Screening-detected refractive errors in infants born after assisted reproductive technology: a prospective cohort study.

ART conception may not be associated with an increased risk of screening-detected refractive errors at one year of age and protocol-related differences observed within the ART population need to be interpretated with caution.

This study was based on data from the Jiangsu Birth Cohort Study (JBC), a prospective longitudinal cohort previously described [ 18 ], designed to investigate health outcomes in women and their offspring following ART and spontaneous conception. Couples undergoing ART treatment were recruited from ART clinics, while spontaneously conceived pregnancies were enrolled during early pregnancy (8–14 weeks of gestation) at obstetric clinics. Clinical data, such as infertility diagnosis, ART procedures, and pregnancy health status, were extracted from medical records. Both groups were followed using the same regimen throughout pregnancy. Mothers were interviewed about their offspring’s health status at approximately 42 days and six months postpartum via telephone. At one year of age, all infants were invited for a standardized physical examination conducted by trained health professionals. Between September 2015 and December 2022, the JBC included 17,849 live birth pregnancies (5,012 ART-conceived and 12,837 spontaneous conceptions). A total of 11,536 infants were excluded due to loss to one-year follow-up ( n  = 7130), no eye examinations ( n  = 1799), eye examinations conducted outside the 12 ± 1-month window ( n  = 1204), or instrument non-harmonization (n = 1403). The final analysis included 2739 ART-conceived births (including 1066 twins) and 3574 spontaneously conceived births (including 72 twins) (Supplemental Fig. 1). This study was based on an established prospective cohort, and the sample size was determined by the number of eligible infants with completed 1-year vision screening. We conducted a post hoc detectable-effect assessment (two-sided α = 0.05; power = 80%). Based on an RE prevalence of 11.2% in the spontaneous conception group and group sizes of 3,237 (ART) and 3610 (spontaneous), the minimum detectable relative risk was approximately 1.19 (or 0.81 for a protective association). Written informed consent was obtained from all participants, and the study protocol was approved by the Human Investigation Committees at Nanjing Medical University (NJMUIRB[2014]248). The exposure in this study was ART, defined as IVF with or without ICSI. Detailed information on infertility diagnosis and specific ART procedures was extracted from medical records. The infertility diagnoses included polycystic ovary syndrome (PCOS), ovarian dysfunction, endometriosis, pelvic or tubal factors, uterine factors, thyroid disorders, and sperm disorders. The variables for specific ART procedures included cycle type (fresh or frozen), number of embryos transferred (one versus two or more), fertilization method (IVF or ICSI), stage of embryo transfer (cleavage stage or blastocyst), preimplantation genetic diagnosis (PGD) (yes or no), embryo quality (good, mixed, or poor), and ovulation induction protocols (GnRH-agonist, GnRH-antagonist, or microstimulation). Embryo quality was assessed morphologically by highly trained and experienced embryologists. Embryos were classified as ‘good’ or ‘poor’ based on the grading criteria outlined in the Istanbul consensus and the Association of Clinical Embryologists (ACE) grading scheme [ 19 ]. Both cleavage-stage embryos [ 20 ] and blastocyst quality [ 21 ] were assessed and categorized as ‘good’ or ‘poor’ quality. If a participant received both ‘good’ and ‘poor’ quality embryos within a single cycle, the transfer was classified as ‘mixed’. Detailed definitions for cleavage-stage and blastocyst-stage grading are provided in the Supplementary Method1. All infants were invited to undergo the same standardized refraction screening at one year of age regardless of conception mode. REs were assessed using the Spot Vision Screener (Welch Allyn, Skaneateles Falls, NY) at one year of age. The Spot Vision Screener is a validated tool designed to detect REs in children as young as six months. The examiner selected the appropriate age range on the device’s home screen before positioning the device approximately three feet from the child. The device used twinkling lights and sounds to attract the child’s attention as a fixation target. In a dimly lit room, readings were typically obtained within