Behavioural disorders after prenatal exposure to anaesthesia for maternal surgery

Abstract

Objective:

To evaluate the association between prenatal exposure to general anaesthesia due to maternal surgery during pregnancy and subsequent risk of disruptive or internalizing behavioural disorder (DIBD) diagnosis in the child.

Methods:

A nationwide sample of pregnant women linked to their liveborn infants was evaluated using the Medicaid Analytic eXtract (MAX, 1999–13). Multivariate matching was used to match each child prenatally exposed to general anaesthesia due to maternal appendectomy or cholecystectomy during pregnancy with five unexposed children. The primary outcome was a DIBD diagnosis in children. Secondary outcomes included diagnoses for a range of other neuropsychiatric disorders.

Results:

34,271 prenatally exposed children were matched with 171,355 unexposed children. Prenatally exposed children were significantly more likely than unexposed children to receive a diagnosis for a DIBD (hazard ratio [HR], 1.31; [95% confidence interval (CI), 1.23–1.40]). For secondary outcomes, increased hazards of disruptive (HR, 1.32; [95% CI, 1.24–1.41]) and internalizing (HR, 1.36; [95% CI, 1.20–1.53]) behavioural disorders were identified, as well as increased hazards of attention deficit hyperactivity disorder (HR, 1.32; [95% CI, 1.22–1.43]), behavioural disorders (HR, 1.28; [95% CI, 1.14–1.42]), developmental speech or language disorders (HR, 1.16; [95% CI, 1.05–1.28]), and autism (HR, 1.31; [95% CI, 1.05–1.64]).

Conclusions:

Prenatal exposure to general anaesthesia is associated with a 31% increased risk for a subsequent DIBD diagnosis in the children. Caution however is advised when making any clinical decisions regarding care of pregnant women, as avoidance of necessary surgery during pregnancy can have detrimental effects on mothers and their children.

Introduction

Millions of children receive anaesthesia each year in the United States,^1,2 and an estimated 80,000 pregnant women receive anaesthesia for non-obstetric surgery.^3,4 Questions have emerged about the safety of anaesthetics based on preclinical,^5–7 and clinical studies.^8–10 Given these concerns, the US Food and Drug Administration (FDA) issued a Drug Safety Communication (DSC) in 2016 stating that the use of general anaesthetic drugs in “children younger than 3 years or in pregnant women during their third trimester”, may “affect the development of children’s brains.”¹¹ There is controversy regarding this warning as sufficiently powered randomized controlled trials establishing a causal effect of anaesthesia are challenging to perform, and observational studies in children needing surgery and anaesthesia to address a medical condition may be biased due to confounding by indication.¹² This study evaluates whether prenatal exposure to general anaesthesia is associated with increased incidence of subsequently developing a disruptive or internalizing behavioural disorder (DIBD). Prenatal exposures are particularly relevant because general anaesthetic drugs cross the placenta,¹³ and this neurodevelopmental period is associated with peak brain vulnerability to anaesthetic agents in preclinical studies.^7,14 In addition, prenatal exposures are typically due to medical disease in the mother, not the child, thus potentially dissociating the indication for surgery from the child being evaluated and ameliorating bias due to confounding by indication.

Methods

Data Source and Study Cohort

Medicaid offers health insurance to low-income pregnant women and children, covering 48% of all pregnancies and 39% of all children in the United States.^15,16 The study cohort was drawn from Medicaid Analytic eXtract (MAX) claims data from 47 states and Washington D.C., for the period of 1999 to 2013.¹⁷ Maine, Montana, and Connecticut were excluded due to the inability to link data from mothers and infants. The data were accessed using the Centers for Medicare & Medicaid Services Virtual Research Data Center (VRDC). Linkages between deliveries and live-born infants were performed using previously described methods,^18–21 and last menstrual period (LMP) was estimated using a validated algorithm.²² After linkage, all available claims from mothers were evaluated from 90 days prior to the estimated LMP date until the delivery date. In children with preterm delivery, maternal claims were evaluated for a maximum observation period of 335 days before infant date of birth (DOB), and 360 days before infant DOB for all others. In mothers eligible for Medicaid for less than these maximum observation periods, observation began at the time of Medicaid enrolment during pregnancy.

