Global prevalence of preterm birth among Pacific Islanders: A systematic review and meta-analysis

The epidemiology of preterm birth among Pacific Islanders is minimally understood. The purpose of this study was to estimate pooled prevalence of preterm birth among Pacific Islanders and to estimate their risk of preterm birth compared to White/European women. We searched MEDLINE, EMBASE, Web of Science Core Collection, Cochrane Library, CINAHL, Global Health, and two regional journals in March 2023. Observational studies were included if they reported preterm birth-related outcomes among Pacific Islanders. Random-effects models were used to estimate the pooled prevalence of preterm birth with 95% confidence interval (CI). Bayes meta-analysis was conducted to estimate pooled odds ratios (OR) with 95% highest posterior density intervals (HPDI). The Joanna Briggs Institute checklists were used for risk of bias assessment. We estimated preterm birth prevalence among Pacific Islanders in the United States (US, 11.8%, sample size [SS] = 209,930, 95% CI 10.8%-12.8%), the US-Affiliated Pacific Islands (USAPI, SS = 29,036, 6.7%, 95% CI 4.9%-9.0%), New Zealand (SS = 252,162, 7.7%, 95% CI 7.1%-8.3%), Australia (SS = 20,225, 6.1%, 95% CI 4.2%-8.7%), and Papua New Guinea (SS = 2,647, 7.0%, 95% CI 5.6%-8.8%). Pacific Islanders resident in the US were more likely to experience preterm birth compared to White women (OR = 1.45, 95% HPDI 1.32–1.58), but in New Zealand their risk was similar (OR = 1.00, 95% HPDI 0.83–1.16) to European women. Existing literature indicates that Pacific Islanders in the US had a higher prevalence of preterm birth and experienced health inequities. Learning from New Zealand’s culturally-sensitive approach to health care provision may provide a starting point for addressing disparities. The limited number of studies identified may contribute to higher risk of bias and the heterogeneity in our estimates; more data is needed to understand the true burden of preterm birth in the Pacific region.


Introduction
In light of persistent disparities in perinatal health between minority and majority populations globally, greater attention is being paid to understanding unique risks that explain minority populations increased risk of preterm birth (live birth before 37 weeks gestation [1]). Despite efforts over the past several decades to intervene, preterm birth remains the leading cause of both neonatal and under five year death [2,3]. In 2014, which is the most recent estimate from the World Health Organization, the global prevalence of preterm birth was 10.6% (uncertainty interval 9.0%-12.0%) [4].
Pacific Islanders are particularly underrepresented in perinatal health research, and little is known about the prevalence of preterm birth among this group. Data from the Pacific Islands themselves is sparse, both as a result of geographic isolation (~1000 islands across 300,000 square miles [5]) and nascent research infrastructure. Pacific Islander migrants are, however, among the fastest growing minority groups in the United States (US), New Zealand, and Australia. In the US, 1.2 million people identified as Native Hawaiian or Other Pacific Islanders (NHOPI) in 2010 [6]. In the 2018 census 24.6% of New Zealanders identified as Māori or Pacific Islander [7], and in Australia (2016),~250,000 people reported Pacific Islander ethnicity [8]. While data from these settings should allow for ethnicity-specific examination of health outcomes, Pacific Islanders continue to be aggregated with other minority groups; in the US with Asian or Native Americans (or Pacific Islanders are omitted from analyses because of small sample size), and in Australia with Indigenous Australian groups.
Health challenges common among Pacific Islanders, such as, disproportionately high prevalence of obesity and obesity-related complications [5], may put them at a higher risk of preterm birth [5,9]. Obesity is a significant risk factor for pre-eclampsia and pre-pregnancy diabetes [10,11], which have been associated with indicated preterm births [12]. Endemic tropical illnesses may also increase risk [12,13]: Papua New Guinea, for example, still has a high rate of malaria [14]. Furthermore, Pacific Islander migrants in developed countries may have limited access to social services [15][16][17][18] and reportedly experience discrimination [19], which may worsen their perinatal health outcomes.
To better understand the epidemiology of preterm birth among Pacific Islanders and the need for perinatal health intervention, this systematic review and meta-analysis aims to: (a) estimate the pooled prevalence of preterm birth among Pacific Islanders globally; and (b) identify whether Pacific Islanders were more likely to experience preterm birth compared to non-Hispanic White/European women.

