Publications from its establishment to October 19, 2023 were searched in PubMed, Embase and Web of Science databases, and the search was not limited by language. In addition, reference tracks (Federal Bureau of Agriculture), abstracts,¹ and Google Scholar² were used to supplement the search. The search strategies used were: (hypocalcemia OR hypocalcemias OR parturient paresis OR parturient pareses OR animal milk fever) AND (cattle OR Bos indicus OR zebu OR zebus OR Bos taurus OR domestic cow OR domestic cows OR Bos grunniens OR yak OR yaks) AND (calcium OR calcium 40).

According to the PRISMA guidelines for Systematic Reviews and Meta-Analysis (33), following the principles of PICO (Problem/Patient/Population, Intervention/Indicator, Comparison, Outcome), the inclusion standards are pre-defined as follows: (i) question: healthy postpartum cows; (ii) intervention: oral calcium; (iii) comparison: standard interventions except for oral calcium; (iv) results: production-related diseases and blood indicators, including clinical and subclinical hypocalcemia, and Ca, Mg, and P concentrations in the blood, pregnancy risk and milk yield; and (v) study design: control in vivo experiment. In contrast, the exclusion criteria were as follows: (i) there are no restrictions on specific cow breed, parity; (ii) no limit to the sources, doses, forms and frequency of oral calcium; (iii) case reports, review articles, and letters; (iv) unreliable or incomplete data.

Two authors collected the details of included studies independently, and a third person adjudicated when they disagreed. These details include: (i) first author name and year of publication; (ii) characteristics of animal models (breed, parity, milk production and feed); (iii) information on treatment/control group, including the treatment drug and control, sample size, calcium form, dosage, time, and number of administrations; and (iv) outcome indicators (incidence of clinical and subclinical hypocalcemia, blood Ca concentrations; Mg and P serum concentrations; the incidence of ketosis, mastitis, metritis, displacement of the abomasum, retained placenta; pregnancy risk at first service, overall pregnancy rate and milk yield). If there are multiple groups of experimental group data with the same intervention in the literature, each group will be considered independent data and included in the analysis. In addition, the data in the graph were measured by using a digital ruler software (Getdata Graph Digitizer, version 2.25, Russia).

The Cochrane Collaboration developed the RoB tool to assess the quality of Randomized Controlled Trial (RCT) to avoid the risk of bias in animal experiments, and the SYRCLE RoB tool is an adapted version of the Cochrane RoB tool. The SYstematic Review Center for Laboratory Animal Experimentation (SYRCLE) RoB tool was used to assess the risk of bias in animal studies by using the 10 items: sequence generation (selection bias), baseline characteristics (selection bias), allocation concealment (selection bias), random housing (performance bias), blinding (performance bias), random outcome assessment (detection bias), blinding (detection bias), incomplete outcome data (attrition bias), selective outcome reporting (reporting bias), other sources of bias (other). The evaluation results of each project classify the research into high risk, low risk or unclear risk according to PRISMA guidelines. In the result chart, red, green and yellow represent “high risk,” “low risk” and “unclear risk” respectively (34).

This study, used RStudio software (version 5701.9.1.0) for data analysis. The mean difference (MD) with 95% confidence interval (CI) for continuous variables and relative risk (RR) with 95% CI for dichotomous variables were calculated. Heterogeneity was assessed using I² and Π² statistical tests, and heterogeneity was considered high, if I² was greater than 50%. Sensitivity analyses were performed for outcome measures with moderate heterogeneity (heterogeneity I² of nearly 50%) to assess the robustness of the results. Statistical significance was set at p < 0.05. Publication bias assessment was performed using Egger’s test.

A total of 2,110 potentially relevant studies were identified through a database searching and supplementary searches. 343 studies were considered duplicated, while 1,402 studies were excluded since they did not meet the inclusion criteria. Finally, 22 eligible studies were included, as shown in Figure 1 (English = 21, German = 1).