two seconds for cooperative children. If the initial attempt was unsuccessful, multiple "retry" attempts were made. Upon successful measurement, the device immediately generated a report detailing pupillary diameter, ocular alignment, and binocular refraction [ 22 ]. Refractive errors were classified post hoc using a uniform definition based on the American Association for Pediatric Ophthalmology and Strabismus (AAPOS) 2021 guidelines, applied to the refractive measurements recorded at the one-year examination. According to these guidelines [ 23 ], amblyopia risk factors (ARFs) or other visually significant RE failure levels were defined using meridional refractive power. Hyperopia meridional refractive power was calculated as the spherical equivalent plus half of the cylinder power, while myopia meridional refractive power was calculated as the spherical equivalent minus half of the cylinder power. Anisometropia was defined as the difference in refractive power between the two eyes in the lesser meridian (the spherical equivalent minus half of the cylinder power). This structured approach aimed to ensure standardized assessment and improve the reliability of screening-detected RE detection in our study population. Potential confounders were defined as covariates associated with both the exposure and the outcome but not lying in the potential causal pathway, using a directed acyclic graph (DAG) (Supplemental Fig. 2). We first adjusted for maternal age at delivery (continuous), parity (nulliparous/multiparous), household income ( 200,000 CNY), maternal education level (less than high school/high school), maternal pre-pregnancy body mass index (BMI) (continuous), and area of residence (urban/rural). Considering the potential impact on offspring visual development [ 24 , 25 ], we further adjusted for pregnancy-related complications, including diabetes (pre-pregnancy/gestational) and hypertensive disorders (chronic hypertension/gestational hypertension/preeclampsia), as well as child sex (male/female) and breastfeeding duration (months, continuous). Descriptive analysis was performed to summarize the baseline characteristics of the participants. Continuous variables were presented as means ± standard deviation, while categorical variables were reported as counts and percentages. We employed a three-step analytical strategy. First, associations between ART conception and screening-detected refractive outcomes were evaluated by estimating risk ratios (RRs) and 95% confidence intervals (CIs) using Poisson regression models with a log link fitted within a generalized estimating equations (GEE) framework, with robust variance estimates to account for within-pregnancy clustering, particularly among twins. To reduce potential selection bias arising from loss to follow-up and missing eye screening data, the primary analyses incorporated inverse probability weighting (IPW). The probability of being included in the final analytic sample was estimated using baseline maternal and pregnancy-related characteristics measured before follow-up, and the inverse of this probability was used as the analytic weight. Second, subgroup analyses were conducted to compare subgroups of ART-conceived infants with the spontaneous conception group. Third, we conducted within-group analyses only among ART-conceived infants to explore the differences in the risk of screening-detected REs resulting from specific infertility diagnoses and ART procedures. For the comparison of ovarian stimulation protocols, analyses were limited to the GnRH-antagonist and GnRH-agonist groups to improve clinical comparability. Because protocol selection in routine clinical practice is based on patient characteristics rather than random allocation, 1:1 propensity score matching was additionally used to improve balance in baseline maternal and infertility-related characteristics before estimating the association between protocol type and outcome. Then, we conducted a multivariable Poisson regression that included variables associated with screening-detected REs in subgroup analyses and covariates. We also performed stratified analyses by child sex to examine potential effect modification. In addition, sensitivity analyses were conducted by excluding pregnancy complications (diabetes and hypertensive disorders during pregnancy), preterm birth (PTB), and low birth weight (LBW) to test the robustness of our findings. All analyses were performed using R software (version 4.1.3; R Foundation for Statistical Computing, http://www.cran.r-project.org/ ).

