Investigating the neurodevelopmental consequences of anesthesia and surgical interventions on cognitive and behavioral A comprehensive Mendelian randomization study

Authors: Wen Wu, Yuchao Fan

Abstract

Anesthesia is indispensable to modern surgical practice and clinical care; however, its potential impact on neurodevelopment and behavioral outcomes has raised growing concerns. Discrepancies among previous studies may partly arise from difficulties in disentangling the effects of anesthesia from those of surgical interventions. Mendelian randomization (MR) offers an approach to explore these associations while minimizing confounding and reverse causation. We analyzed genetic instruments related to exposure to different types of anesthesia and the number of surgical operations, in conjunction with 39 genome-wide association study datasets covering a broad range of neurodevelopmental, cognitive, and behavioral outcomes. Both univariable and multivariable MR analyses were conducted using inverse-variance weighted, MR-Egger, and weighted median and mode methods. Sensitivity analyses were performed to assess robustness, and the Steiger test was applied to evaluate potential reverse causality. No significant causal associations were observed between exposure to either general or regional anesthesia and neurodevelopmental outcomes. In contrast, a higher number of surgical operations was causally associated with fewer years of schooling (odds ratio [OR] 0.9088, 95% confidence interval [CI] 0.8569–0.9639, P = .0015), a younger age at completion of full-time education (OR 0.8911, 95% CI 0.8346–0.9516, P = 5.77 × 10^−4^), an increased risk of neuroticism (OR 1.5962, 95% CI 1.2727–2.0018, P = 5.18 × 10^−5^), depression (OR 1.0951, 95% CI 1.0544–1.1374, P = 2.58 × 10^−6^), and emotionally unstable personality disorder (OR 2.4124, 95% CI 1.2408–4.6906, P = .0094). Multivariable MR analyses demonstrated that adjusting for anesthesia exposure did not attenuate the associations between surgical procedures and educational outcomes, neuroticism, or depression. Notably, adjustment for inhalational or intravenous anesthesia reduced the effect of surgery on emotionally unstable personality disorder, whereas adjustment for local or epidural anesthesia did not. These findings suggest that anesthesia exposure itself may not exert a significant causal effect on neurodevelopmental outcomes. However, surgical procedures appear to be associated with long-term consequences related to academic attainment, emotional well-being, and personality traits.

1. Introduction

Anesthesia is an essential component of contemporary medical care, particularly in pediatric practice, where general anesthesia is routinely used for surgical procedures, distressing medical interventions, and diagnostic imaging examinations.^[1]^ In the United States, approximately 1.5 to 2 million children under the age of 3 undergo anesthesia each year, with a substantial proportion of these cases – around 43,000 – occurring at Texas Children’s Hospital alone.^[2]^ Among these, approximately 13,000 involve children younger than 3 years, and nearly 1300 cases require anesthesia durations exceeding 3 hours.^[2]^ Similarly, in the United Kingdom, an estimated 200,000 children under 6 years of age receive general anesthesia annually.^[3]^

Despite its indispensable role, growing concerns have emerged within both academic and public spheres regarding the potential impact of anesthetic exposure on neurodevelopment. These concerns center on the possibility that anesthetic agents may contribute to adverse intellectual, psychological, or behavioral outcomes, particularly in children exposed to multiple or prolonged episodes of general anesthesia.^[4]^ Consequently, cautionary statements issued by regulatory authorities may exert far-reaching effects on healthcare professionals, expectant mothers, and parents of young children, potentially influencing clinical decision-making and broader public health perceptions.^[2]^

Evidence from animal studies, including investigations in nonhuman primates, suggests that commonly used intravenous and inhalational anesthetics can disrupt brain development in immature animals, especially following prolonged or repeated exposures.^[5]^ However, translating these findings to human clinical settings requires careful interpretation, given substantial interspecies differences in neurodevelopment. Moreover, experimental conditions in animal studies often differ markedly from real-world clinical scenarios, including the use of supraphysiological anesthetic doses, extended exposure durations, and limited simulation of underlying diseases or surgical stressors.^[6]^ Although numerous high-quality clinical studies have explored this issue, their findings remain inconsistent.^[7]^ Several studies indicate that single or multiple anesthesia exposures are associated with relatively low risk and do not result in long-term impairments in general intelligence or global cognitive developmentt.^[8–10]^ In contrast, other investigations suggest that repeated anesthesia exposure may adversely affect specific behavioral patterns or psychological development.^[11,12]^ Notably, a recent study of infants undergoing corrective or palliative surgery for major congenital heart disease reported no significant association between cumulative volatile anesthetic exposure and neurodevelopmental scores at 18 months of follow-up.^[13]^

