivery, which dampens the true effect of race and ethnicity on adverse perinatal outcomes.9 
Our finding of disparities in perinatal outcomes across regions and over time in underserved racial and ethnic groups highlights the global need to address the structural, interpersonal, and internalised barriers faced by these women. In many countries, poor maternal and perinatal outcomes have been linked to structural racism7,107–110—a system where public policies, institutional practices, cultural depiction, and other means contribute to and reinforce racial inequity.111 The recent inquiry into racial justice and human rights in UK maternity care found that systemic factors such as negative stereotyping, microaggressions, race-based risk assumptions, and dehumanisation of women from underserved racial and ethnic groups contributed to their poor pregnancy outcomes.107 Structural racism was also highlighted as a key contributing factor to poor outcomes in Black mothers in the testimonies submitted to the US House Oversight and Reform Committee for their hearing, Birthing While Black.112 
Multifaceted antenatal interventions are urgently needed across all regions and countries to reduce the racial and ethnic inequities in pregnancy care and outcomes. Central to any such effort should be the removal of organisational and policy-level structural barriers contributing to poor perinatal outcomes.113 Interventions should focus on understanding why Black and south Asian babies die or develop complications at a disproportionate rate to White babies, and avoid clinical decisions guided by race and ethnicity that could exacerbate such inequalities.114 In the UK, the Royal College of Obstetricians and Gynaecologists has launched the Race Equality Taskforce to tackle racial disparities in women’s health care, including pregnancy outcomes.115 This is supported by national strategies such as the Race and Health Observatory, the NHS England Equality strategy, and the Core20Plus5 approach.116–118 Similar efforts are underway in other countries.119,120 
The window of opportunity available to maternity services to tackle these disparities is brief but substantial, and requires resource-intensive and time-consuming changes in social and maternity care.105 These efforts need to be complemented by a life course approach to optimising the health of, and underpinning determinants in, girls and women from underserved and under-represented groups. The curriculum and training offered to midwifery and medical students should integrate strategies to identify explicit and implicit racial biases in health-care settings and provide the tools to improve communication while caring for women from various backgrounds.121 
Despite race and ethnicity being a risk factor for adverse health outcomes, particularly in pregnancy, there are no comprehensive research strategies or initiatives to address this problem. In addition to encouraging women from various racial and ethnic backgrounds to participate in research,122 funding bodies need to prioritise topics that directly address the disparities in pregnancy outcomes that are related to race and ethnicity. The voices of women from relevant backgrounds should be central to lead and guide the efforts in this area. Given that race and ethnicity are key demographic variables, studies should aim to comprehensively collect and report these data in line with current recommendations at all stages of a woman’s life.19 This will allow us to not only map the magnitude of disparities at various timepoints, such as childhood, adolescence, pre-pregnancy, and pregnancy, contributing to poor pregnancy outcomes and their long-term effects in later years, but also plan targeted interventions at crucial timepoints to improve the health of babies in the short and long term, with the impact spanning generations.