Study Conduct

This study was approved by the Institutional Review Board at Columbia University Medical Center (New York, NY) as exempt from requiring written/informed consent.

Exposure

Prenatal general anaesthetic exposure was identified using maternal claims with International Classification of Disease, Ninth revision, Clinical Modification (ICD-9-CM), or Current Procedural Terminology (CPT) procedure codes occurring between the estimated LMP and infant DOB for either an appendectomy or cholecystectomy (eTable 1), which are the two most common non-obstetric procedures performed during pregnancy.²³ Exposures were classified as occurring in the first, second, or third trimester of pregnancy. For mothers with both an appendectomy and cholecystectomy during pregnancy, if the procedures were performed on separate days, the trimester of exposure and procedure type were based on the first procedure. If both procedures were performed on the same day, the procedure was considered a cholecystectomy. The deliveries and linked infants for the remaining mothers were classified as unexposed potential controls for matching.

Neuropsychiatric Outcomes

The claims of linked children were used to evaluate neuropsychiatric outcomes. The observation period in children started at birth and continued until censoring defined as December 31^st, 2013, or any period where a child had 3 consecutive months without Medicaid eligibility. The primary outcome was the age at first presence of any ICD-9-CM diagnosis for a DIBD, a composite outcome consisting of any diagnosis for a disruptive behavioural disorder (attention-deficit/hyperactivity disorder (ADHD) or conduct, impulse control, or oppositional defiant disorders) or an internalizing behavioural disorder (bipolar disorder, depression, or anxiety) (eTable 2). As secondary outcomes, disruptive behavioural disorders and internalizing behavioural disorders were evaluated independently. Diagnoses for autism, ADHD, learning difficulty, developmental speech or language disorder, intellectual disability, and behavioural disorder were identified using claims-based algorithms validated as having high positive predictive values.²⁴

Covariates

A broad range of potential confounders were evaluated including sociodemographic covariates (age, race, ethnicity, median income of the zip code of residence, and presence in Medicaid for disability vs. poverty) and healthcare utilization prior to pregnancy (number of inpatient admissions, and outpatient and emergency department visits). Variables identifying total months of Medicaid eligibility, types of managed care coverage, restricted benefits, and private insurance coverage were also recorded. Maternal medical comorbidities associated with maternal perinatal morbidity,²⁵ psychiatric comorbidities, and comorbidities associated with appendicitis and cholelithiasis (hypercholesterolemia, gout, irritable bowel syndrome, thalassemia, or urinary calculus) were identified (eTable 3), as well as a total count of the listed medical and psychiatric comorbidities. Filled prescriptions for psychotropic medications (eTable 4), folate, progestin, and corticosteroids were also identified. Post-exposure variables collected between birth and 3 months following delivery were recorded and included characteristics of labour and delivery in the mother (eTable 5), and conditions in the child including birth weight and congenital malformations.

Matching Methods

Each prenatally exposed child was matched with five unexposed children. Some covariates, such as mother’s race or child’s sex were fixed at conception and did not vary; however, other covariates, such as prescriptions filled or Medicaid coverage often varied during the pregnancy. Also, exposed mothers were exposed to anaesthesia at different times during the pregnancy. By analogy with an experiment, we wanted to match so that exposed and control mothers were similar prior to treatment. Mothers of exposed children were therefore matched to mothers of unexposed children who were similar up to the month of gestation immediately prior to the exposed mother’s month of exposure to anaesthesia. Matching was exact for the state of residence, year of birth, and child sex, and as close as possible for other covariates, using the multivariable Mahalanobis distance.²⁶ Matching also controlled for the estimated risk of a childhood diagnosis of a neuropsychiatric disorder diagnosis, or prognostic score, as follows: Before matching, a 10% random sample of unexposed mothers and linked children was excluded from the study cohort and used to estimate risk of a childhood diagnosis of neuropsychiatric disorder, based on Cox’s time-dependent proportional hazards model.²⁷ This was a model fit on a random external set-aside 10% population not part of the final study population. Matching controlled for the estimated risk in the month prior to exposure. Therefore, matching on the prognostic score gave emphasis to covariates that predict later diagnosis. Optimal matching was implemented through the NETFLOW procedure in SAS.^28,29 All matching was completed before examining any outcomes.³⁰ We aimed for standardized differences in covariate means after matching of less than 0.05 in the nationwide cohort, below the traditional standard of 0.2.^31,32 Even more stringently, we also aimed for standardized differences below 0.2 within each of the 47 separate states so, for example, treated mothers in Arkansas were similar to control mothers in Arkansas. The matching algorithm is further described in the Supplemental Methods.