Of the 22 eligible studies, 15 studies used Holstein cows, three studies used Jersey cows, and five studies adopted crossbred cows. One study did not report the breed used for their experiment. Regarding parity, 12 studies used multiparous cows, eight used primiparous and multiparous cows, and two did not report parity. Seventeen studies reported milk production in cows, and 18 reported DCAD. The major outcome indicators were the incidence of clinical and subclinical hypocalcemia. Secondary outcome indicators include the concentration of Ca, Mg, and P in the blood; incidence of ketosis, mastitis, metritis, displacement of the abomasum, retained placenta; pregnancy risk at first service, overall pregnancy rate and milk yield. The overall characteristics of the included studies are summarized in Table 1.

The definitions of clinical diseases in the included studies are not consistent. Clinical hypocalcemia is defined as lying flat within 24 h after delivery, corrected by intravenous injection of calcium, or as the symptoms of lying weak or cold limbs without signs of physical injury within 72 h after delivery (28, 29). Subclinical hypocalcemia is defined as serum Ca ≤ 2.12 mmol/L, or blood total Ca < 2.125 mM, or ≤ 1.00 mmol/L on calving day (23, 35, 38). Ketosis is defined as a sharp drop in milk production, with acetoacetic acid in urine, or BHBA ≥1.2 mmol/L on BHBA test (24, 25, 38). Metritis is defined as foul-smelling uterine secretions that are red to brown within 14 days after delivery, or body temperature ≥ 39.5\u00B0C, lethargy, anorexia and vaginal discharge (29, 35). Displacement of the abomasum is defined as the decrease of milk production, and at the same time, when tapping and auscultating the left or right abdominal wall between the ninth and twelfth rib spaces, high-pitched tympanic resonance can be heard, with or without colic (24). Retained placenta is defined as the fetal membranes or placenta were not discharged within 24 h after delivery (28, 29, 35).

The characteristics of the interventions are presented in Table 2. Table 2 is as simple as possible, but due to the characteristics of study design or intervention, some studies have extracted multiple sets of data. The sample sizes of the included studies varied widely, ranging from 10 to 3,999 animals, with only 10 studies having the same sample size for the experimental and control groups. Accurate data of cows assigned to each experimental group could not be obtained from one study after contacting the authors (46). The treatment group cows in the included studies used different forms of Ca, including Ca gel, Ca tube, Ca bolus or Ca liquid preparation, among which Ca bolus was the most commonly used (18 studies). Cows in the control group usually have no calcium supplement, and only one study used a placebo made of test tubes filled with gelatin. There are varying many sources of Ca, and in most studies, calcium chloride (CaCl₂) was used as the primary calcium source, one study did not report Ca sources. Eighteen studies used commercial brands of calcium. Moreover, the treatment times and frequencies in the included studies were inconsistent. In most studies, the time for oral calcium administration was 0–6, 12, and 24 h after calving. The times for the treatment regimen was 1 to 3, only two articles administer oral calcium 5 times. The dose was also inconsistent (25–110 g).

The quality assessment of the included studies using SYRCLE’s RoB tool is show in Figure 2, in which all 22 studies were considered as controlled studies. Sixteen studies used random methods to generate sequences (low risk, green), however in four studies, cows were alternately assigned to treatment and control groups, based on the calving date or lactation order (high risk, red). Two studies did not describe the grouping method (unclear risk, yellow). The number of animals in the experimental and control groups in the 12 studies was also inconsistent; therefore, the risk of bias assessment of their “baseline characteristics” was unclear (unclear risk, yellow), and 19 studies adjusted for confounders in the analysis (low risk, green). Eleven studies described the allocation concealment method (low risk, green), and during the experiment implementation stage, seven studies mentioned random housing of the animals (low risk, green), while, 15 studies did not describe them (unclear risk, yellow). Lighting, humidity, temperature and other housing conditions are known to affect the research results, so it is necessary to randomize the feeding of these animals to reduce the performance bias (48). Only seven studies mentioned blinding measures used to blind trial investigators, study designers, and the outcome assessors from avoiding know which intervention each cow had received (low risk, green). Moreover, nine studies mentioned random selection of animals for assessment (low risk, green). Twenty-two studies described complete data for each primary outcome (low risk, green). In the assessment of “Is the study report irrelevant to selective outcome reporting,” all studies were considered low risk of bias (low risk, green). Eight studies mentioned other problems that could have resulted in an unclear risk of bias (low risk, green).