The heterogeneity of clinical findings may largely stem from methodological limitations. Although randomized controlled trials are considered the gold standard for causal inference, they face substantial challenges in this context. In particular, it is difficult to disentangle the effects of anesthesia from those of surgery itself, as underlying disease severity and surgical intervention may independently influence neurodevelopmental outcomes.^[1]^ These challenges raise both ethical and practical barriers to trial design. Furthermore, randomized controlled trials capable of capturing long-term neurodevelopmental outcomes are difficult to conduct due to financial constraints and participant attrition.^[7]^ As a result, most existing evidence is derived from retrospective observational studies, which require extensive adjustment for confounding variables. However, excessive adjustment may inadvertently reduce the precision of estimated anesthetic effects on neurodevelopment.^[14]^

In recent years, genetic approaches to causal inference have gained increasing attention in epidemiological research.^[15]^ Mendelian randomization (MR) leverages genetic variants as instrumental variables (IVs) to assess causal relationships between exposures and outcomes. Because genetic variants are randomly allocated at conception, MR analyses are inherently less susceptible to confounding and reverse causation, thereby minimizing both measured and unmeasured confounders.^[16–18]^ Single nucleotide polymorphisms (SNPs), which serve as proxies for exposures in MR studies, are distributed according to Mendelian inheritance principles, creating conditions analogous to those of randomized controlled trials.^[16,19]^ Importantly, MR avoids many ethical and logistical challenges associated with interventional studies, making it a particularly suitable framework for investigating the potential neurodevelopmental effects of anesthesiat.^[1,17]^

In this study, we 1st examined the causal associations between several anesthesia-related exposures – including inhalational anesthesia, intravenous anesthesia, local anesthetic nerve blocks, lumbar epidural anesthesia, and the number of surgical operations – and 39 neurobehavioral outcomes using univariable Mendelian randomization (UVMR). These outcomes were categorized into 5 intellectual and cognitive performance, neurodevelopmental disorders, personality traits and disorders, clinically diagnosed mental health outcomes, and brain structural volumes related to cognition and emotional regulation. Recognizing that surgical intervention represents an unavoidable and potentially influential factor in the context of anesthesia exposure, we further conducted multivariable Mendelian randomization (MVMR) analyses. These models jointly incorporated specific anesthetic exposures and surgical burden to assess the independent effects of anesthesia on neurobehavioral outcomes. When a particular anesthetic exposure showed a significant association with an outcome, the number of surgical operations was included as an adjusting variable in the MVMR analysis. Conversely, when the number of operations demonstrated a significant association with an outcome, individual anesthetic or anesthesia modalities were incorporated into the corresponding MVMR models.

2. Methods

2.1. Study design

In this study, we applied a UVMR framework to investigate the potential causal associations between exposure to different anesthetics, anesthesia techniques, or the number of surgical operations and a range of neuropsychological and behavioral outcomes, with each outcome analyzed independently.

The evaluated neuropsychological and behavioral outcomes were categorized into 5 primary intelligence and cognitive performance, neurodevelopmental disorders, personality disorders, clinically diagnosed mental health outcomes, and brain structural volumes related to cognitive and emotional functions.

To further disentangle the independent effects of anesthesia exposure from surgical burden, we implemented MVMR, a recently developed extension of the MR framework. MVMR leverages genetic variants associated with multiple exposures to estimate their individual causal effects on a single outcome.^[20]^ In this analysis, overlapping SNPs shared across exposures and potential confounding variables were selected as IVs.

Through mutual adjustment within the MVMR framework, we aimed to identify exposures that demonstrated causal associations with neuropsychological and behavioral outcomes independent of other related factors. Specifically, when a particular anesthetic or anesthesia method showed a significant association with an outcome in UVMR analyses, the number of surgical operations was included as an adjusting variable in the corresponding MVMR model. Conversely, when the number of operations demonstrated a significant association with an outcome, individual anesthetic or anesthesia modalities, as well as all anesthetic exposures combined, were incorporated as covariates in the MVMR analyses.