Statistical Analyses

The primary outcome was age at first diagnosis of DIBD, as identified by claims in children. Evaluation of an interaction between exposure and time was performed to confirm the appropriateness of the proportional hazards assumption. Hazard ratios associated with exposure status were modelled using the matched version of the proportional hazards model.³³ Subgroup analyses to evaluate risk differences in various groups of children were performed, comparing exposed boys vs. exposed girls, appendectomy vs. cholecystectomy, trimester of exposure, region of the country,³⁴ or year of birth. To determine whether risk differences by subgroup were statistically significant, we tested for effect modification – that is, for different effects of exposure in different subgroups of children – by testing an exposure-subgroup interaction in the matched version of the proportional hazards model. The robustness of our results to unmeasured confounding was evaluated using an asymptotic separable calculation.³⁵ Secondary outcomes were also evaluated, using the matched version of the proportional hazards model. Adjustment for multiple comparisons was not performed when evaluating secondary outcomes or differences between subgroups.

Additional Analyses

To further assess the robustness of our findings, we excluded all matched sets of mothers where the exposed mother had selected characteristics. All unexposed mothers within the remaining matched sets with that characteristic were then also excluded. To evaluate the association between exposure and DIBD among mothers with no pre-existing psychiatric disorders for example, all mothers with a psychiatric disorder diagnosis or psychotropic medication use prior to exposure were excluded. In a separate analysis, we excluded mothers who lacked continuous enrolment throughout pregnancy, or had restricted benefits, to compare our cohort with more restrictive cohorts of Medicaid enrolled mothers used in other studies.^18–21

To explore the contribution of obstetrical complications known to be associated with surgery during pregnancy,²³ we performed an additional analysis that also matched on selected post-exposure variables (preterm birth, birth weight categories, and caesarean birth). In this analysis, preterm birth was matched exactly, permitting separate subgroup analysis of preterm and non-preterm infants.

Results

Characteristics of the Study Cohort

The study cohort consisted of 16,778,231 deliveries linked to an infant born between 1999 and 2013 with known sex (Figure 1). 34,271 children who were exposed to general anaesthesia due to maternal appendectomy or cholecystectomy during pregnancy were identified after excluding 11 children from Washington D.C. where a small sample size did not allow for adequate matching. 148 mothers had both an appendectomy and cholecystectomy, while the remaining 34,123 had one or the other. The contribution of subjects from individual states can be seen in eTable 6.

Figure 1.

Flowchart of Cohort Generation

Quality of the Matched Sets

The matching was successful in producing exposed and unexposed mothers who were similar prior to exposure, see Table 1 for selected covariates and eTable 7 for all covariates. We observed no standardized differences larger than 0.05 standard deviations in the nationwide match, and none larger than 0.2 standard deviations in the individual state-specific matches. The standardized differences in covariates prior to matching as compared to after matching can be seen in eFigures 1 and 2, with standardized differences below the 0.05 threshold seen in all pre-exposure covariates after matching, but above the threshold in some post-exposure covariates, such as preterm birth.

Table 1.