A total of 3237 ART infants and 3610 spontaneously conceived infants were included in the analysis. Twin pregnancies were more prevalent among ART pregnancies (21.0%; n  = 569) compared to spontaneous pregnancies (1.0%; n  = 38) (Table  1 ). Women who received ART were generally older and more likely to be nulliparous, to reside in rural areas, and to have lower educational attainment and annual household income. Furthermore, ART pregnancies showed higher rates of complications, and infants born following ART were more likely to be preterm and to have a higher incidence of low birth weight. The distribution of sex was comparable between ART-conceived and spontaneously conceived infants (Table  1 ). Additionally, we compared baseline characteristics between participants included in the analytic sample and those lost to follow-up. While some sociodemographic characteristics differed, key clinical characteristics were generally similar between the two groups (Supplemental Table S1). The overall prevalence of screening-detected REs in the study population was 11.3%, with anisometropia being the most common type (9.1%), followed by hyperopia (2.8%), astigmatism (1.5%), and myopia (0.3%) (Supplemental Table S2). Due to the limited number of spontaneously conceived twins, statistical comparisons between ART and spontaneously conceived twins were not feasible. Nonetheless, a higher prevalence of screening-detected REs was observed among ART-conceived twins than among singletons, irrespective of conception mode. Table 1 Characteristics of participants according to the mode of conception a Characteristics no. (%) ART-conception group Spontaneous-conception group P value Maternal characteristics  No 2739 3574 Maternal age, year  Mean (SD) 32.14 (3.7) 30.18 (3.7)  < 0.001 Parity  < 0.001  Nulliparous 2477 (91.2) 2556 (71.6)  Multiparous 240 (8.8) 1016 (28.4) Maternal pre-pregnancy BMI, kg/m 2 ,  < 0.001 < 18.5 221 (8.1) 447 (12.5)  18.5–23.9 1772 (64.7) 2499 (69.9)  24–27.9 563 (20.6) 502 (14.0) ≥ 28 183 (6.7) 124 (3.5) Household income, CNY  < 0.001  200,000 851 (31.1) 1411 (39.5) Maternal education, year,   12 2055 (75.0) 3172 (88.8) Area of residence 0.346  Rural 209 (7.6) 297 (8.3)  Urban/suburban 2530 (92.4) 3276 (91.7) Complications of pregnancy  Diabetes b 860 (31.4) 850 (23.8)  < 0.001  Hypertension c 365 (13.3) 140 (3.9)  < 0.001 Plurality  < 0.001  Singleton 2170 (79.2) 3536 (98.9)  Twins 569 (20.8) 38 (1.1) Infant characteristics No. 3237 3610 Gestational age at delivery week, mean (SD), week 38.14 (2.0) 39.45 (1.3)  < 0.001 Preterm birth (< 37 week) 737 (22.8) 138 (3.8)  < 0.001 Birth weight, mean (SD), gram 3085.64 (604.4) 3343.68 (446.9)  < 0.001 Low birth weight (< 2500 g) 507 (15.7) 111 (3.1)  < 0.001 Child sex 0.015  Boy 1767 (54.6) 1863 (51.6)  Girl 1470 (45.4) 1747 (48.4) Breastfeeding duration, month  < 0.001 < 6 1792 (49.6) 2029 (62.7) ≥ 6 1698 (47.0) 1154 (35.7) ART assisted reproductive technology, SD standard deviation, BMI body mass index, CNY Chinese yuan a Missing data were less than 5% for all characteristics b Includes chronic and gestational diabetes c Includes chronic and pregnancy-induced hypertension Characteristics of participants according to the mode of conception a ART assisted reproductive technology, SD standard deviation, BMI body mass index, CNY Chinese yuan a Missing data were less than 5% for all characteristics b Includes chronic and gestational diabetes c Includes chronic and pregnancy-induced hypertension After adjusting for conventional covariates, including maternal age, parity, household income, maternal education, area of residence, and maternal pre-pregnancy BMI, no overall association was observed between ART and screening-detected REs in offspring infants (aRR, 0.93; 95% CI 0.80–1.09). The estimates remained similar after further adjustment for additional covariates, including diabetes during pregnancy, hypertensive disorders during pregnancy, child sex, and breastfeeding duration (Table  2 ). Table 2 Associations of ART with overall screening-detected RE and specific types of screening-detected RE Outcomes ART-conceived Spontaneous-conceived Crude RR (95%CI) P value aRR (95% CI) a P value aRR (95% CI) b P value All ART-conceived births vs. All Spontaneous-conceived births  No 3237 3610  Overall RE 369 (11.4) 405 (11.2) 0.95(0.82, 1.09) 0.454 0.93(0.80, 1.09) 0.397 0.91(0.77, 1.07) 0.250  Myopia 10 (0.3) 13 (0.4) 0.80(0.34, 1.89) 0.610 0.83(0.33, 2.08) 