Protocol and registration
This review (PROSPERO ID: CRD42021283377, protocol [20]) followed Preferred Reporting Items for Systematic Reviews and Meta Analyses (PRISMA) [21] and Meta-analysis of Observational Studies in Epidemiology (MOOSE) reporting guidelines [22] and received Institutional Review Board exemption.
[RMI], and Palau), Kiribati, Nauru, Papua New Guinea, Solomon Islands, Fiji, New Caledonia, Vanuatu, Tonga, Tuvalu, Tokelau, Niue, French Polynesia, New Zealand (Māori, the indigenous Polynesian people of New Zealand), Samoa, and the Cook Islands. Studies from the US (Hawai'i and the contiguous US) and Australia (individuals from Ni-Vanuatu, Tahiti, and the Pitcairn islands; Aboriginal and Torre Strait Islanders were not included) were included, since there are large migrant populations in both countries.

Information sources
We searched MEDLINE ALL (Ovid), EMBASE (Ovid), Web of Science Core Collection (as licensed at Yale [20]), Cochrane Library, CINAHL (EBSCOhost), and Global Health. The Pacific Journal of Reproductive Health and Pacific Health Dialog were searched manually since they are not well indexed in major bibliographic databases. Backwards and forwards citation chaining was searched by hand via Google Scholar. Reports from international, national, state-level, and territorial government agencies were searched manually.

Search strategy
The search strategy was developed by the first author in consultation with all co-authors, including a medical librarian (KN). The search used two concepts: (1) Pacific Islanders and (2) preterm birth outcomes. Appropriate controlled vocabulary terms and keyword search terms were used (MEDLINE ALL example, S1 Table) and archived [24].

Study selection
Peer-reviewed observational studies and agency reports published before March 17 th 2023 were included. Case reports and case-control studies were excluded due to inability to estimate prevalence. While doctoral theses were included, conference abstracts and master's theses were not since final study outcomes may not have been available/reported.
Using Covidence data management software, each article was screened by two authors independently at the title-abstract and full-text screening stages. Authors met to reach consensus on inclusion and reasons for exclusion. Screening questions are presented in S2 Table. Studies using the same datasets with the same study period were examined for potential sample overlap. Where overlap was identified, only the study with the largest sample size was retained in analyses.

Data extraction
Data extraction was completed by the first author and checked by the senior author. Extracted information included preterm birth-related outcomes (prevalence among Pacific Islander women and non-Hispanic White/European women), data source, publication date, data collection period, study country/setting, study design, gestational age (GA) measurement method, Pacific Islander ethnicity, and sample size.
For the prevalence estimate, if a study reported prevalence but not the absolute number of events, this number was calculated with N(Preterm Births) = P*N(total Pacific Islanders) and rounded to the next whole number, where N represented sample size and P represented prevalence of preterm birth. To compare risk among Pacific Islander and White/European women, the prevalence among the two groups and the modelling outcomes (odds radio [OR], modelling method, and adjusted confounders) were recorded. For studies reporting prevalence, OR were calculated with OR = ad/bc.

Risk of bias assessment
The Joanna Briggs Institute (JBI) [25] checklist for prevalence studies [26] was used for the assessment in prevalence meta-analyses; the checklist for cross-sectional studies [27] was used for meta-analyses comparing risk. Two authors (BW and EI) completed appraisals and disagreements were discussed to reach consensus. Total scores represented the proportion of "checks" with 100% the maximum possible score. Egger's test [28] (for prevalence meta-analyses) and weight-function models [29] (for Bayesian risk comparison meta-analyses) were used to assess publication bias.

Synthesis of results
Stratified by study country/territory, we conducted prevalence meta-analyses of preterm birth among Pacific Islander women and comparison of preterm birth risk meta-analyses between Pacific Islander women and non-Hispanic White women in the US and New Zealand. As random-effects models (DerSimonian and Laird method) [30] consider the included studies to be a representative sample of possible articles on the research question of interest, these models were used for the pooled prevalence with 95% confidence intervals (CI). I 2 is usually high in prevalence meta-analysis but may not indicate the data is inconsistent [31,32], and it can be biased in small meta-analyses [33], so we reported tau with both the inverse variance (IV) method (weighting more to studies with a larger sample size) and the generalized linear mixed models (GLMM) method (weighting more to studies with smaller sample size) for the heterogeneity assessment. We compared the estimated tau 2 [IV] to the within-study variances, with tau 2 [IV] larger than within-study variance indicating the weights of any two studies are approximately equal [34]. A large proportion of equal weights in a meta-analysis indicates the heterogeneity exists. Prediction intervals were presented in which future studies effects may fall based on present evidence. Subgroup analyses by Pacific Islander ethnicity were conducted to understand the between-study heterogeneity.
For the pooled risk comparison estimate, we used a Bayesian meta-analysis method [35] due to the relatively small number of studies identified by our search. Pooled ORs with 95% highest posterior density intervals (HPDI) were reported. We assumed no association, so we restricted the effect ln(OR) to normal prior centered at ln(OR) p = 0 (no effect), and the prior standard deviation was restricted to σ p = 4; the priori expected heterogeneity was restricted to tau � 0.98 with 95% probability as half-normal priori with scale 0.5. Tau with 95% HPDI were used for the heterogeneity assessment. Prediction intervals and subgroup analyses by ethnicity were also provided.
All analyses were performed using RStudio (RStudio, Inc., Boston, MA, USA). The R package meta [36] was used to conduct prevalence meta-analyses and to generate Forest plots and package dmetar [37] was used to perform Egger's test for the corresponding publication bias assessment; package bayesmeta [38] was used to conduct Bayes meta-analyses, and package RoBMA [39] was used to perform the corresponding publication bias assessment (Bayes factor [BF]<3 indicating weak evidence [40]). R code is provided in S1 Appendix.