Selected List of Covariates and Standardized Differences from the Nationwide Match. (The full list of covariates is available in eTable 7 in the Supplementary Appendix)

	Matching Method	Exposed (n=34271) Mean or %	Unexposed (n=171355) Mean or %	Standardized Difference
Children year of birth (mean year)	E	2006.94	2006.94	0
Child sex (%)
Male	E	17363 (50.7)	86815 (50.7)	0
Female		16908 (49.3)	84540 (49.3)	0
Maternal age (mean years)	D	24.55	24.43	0.021
Maternal race (%)
African American/Black	D	3415 (10)	18892 (11)	−0.032
American Indian/Alaskan Native	D	996 (2.9)	3867 (2.3)	0.044
Asian	D	262 (0.8)	1224 (0.7)	0.004
Hawaiian/Pacific Islander	D	144 (0.4)	609 (0.4)	0.009
White	D	13443 (39.2)	67449 (39.4)	−0.003
Unknown	D	16011 (46.7)	79314 (46.3)	0.009
Maternal ethnicity (%)
Hispanic	D	7015 (20.5)	33934 (19.8)	0.016
Non-Hispanic	D	16371 (47.8)	82572 (48.2)	−0.008
Unknown	D	10885 (31.8)	54849 (32)	−0.005
Income by zip code (%)
Quartile 1	D	7386 (21.6)	35950 (21)	0.014
Quartile 2	D	8421 (24.6)	42582 (24.9)	−0.007
Quartile 3	D	8691 (25.4)	43749 (25.5)	−0.004
Quartile 4	D	8300 (24.2)	42190 (24.6)	−0.009
Unknown		1473 (4.3)	6884 (4)	0.013
Pre-pregnancy healthcare utilization (%)
No outpatient visits	D	23239 (67.8)	116117 (67.8)	0.001
1 outpatient visit	D	3418 (10)	16782 (9.8)	0.006
2 outpatient visits	D	2205 (6.4)	12632 (7.4)	−0.042
3 outpatient visits	D	1680 (4.9)	8834 (5.2)	−0.013
4 outpatient visits	D	1138 (3.3)	5435 (3.2)	0.01
5 or more outpatient visits	D	2591 (7.6)	11555 (6.7)	0.037
No inpatient visits	D	33067 (96.5)	165980 (96.9)	−0.024
Any inpatient visits	D	1204 (3.5)	5375 (3.1)	0.024
No emergency room visits	D	31758 (92.7)	160087 (93.4)	−0.034
Any emergency room visits	D	2513 (7.3)	11268 (6.6)	0.034
Enrolled in Medicaid for disability (%)	D	662 (1.9)	2464 (1.4)	0.039
Medicaid eligibility prior to exposure (mean months)
Not eligible for Medicaid	D	3.23	3.2	0.01
Comprehensive plan only	D	0.82	0.88	−0.016
Dental plan only	D	0.19	0.17	0.01
Behavioural plan only	D	0.31	0.31	0.002
Primary care case management plan only	D	0.36	0.37	−0.003
Other managed care plan only	D	0.23	0.22	0.004
Comprehensive and dental plan	D	0.3	0.32	−0.009
Comprehensive and behavioural plan	D	0.3	0.32	−0.007
Comprehensive and other managed care plan	D	0.09	0.09	−0.005
Comprehensive, dental, and behavioural plan	D	0.06	0.07	−0.009
Primary care case management and dental plan	D	0.03	0.03	−0.001
Primary care case management and behavioural plan	D	0.12	0.12	−0.001
Primary care case management and other managed care plan	D	0.11	0.11	−0.004
Primary care case management, dental, and behavioural plan		0	0	0.004
Dental and behavioural plan	D	0.02	0.01	0.007
Other combinations	D	0.06	0.06	−0.002
Fee for service	D	2.11	2.05	0.015
Managed care plan status is unknown	D	0	0	0
Maternal medical comorbidities (%)
Asthma	D	1258 (3.7)	6365 (3.7)	−0.002
Cardiac valvular disease	D	94 (0.3)	517 (0.3)	−0.003
Chronic congestive heart failure	D	(−) (0)	(−) (0)	−0.001
Chronic ischemic heart disease	D	27 (0.1)	117 (0.1)	0.002
Chronic liver disease	D	299 (0.9)	488 (0.3)	0.046
Chronic renal disease	D	185 (0.5)	578 (0.3)	0.014
Congenital heart disease	D	85 (0.3)	448 (0.3)	−0.001
Human immunodeficiency virus	D	21 (0.1)	193 (0.1)	−0.013
Overweight or obesity	D	991 (2.9)	4036 (2.4)	0.021
Pre-existing diabetes	D	519 (1.5)	2480 (1.5)	0.004
Pre-existing hypertension	D	755 (2.2)	3243 (1.9)	0.013
Pulmonary hypertension	D	(−) (0)	30 (0)	0