0.697 0.72(0.30, 1.71) 0.453  Hyperopia 103 (3.2) 91 (2.5) 1.23(0.91, 1.65) 0.173 1.12(0.79, 1.59) 0.513 1.15(0.82, 1.61) 0.423  Astigmatism 49 (1.5) 53 (1.5) 1.04(0.69, 1.54) 0.865 0.96(0.63, 1.45) 0.836 0.92(0.60, 1.42) 0.720  Anisometropia 290 (9.0) 333 (9.2) 0.89(0.76, 1.04) 0.152 0.88(0.74, 1.06) 0.180 0.85(0.71, 1.02) 0.087 ART-conceived singletons vs. Spontaneous-conceived singletons  No 2171 3538  Overall RE 240 (11.1) 400 (11.3) 0.93(0.78, 1.11) 0.412 0.87(0.72,1.05) 0.142 0.85(0.70, 1.03) 0.100  Myopia 6 (0.3) 13 (0.4) 0.62(0.19, 1.98) 0.419 0.69(0.20,2.40) 0.561 0.65(0.19, 2.26) 0.497  Hyperopia 63 (2.9) 90 (2.5) 1.16(0.82, 1.63) 0.415 1.03(0.71,1.51) 0.859 1.05(0.72, 1.54) 0.802  Astigmatism 34 (1.6) 53 (1.5) 1.03(0.63, 1.69) 0.893 0.95(0.56,1.62) 0.847 0.95(0.55, 1.64) 0.862  Anisometropia 187 (8.6) 328 (9.3) 0.87(0.72, 1.06) 0.177 0.82(0.67,1.01) 0.067 0.79(0.64, 1.01) 0.063 ART assisted reproductive technology, RE refractive error, CI confidence interval, RR risk ratio a Analyses were adjusted for maternal age, parity, household income, maternal education, area of residence, and maternal pre-pregnancy BMI. These covariates are recommended by directed acyclic graph (DAG) b Analyses were additionally adjusted for diabetes during pregnancy, hypertensive disorders during pregnancy, child sex, and breastfeeding duration Associations of ART with overall screening-detected RE and specific types of screening-detected RE ART assisted reproductive technology, RE refractive error, CI confidence interval, RR risk ratio a Analyses were adjusted for maternal age, parity, household income, maternal education, area of residence, and maternal pre-pregnancy BMI. These covariates are recommended by directed acyclic graph (DAG) b Analyses were additionally adjusted for diabetes during pregnancy, hypertensive disorders during pregnancy, child sex, and breastfeeding duration Similar patterns were observed for specific types of screening-detected REs. The aRRs for myopia and hyperopia were 0.72 (95% CI 0.30–1.71) and 1.15 (95% CI 0.82–1.61), respectively. Similarly, no associations were found for astigmatism (aRR, 0.92; 95% CI 0.60–1.42) or anisometropia (aRR, 0.85; 95% CI 0.71–1.02). No differences in the risk of screening-detected REs were observed between ART-conceived and spontaneously conceived singletons (Table  2 ). Stratified analyses by household income and area of residence yielded results consistent with the primary findings (Supplemental Table S3). Stratified analyses by child sex showed no heterogeneity between boys and girls in the association between ART and screening-detected REs (Supplemental Table S4). Sensitivity analyses excluding pregnancies complicated by preterm birth, diabetes, hypertensive disorders, and low birth weight yielded similar results to those in the primary analyses (Supplemental Table S5). To further explore the potential influence of plurality, we examined the association between twinning and screening-detected REs among ART-conceived offspring. Among ART-conceived infants (infant-level analysis), we observed comparable rates of overall screening-detected REs between singleton infants and twin infants. Likewise, no differences were observed across specific types of screening-detected REs (Supplemental Table S6). To further assess whether specific infertility diagnoses or ART procedures were associated with the risk of screening-detected REs, we first compared subgroups of the ART population to those spontaneously conceived. None of the infertility-related diagnoses, such as PCOS, endometriosis, or tubal factors, were associated with an increased risk of screening-detected REs compared to those spontaneously conceived. Similarly, no associations were observed across different ART procedural subtypes, including fertilization method, cycle type, embryo transfer stage, and ovulation induction protocols, compared to those spontaneously conceived (Table  3 ). Table 3 Associations of infertility diagnosis and ART treatment procedures with the risk of overall screening-detected RE compared to spontaneously conceived infants Characteristics Number Rate aRR (95% CI) a P value Spontaneously conceived 3610 405 (11.2) 1.00 —— Infertility diagnosis  Duration of infertility, month < 36 1492 173 (11.6) 0.69 (0.33, 1.46) 0.330 ≥ 36 1724 195 (11.3) 0.71 (0.34, 1.50) 0.373  PCOS No 2970 349 (11.8) 0.72 (0.34, 1.50) 0.376 Yes 267 20 (7.5) 0.51 (0.22, 1.20) 0.123  Ovarian dysfunction No 2735 322 (11.8) 0.73 (0.35, 1.52) 0.398 Yes 502 