The inclusion criteria for this study include: the protocol for this review was recorded in the International Prospective Register of Systematic Reviews (PROSPERO) with the ID CRD42024535950. The research adheres to the guidelines outlined in the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement (12). The search strategy was based on the PICOS (P: Population; I: Intervention; C: Comparison; O: Outcome; S: Study design) methodology. The PICOS framework is as follows: Population—singleton pregnant women; Intervention—vitamin D deficiency (25-OH-vitD3 < 20 ng/ml) (9); Comparison—adequate levels of vitamin D (25-OH-vitD3 ≥ 30 ng/ml) (9) in the control group; Outcome—occurrence of preterm labor (<37 weeks' gestation), LBWI (less than 2,500 g), or SGA (birth weight less than 10th percentile for gestational age) (13); and Study design—cohort or case-control study.

The exclusion criteria for this study include: (1) incomplete data where the full text is unavailable; (2) conference abstracts, systematic evaluations, reviews, case reports of animal experiments, expert consensus, and master's and doctoral degree papers; (3) duplicated publications; (4) literature not in Chinese or English; and (5) low-quality literature with questionable data authenticity.

A systematic review was conducted to examine the impact of vitamin D deficiency during pregnancy on preterm delivery, low birth weight infants, and small for gestational age infants. Literature searches were conducted in databases including PubMed, Embase, Web of Science, Cochrane Library, VIP Chinese Science and Technology Journal Database (VIP), China National Knowledge Internet (CNKI), Wanfang Database, and Sinomed, with search terms in both Chinese and English languages. The search strategy employed for this meta-analysis included keywords such as “Pregnancy” OR “Pregnancies” OR “Gestation” AND “Vitamin D Deficiency” OR “Vitamin D Deficiencies” OR “Deficiencies, Vitamin D” AND “newborn” OR “infants” OR “neonate” up until August 2023. A combination search approach utilizing MeSH subject terms and free terms was employed, with references being systematically traced back to the included literature. The detailed search strategy for PubMed is outlined in Supplementary Table S1.

Two researchers (You ZY and Zhang YY) conducted independent screening of the literature for data extraction, cross-checking their findings. Any discrepancies were resolved through consensus discussion or consultation with a third party. The data extraction process encompassed: basic information of the included studies, including study title, first author, year of publication, study type, baseline characteristics, diagnostic criteria of study subjects, risk of bias evaluation, outcome indicators, and outcome measures of interest.

The quality of the literature was evaluated using the Newcastle-Ottawa Scale (NOS) for cohort and case-control studies. The evaluation of cohort studies encompassed three primary dimensions: selection of study population, comparability between study groups, and measurement of outcomes, with scores on the NOS ranging from 0–9 points. Two researchers (You ZY and Zhang YY) independently conducted assessments, with a third researcher available to resolve any discrepancies. Studies receiving a score of 7–9 were deemed to be of high quality, indicating a low risk of bias.

A meta-analysis was conducted using RevMan54 and Stata 14 software to systematically assess the impact of vitamin D deficiency during pregnancy on adverse birth outcomes, specifically preterm infants, LBWI, and SGA infants. Mean ± standard deviation (MD) was utilized as the effect indicator for continuous data, while odds ratio (OR) was used for categorical data, with each effect size reported with a 95% confidence interval (CI). Heterogeneity was assessed through the application of the Chi-square test and I² statistic. When I²> 50% and P ≤ 0.1, it signified heterogeneity among the studies. In such cases, the random-effects model was employed to compute the combined statistic, and subgroup or sensitivity analyses were conducted to explore the underlying sources of heterogeneity. Conversely, if these criteria were not met, the fixed-effects model was utilized. Additionally, funnel plots were generated for the outcome measures, and publication bias was evaluated through the integration of Egger's and Begg's tests. There was significant difference between the two group when P ≤ 0.05.

A total of 2,720 articles were obtained, comprising 568 articles in Chinese (256 articles in CNKI, 118 articles in Wanfang database, 22 articles in VIP database, and 172 articles in Sinomed) and 2,152 articles in English (890 articles in PubMed, 634 articles in Embase, 142 articles in Cochrane library, 486 articles in Web of Science). 876 duplicates were removed, and 579 articles were excluded due to their meta, systematic evaluation, case report, expert consensus, animal test, etc. After reviewing titles, and abstracts, a total of 1,130 articles were excluded for non-compliance with intervention and control measures, study content, and outcome indicators. After reviewing full texts, an additional 122 articles were excluded for reasons such as inability to locate full text, incomplete data, non-case-control or cohort study design, and low quality. Ultimately, 13 cohort studies (8, 14–25) were included in the literature screening process, as depicted in Figure 1. The 13 publications were published from 2012–2021, covering a total of 8 countries and regions, involving 28,155 cases in the group with vitamin D deficiency in pregnancy (experimental group) and 8,580 cases in the group with adequate levels of vitamin D (control group). The basic information of the literature is shown in Supplementary Table S2.