Sickle cell disease	D	32 (0.1)	68 (0)	0.014
Smoking/Tobacco Use	D	1452 (4.2)	7869 (4.6)	−0.011
Underweight	D	(−) (0)	23 (0)	−0.002
Maternal medical comorbidities associated with appendectomy or cholecystectomy (%)
Diverticular disease	D	33 (0.1)	61 (0)	0.016
Gout	D	(−) (0)	(−) (0)	0.004
Hypercholesterolemia	D	82 (0.2)	362 (0.2)	0.005
Irritable bowel syndrome	D	121 (0.4)	345 (0.2)	0.023
Thalassemia	D	(−) (0)	23 (0)	−0.001
Urinary calculus	D	407 (1.2)	1607 (0.9)	0.017
Maternal psychiatric comorbidities (%)
Adjustment disorders	D	346 (1)	1661 (1)	0.003
Alcohol related disorders	D	197 (0.6)	874 (0.5)	0.007
Anxiety disorder	D	1242 (3.6)	5906 (3.5)	0.008
Bipolar disorder	D	424 (1.2)	1816 (1.1)	0.014
ADHD and other disruptive behavioural disorders	D	191 (0.6)	1179 (0.7)	−0.015
Depression	D	1689 (4.9)	8188 (4.8)	0.006
Eating disorder	D	13 (0)	78 (0.1)	−0.003
Schizophrenia	D	67 (0.2)	382 (0.2)	−0.005
Substance related disorders	D	593 (1.7)	3736 (2.2)	−0.022
Other mental disorders	D	477 (1.4)	2319 (1.4)	0.003
Maternal psychotropic medication use (%)
ADHD medication	D	145 (0.4)	696 (0.4)	0.003
Antidepressants	D	2330 (6.8)	9957 (5.8)	0.038
Antipsychotics	D	317 (0.9)	1347 (0.8)	0.015
Mood Stabilizers	D	378 (1.1)	1550 (0.9)	0.02
Sedative/anxiolytics	D	1153 (3.4)	5568 (3.3)	0.006
Conditions during pregnancy prior to exposure (%)
Chorioamnionitis	D	21 (0.1)	123 (0.1)	−0.002
Early or threatened labour	D	1017 (3)	5438 (3.2)	−0.004
Foetal heart rate abnormality	D	229 (0.7)	1331 (0.8)	−0.003
Gestational diabetes	D	281 (0.8)	1388 (0.8)	0
Gestational hypertension	D	85 (0.3)	413 (0.2)	0
Multiple gestation	D	359 (1.1)	1762 (1)	0.001
Oligohydramnios	D	54 (0.2)	362 (0.2)	−0.003
Placenta accreta	D	(−) (0)	15 (0)	−0.001
Placenta previa	D	314 (0.9)	1613 (0.9)	−0.001
Placental abruption	D	40 (0.1)	177 (0.1)	0.001
Polyhydramnios	D	28 (0.1)	208 (0.1)	−0.003
Premature rupture of membranes	D	(−) (0)	76 (0)	−0.002
Preeclampsia- Mild	D	45 (0.1)	243 (0.1)	−0.001
Preeclampsia/eclampsia- Severe	D	(−) (0)	75 (0)	−0.001
Corticosteroid use	D	2516 (7.3)	10090 (5.9)	0.044
Folate Supplementation	D	8992 (26.2)	43223 (25.2)	0.02
Progestin use	D	623 (1.8)	2083 (1.2)	0.044
Conditions at the time of delivery (%)
Caesarean section	X	10395 (30.3)	46274 (27)	0.074
Labour epidural	Z	11188 (32.7)	56013 (32.7)	−0.001
Pyrexia during labour	Z	353 (1)	1495 (0.9)	0.016
Any other surgical procedures during pregnancy	Z	511 (1.5)	2049 (1.2)	0.025
Conditions in the child (%)
Preterm birth	X	7122 (20.8)	24674 (14.4)	0.168
Low Birth Weight (<2500g and 1500g)	X	1801 (5.3)	5337 (3.1)	0.107
Very Low Birth Weight (<1500g and 1000g)	X	372 (1.1)	942 (0.6)	0.058
Extremely Low Birth Weight (<1000g)	X	312 (0.9)	578 (0.3)	0.07
Intrauterine hypoxia	Z	360 (1.1)	1766 (1)	0.002
Intrauterine exposure to noxious agent	Z	368 (1.1)	1689 (1)	0.009
Intrauterine placental insufficiency	Z	581 (1.7)	2554 (1.5)	0.016
Congenital anomalies (%)
Chromosomal anomalies	Z	46 (0.1)	239 (0.1)	−0.001
Heart and circulatory system	Z	1540 (4.5)	5666 (3.3)	0.061
Central nervous system	Z	159 (0.5)	596 (0.4)	0.018
Eye	Z	256 (0.8)	1218 (0.7)	0.004
Genital organs	Z	373 (1.1)	1755 (1)	0.006
Digestive system	Z	582 (1.7)	2456 (1.4)	0.022
Musculoskeletal	Z	683 (2)	3371 (2)	0.002
Respiratory system	Z	212 (0.6)	777 (0.5)	0.023
Skin	Z	511 (1.5)	2408 (1.4)	0.007
Urinary system	Z	221 (0.6)	924 (0.5)	0.014
Other anomalies	Z	127 (0.4)	539 (0.3)	0.01