47 (9.4) 0.51 (0.23, 1.15) 0.104  Endometriosis No 2964 337 (11.4) 0.69 (0.33, 1.44) 0.319 Yes 273 32 (11.7) 0.81 (0.36, 1.82) 0.609  Pelvic or tubal factor No 1058 122 (11.5) 0.70 (0.33, 1.49) 0.357 Yes 2179 247 (11.3) 0.70 (0.33, 1.47) 0.347  Uterine factor No 2312 264 (11.4) 0.68 (0.32, 1.43) 0.309 Yes 925 105 (11.4) 0.74 (0.35, 1.57) 0.434  Thyroid disorder No 3054 342 (11.2) 0.69 (0.33, 1.44) 0.322 Yes 183 27 (14.8) 0.94 (0.41, 2.19) 0.891  Sperm disorder No 1906 227 (11.9) 0.74 (0.35, 1.55) 0.418 Yes 1331 142 (10.7) 0.64 (0.30, 1.37) 0.252 ART procedure  Cycle type Fresh 729 95 (13.0) 0.88 (0.41, 1.88) 0.741 Frozen 2486 272 (10.9) 0.65 (0.31, 1.36) 0.250  Number of embryo transferred One 1267 145 (11.4) 0.73 (0.35, 1.55) 0.418 Two or more 1954 223 (11.4) 0.67 (0.32, 1.42) 0.297  Fertilization method IVF 2310 260 (11.3) 0.69 (0.33, 1.44) 0.320 ICSI 907 107 (11.8) 0.74 (0.34, 1.57) 0.429  Stage of embryo transferred Cleavage 1492 180 (12.1) 0.75 (0.35, 1.58) 0.447 Blastocyst 1730 188 (10.9) 0.66 (0.32, 1.40) 0.284  PGD No 3103 355 (11.4) 0.70 (0.34, 1.48) 0.352 Yes 137 14 (10.2) 0.63 (0.25, 1.57) 0.318  Embryo quality Good 2130 250 (11.7) 0.74 (0.35, 1.56) 0.429 Mix 617 68 (11.0) 0.66 (0.30, 1.43) 0.289 Poor 468 49 (10.5) 0.53 (0.24, 1.19) 0.122  Ovulation induction protocols GnRH-a 1300 191 (14.7) 0.95 (0.45, 2.00) 0.886 Microstimulation 506 52 (10.3) 0.64 (0.29, 1.40) 0.265 GnRH-ant 1355 111 (8.2) 0.51 (0.24, 1.08) 0.077 Analyses were adjusted for maternal age, parity, household income, maternal education, area of residence, maternal pre-pregnancy BMI, diabetes during pregnancy, hypertensive disorders during pregnancy, child sex, and breastfeeding duration ART assisted reproductive technology, BMI body mass index, RE refractive error, ICSI intracytoplasmic sperm injection, IVF in vitro fertilization, GnRH-a Gonadotropin-releasing hormone agonist, GnRHant GnRH antagonist, PGD preimplantation genetic diagnosis, PCOS polycystic ovary syndrome, CI confidence interval, RR risk ratio Associations of infertility diagnosis and ART treatment procedures with the risk of overall screening-detected RE compared to spontaneously conceived infants Analyses were adjusted for maternal age, parity, household income, maternal education, area of residence, maternal pre-pregnancy BMI, diabetes during pregnancy, hypertensive disorders during pregnancy, child sex, and breastfeeding duration ART assisted reproductive technology, BMI body mass index, RE refractive error, ICSI intracytoplasmic sperm injection, IVF in vitro fertilization, GnRH-a Gonadotropin-releasing hormone agonist, GnRHant GnRH antagonist, PGD preimplantation genetic diagnosis, PCOS polycystic ovary syndrome, CI confidence interval, RR risk ratio To examine whether differences in infertility-related diagnoses or specific treatment procedures may contribute to variations in the risk of screening-detected REs, we conducted additional analyses examining the role of parental infertility diagnoses and specific ART procedures only among ART-conceived infants (Fig.  1 ). Baseline maternal and infant characteristics, infertility diagnoses (including PCOS), and ART procedure characteristics across ovarian stimulation protocols are summarized in Supplementary Table S7. The duration of infertility was not associated with an increased risk of screening-detected REs (aRR, 1.00; 95% CI 0.81–1.24). Compared with infants born to women without PCOS, offspring born to women diagnosed with PCOS exhibited a lower risk of screening-detected REs (aRR, 0.62; 95% CI 0.39–0.98) (Fig.  1 ). Other infertility diagnoses, including ovarian dysfunction, endometriosis, pelvic or tubal factors, uterine factors, thyroid disorders, and sperm disorders, were not associated with screening-detected REs in ART-conceived offspring (Fig.  1 ). Fig. 1 Associations of infertility diagnosis and ART treatment procedures with risk of overall screening-detected RE among ART-conceived infants. ART assisted reproductive technology, BMI body mass index, RE refractive error, ICSI intracytoplasmic sperm injection, IVF , in vitro fertilization, GnRH-a gonadotropin-releasing hormone agonist, GnRHant GnRH antagonist, PGD , preimplantation genetic diagnosis, PCOS polycystic ovary syndrome, CI confidence interval, RR risk ratio. Analyses were adjusted for maternal age, parity, household income, maternal education, area of residence, maternal pre-pregnancy BMI, diabetes during pregnancy, hypertensive disorders during pregnancy, child sex, and