Thirteen cohort studies (8, 14–25) were assessed for quality using the NOS, with two articles receiving a score of 7 (18, 25), seven articles receiving a score of 8 (14, 16, 17, 19, 21–23), and four articles receiving a score of 9 (8, 15, 20, 24), indicating high-quality literature. The definitions of vitamin D deficiency in pregnancy and gestational staging in these studies were based on standardized criteria. The quality scores of the included articles can be found in Supplementary Table S3.

Effect of vitamin D deficiency in pregnancy on preterm infants:A total of nine papers (8, 14, 15, 17–20, 22, 23) reported the effect of vitamin D deficiency in pregnancy on preterm labor. Statistical heterogeneity existed among the studies (P < 0.00001, I²= 82%). Meta-analysis was performed using a random-effects model, which showed that in the vitamin D deficiency group during pregnancy, compared with the adequacy group, the Preterm birth rate was not statistically significant (OR = 1.25, 95% CI 0.78–1.99, P = 0.36); subgroup analysis was performed according to the gestational week of maternal vitamin level measurement: there was no statistically significant effect of vitamin D deficiency on the occurrence of preterm birth in the first (OR = 1.42, 95% CI 0.97–2.07, P = 0.07), second (OR = 0.94, 95% CI 0.55–1.62, P = 0.83), third (OR = 0.71, 95% CI 0.09–5.90, P = 0.75) trimester of pregnancy, and the whole (OR = 1.83, 95% CI 0.52–6.45, P = 0.34) pregnancy (Figure 2).

Effect of vitamin D deficiency in pregnancy on LBWI: A total of five papers (14, 16, 18, 21, 25) reported the effect of vitamin D deficiency in pregnancy on LBWI, and there was statistical heterogeneity among the studies (P < 0.001, I²= 90%). Meta-analysis was performed using a random-effects model, and the results showed that neonates in the deficient group were more likely to develop LBWI than those in the vitamin adequate group during pregnancy, and the difference was statistically significant(OR = 5.52, 95% CI = 1.31–23.22, P = 0.02); subgroup analyses were performed according to the gestational week in which the vitamin levels of pregnant mothers were measured: the whole pregnancy (OR = 12.58, 95% CI = 4.58–34.59, P < 0.001) found that vitamin D level deficiency increased the incidence of LBWI, while there was no statistically significant difference in the effect of vitamin D levels on LBWI in the first (OR = 12.61, 95% CI = 0.10–1,553.12, P = 0.30), second (OR = 1.60, 95% CI = 0.70–3.65, P = 0.27) and third (OR = 2.65, 95% CI = 0.35–20.60, P = 0.35) trimesters (Figure 3).

Effect of vitamin D deficiency in pregnancy on SGA: Nine studies (14–17, 20–24) reported the effect of vitamin D deficiency in pregnancy on SGA, and there was statistical heterogeneity among studies (P < 0.001, I²= 89%), a random-effects model was used for meta-analysis, and the results showed that there was no statistically significant difference in the incidence of SGA in newborns between the vitamin D deficiency group and the adequate group during pregnancy (OR = 1.47, 95% CI = 0.81–2.68, P = 0.21). Subgroup analyses were conducted according to vitamin D deficiency included in gestational age: neonates with vitamin D deficiency were more likely to develop SGA throughout pregnancy (OR = 6.66, 95% CI = 4.47–9.94, P < 0.001), and the effect of vitamin D levels on SGA in the first (OR = 1.05, 95% CI = 0.82–1.36, P = 0.69) and second (OR = 1.11, 95% CI = 0.69–1.78, P = 0.68) trimesters was statistically insignificant (Figure 4).

We use inverted funnel plots analysis of each outcome to evaluate the publication bias among the studies (Figures 5A,C,E). The results showed that the dots in the funnel plot were basically symmetrical, indicating that the possibility of publication bias was small. Egger test results showed that no significant publication bias was found (all P > 0.05) (Figures 5B,D,F). Sensitivity analyses were conducted on birth outcomes. Deleting individual studies had minimal impact on the overall results, indicating the differences between studies are small and the analysis is highly reliable (Figures 6A–C).