Notes:

In the Matching Method Column:

E denotes that exposed and unexposed mothers and children were matched exactly for the variable.

D denotes that the distance between exposed and unexposed mothers and children was minimized for the variable. The distance was adjusted by a penalty when adequate balance was not achieved by distance matrix alone and varied by state. Some variables, such as specific Medicaid eligibility classes or rare medical conditions, were not present in all states and therefore had to be removed from the match in some states.

X denotes variables that were measured after exposure and therefore not included in our primary match, but were matched in an additional analysis that matched selected post-exposure covariates.

Z denotes variables that were measured after exposure and not matched.

In the Standardized Differences Column:

Green shading identifies cells where standardized differences were greater than +/− 0.05 but less than +/− 0.1.

Red shading identifies cells where standardized differences were greater than +/− 0.1.

(−) denotes a number <11 that cannot be displayed.

Exposed mothers had eligibility for the full scope of Medicaid benefits for a mean of 4.01 (SD=3.31) months prior to the month of gestation in which they had surgery for appendectomy or cholecystectomy. Unexposed mothers were matched to exposed mothers on that month of gestation and had a mean of 4.08 months of eligibility (SD=3.34) prior to that month. When following exposed and unexposed children after birth, the exposed children were eligible for Medicaid for a mean of 1,056 days and a median of 641 days, while matched unexposed children were eligible for a mean of 1,066 days and a median of 646 days.

Comparing DIBD between Exposed and Unexposed Children

Cumulative probabilities of DIBD diagnosis between exposed and unexposed children were similar in the first three years of life, but diverged thereafter (Figure 2). The 95% confidence intervals of the cumulative probabilities began to overlap after 13 years of age, which is likely to be due to a reduced number of children with continued follow-up beyond that age. Using a Cox proportional hazards model, prenatally exposed children were 31% more likely than unexposed children to receive a DIBD diagnosis (HR, 1.31; [95% confidence interval (CI), 1.23–1.40]) (Figure 3).

