The systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. PubMed, Embase, Cochrane Library, and Web of Science databases were searched for relevant studies published up to April 2023. The subject terms included ‘Miscarriage,’ ‘Pregnancy loss,’ ‘Abortion, Spontaneous,’ ‘Zinc,’ ‘Copper,’ ‘Lead,’ and ‘Cadmium.’ Random combinations of these subject terms and their synonyms were used for retrieval. The detailed literature search strategy is provided in the Supplementary Material . We reached the corresponding authors when the data were missing.

Studies meeting the following criteria were included in the meta-analysis. (a) Study population: Pregnant women without internal and obstetric diseases that impair the normal process of pregnancy, including infectious diseases, gestational hypertension, gestational diabetes mellitus, and infertility. (b) Measurement: Female serum, plasma, or whole blood metal concentrations. (c) Observation group: Women who had experienced abortion, including SA and RPL. SA is defined as the involuntary loss of a fetus before the 28th GW (including the 20th and 24th GW) (2). RPL is defines as two or more abortions, irrespective of whether they are consecutive (5). (d) Control group: Healthy pregnant women with normal pregnancy or delivery. (e) Study type: Observational study.

The exclusion criteria were as follows: (a) article type: review, meta-analysis, meeting, case report, letter, comment, editorial, note, trial registry record, or protocol, (b) studies that focused on non-human cases (e.g., animal studies), (c) unclear definition of SA or RPL, (d) insufficient data on metal concentration; and (e) unavailable full text.

The studies that met our inclusion criteria after the initial search were case–control, nested case–control, and cross-sectional studies. Therefore, the Newcastle-Ottawa Scale (NOS) was used to assess the quality of the studies (25). Each included article was independently appraised by two authors (MR and LiqW). Based on the NOS, studies were categorized as high- (8, 9), moderate- (6, 7), or low-quality (<6). Any disagreements regarding the assessment of the studies were discussed with the third author (LiaW).

Two investigators independently extracted the relevant data from the included studies (MR and LiaW). All data were double-checked by the third author (LiqW). The following information was extracted from the selected studies: first author, publication year, country and continent of the study population, type of detected sample, type of article, type of heavy metal, type of abortion, follow-up endpoint, sample size, concentrations of heavy metals, and analytical method employed.

Meta-analysis was performed using Stata 16.0 (Stata Corp, College Station, TX, USA). The standard mean difference (SMD) was adopted to integrate the data on metal concentration, as it is a continuous variable with different units across various studies. The 95% confidence intervals (CIs) were computed and presented as forest plots. For each study, statistical heterogeneity was assessed using Cochran’s Q-test and I² statistics, and a random effects model was used to estimate the relationship between metal concentrations and abortion, as there was significant heterogeneity (p <0.05, I² >50%). To investigate the impact of metal concentration on the different types of abortions (SA and RPL), a subgroup analysis was performed. To investigate the origin of the heterogeneity, four additional subgroup analyses were performed based on the follow-up endpoints (ongoing pregnancy and live birth) of participants, continent of the study population (Africa, Asia, North America, Oceania, and Europe), type of article (case–control study, cross-sectional study, and nested case–control study), and type of detected sample (serum, plasma, and whole blood). An influence analysis (sensitivity analysis) was conducted to improve the reliability of the meta-analysis results. A funnel plot and Begg’s and Egger’s tests were used to detect potential publication bias; p-values <0.05 represented significant statistical publication bias for Begg’s and Egger’s tests.

Figure 1 illustrates the PRISMA flow diagram for the selection of studies for inclusion in the systematic review and meta-analysis. A total of 4,222 potential studies were identified through database search. Among them, 136 articles were removed for duplicates, 1,209 articles were not observational studies (including reviews, meta-analyses, meetings, case reports, letters, comments, editorials, notes, trial registry records, and protocols), and 2,829 articles were not relevant to our study based on screening of their titles and abstracts by two authors. The literature screening results were double-checked to ensure that the relevant documents were not missed and did not need to be retrieved. After an independent review of the full texts by three authors (MR, LiaW, and LiqW), 12 studies were excluded because they did not meet the inclusion criteria, and eight studies were excluded because they had insufficient data (they only reported mean values without standard deviation of the metal concentration). Twenty-eight relevant studies were subjected to a final quantitative assessment based on the exclusion and inclusion criteria. Among the 28 studies, 14 investigated Zn and Cu, 15 investigated Pb, and eight investigated Cd.