Genetic variants were used as IVs to estimate the causal effects of exposures on outcomes. Because SNPs are randomly allocated during meiosis, they are less susceptible to common biases encountered in observational studies, including confounding, reverse causation, and measurement error. When specific assumptions are satisfied, SNPs can therefore be effectively employed to infer causal relationships.

The core assumptions required for genetic variants to serve as valid IVs in this study were as the genetic variant is robustly associated with the exposure of interest; the genetic variant is not associated with potential confounders of the exposure–outcome relationship; and the genetic variant influences the outcome solely through its effect on the exposure and not through alternative pathways. The overall study design and MR assumptions are illustrated in Figure 1.

2.2. Data sources of exposures and outcomes

Genome-wide association study (GWAS) summary statistics used in this study were obtained from the IEU Open GWAS Project (https://gwas.mrcieu.ac.uk/). These datasets integrate data from multiple well-established sources, including the UK Biobank, FinnGen, Neale Lab, Enhancing Neuroimaging Genetics through Meta-Analysis, the Medical Research Council Integrative Epidemiology Unit (MRC-IEU), the Social Science Genetic Association Consortium, the Within-Family GWAS Consortium, and large-scale genome-wide association meta-analyses.^[21,22]^

Most GWAS datasets included in this study were derived from large population-based cohorts, such as the UK Biobank, FinnGen, and related consortia, which predominantly consist of adult participants. Accordingly, the neuropsychological and behavioral outcomes analyzed in this study mainly reflect long-term cognitive, educational, personality, and psychiatric traits assessed in adolescence or adulthood, rather than early childhood developmental milestones. This approach enables the evaluation of the long-term neurodevelopmental consequences potentially associated with anesthesia exposure and surgical interventions.

To minimize the influence of confounding, genetic associations were derived from independent GWAS datasets with shared ancestral backgrounds. Because this study exclusively used publicly available summary-level data, formal ethical approval was not required. Detailed descriptions of all included GWAS datasets are provided in Table S1, Supplemental Digital Content 1.

2.3. Exposures

Genetic associations for anesthesia-related exposures were obtained from the Medical Research Council Integrative Epidemiology Unit database, which includes 4 independent GWAS datasets derived from populations of European ancestry. These datasets represented distinct anesthetic agents or anesthesia modalities, including inhalational anesthesia, intravenous anesthesia, local anesthetic nerve blocks, and lumbar epidural anesthesia. In addition, the number of surgical operations was included as a separate exposure variable.

2.4. Outcomes

A total of 39 GWAS datasets related to neuropsychological and behavioral traits were included as outcome variables in this study. These outcomes were categorized into the following 5

2.4.1. Intelligence and cognitive performance

Childhood intelligence, fluid intelligence score, number of fluid intelligence questions attempted within the time limit, years of schooling, age at completion of full-time education, cognitive performance, and cognitive function.

2.4.2. Neurodevelopmental disorders

Hyperkinetic disorders (excluding attention-deficit/hyperactivity disorder), autism spectrum disorder, disturbance of activity and attention, other symptoms and signs involving cognitive functions and awareness, speech and linguistic disorders, specific developmental disorders of speech and language, other behavioral and emotional disorders with onset usually occurring in childhood and adolescence, behavioral and emotional disorders with onset usually occurring in childhood and adolescence, emotional disorders and disorders of social functioning with onset specific to childhood, emotional disorders starting during childhood or adolescence, social disorders starting during childhood or adolescence, and outcomes related to negative life events in childhood.

2.4.3. Personality disorders

Anankastic personality disorder, anxious personality disorder, dependent personality disorder, dissocial personality disorder, emotionally unstable personality disorder, histrionic personality disorder, mixed and other personality disorders, paranoid personality disorder, schizoid personality disorder, and neuroticism score.

2.4.4. Mental health outcomes diagnosed by a professional

Depression, anxiety, panic attacks, social anxiety or social phobia, and other specific phobias (e.g., fear of heights or spiders).

2.4.5. Brain structure volumes related to cognitive and emotional functions

Cerebral white matter volume, cortical volume, amygdala volume, and hippocampal volume.