A meta-analysis of the five studies showed that oral Ca significantly reduced the incidence of clinical hypocalcemia, compared to the control (n = 1,629; RR = 0.67; 95% CI = [0.52, 0.87], p = 0.003; heterogeneity I² = 49%; Figure 3A), and the difference between the two groups was statistically significant. The meta-analysis of six studies shown that oral Ca significantly reduced the incidence of subclinical hypocalcemia (n = 4,903; RR = 0.81; 95% CI = [0.72, 0.91], p = 0.0005; heterogeneity I² = 66%; Figure 3B). The difference between the two groups was statistically significant.

A meta-analysis of 14 studies showed that oral Ca significantly improves Ca concentration in blood compared to the control (n = 1797; MD = 0.08; 95% CI = [0.04, 0.11]; p < 0.0001; heterogeneity I² = 73%; Figure 4A), the difference between the two groups was statistically significant. The meta-analysis of serum Mg and P concentrations illustrated in Figures 4B,C, the MD of serum Mg concentrations was −0.04 (95% CI = [−0.10, 0.02]; p = 0.2068; heterogeneity I² = 87%), the result was not statistically significant; the MD of serum P concentrations was 0.05 (95% CI = [−0.10, 0.21]; p = 0.4944; heterogeneity I² = 79%), this result was not statistically significant.

In this study, other calving-related clinical disorders including: retained placenta, metritis, ketosis, mastitis, displacement of the abomasum were included. The results of meta-analysis are shown in Figures 5, 6. A meta-analysis of the six studies showed that oral Ca significantly reduced the incidence of retained placenta, compared to the control (n = 4,523; RR = 0.77; 95% CI = [0.62, 0.95], p = 0.0151; heterogeneity I² = 23%; Figure 5A), the difference between the two groups was statistically significant. The RR of incidence of metritis was 1.06 (n = 6,112; 95% CI = [0.94, 1.19]; p = 0.3567; heterogeneity I² = 36%; Figure 5B); the RR of incidence of ketosis was 1.04 (n = 1,053, 95% CI = [0.91, 1.18]; p = 0.5933; heterogeneity I² = 0%; Figure 6A); the RR of incidence of mastitis was 1.02 (n = 5,854; 95% CI = [0.86, 1.21]; p = 0.8436; heterogeneity I² = 1%; Figure 6B); the RR of incidence of displacement of the abomasum was 0.81 (n = 3,357; RR = 0.81; 95% CI = [0.57, 1.16]; p > 0.2573; heterogeneity I² = 28%; Figure 6C); however, these results are not statistically significant (incidence of metritis, ketosis, mastitis and displacement of the abomasum).

In addition, we performed a meta-analysis of the number of pregnancies and milk yield, the RR of pregnancy risk at first service was 0.99 (n = 7,481; 95% CI = [0.94, 1.05]; p = 0.8309; heterogeneity I² = 42%; Figure 7A), the result was not statistically significant; the RR of overall pregnancy rate was 1.03 (n = 3,411; 95% CI = [0.98, 1.08]; p > 0.2030; heterogeneity I² = 7%; Figure 7B), this result was not statistically significant; the MD of milk yield was 0.44 (95% CI = [−0.24, 1.13]; p = 0.2022; heterogeneity I² = 99%; Figure 7C), the result was still not statistically significant.

Publication bias was analyzed using funnel plots and Egger’s test. The distribution of the funnel plot was symmetrical (Supplementary Figure S1) the p-values of incidence of subclinical hypocalcemia, Ca concentration in blood, incidence of metritis, and number of pregnant cows at first service in Egger’s test were 0.8667, 0.5329, 0.7050, and 0.8959 respectively, and the results suggested that there was no publication bias between oral Ca and the incidence of these diseases. Sensitivity analyses were performed for meta-analysis of the incidence of clinical hypocalcemia, subclinical hypocalcemia and number of pregnant cows at first service to assess the robustness of the results, and the results suggested that the results of meta-analysis are robust (Supplementary Figures S2–S6).