Stillbirth Among Patients With Diabetes in Pregnancy in Ethiopia: Systematic Review and Meta-Analysis

Purpose: Maternal diabetes mellitus and the resulting adverse fetal outcomes including stillbirth in low- and middle-income countries (LMICs) are high. Thus, setting specific evidence is pivotal to plan, evaluate, and improve national preventive measures and to achieve international sustainable development goals. Therefore, this systematic review and meta-analysis was the first of its kind to estimate the pooled prevalence of stillbirth and its determinants among diabetic mothers in Ethiopia. Methods: Primary studies were exhaustively searched using PubMed, ScienceDirect, Web of Science, SCOPUS, and Google Scholar databases, and gray literature found in Addis Ababa and Haramaya University online repositories was accessed. Eligible studies were selected and critically appraised for quality using the Joanna Briggs Institute (JBI) quality appraisal checklist. The overall prevalence of stillbirth among diabetic mothers was estimated using a weighted inverse random-effect model. I2 statistic was used for evidence of heterogeneity. Egger's test and funnel plot were used to check the presence of publication bias. Results: The pooled prevalence of stillbirth among diabetic mothers was 2.39 [95% confidence interval (CI): −0.20, 4.97]. Being a housewife [adjusted odds ratio (AOR) = 2.25; 95% CI: 1.26, 3.23], maternal age of <30 years [AOR = 2.08 (95% CI: 1.02, 3.13)], and gestational age of <37 completed weeks [AOR = 9.76 (95% CI: 7.83, 11.70)] increased the risk of stillbirth among diabetic mothers. Conclusions: The national pooled prevalence of stillbirth among diabetic mothers was 2.39%. Maternal age of <30 years, gestational age of <37 completed weeks, and being a housewife were significantly associated with stillbirth. Trial registration: PROSPERO 2020: CRD4202016774.


INTRODUCTION
Diabetes mellitus is a metabolic syndrome characterized by hyperglycemia due to either deficiency in insulin secretion or reduced insulin action (1). Currently, diabetes mellitus is a severe public health problem worldwide. Maternal diabetes mellitus in low-and middle-income countries (LMICs) is high, which accounts for 90% (2). Thus, maternal hyperglycemia complicates 17% of pregnancy, which results in different adverse birth outcomes for both the mothers and newborns (3)(4)(5). Maternal hyperglycemia has many adverse effects on embryogenesis and fetal development (6) like stillbirth (7). According to the World Health Organization (WHO), stillbirth can be defined as a baby born with no sign of life at or after 28 weeks of gestation (8).
Pre-existing maternal diabetes during pregnancy had four to five times higher risk of stillbirth (9). On the other hand, the rate of stillbirth among type 1 and type 2 diabetic pregnant women were 16.1 and 22.9 per 1,000 births, respectively (10). Similarly, a systematic review conducted in LMICs indicated that the incidence of stillbirth was 6.3% higher among diabetic pregnant mothers than non-diabetic pregnant mothers (11). Elsewhere, the rate of stillbirth among diabetic pregnant mothers in different parts of the world is as follows: 13.9 per 1,000 births in England (12), 25% in Australia (13), 3.66% in Saudi Arabia (14), 2.6% in King Khalid University Hospital (4), 2% in Nigeria (15), 5.2% in Ghana (16), and 2.3% in Cape Town (17). Moreover, studies conducted in different parts of Ethiopia showed that the magnitude of stillbirth among diabetic pregnant mothers ranges between 2.6 and 16.05% (18)(19)(20)(21)(22)(23).
Stillbirth can result in long-lasting disability, which may have significant burdens on parents, families, healthcare providers, and societies worldwide (24). Elsewhere, the loss of life and the psychological and financial costs for women, families, and societies are severe and long-lasting (25). According to a 2015 WHO report, more than 7,178 deaths per day resulted from stillbirth. The majority (98%) of this death occurred in LMICs (26).
Factors such as higher HbA1c and maternal overweight/obesity increase the risk of stillbirth among diabetic pregnant mothers (10). In addition, lack of antenatal, medical, and preconception cares are factors affecting stillbirth among diabetic pregnant mothers (2,27,28).
Prevention of stillbirth has been recognized as an essential part of the Sustainable Development Plan (SDP), intended to end preventable stillbirth-related death by 2030 (29). The Every Newborn Action Plan aims to end preventable deaths, setting a stillbirth target of 12 per 1,000 births or less by 2030 (8). High-income and upper-middle-income countries have already met this target, but 56 countries particularly in Africa will have to more than double the present progress to reach this target (24). In addition, the International Federation of Obstetrics and Gynecology (FIGO) and the International Federation of Diabetes (IDF) issued a joint statement and declaration outlining the public health challenge posed by hyperglycemia in pregnancy and calling for urgent attention and action that is helpful to prevent the resulting adverse birth outcomes together with stillbirth (30).
However, there is no specific preventive strategic plan to reduce and control the existence of stillbirth; Ethiopia set different maternal and child healthcare plans like SDP (31) to combat such problems (32). Studies to identify the magnitude and determinants of stillbirth among diabetic pregnant mothers in Ethiopia had great discrepancy and inconsistent finding. Therefore, this systematic review and meta-analysis was the first in its kind to assess the pooled prevalence of stillbirth and its determinants among diabetic pregnant mothers in Ethiopia, which is pivotal to plan, evaluate, and improve national preventive measures of stillbirth among diabetic pregnant mothers.