breastfeeding duration Associations of infertility diagnosis and ART treatment procedures with risk of overall screening-detected RE among ART-conceived infants. ART assisted reproductive technology, BMI body mass index, RE refractive error, ICSI intracytoplasmic sperm injection, IVF , in vitro fertilization, GnRH-a gonadotropin-releasing hormone agonist, GnRHant GnRH antagonist, PGD , preimplantation genetic diagnosis, PCOS polycystic ovary syndrome, CI confidence interval, RR risk ratio. Analyses were adjusted for maternal age, parity, household income, maternal education, area of residence, maternal pre-pregnancy BMI, diabetes during pregnancy, hypertensive disorders during pregnancy, child sex, and breastfeeding duration No differences in the risk of screening-detected REs were observed across specific ART procedures, including cycle type, number of embryos transferred, fertilization method, PGD, or transfer stage (Fig.  1 ). However, among ovulation induction protocols, the GnRH-antagonist protocol appeared to be associated with a reduced risk of screening-detected REs compared to the GnRH-agonist protocols (aRR, 0.56; 95% CI 0.44–0.71). The microstimulation protocol was also associated with a modest reduction in risk of screening-detected REs (aRR, 0.73; 95% CI 0.54–1.00). In multivariable models adjusting for maternal characteristics and other potential confounders, we observed a lower risk of screening-detected REs among ART-conceived infants conceived using the GnRH-antagonist protocol (aRR, 0.56; 95% CI 0.44–0.72). The previously observed inverse association between PCOS and screening-detected REs risk was attenuated and no longer statistically significant after full adjustment (Table  4 ). Among ART-conceived singletons, we observed a similar association for the GnRH-antagonist protocol (aRR, 0.58; 95% CI 0.43–0.79) (Supplemental Table S8). Table 4 Multivariable analysis of the associations of ART use and the covariates with overall screening-detected RE among all ART-conceived births Characteristics Category RR (95%CI) P value Maternal age – 0.98 (0.95, 1.01) 0.183 Maternal pre-pregnancy BMI – 1.02 (0.98, 1.05) 0.363 Parity Multiparous Ref Nulliparous 0.97 (0.64, 1.48) 0.896 Household income 50,000–100,000 Ref   200,000 0.85 (0.63, 1.14) 0.271 Maternal education  ≥ 12 Ref  < 12 1.03 (0.80, 1.35) 0.801 Area of residence Rurual Ref Urban/suburban 1.10 (0.73, 1.65) 0.651 Child sex Boy Ref Girl 1.10 (0.89, 1.36) 0.382 Hypertensive disorders during pregnancy No Ref Yes 1.29 (0.98, 1.71) 0.070 Diabetes during pregnancy No Ref Yes 1.05 (0.84, 1.32) 0.667 Breastfeeding duration  ≥ 6 Ref  < 6 1.13 (0.90, 1.42) 0.277 Ovulation induction protocols GnRH-a Ref Microstimulation 0.75 (0.55, 1.02) 0.069 GnRHant 0.56 (0.44, 0.72)  < 0.001 PCOS No Ref Yes 0.67 (0.42, 1.06) 0.087 ART assisted reproductive technology, BMI body mass index, RE refractive error, GnRH-a Gonadotropin-releasing hormone agonist, GnRHant GnRH antagonist, PCOS polycystic ovary syndrome, CI confidence interval, RR risk ratio Multivariable analysis of the associations of ART use and the covariates with overall screening-detected RE among all ART-conceived births ART assisted reproductive technology, BMI body mass index, RE refractive error, GnRH-a Gonadotropin-releasing hormone agonist, GnRHant GnRH antagonist, PCOS polycystic ovary syndrome, CI confidence interval, RR risk ratio Because some heterogeneity was observed according to ovarian stimulation protocol (Supplementary Table S7), we further restricted the comparison to the GnRH-antagonist and GnRH-agonist groups to improve clinical comparability. Baseline maternal and infertility-related characteristics before and after 1:1 propensity score matching are presented in Supplementary Table S9. Before matching, differences were observed between the two protocol groups in several characteristics, including maternal age, pre-pregnancy BMI, household income, area of residence, PCOS, ovarian dysfunction, endometriosis, and sperm abnormalities. After matching, covariate balance was substantially improved, with most absolute standardized mean differences below 0.10. In the matched cohort of 2216 ART pregnancies (1108 in each group), the GnRH-antagonist protocol remained associated with a lower risk of screening-detected REs than the GnRH-agonist protocol (RR, 0.56; 95% CI 0.44–0.72). The estimate was materially unchanged after additional covariate adjustment in the matched sample (RR, 0.57; 95% CI 0.45–0.73) (Supplementary Table S10).