Table 1 lists the baseline characteristics of the included articles. The included articles were observational studies published between 1979 and 2023 and involved 1,377 abortion cases (including 1,159 females with SA and 218 females with RPL), together with 3,289 normal pregnant females. Of the 28 articles included in this meta-analysis, six were completed in China (37, 38, 40, 42, 45, 51), three in Egypt (26, 30, 31), three in Saudi Arabia (29, 36, 47), three in Iran (48–50), two in Mexico (33, 39), two in Turkey (35, 52), two in Poland (19, 41), one in Nigeria (27), one in Croatia (28), one in Italy (32), one in Australia (34), one in Serbia (43), one in India (44), and one in Russia (46). The sample size of the included studies ranged from 38 to 1,447.

The quality assessment results for all studies are shown in Supplementary Table 1 . Studies with quality scores higher than 6, which is the cut-off NOS score for low quality, were considered credible. All of the included articles had quality scores above 6.

Fourteen studies investigated the association between Zn concentrations and abortion. The pooled effect size showed that the Zn concentration was negatively associated with abortion (SMD = −1.05, 95% CI: −1.74, −0.36, p = 0.003, I² = 96.9%; Figure 2A ). Subgroup analysis showed that women with RPL had significantly lower Zn concentrations than healthy controls (SMD = −3.44, 95% CI: −5.01 to −1.87, p <0.001), whereas the Zn concentrations of women with SA and healthy controls were not significantly different (SMD = −0.14, 95% CI: −0.86–0.58, p = 0.710). Significant heterogeneity was observed in each subgroup (SA, p <0.001, I² = 96.6%; RPL, p <0.001, I² = 96.6%; Figure 2B ). To investigate the origin of the high heterogeneity, subgroup analyses based on follow-up endpoint, continent, type of article, and type of detected sample were performed. Subgroup analyses revealed persistently high heterogeneity ( Supplementary Figure 1 ). Sensitivity analysis showed that omission of any study did not change the overall effect ( Figure 2C ). There was no evidence of publication bias among the included studies (Begg, p = 0.511; Egger, p = 0.335; Figure 2D ).

Comparisons of Cu concentrations in women with and without abortion were reported in 14 studies. The pooled effect size of the 14 studies revealed significantly lower Cu concentrations in the abortion group than in the control group (SMD = −1.42, 95% CI: −1.97, −0.87, p <0.001, I² = 95.4%) ( Figure 3A ). Subgroup analysis stratified by the type of abortion (SA and RPL) showed that patients with SA and RPL had lower Cu concentrations than the healthy controls ( Figure 3B ); the SMD was −0.97 (95% CI: −1.47, −0.48, p <0.001) for women with SA and −3.92 (95% CI: −6.97, −0.87, p = 0.012, respectively). However, subgroup analysis based on abortion type showed obvious heterogeneity (SA, p <0.001, I² = 94.0%; RPL, p <0.001, I² = 97.8%). Subgroup analyses for the follow-up endpoint, continent, type of article, and type of detected sample showed high heterogeneity ( Supplementary Figure 2 ). Sensitivity analysis showed that omission of any single study did not change the overall effect ( Figure 3C ). Publication bias was detected in the studies that included Cu (Begg: p = 0.037; Egger: p = 0.012). Visual inspection of funnel plots showed asymmetry ( Figure 3D ).

The meta-analysis of the association between Pb concentration and abortion included 15 studies ( Figure 4A ). The pooled circulating Pb concentration was significantly higher in women who had experienced an abortion than in those with normal pregnancies (SMD = 1.47, 95% CI: 0.89–2.05, p <0.001, I² = 96.0%). Subgroup analysis for SA and RPL showed significantly higher Pb concentrations in women with SA (SMD = 0.33, 95% CI: 0.12–0.55, p = 0.002) and RPL (SMD = 8.19, 95% CI: 4.52–11.85, p <0.001) than in healthy pregnant women ( Figure 4B ). However, significant heterogeneity was observed (SA: p <0.001, I² = 68.6%; RPL: p <0.001, I² = 98.2%). Further subgroup analyses based on the follow-up endpoint, continent, type of article, and type of detected sample also showed high heterogeneity ( Supplementary Figure 3 ). Sensitivity analysis showed that omission of any single study did not change the overall effect ( Figure 4C ). Visual inspection of the funnel plots ( Figure 4D ) and Begg’s and Egger’s tests showed publication bias (Begg: p = 0.002; Egger: p = 0.001).