Reporting
The review protocol has been registered in the international prospective register of systematic reviews (PROSPERO) with registration number (PROSPERO 2020: CRD2020167734), and the result of the review was presented based on standard Preferred Reporting Items for Systematic review and Metaanalysis (PRISMA) checklist (33) (Supporting Information 1).

Inclusion and Exclusion Criteria
Cross-sectional, case-control, and prospective cohort studies that reported the magnitude of stillbirth and/or at least one determinant among diabetic pregnant mothers were included. Similarly, both published and unpublished studies were involved. Studies that were conducted among pregnant women with preexisting (type 1 and 2 diabetes mellitus) and/or gestational diabetes mellitus confirmed by WHO were included. However, articles without full-text or abstract were excluded.

Outcomes of Interest
Stillbirth among diabetic pregnant mothers is the primary outcomes of interest; it can be defined as the birth of an infant at or after 28 weeks of gestation who at the time of delivery did not breathe or show signs of life (8).

Electronic Search
Electronic databases such as PubMed, Web of Science, Scopus, Google Scholar, and ScienceDirect were searched using Population, Intervention, Comparison and Outcomes (PICO) format. The included search terms used a combination of relevant Medical Subject Heading (MeSH) Boolean operators, such as "AND" and "OR, " to combine search terms. The search strategy was started on October 16, 2020, and an updated search was conducted on November 2, 2020, using the following search terms: (pregnancy OR pregnant) AND (women OR mothers) AND (diabetes mellitus OR type 1 diabetes mellitus OR juvenile diabetes mellitus OR insulin-dependent diabetes mellitus OR gestational diabetes mellitus OR type 2 diabetes mellitus OR adult-onset diabetes mellitus OR non-insulin dependent diabetes mellitus OR pre-existing diabetes mellitus OR pre-gestational diabetes mellitus) AND (fetal loss OR still birth) AND (Ethiopia). Moreover, gray literature found in Addis Ababa and Haramaya University online repositories was used. In addition, reference lists and citations of included papers were cheeked to identify other potentially relevant articles. Finally, the result of search strategy for PubMed database had been presented as an example ( Table 1). Population, Intervention, Comparison, and Outcomes (PICO) were as follows: P: pregnant women, I: diabetes, C: non-diabetes, O: stillbirth.
Research question: Do pregnant women who have diabetes compared to non-diabetic pregnant women have higher stillbirth?

Study Selection and Quality Assessment
EndNote version 9 (Thomson Reuters, London) reference manager (34) was used to remove duplicate studies. After screening titles and abstracts, a full-text review was done to determine the eligibility of each study by two independent authors (DM and WA). Any discrepancy was resolved by a third author (WA). Elsewhere, two independent authors (DM and DK) evaluated the eligibility of all retrieved studies using the Joanna Briggs Institute (JBI) quality appraisal checklist (35). Hence, one prospective cohort study (18), four cross-sectional studies (19)(20)(21), and two case-control studies (23,36) were appraised for quality using the JBI checklist. Therefore, studies are considered high quality whenever they meet 50% and/or above quality assessment criteria. Accordingly, none of these studies were excluded based on the quality assessment criteria as the quality assessment result ranges from 62.5 to 100% ( Table 2).