2.5. Selection of instrumental variables

The investigation meticulously adhered to strict standards in the selection of IVs to guarantee their credibility and dependability.

The selection of IVs was based on a suggestive genome-wide association threshold (P < 1 × 10^−5^),^[23]^, a minor allele frequency >0.01, and linkage disequilibrium clumping with an r^2^ < 0.001 within a 10,000 kb window. This threshold was adopted to ensure an adequate number of independent genetic instruments while maintaining sufficient statistical power.

To assess instrument strength and minimize weak instrument bias, the F statistic was calculated for each individual SNP. SNPs with an F statistic below 10 were considered weak instruments and were excluded from subsequent analyses.^[24]^ The F statistic was calculated using a standard formula based on the SNP effect size and its corresponding standard error.

The proportion of variance explained (R^2^) by each SNP was estimated using a conventional formula incorporating the SNP effect size, effect allele frequency, standard error, and GWAS sample size. Detailed formulas for F statistic and R^2^ calculations are provided in the Supplementary Methods, Supplemental Digital Content 2.

2.6. UVMR and MVMR analyses

UVMR analyses were 1st conducted to estimate the total causal effects of anesthesia exposure and the number of surgical procedures on neuropsychological and behavioral outcomes. The inverse-variance weighted (IVW) method was used as the primary analytical approach, providing efficient causal estimates under the assumption of valid instrumental variables.^[25]^ Fixed- or random-effects models were selected according to the presence of heterogeneity.

Complementary sensitivity analyses, including MR-Egger regression,^[26]^ weighted median,^[27]^ and weighted mode methods,^[28]^ were applied to evaluate the robustness of causal estimates under different assumptions regarding horizontal pleiotropy and instrument validity. Consistency in effect direction across these methods was considered supportive of robust causal inference.

The directionality of causal associations between exposures and outcomes was evaluated using the MR Steiger test.^[29]^ To account for multiple testing, a Bonferroni-corrected significance threshold of P < .01 (.05/5, corresponding to 5 exposure traits related to anesthetic types or number of operations) was applied. Associations with P values between .01 and .05 were interpreted as suggestive and requiring further validation.

MVMR analyses were subsequently performed to disentangle the independent effects of anesthesia exposure and surgical procedures by mutually adjusting for correlated exposures. MVMR leverages genetic variants associated with multiple exposures to estimate their direct effects on outcomes.^[20]^ The IVW method was used as the primary estimator in MVMR, with fixed- or random-effects models selected based on heterogeneity, and MVMR-Egger was employed as a supplementary sensitivity analysis. Concordant effect directions between IVW and MVMR-Egger were considered indicative of result robustness.

2.7. Sensitivity analyses

Within the UVMR framework, heterogeneity across instrumental variables was assessed using Cochran Q statistic in combination with the MR-Egger approach, as described by Bowden et al and Burgess and Thompson.^[26,30]^ A P value >.05 was interpreted as evidence of no significant heterogeneity.

Potential horizontal pleiotropy was evaluated using the MR-Egger intercept test and the Mendelian randomization pleiotropy residual sum and outlier (MR-PRESSO) global test, following the methodology proposed by Verbanck et al.^[31]^ In these analyses, a P value exceeding .05 indicated no evidence of directional pleiotropy.

The MR-PRESSO method was additionally applied to identify potential outlier SNPs. After the removal of detected outliers, the results remained consistent with those obtained from the IVW method, further supporting the robustness of the primary findings.^[31]^

To examine the influence of individual SNPs on the estimated causal effects, leave-one-out analyses were performed. For robustness assessment within the multivariable MR framework, the MVMR-Egger method was employed, which is capable of accounting for both measured and unmeasured pleiotropic effects.

All causal effect estimates are reported as odds ratios (ORs) with corresponding 95% confidence intervals (CIs). Statistical analyses were conducted using R software (version 4.2.2; University of Auckland), with the primary packages including TwoSampleMR, MendelianRandomization, psych, and MR-PRESSO. Forest plots were generated using the ggplot2 package (version 3.3.6) to visually summarize the MR results.

3. Results

An overview of the GWAS datasets included in this MR analysis – including sample size, population ancestry, exposure and outcome definitions, and data sources – is provided in Table S1, Supplemental Digital Content 1.

