The literature searches identified 2503 studies, with the addition of 32 studies identified through reference screening (Fig. 1). Following a review of 96 full-text articles, 57 were excluded for failing to meet the inclusion criteria. Five studies were excluded because they focused on infants under 1 year of age (1 study) and reported duplication of cohort and data in four other studies. Only 34 studies [16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49] were ultimately included in the meta-analyses. The quality of the included studies that were assessed by the Newcastle–Ottawa Scale is shown in Supplemental Table S1 for cohort studies reporting intelligence outcomes, Supplemental Table S2 for case–control studies reporting intelligence outcomes, Supplemental Table S3 for cohort studies reporting motoric outcomes, Supplemental Table S4 for case–control reporting motoric outcome, Supplemental Table S5 for cohort studies reporting behavioral outcomes.

Flow diagram of included study in meta-analysis

Table 1 shows pooled analysis of the background characteristics of the children. There were no differences in children’s gender proportion, mother education level, and family socioeconomic background in all age groups born from all types of ART conceptions compared to naturally conceived control.

Four studies used Bayley’s Mental Development Index to measure cognitive development in the toddler age group [16, 19, 20, 26]. There were no significant differences in the mental development of assisted reproductive technology (ART)-born compared to naturally conceived (NC) toddlers (p = 0.16). There was no evidence of publication bias (p-Egger = 0.506), and the data exhibited good homogeneity (I 2 = 0%, p = 0.94) (Fig. 2A). Supplemental Table S6 summarizes the statistics for the meta-analysis.

Intelligence outcome of ART-born compared to naturally conceived (NC) toddlers as assessed with A Mental Development Index, B Language Development (Verbal Intelligence) Score, and C Non-verbal Intelligence Score

The McArthur Bates Language Inventory [17], British Naming Ability [16], Receptive Expressive Emergent Language-II (REEL-2) [19], McArthur Communicative Developmental Inventories (N-CDI) [21], Brunet-Lezine language sub-scores [22, 25], and Griffith hearing and speech sub-scores [23, 24] were used to measure language development or verbal intelligence. There were no significant differences in language development between ART-born and NC toddlers (p = 0.76). Although there was significant heterogeneity (I 2 = 71%, p = 0.0003), the pooled analysis did not indicate publication bias (p-Egger = 0.118) (Fig. 2B). Since the method of conception might affect  heterogeneity, separate subgroup analyses were performed. Good homogeneities were identified in the analyses on IVF vs NC and ICSI vs NC (p > 0.05); high heterogeneity was only detected in the analysis on ART vs NC group which included studies that did not specify the mode of conception (I 2 = 92%, p = 0.00005). The language development score of toddlers born after IVF was significantly lower than NC toddlers (p = 0.02); meanwhile, ICSI toddlers' score was significantly higher compared to NC toddlers (p = 0.005).

Non-verbal intelligence was reported in 3 studies that used Bayley-III cognitive [15] and Griffith performance sub-scores [23, 24]. Pooled analyses showed that non-verbal intelligence in ART toddlers is significantly lower compared to the NC toddlers  (p = 0.047) (Fig. 2C). Good homogeneity (I 2 = 10%, p = 0.34) and lack of publication bias (p-Egger = 0.703) were both displayed in these studies.

Bayley-II Psychomotoric Development Index (PDI) [16, 19, 20], Bayley-III motor composite score [15], Brunet-Lezine posture and coordination [22, 25], and Griffith locomotor and eye-hand coordination [22, 24] were utilized to assess the total motor skill outcome. Pooled analysis showed no significant difference in total motor score between toddlers born via ART and naturally conceived toddlers (p = 0.27) (Fig. 3A). There were no evidence of data heterogeneity (I2 = 6%, p = 0.38) and publication bias (p-Egger = 0.575). Similarly, subgroup analyses also revealed the insignificant differences of total motor score in toddlers born from ART, IVF, or ICSI compared to NC toddlers, with good homogeneity and no publication bias (p > 0.05).

Motoric outcome of ART-born compared to naturally conceived (NC) toddlers as assessed with A Total Motor Score, B Gross Motor Score, and C Fine Motor Score

The gross motor score was obtained from the Griffith locomotor [22, 25] and the Brunet-Lezine posture subtests [23, 24], in both pooled analysis (p = 0.79) and subgroup analyses based on the method of conception (IVF, p = 0.93; ICSI, p = 0.83) (Fig. 3B). Significant heterogeneity between studies was identified (I 2 = 61%, p = 0.03), especially in the ICSI subgroup (I 2 = 76%, p = 0.006), suggesting that factors other than conception mode might also influenced how children developed their motor skills.

The fine motor score in ART and NC toddlers was similar (p = 0.055) based on Brunet-Lezine's coordination [, ] and Griffith's ey-hand coordination [, ] assessments. The analyses showed low heterogeneity (I 2 = 35%, p = 0.17) and no publication bias (p-Egger = 0.322). The subgroup analyses revealed that toddlers born after IVF had a noticeably lower fine motor score (p = 0.01) than NC toddlers. No significant disparity was noticed in ICSI toddlers compared to NC toddlers (p = 0.28) (Fig. 3C).

According to three studies, NC mothers reported behavioral issues more frequently than ART mothers, as assessed using Achenbach’s Child Behavioral Checklist (CBCL) [20, 21, 26]. Compared to ART children, NC children showed higher total (p = 0.01) and externalizing behavior scores (p = 0.001) (Fig. 4A, C). No significant difference was noted in internalizing behavior score between the two groups (p = 0.09) (Fig. 4B). The data showed good homogeneities (I 2 = 0%, p > 0.05) and no publication biases.

Behavior and social outcomes of ART-born toddlers compared to naturally conceived (NC) toddlers as assessed using A Total Behavior Problems Score, B  Internalizing Behavior Score, C Externalizing Behavior Score, and D Social Score

There was no statistically significant difference observed in the social skills of ART and NC toddlers as assessed using Griffith’s social [23, 24], Brunet-sociability Lezine’s [21, 25], and Vineland Adaptive Behavior socialization [16, 19] (p = 0.22) (Fig. 4D). Likewise, there were no significant differences observed in the subgroup analyses between the IVF (p = 0.57) and ICSI (p = 0.28) toddlers compared to NC toddlers. No heterogeneity (I2 = 0%, p >0.05), and publication bias (p-Egger > 0.05) were found in the analyses.

Weschler Preschool and Primary School Intelligence-Revised version (WPPSI-R) [22, 27, 31, 33, 39, 40, 42], Weschler Abbreviated Scale of Intelligence (WASI) [30], Weschler Intelligence Scale for Children (WISC) [36, 37], Kauffman Assessment Battery for Children (K-ABC) [38, 41], and Revised Amsterdam Child Intelligence Test (RAKIT) [35] were used to measure intelligence. There was no significant difference in the overall full-scale IQ of ART schoolers compared to NC schoolers (p = 0.31). There was significant heterogeneity observed among the studies (I 2 = 50%, p = 0.01), but no evidence of publication bias was detected (p-Egger = 0.438). ICSI subgroup analysis also demonstrated significant heterogeneity (I2 = 59%, p = 0.01). Across the three subgroups, the results consistently indicated that there was no significant differencebetween ART and NC schoolers.

The verbal intelligence quotient was calculated from WPPSI-R [