Risk of Bias Assessment
The risk of bias tool developed by Hoy et al. (37) was used to assess the risk of bias for each study. The tool comprises 10 items that address four areas of bias. Items 1-4 assess the external validity of the studies (domains are selection and non-response bias) and items 5-10 assess the internal validity of the studies (items 5-9 assess the domain of measurement bias, and item 10 assesses bias related to analysis). Accordingly, studies were classified as having a low risk of bias when 8 or more of the 10 questions were answered "yes, " a moderate risk of bias when 6 to 7 of the 10 questions were answered "yes, " and a high risk of bias when 5 or fewer questions were answered "yes" ( Table 3).

Data Extraction
After collecting the required articles from the entire database, all important data were extracted by the authors (AY and AS) using the standardized data extraction form and crosschecked to ensure consistency. Three authors (DM, WA, and DK) extracted data independently on author/s name, year of publication, study area/region, study design, sample size, and prevalence of stillbirth with 95% CI and determinants. Any dissimilarity and inconsistencies among the authors were resolved by discussion and repeating the procedure. The reviewer contacted the corresponding author for further information whenever pertinent data were missed from the included studies.

Data Analysis
The extracted data were exported to Stata version 14 statistical software for meta-analysis. I 2 statistic and Cochran's "Q" test was used to calculate the percentage of total variation in the study estimated due to heterogeneity. I 2 statistic ranged between 0 and 100%; the value of 0, 25, 50, and 75% represented no, low, moderate, and high heterogeneity, respectively. In addition, a pvalue of the I 2 statistic of <0.10 was used to declare significant heterogeneity (38,39). The pooled prevalence of stillbirth with

Search Results and Study Characteristics
Our searches from different electronic databases yielded a total of 1,982 articles. After removing 100 duplicate articles, 1,880 articles remained. From the remaining 1,880 articles, 1,080 articles were excluded after reviewing titles and abstracts. From the remaining 800 articles, 195 articles were omitted because their full text was not available. Lastly, 605 articles were suitable for full-text review. Still, 598 articles were excluded based on the predetermined eligibility criteria. Finally, seven articles (18)(19)(20)(21)(22)(23)36) were included for meta-analysis (Figure 1). The total sample size of the included studies was 1,225 diabetes mellitus in pregnancy, where the sample size of the individual studies ranges from 45 in Hiwot Fana and Dilchora Hospital (23) to 346 in Mettu Karl Hospital (19). From the included studies, four were done in Addis Ababa (18,(20)(21)(22), one study was carried out in Amhara region (36), and the remaining two were conducted in the Oromoia region (19,42). All studies were conducted at different hospitals of Ethiopia using different study designs: four cross-sectional (19)(20)(21), two casecontrol (23,36), and one prospective cohort (18). Moreover, all the included studies were studied from the period 2013-2018. Among the included studies, six studies contributed data for the prevalence of stillbirth (18)(19)(20)(21)(22)(23) and five studies (18,19,21,22,36) contributed data regarding determinants of stillbirth with response rate ranges from 80.2 to 100%. Accordingly, the prevalence of stillbirth among diabetic pregnant women ranged from 2.6 to 16.05% (19,22). Moreover, the included studies showed that the presence of glucometer at home, preconception care, being a housewife, high blood glucose level, antenatal care (ANC) visit, maternal age, rural residency, short birth interval, level of education, and history of previous adverse birth outcomes were significantly associated with stillbirth (18,19,21,22,36) ( Table 4).

Publication Bias
Asymmetrical distribution of the funnel plot implies the presence of publication bias among the included studies (Figure 3). Furthermore, Egger's test with a p-value of 0.006 shows the presence of publication bias ( Table 5).

Trim and Fill Analysis
We have performed trim and fill method analysis. A bias-adjusted effect estimate of stillbirth among diabetic pregnant mothers was found to be 1.104 at a p-value of 0.420, assuming there are missing studies ( Table 6).

Investigation of Heterogeneity
There was no significant heterogeneity among studies to show the pooled prevalence of stillbirth (I 2 = 31%, p = 0.19).