3.1. Genetic instrument selection

Based on predefined instrument selection criteria, we identified 18, 9, 7, 8, and 130 independent SNPs as IVs for inhalational anesthesia, intravenous anesthesia, local anesthetic nerve block, lumbar epidural anesthesia, and the number of surgical operations, respectively, in the UVMR analyses (Table S2, Supplemental Digital Content 3). The corresponding F statistics ranged from 19.53 to 53.99, all exceeding the conventional threshold of 10, indicating that the selected IVs were sufficiently strong (Table S2, Supplemental Digital Content 3).

3.2. UVMR analyses of different anesthetics or anesthesia methods, as well as the number of operations, on the risk of neuropsychological and behavioral outcomes

Using the selected genetic instruments, UVMR analyses provided suggestive evidence that genetic liability to inhalational anesthesia exposure was associated with an increased risk of social anxiety or social phobia (IVW: OR = 1.4475, 95% CI 1.0662–1.9651, P = .0177; Table 1). No significant causal associations were observed between inhalational anesthesia and the remaining 38 neuropsychological and behavioral outcomes (Table S3, Supplemental Digital Content 4).

Genetic liability to intravenous anesthesia exposure showed suggestive associations with fewer years of schooling (IVW: OR = 0.0828, 95% CI 0.0102–0.2327, P = .0199), lower fluid intelligence scores (IVW: OR = 6.71 × 10^−5^, 95% CI 1.94 × 10^−8^–0.2327, P = .0209), and a reduced risk of emotional disorders starting during childhood or adolescence (IVW: OR = 8.41 × 10^−26^, 95% CI 3.78 × 10^−50^–0.1871, P = .0435; Table 1). No significant associations were detected for the remaining 36 outcomes.

No evidence of causal associations was identified between either local anesthetic nerve block or lumbar epidural anesthesia and any of the neuropsychological or behavioral outcomes examined (Table S3, Supplemental Digital Content 4).

In contrast, a higher genetic predisposition to an increased number of surgical operations was causally associated with fewer years of schooling (IVW: OR = 0.8979, 95% CI 0.8287–0.9730, P = .0086) and a younger age at completion of full-time education (IVW: OR = 0.8615, 95% CI 0.8041–0.9229, P = 2.22 × 10^−5^). Additionally, it was associated with a higher neuroticism score (IVW: OR = 1.8033, 95% CI 1.3918–2.3364, P = 8.12 × 10^−6^), an increased risk of depression (IVW: OR = 1.0951, 95% CI 1.0544–1.1374, P = 2.57 × 10^−6^), and emotionally unstable personality disorder (IVW: OR = 2.4124, 95% CI 1.2408–4.6906, P = .0094; Table 1).

Sensitivity analyses were conducted for exposure–outcome pairs with IVW P values <.05. The estimates obtained using alternative MR methods were directionally consistent with the IVW results. Significant heterogeneity was detected in analyses of the number of operations with years of schooling (MR-Egger P = 4.75 × 10^−12^), age at completion of full-time education (MR-Egger P = 3.63 × 10^−11^), and neuroticism score (MR-Egger P = 4.01 × 10^−8^; Table 2); therefore, random-effects IVW models were applied for these associations.

MR-PRESSO analyses identified potential pleiotropic outliers among IVs for the number of operations. After removal of these variants, the associations remained years of schooling (OR = 0.9088, 95% CI 0.8569–0.9639, P = 1.45 × 10^−3^), age at completion of full-time education (OR = 0.8911, 95% CI 0.8346–0.9516, P = 5.77 × 10^−4^), and neuroticism score (OR = 1.5962, 95% CI 1.2727–2.0018, P = 5.18 × 10^−5^; Table S4, Supplemental Digital Content 5). MR-Egger intercept tests did not indicate significant horizontal pleiotropy. No substantial heterogeneity was observed for other exposure–outcome pairs, for which fixed-effects IVW models were applied.

The MR Steiger test confirmed the directionality of all significant and suggestive associations (Table 2). Leave-one-out analyses further demonstrated that no single SNP disproportionately influenced the observed causal estimates (Fig. S1, Supplemental Digital Content 6).

3.3. Multivariable MR analyses

Multivariable MR analyses were performed for exposures that demonstrated suggestive or significant associations in UVMR analyses.