Study selection
We identified 10,077 articles from six databases and 15 articles from gray literature on December 3 rd 2021, and updated our search (2nd search) in these six databases on March 17 th 2023 (Fig 1). After removing duplicates, title-abstract and full-text screening, and adding articles through citation chaining, we found 118 articles reporting preterm birth-related outcomes among Pacific Islanders among which 55 articles reported results related to our objectives.
After removing studies with overlapping data (S3 Table), we included 33 articles that reported preterm birth prevalence, and 15 articles that compared risk between Pacific Islander and White/European women.

Heterogeneity assessment
Checking the proportion of the studies with a smaller variance than the tau 2 [IV] suggests that no heterogeneity existed in the prevalence meta-analysis in the US (0%) and New Zealand (0%), however, meta-analyses conducted for the USAPI (100.0%), Australia (100.0%) and Papua New Guinea (100.0%) all showed strong evidence of heterogeneity. When using the same method to check the proportion of the studies with a smaller variance of log scale of the ORs (Varlog[OR]) than the tau 2 [IV] there was strong evidence of heterogeneity in meta-

Main findings
To our knowledge, this is the first systematic review and meta-analysis to summarize data from both Pacific Islander immigrants in high income settings and those resident in the Pacific Islands. Results indicate that US-resident Pacific Islanders had a relatively higher prevalence of preterm birth than in New Zealand and Australia and experienced poorer birth outcomes than White women, which was an inequity not observed in New Zealand. Estimates in other settings were limited by sparse data. Our Pacific Islander-specific findings mirror the most recent country level estimates reported by the WHO. In 2014, among the three developed nations, the US had the highest overall preterm birth prevalence (9.6%, uncertainty interval [UI] 10.3%-14.0%), while the PLOS GLOBAL PUBLIC HEALTH estimates in New Zealand (7.5%, UI 7.0%-9.8%) and Australia (8.6%, UI 6.9%-9.5%) were similar [82]. Notably, in our analysis we identified disparities between Pacific Islander and white women that were present in the US, but not in New Zealand. In New Zealand, while Pacific Islanders have become a minority group since European occupation, they make up a far greater proportion (24.6% [7]) of the population than in the US (0.4%) [83]. As such, supportive and culturally-based health policies have been prioritized in New Zealand to care for populations with high need (including Māori and Pacific Islanders) [84], such as Very Low Cost Access (VLCA), Te Tiriti o Waitangi (the Treaty of Waitangi) [85], which protects Māori resources, and iwi (tribally)-based primary care consistent with Māori values, attitudes, and aspirations [86,87]. In contrast, in the US Pacific Islander women have reported limited health care access related to citizenship [16,17], discrimination [19], and mistrust of health professionals [15][16][17][18]. Replicating New Zealand's focus on primary health care, support for culturally-sensitive initiatives, and efforts to decrease structural racism [87][88][89] may be important if US disparities are to be addressed.
While Australia has among the largest absolute number of Pacific Islander migrants, they make up a small proportion (0.9%) of the population 12 , which likely explains the limited number of Pacific Islander-related studies reporting preterm birth. Australia has continuously supported seven surrounding Pacific Island countries to promote universal health coverage since 1995, including Fiji, Kiribati, Nauru, Samoa, Solomon Islands, Tonga and Vanuatu [90], but did not systematically record maternal ethnicity other than Caucasian and Indigenous (Aboriginal and Torres Strait Islander) in their own health system before 1998 [91]. Studies indicate that language barriers and inadequate health care access may increase the burden of adverse pregnancy outcomes among Pacific Islander migrants to Australia [66], but further data, with improved specificity in race/ethnicity reporting, is needed to better understand their perinatal outcomes.
Data from the Pacific Islands themselves were sparse and limited conclusions about the prevalence of preterm birth. While Pacific Islander women in the USAPI had a lower preterm birth estimate than the US, the wider prediction interval should be noted. All of the USAPIs are medically underserved [92] and, although prenatal/obstetric care is available [93,94], coverage in different settings varies [94,95] limiting opportunities for systematic data collection. Further data is sorely needed to adequately understand the needs of these territories, which are heavily dependent on US grant funding to sustain their health care systems.
Although Papua New Guinea is the largest of the Pacific Islands, with 8.9 million residents [96], studies were limited by design and response rate. For several reasons, the estimated prevalence may not accurately reflect the national burden of preterm birth. First, most of the included studies were from Madang Province, where much research is concentrated since the province is home to one of the country's two medical schools and the Papua New Guinea Institute of Medical Research. The relatively higher prenatal care coverage (63%) in this setting [97,98] compared to other regions may reduce maternal morbidity during or after pregnancy, meaning that estimates for preterm birth prevalence may be lower than the whole nation. Second, although malaria-a known risk factor for preterm birth-is endemic to Papua New Guinea (86% of cases in WHO Western Pacific Region were from Papua New Guinea [14]), most women in the included studies had either taken malaria prophylaxis before the recruitment or were cured before giving birth [68,69,71] (one study did not report malaria related information [70]). This also may not be the case for women residing in more rural and less medically served settings. Finally, in one of the largest samples of women among the Papua New Guinea studies [68], preterm birth was defined as less than 38 weeks, which may also have influenced the estimate.