The pooled results of the meta-analysis of eight studies on Cd and abortion showed significantly higher Cd concentrations in women who underwent abortion than in normal pregnant women (SMD = 1.15, 95% CI: 0.45–1.85, p = 0.001, I² = 93.7%) ( Figure 5A ). Subgroup analysis based on abortion type showed that women with SA and RPL had significantly higher Cd concentrations than the normal controls (SA: SMD = 0.42, 95% CI: 0.02–0.82, p = 0.040; RPL: SMD = 2.45, 95% CI: 0.85–4.04, p = 0.003). However, the heterogeneity was significant in each subgroup (SA: p = 0.003, I² = 75.1%; RPL: p <0.001, I² = 94.2%) ( Figure 5B ). Further subgroup analyses based on the follow-up endpoint, continent, and type of detected sample also showed high heterogeneity ( Supplementary Figure 4 ). All studies on Cd and abortion were case–control studies, and subgroup analysis for different article types could not be performed. The sensitivity analysis showed that the exclusion of any single study could change the overall effect ( Figure 5C ). Both funnel plots and Begg’s and Egger’s tests showed no publication bias for Cd (Begg: p = 0.174; Egger: p = 0.113; Figure 5D ).

In our study, we found that Zn inadequacy tended to increase the chances of abortion, especially for RPL; however, it may not increase the incidence of SA. The underlying mechanism of Zn inadequacy-related RPL remains unknown. However, it may also be associated with endocrine dysfunction caused by Zn deficiency.

Zn deficiency has been reported to decrease insulin sensitivity and cause insulin resistance (IR) (57, 58), whereas Zn supplements can decrease IR (59, 60). IR, defined clinically as a decreased biological response to exogenous or endogenous insulin, can cause mitochondrial dysfunction in the placenta, diminished trophoblast invasion, a subclinical inflammatory state, and oxidative stress. These factors are all considered crucial in the pathophysiology of RPL (61–64). Zn can reinforce glucose transport into cells and potentiate insulin-induced glucose transport via the insulin signaling pathway (65). Zn can also act as an insulin mimetic to maintain glucose homeostasis, which may also be a mechanism underlying Zn deficiency-induced IR (66).

Apart from IR, Zn deficiency is also closely related to vitamin D deficiency, as Zn regulates the transcriptional activation of hormone-related genes via a cysteine-rich Zn-finger region in vitamin D receptors (VDRs) (67–69). Vitamin D plays a vital role in maintaining normal biological functions, such as calcium homeostasis, and cell proliferation, differentiation, and apoptosis, all of which are crucial for immunomodulation and normal pregnancy (70). Vitamin D inadequacy was reported to be associated with SA and RPL in a recent meta-analysis (71, 72). Supplementation with vitamin D can suppress inflammatory cytokine production and elevate the secretion of cathelicidin in decidual cells and trophoblasts, which can reduce the risk of abortion (73–75).

Zn deficiency appears to interfere with sex hormone synthesis and further causes RPL. Zn can affect the biosynthesis and function of sex hormones, such as progesterone and prolactin, by altering LH and FSH levels and inducing oxidative stress (17, 76–78). Zn may also promote estrogen release by forming ligand bonds with metal-binding sites on the estrogen receptor (ER) (79). Insufficient secretion of sex hormones, such as progesterone, testosterone, estrogen, and prolactin, can reduce endometrial receptivity and oocyte quality in women, which is related to RPL (55, 80, 81).

Furthermore, Zn is an essential trace element for thyroid function and homeostasis (82), and its deficiency can lead to hypothyroidism (82–84). Hypothyroidism and subclinical hypothyroidism can lead to poor pregnancy outcomes such as SA and RPL (85, 86). Zn supplements can elevate thyroxine (T4) concentrations and reduce triiodothyronine (T3) concentrations by altering the expression of key genes (nis, tpo, thrα, dio1, dio2, and ugt1ab) in the hypothalamic–pituitary–thyroid (HPT) axis (87).