In conclusion, our prospective cohort study found no strong evidence that ART conception is associated with an increased risk of screening-detected REs at one year of age. Neither infertility-related diagnoses nor ART treatment procedures were independently associated with the risk of screening-detected REs when compared to spontaneous conception. However, among ART-conceived infants, those conceived using the GnRH-antagonist protocol might have a lower risk of screening-detected REs compared with those born with the GnRH-agonist protocol. Because protocol assignment is clinically indicated and longitudinal hormone measurements were unavailable, protocol-specific findings should be interpreted as associative and hypothesis generating, and further studies are needed to rule out potential residual confounding. Our findings underscore the importance of considering treatment heterogeneity and selection patterns when interpreting the associations between ART procedures and the visual outcomes of offspring.

In this prospective cohort study, we found no association between ART conception and the risks of overall or specific screening-detected REs in infants at one year of age, neither in the full cohort nor among singletons. The same findings were observed when comparing ART subgroups defined by parental infertility diagnoses and treatment procedures to the spontaneously conceived group. These findings remained robust after adjustment for potential confounders and in the sensitivity analyses. In the analyses restricted to the ART-conceived population, the use of the GnRH-antagonist protocol was associated with a lower risk of screening-detected REs compared to the GnRH-agonist protocol. Our study found no difference in the risk of screening-detected REs between ART-conceived and spontaneously conceived infants at one year of age, which is consistent with the findings in a Swedish study [ 26 ]. However, the previous study focused exclusively on ICSI and REs at a later developmental stage of age five, while our study included all ART types and evaluated REs at one year of age, a period critical to early visual maturation [ 8 ]. Similarly, several earlier studies found no differences in ocular biometry or visual function between ART and non-ART offspring [ 10 , 14 , 27 ]. Yet, those studies focused on outcomes like axial length or visual acuity rather than clinically defined REs. In contrast, a few population-based registry studies have reported increased risks of clinically diagnosed visual impairment in ART-conceived children [ 11 , 15 ] and another Swedish study identified altered retinal vessel morphology in ICSI-conceived boys [ 17 ]. These discrepancies may be due to differences in phenotypes assessed, follow-up timing, or heterogeneity within ART groups. To address this heterogeneity, we further examined ART subgroups. In our stratified comparison using the spontaneous conception group as the reference, we found no associations between specific infertility diagnoses or ART procedures and the risk of screening-detected REs. However, analyses restricted to the ART population showed that the GnRH-antagonist protocol was associated with a reduced risk of screening-detected REs. Prior evidence suggests that the ovarian stimulation regimen may be relevant to offspring RE risk. For example, Kuiper et al. reported a lower risk of REs in modified natural cycle IVF or subfertile spontaneous conception compared with controlled ovarian hyperstimulation IVF (COH-IVF) cycles [ 16 ]. In our study, overall screening-detected REs’ risk was comparable between ART-conceived and spontaneously conceived infants at one year of age, while protocol-related heterogeneity within ART was observed. This finding should be interpreted cautiously. According to Chinese clinical guidance for controlled ovarian stimulation, the choice between GnRH-agonist and GnRH-antagonist protocols is guided primarily by ovarian reserve/response status and related clinical characteristics, including age, ovarian reserve markers, PCOS, previous ovarian response, endocrine profile, and anticipated risk of ovarian hyperstimulation syndrome [ 28 ]. Thus, the differences between protocol groups may partly reflect underlying patient characteristics and treatment indications rather than the protocol itself. Although we used propensity score matching to improve comparability between groups, residual confounding by indication cannot be excluded. Several biological mechanisms could plausibly underlie the observed association. Embryonic eye development occurs primarily between weeks 3 and 10 of gestation [ 29 ], a period when ocular tissues are susceptible to hormonal signals [ 30 ]. GnRH-agonist protocols are known to cause greater fluctuations in estrogen levels, whereas GnRH-antagonist protocols offer more stable hormonal profiles, which may help preserve normal developmental processes [ 31 – 33 ]. However, because longitudinal hormonal measurements were not available in our cohort (only baseline E2 was recorded), we could not directly evaluate estrogen fluctuation metrics across protocols; therefore, this hormonal explanation should be interpreted as hypothesis-generating. In