Determinants of Stillbirth
To identify the pooled determinants for the prevalence of stillbirth among diabetic pregnant mothers, meta-analysis was computed. We only included adjusted factors that were investigated in at least two studies, and the definition of the same factors has to be similar across all the included studies. As a result, a maternal age of <30 years, a gestational age of <37 completed weeks, and being a housewife were the criteria involved in the meta-analysis. Consequently, being a housewife [AOR = 2.25 (95% CI: 1.26, 3.23)] (Figure 4), maternal age of <30 years [AOR = 2.08 (95% CI: 1.02, 3.13)] (Figure 5), and gestational age of <37 completed weeks [AOR = 9.76 (95% CI: 7.83, 11.70)] (Figure 6) were significantly associated with stillbirth among diabetic pregnant mothers.

DISCUSSION
Increased risk of stillbirth in diabetes pregnancies has been a wellknown and recognized complication for decades. This systematic review and meta-analysis estimated the pooled prevalence and determinants of stillbirth among diabetic pregnant mothers in Ethiopia. Hence, the pooled prevalence of stillbirth among diabetic pregnant mothers was 2.39 [95% CI: −0.20, 4.97], which is consistent with the study conducted in King Khalid University Hospital (2.6%) (17), Saudi Arabia (3.66%) (14), Delhi (2.6%) (43), and Vietnam (1.43%) (44). This consistency could be due to a similar healthcare package and system toward maternal and newborn health service between countries. However, our finding is lower than the finding from Ghana 5.55% (16), Canada (13.3%) (45), and China (42.7%) (46). Differences in universal screening strategies of diabetes mellitus during pregnancy between the countries contributed for the discrepancy; Ethiopia has been using WHO diabetes screening strategies, whereas Ghana and China has been using American Diabetes Association criteria to screen diabetes mellitus during pregnancy. Besides, the difference in the study participants between Canada (only women with preexisting diabetes mellitus) and Ethiopia (both pre-existing and gestational diabetes mellitus) might have played a role for the variation.
However, our finding is higher than the study conducted in Bangladesh (0.014%) (47) and England (13.9 per 1,000 births) (12). The possible reason for the variation could be due to the socio-demographic difference of the study participants. The study participants in Ethiopia were from both rural and urban residency. In contrast, the study participants in Bangladesh were only from urban areas. Furthermore, the outcome of interest in Ethiopia was after 28 weeks of gestation whereas that in England was after 32 weeks of gestation.
Based on this review, the risk of stillbirth among diabetic pregnant housewife mothers was 2.25 times higher than their counterpart. This is due to the fact that housewife mothers have a low socio-economic status and have limited information about the importance of ANC to reduce the risk of adverse birth outcomes like stillbirth.
Likewise, the likelihood of stillbirth among diabetic mothers <30 years of age was two times higher than their counterpart. This is due to null parity being a risk factor for stillbirth. Thus, a low maternal age mother probably gives stillbirth due to null parity.
Moreover, the odds of stillbirth among diabetic mothers with a gestational age of <37 completed weeks was 9.76 times higher than the respective counterpart. This is due to the fact that when gestational age decreases, the maturity of the fetus and the chance to survive decrease, which results in stillbirth.
It is recommended to offer concern for mothers with low maternal age during pregnancy. It is also highly recommended to increase the coverage of ANC for all mothers with   hyperglycemia for early detection and to control glycemia to normal levels. Furthermore, community awareness about the related adverse birth outcomes due to maternal hyperglycemia should be promoted.

STRENGTH AND LIMITATIONS OF THE STUDY
This is the first systematic review and meta-analysis to estimate the national prevalence of stillbirth and its determinants among diabetic mothers. The limited number of primary studies used to investigate the pooled prevalence of stillbirth in the nation is a limitation. Besides, all the primary studies are hospital-based studies, which affects representativeness to the general community.

CONCLUSIONS
The national pooled prevalence of stillbirth among diabetic pregnant mothers was 2.39%, which is higher than the prevalence of stillbirth among the general population (1.18%). Low maternal FIGURE 6 | Forest plot to show the association between gestational age of <37 completed weeks and pooled prevalence of stillbirths among diabetic pregnant mothers in Ethiopia, 2020.
age, gestational age of <37 completed weeks, and being a housewife were significant determinants of stillbirth.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author/s.