We found that women undergoing abortion (both SA and RPL) had lower Cu concentrations, indicating that Cu deficiency may be closely related to the incidence of abortion (SA and RPL). However, the underlying mechanism remains unknown. However, previous studies have reported that Cu deficiency can induce endocrine dysfunction, such as IR, vitamin D insufficiency, and abnormal thyroid and sex hormone concentrations, which may be involved in the pathogenesis of SA and RPL. Insufficient Cu can cause IR by upregulating cytochrome c oxidase 1 (SCO1) and vascular adhesion protein-1 (VAP-1) (88–90); it can also reduce progesterone synthesis by regulating the expression of steroidogenic factor 1 (SF-1) (91). Cu deficiency can also reduce Cu/Zn superoxide dismutase (Cu/Zn-SOD) and cause oxidative stress in the ovary, ultimately leading to dysfunctional luteal formation and insufficient progesterone secretion (17). In addition, Cu deficiency can decrease the expression of estrogen synthetases such as aromatase (CYP19A1) and 17β-hydroxysteroid dehydrogenase (17β-HSD) (92). Furthermore, Cu deficiency can lead to hypothyroidism (82–84, 93) by inducing oxidative stress and decreasing thyroxine synthesis by limiting tyrosinase availability (82, 93–95).

Our study found that women who experienced abortion (SA and RPL) had higher Pb concentrations, suggesting that Pb exposure could increase the risk of abortion (SA and RPL). Our results are consistent with those of the meta-analysis by Kaur et al. (23). Pb can substitute polyvalent cations, such as calcium (Ca²⁺), and affect various cellular processes, such as apoptosis, cell adhesion, and cell signaling (96). However, the mechanism underlying Pb-induced abortion remains unclear. Animal studies have shown that Pb exposure can downregulate IR-related genes in the P13K and Akt signaling pathways, which are involved in hepatic gluconeogenesis and glucose production (97, 98). Low-level Pb exposure promotes the gene expression of key enzymes involved in hepatic gluconeogenesis and eventually induces hyperglycemia and impaired fasting plasma glucose, which is known as hepatic insulin resistance (99). Additionally, Pb appears to be involved in the pathology of vitamin D deficiency. Pb can diminish the activity of vitamin D by blocking the normal renal synthesis of active 1,25-dihydroxy vitamin D (1,25(OH)2D) and reduce the generation of vitamin D binding protein (DBP) (56, 100, 101). Pb can also promote degradation and block the synthesis of 1,25(OH)2D3 by upregulating the hepatic expression of Cyp24a1 enzymes and inhibiting 25-hydroxylase (CYP2R1) and 1-α-hydroxylase (CYP27B1) at the gene and protein levels (100, 102). Previous studies have suggested that Pb may be closely associated with luteal phase deficiencies. Pb can directly inhibit the expression of several key enzymes involved in progesterone synthesis, such as StAR, CYP11A1, and 3β-HSD (103, 104). Pb also appears to indirectly interfere with progesterone synthesis by inhibiting the cAMP-PKA-dependent signaling pathway that regulates the expression of these key enzymes (104–106). Pb has adverse effects on sex hormone concentration. Pb exposure is associated with increased testosterone and prolactin concentrations and appears to reduce estrogen concentrations by decreasing the expression of estrogen synthases such as 17β-HSD (103, 107, 108). Furthermore, Pb accumulation negatively affects thyroid function, which is also related to abortion. Excessive exposure to Pb may lead to hypo- or hyperthyroidism (109). As an oxidant, Pb can negatively impact thyroid cells by promoting oxidative stress, and it can also interact with other essential elements such as Cu, Zn, and Fe to indirectly affect thyroid function (82, 109).