After adjustment for the number of surgical operations, the suggestive associations between inhalational anesthesia and social anxiety or social phobia (IVW: OR = 1.3180, 95% CI 0.9662–1.7978, P = .0813), as well as between intravenous anesthesia and years of schooling, fluid intelligence score, and emotional disorders starting during childhood or adolescence, were no longer evident (Table S5, Supplemental Digital Content 7).

In contrast, after adjusting for inhalational anesthesia, intravenous anesthesia, local anesthetic nerve block, lumbar epidural anesthesia, and all anesthesia modalities combined, the negative associations between genetic liability to an increased number of operations and both years of schooling and age at completion of full-time education remained significant. Similarly, the positive associations between the number of operations and increased risks of depression and higher neuroticism scores persisted across all adjustment models (Table S5, Supplemental Digital Content 7, Fig. 2).

When adjusting for local anesthetic nerve block and lumbar epidural anesthesia, the causal association between the number of operations and emotionally unstable personality disorder remained significant. However, this association was attenuated and no longer statistically significant after additional adjustment for inhalational anesthesia, intravenous anesthesia, or all anesthesia modalities combined (Table S5, Supplemental Digital Content 7, Fig. 2). Consistently, when adjusting for the combined effects of inhalational and intravenous anesthesia, the association between the number of operations and emotionally unstable personality disorder was no longer observed (IVW: OR = 2.4256, 95% CI 0.9239–6.3679, P = .0720).

Sensitivity analyses using the MVMR-Egger approach yielded consistent results, and no evidence of horizontal pleiotropy was detected based on the MVMR-Egger intercept test. Nonetheless, heterogeneity was observed in the corresponding MVMR analyses (Table S6, Supplemental Digital Content 8).

4. Discussion

Anesthesia is an indispensable component of modern medical practice; however, concerns persist regarding its potential long-term neurodevelopmental and behavioral consequences, particularly in pediatric populations exposed to prolonged or repeated procedures.^[4]^ Despite extensive experimental and clinical investigations, existing evidence remains inconclusive, largely due to the intrinsic challenges of disentangling anesthetic effects from surgical indications, disease burden, and other clinical confounders. In this context, MR, leveraging genetically randomized variation, provides an opportunity to explore potential causal relationships while minimizing residual confounding.^[1]^

In the present study, we observed suggestive associations between exposure to inhalational anesthetics and an increased risk of social anxiety or social phobia, as well as between intravenous anesthetic exposure and reduced educational attainment and fluid intelligence scores. However, these associations did not persist after accounting for surgical exposure in multivariable MR analyses. In contrast, the number of operations demonstrated consistent causal associations with reduced years of schooling and age at completion of full-time education, alongside increased risks of depression, higher neuroticism scores, and emotionally unstable personality disorder. Importantly, these associations largely remained robust even after adjustment for multiple anesthetic modalities, indicating that cumulative surgical exposure, rather than anesthetic agents alone, may play a more prominent role in shaping long-term educational and emotional outcomes.

Our findings align with prior clinical and epidemiological studies reporting largely null effects of single or brief anesthetic exposures on global cognitive outcomes. Multiple prospective and observational studies have shown no significant differences in full-scale intelligence following general anesthesia in childhood.^[8–10,32]^ Similarly, controlled studies of short-duration sevoflurane anesthesia in preschool-aged children failed to detect meaningful neurocognitive deficits during follow-up.^[33]^ These observations are further supported by evidence from pediatric intensive care cohorts, where sedative and anesthetic drug exposure did not independently predict long-term neurocognitive impairment.^[34]^ Collectively, these data are consistent with our MR results demonstrating no causal association between anesthetic exposure and intelligence-related outcomes.

The apparent discrepancy between preserved intelligence and impaired educational or emotional outcomes warrants careful interpretation. Educational attainment and personality traits represent complex, functionally integrated phenotypes that reflect not only cognitive capacity but also emotional regulation, behavioral adaptation, and sustained engagement with learning environments. In this regard, our findings resonate with large epidemiological studies suggesting that long-term neurodevelopmental trajectories are strongly influenced by socio-environmental context, family background, and developmental plasticity, often exceeding the explanatory power of isolated medical exposures.^[35,36]^ Thus, the causal associations observed with the number of operations may reflect the cumulative impact of repeated medical interventions within a broader developmental context, rather than direct neurotoxicity attributable to anesthesia itself.