Strengths and Limitations
Limitations of our study should be noted. Beyond the USAPIs, Papua New Guinea and New Zealand there was little data on perinatal health from the remaining Pacific Islands. We identified only one study from Vanuatu [73] (PTB prevalence: 8.0%), and one study from the Solomon Islands in which the reported preterm birth prevalence was 23.8% in 2011-2013 [72] indicating that there may be broad heterogeneity in preterm birth prevalence across the region that was not captured here. Even within our studies, a large degree of heterogeneity was evident in prevalence meta-analyses for the USAPI, Australia, and Papua New Guinea. Potential explanations could include the lack of consistency in GA measurement methods, the limited publications, and quality of health care provided by local clinics. Similar heterogeneity in the risk comparison meta-analysis in the US may be from different perinatal outcomes among Pacific Islander subgroups, particularly the much higher risk and prevalence of preterm birth among Marshallese women; in New Zealand, the heterogeneity may be from different study designs, varied GA measurement methods (most were unclear), and the unavoidable heterogeneity for smaller meta-analyses [33]. The large proportion of included studies that did not clearly state the methods used to estimated GA also should be noted.
Our study does, however, have several notable strengths. We systematically searched articles reporting preterm birth outcomes among Pacific Islanders, including grey literature and comprehensively conducted backward and forward citation chaining. Second, we used tau for the heterogeneity assessment instead of I 2 statistics, which have been questioned for their reliability especially for prevalence meta-analysis [31] and meta-analyses with a small number of included studies [33]. Additionally, to increase the accuracy of our risk comparison estimates, we used a Bayesian meta-analysis approach that performs better than frequentist meta-analysis approach when a small number of studies are included [35], and the corresponding robust Bayesian modelling method for publication bias assessment.

Interpretation
Our findings have important public health implications. In developed nations like the US and Australia, an improved record of race/ethnicity information and ethnicity-specific analyses are the first steps to understanding the burden of adverse pregnancy outcomes among minority groups. It is vital to identify related, contextually relevant risk factors to inform future polices to decrease health disparities. Lessons may be learned from New Zealand's culturally-sensitive approach. In the USAPI, Papua New Guinea, and the other Pacific nations not represented here, basic health care infrastructure improvements are likely needed before the true burden of preterm birth can be understood.

Conclusion
Existing literature indicates that Pacific Islanders in the US had a higher prevalence of preterm birth than in other global settings and experienced health inequities. Learning from New Zealand's culturally-sensitive approach to health care provision may provide a starting point for addressing disparities. Data from other Pacific settings is sparse, limiting conclusions about prevalence. More data is needed to understand the true burden of preterm birth in the Pacific region.
Supporting information S1 Checklist. PRISMA 2020 checklist. (DOCX) S1  Table. Risk of bias assessment for the preterm birth prevalence meta-analysis using the JBI checklist [26]. (DOCX) S5 Table. Risk of bias assessment for the risk comparison of preterm birth meta-analysis using the JBI checklist [27].