We found that the Cd concentration was significantly higher in women who experienced abortion (SA and RPL) than in normal pregnant women. Our results are in line with those of Kaur et al., who revealed that Cd exposure could increase the risk of abortion (23). Cd is a highly potent environmental pollutant that causes indirect oxidative damage to DNA, leading to the induction of cellular proliferation and inhibition of DNA repair mechanisms, causing cytotoxicity (110). However, research on the mechanisms of Cd exposure-related abortion is lacking. In recent years, an increasing number of studies have found a strong relationship between Cd and endocrine dysfunction, which is the main reason for abortions (both SA and RPL). Epidemiological surveys have shown that Cd can cause IR through perturbations in gluconeogenesis, pancreatic islet dysfunction, and metabolic and mitogen impairments in the liver and adipose tissue (111, 112). Epidemiological studies have also demonstrated that high blood Cd concentrations are negatively correlated with vitamin D concentrations (113, 114), which may be due to the interaction of Cd with renal mitochondrial hydroxylases (115). Cd is also involved in the pathogenesis of luteal phase deficiencies. It can directly or indirectly inhibit the expression of several key enzymes (StAR, CYP11A1, and 3β-HSD) involved in progesterone synthesis by regulating the cAMP-PKA-dependent signaling pathway (24, 116, 117). Cd can also interfere with the balance of sex hormone concentrations. Cd exposure appears to decrease the expression of estrogen synthetases (CYP19A1 and 17β-HSD), and it is also a potent xenoestrogen that can mediate the proliferation of anterior pituitary cells and prolactin secretion by mimicking estrogen (118). Cd can also negatively affect thyroid cells by promoting oxidative stress, ultimately leading to thyroid dysfunction (82, 109).

The strength of our study is that we comprehensively investigated the relationships between blood Zn, Cu, Pb, and Cd concentrations and abortion rates (SA and RPL). In addition, we systematically reviewed previous publications on the endocrine mechanisms of metal exposure-related abortions. We propose that IR, vitamin D insufficiency, and abnormal thyroid and sex hormone concentrations may be involved in Zn and Cu deficiencies and Pb and Cd exposure-associated abortions. Third, the large sample size of 4,666 pregnant women from 14 countries makes our estimates reliable. Fourth, most included studies were of good quality. In the included studies, the definitions of cases and controls were adequate, and the selection of controls and assessment of exposure were consistent. Fifth, a sensitivity analysis was conducted to verify the associations between the four metals and abortion.

This study has some limitations. First, several dated documents that appeared to meet our inclusion criteria were not included because we were unable to reach the authors. Second, despite our best efforts, we were only able to find 28 related papers because of the relatively large number of animal studies and case reports. Third, most meta-analyses included in our study had high heterogeneity. To ascertain its sources, we performed a subgroup analysis based on the type of abortion, follow-up endpoint, continent, type of observational study, and type of detected sample. However, we failed to find sources of heterogeneity by subgroup analysis, as most subgroup analyses showed high heterogeneity. After carefully reviewing the included articles, we found that different diagnoses of SA or RPL may have led to clinical heterogeneity. In addition, regional variations in metal concentrations of the study participants were considered another source of heterogeneity. Local mineral deposits and their exploitation affect the metal concentrations in the environment (air, water, and soil), and different terrains can impact the diffusion of pollutants (119). Thus, participants had different risks of metal exposure. Moreover, participants of different races with different genetic backgrounds have various sensitivities to metal exposure (120). Owing to the limited information regarding the region and race of the investigated subjects in the original literature, we were only able to perform a subgroup analysis based on different continents to investigate the heterogeneity caused by regional differences. The age of the study population and the time point of blood collection in each study were also considered potential sources of heterogeneity, as heavy metals could accumulate in the human body, and older adults may have higher blood metal concentrations. Furthermore, metal concentrations can change during different trimesters and the time points of blood collection may lead to heterogeneity (121, 122). Fourth, the literature regarding Cu and Pb had obvious publication and reporting bias, although the publication and reporting the bias of literature regarding Zn and Cd were acceptable. Language and multiple publication biases were considered primary problems as only the English literature was included, and two studies had outcomes from the same study population.

The findings of the present study broaden our understanding of the effects of toxic and essential metals on the RPL and SA. Endocrine dysfunction can lead to metal exposure and abortions. It will be helpful to screen blood Zn, Cu, Pb, and Cd concentrations in females. However, well-designed prospective cohort studies are needed to clarify the causal relationship between endocrine dysfunction and heavy-metal-induced abortion.