Furthermore, our results extend prior meta-analytic evidence indicating that repeated anesthesia exposure is associated with adverse developmental and academic outcomes, while global cognitive function remains relatively preserved.^[37]^ In that meta-analysis, children with multiple anesthetic exposures exhibited poorer academic performance, language development, and fine motor skills, yet did not demonstrate consistent deficits in intelligence measures. Similarly, in our study, repeated surgical exposure was causally associated with reduced years of schooling and increased emotional vulnerability, whereas intelligence-related outcomes remained largely unaffected. This pattern supports a distinction between preserved core cognitive capacity and vulnerability in functional, emotional, and educational trajectories following repeated medical and surgical interventions.

Several mechanisms may plausibly underlie these associations. Recurrent surgical interventions often entail repeated perioperative stress, prolonged hospitalizations, and disruptions to normal routines, schooling, and social development. Such cumulative psychosocial stressors may influence emotional regulation and personality development over time, particularly during sensitive developmental periods. In parallel, neuroticism – a personality trait characterized by emotional instability, heightened anxiety, and vulnerability to stress – has been robustly linked to increased risks of depression and anxiety disorders,^[38,39]^ consistent with the associations observed in our analyses. While biological mechanisms such as inflammatory responses and stress-related neuroendocrine dysregulation may also contribute,^[40–42]^ our findings suggest that these effects are more closely tied to cumulative surgical exposure than to specific anesthetic agents.

Notably, the association between the number of operations and emotionally unstable personality disorder was attenuated after adjustment for intravenous and inhalational anesthetics, suggesting that general anesthetic exposure may partially contribute to this specific outcome. In contrast, regional anesthesia techniques, including nerve blocks and epidural anesthesia, showed no causal associations with neurobehavioral outcomes in either univariable or multivariable analyses, consistent with prior clinical observations.^[10,43]^ This further supports the notion that the neurobehavioral risks observed are unlikely to be driven by anesthesia exposure alone.

Several limitations warrant consideration. First, the analysis was restricted to individuals of European ancestry, limiting generalizability to other populations. Second, the available datasets did not permit stratification by potentially vulnerable subgroups or detailed characterization of anesthetic agents, dosages, or exposure duration. Third, surgical exposure was captured solely by the number of operations, without accounting for procedural complexity or disease severity. Future studies integrating more granular clinical data and diverse populations may help refine these findings.

5. Conclusion

Our MR analyses suggest that, despite ongoing concerns regarding potential neurodevelopmental effects of anesthetic drugs, exposure to anesthetic agents alone is unlikely to substantially impair long-term cognitive development or core cognitive abilities. In contrast, cumulative surgical exposure was consistently associated with reduced educational attainment, increased risk of depression, and higher neuroticism scores, indicating that surgical-related factors may play a more prominent role in shaping long-term educational and emotional outcomes. Notably, the risk of emotionally unstable personality disorder appeared to be influenced by a combination of general anesthetic exposure and surgical-related factors. These findings underscore the importance of considering the broader perioperative context, suggesting that optimizing surgical decision-making and perioperative management may offer greater benefitsfor pediatric neurodevelopment than focusing exclusively on anesthetic exposure.

Acknowledgments

We appreciate the work of UK Biobank, FinnGen, Neale Lab, Enhancing Neuro Imaging Genetics through Meta-Analysis (ENIGMA), Medical Research Council Integrative Epidemiology Unit (MRC-IEU), Social Science Genetic Association Consortium (SSGAC), Within-family GWAS consortium, and Genome-wide association meta-analysis.

Author contributions

Conceptualization: Yuchao Fan.

**Data ** Yuchao Fan.

**Formal ** Yuchao Fan.

**Funding ** Wen Wu.

Investigation: Yuchao Fan.

Methodology: Yuchao Fan.

**Project ** Yuchao Fan.

Resources: Yuchao Fan.

Software: Yuchao Fan.

Supervision: Yuchao Fan.

Validation: Yuchao Fan.

Visualization: Yuchao Fan.

**Writing – original ** Wen Wu, Yuchao Fan.

**Writing – review & ** Yuchao Fan.