addition, vascular pathways may be involved. Vascular endothelial growth factor (VEGF), essential for retinal vascularization [ 34 , 35 ], has been found at higher levels in individuals undergoing GnRH-antagonist protocols compared to those receiving GnRH-agonists [ 36 , 37 ]. Placental growth factor (PlGF), a member of the VEGF family involved in angiogenesis, is also implicated in fetal microvascular development, especially retinal vasculature formation [ 38 ]. Reduced PlGF levels during pregnancy have been associated with narrower retinal arterioles in childhood [ 39 ], offering a potential mechanistic explanation for the observed association in our study. It remains unclear whether the reduced risk of REs associated with ovulation induction protocols reflects an actual biological effect or underlying differences in patient characteristics and clinical decision-making. Further mechanistic and longitudinal studies are warranted to determine whether specific ART procedures exert lasting effects on visual development and to clarify the clinical significance of our findings. Our study has several strengths. The prospective cohort design, along with detailed data on parental and perinatal characteristics, enabled robust adjustment for potential confounders. Furthermore, this is the first study to investigate how both infertility diagnoses and ART procedures may relate to screening-detected REs at one year of age. In addition, the multicenter design, combined with rigorous quality control measures such as standardized embryo grading and comprehensive ophthalmic screening, further enhanced the internal validity of our findings. Some limitations should be acknowledged. First, the SPOT Vision Screener is a screening instrument rather than a diagnostic test. Nevertheless, utilizing the SPOT Vision Screener in a longitudinal study may provide useful early indications of REs in non-verbal infants and support early identification of visual development concerns [ 40 ]. Second, the infants in our study were relatively young, and early visual development measures may have limited predictive value for long-term outcomes. Continued follow-up of this cohort will be necessary to better assess long-term visual development. Third, longitudinal hormonal measurements during ovarian stimulation were not collected; only baseline E2 was available. Thus, we were unable to directly evaluate estrogen fluctuation metrics across protocols. Fourth, the observational nature of this study precludes causal inference, underscoring the need for further research to better understand the impact of ART procedures on early visual development. Although we incorporated inverse probability weighting to reduce bias related to differential inclusion in the analytic sample, residual selection bias cannot be excluded because weighting can only account for measured predictors of follow-up. Therefore, our findings should be interpreted cautiously as observational associations, and further studies are warranted.

Infants conceived through assisted reproductive technology (ART) account for approximately 2% of live births [ 1 , 2 ]. An increasing body of research has shown potential health risks in ART-conceived children, including higher rates of adverse perinatal outcomes [ 3 ] and various developmental problems in multiple organ systems [ 4 – 6 ]. The eye, a delicate organ that develops in parallel with the nervous system, may also be vulnerable to prenatal influences [ 7 ]. Refractive development, as an early indicator of visual development, is a highly dynamic process, and the first 12 months of life are considered a critical period for establishing normal visual function [ 8 , 9 ]. However, evidence regarding the association between ART and early visual development remains limited and inconsistent [ 10 – 12 ]. While several studies have reported no adverse effects on visual development in ART-conceived children [ 10 ], others have suggested potential risks [ 11 , 12 ]. Moreover, whether the observed association between ART and visual development reflects the effects of the treatment itself or the infertility characteristics [ 13 ] is unclear. Some studies observed no differences in eye abnormalities between in vitro fertilization (IVF)- and spontaneously conceived neonates [ 14 ], while others reported increased risks of poor vision and refractive errors (REs) [ 15 , 16 ], or vascular alterations, specifically in intracytoplasmic sperm injection (ICSI)-conceived children [ 17 ]. In addition to fertilization methods, it has been suggested that infertility-related diagnosis [ 15 ] or treatment protocols [ 16 ] might have effects on the visual development of offspring which require further investigation. We conducted a prospective cohort study to evaluate the association between ART and the risk of REs at one year of age. Furthermore, as infertility-related diagnoses and treatment procedures vary across ART therapies, we explored whether these factors could affect the risk of REs, both in comparison to spontaneous conception and within the ART population.

Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 280 KB) Supplementary file1 (DOCX 280 KB)