Figure 2.

Cumulative Probabilities of Being Diagnosed with a Disruptive or Internalizing Behavioral Disorder for Children Prenatally Exposed and Unexposed to General Anesthesia

Note: Includes the number of subjects at risk at various time points following birth in the exposed and unexposed children as well 95% confidence bands around the cumulative probabilities.

Figure 3.

Hazard of Disruptive or Internalizing Behavioral Disorder Diagnosis and Secondary Outcomes after Prenatal Exposure to General Anesthesia Due to Maternal Appendectomy or Cholecystectomy During Pregnancy

Notes: Exposed and unexposed mothers were exactly matched on male vs. female children, state, and year of birth of the child. Subgroup analyses could therefore be performed on child sex, region of the county, and birth year. Subgroup analyses were also performed on appendectomy vs. cholecystectomy, and trimester of exposure, as these conditions were only present in the exposed children.

In subgroup analyses, the hazard of DIBD following prenatal exposure did not differ based on the sex of the child (exposed boys [HR, 1.37; (95% CI, 1.26–1.48)] vs. exposed girls [HR, 1.23; (95% CI, 1.12–1.36)], p-value for exposure-subgroup interaction = 0.11), maternal procedure type (appendectomy [HR, 1.35; (95% CI, 1.22–1.49] vs. cholecystectomy [HR, 1.29; (95% CI, 1.19–1.40)], p-value for interaction = 0.49), or region of the United States (Northeast [HR, 1.37; (95% CI, 1.10–1.71)] vs. Midwest [HR, 1.33; (95% CI, 1.20–1.47)], p-value for interaction = 0.82, vs. South [HR, 1.28; (95% CI, 1.15–1.41)], p-value for interaction = 0.57, or vs. West [HR, 1.33; (95% CI, 1.15–1.54)], p-value for interaction = 0.84). Hazard of DIBD did however differ based on trimester of exposure with higher hazard in second and third compared to first trimester exposures (first trimester [HR, 1.15; (95% CI, 1.02–1.28)] vs. second trimester [HR, 1.39; (95% CI, 1.27–1.51)], p-value for exposure-subgroup interaction = 0.008, or vs. third trimester [HR, 1.42; (95% CI, 1.23–1.64)], p-value for interaction = 0.02), and year of birth (1999–2003 [HR, 1.47; (95% CI, 1.32–1.64)] vs. 2004–2008 [HR, 1.26; (95% CI, 1.16–1.37)], p-value for interaction = 0.03, or vs. 2009–2013 [HR, 1.17; (95% CI, 0.99–1.39)], p-value for interaction = 0.02). It should be noted that for children born in the 1999–2003 period, the maximum possible follow-up time was nearly 15 years from birth, while for children born in later years, such as the 2009–2013 period, the maximum follow-up time was less than 5 years, resulting in the inclusion of children with ages younger than the full age of risk for initial DIBD diagnoses.

In evaluating the robustness of our results using an asymptotic separable calculation, our primary results may be negated if in a matched pair, if an unobserved confounder triples the odds of prenatal exposure and doubles the odds of DIBD. (Supplemental Analysis in Supplementary Appendix)

Comparing Secondary Outcomes between Exposed and Unexposed Children

As secondary outcomes, increased hazards of disruptive (HR, 1.32; [95% CI, 1.24–1.41]) and internalizing (HR, 1.36; [95% CI, 1.20–1.53]) behavioural disorders were identified. Compared to unexposed children, exposed children also had increased hazards of ADHD (HR, 1.32; [95% CI, 1.22–1.43]), behavioural disorders (HR, 1.28; [95% CI, 1.14–1.42]), developmental speech or language disorders (HR, 1.16; [95% CI, 1.05–1.28]), and autism (HR, 1.31; [95% CI, 1.05–1.64]).

Additional Analyses

After excluding all mothers with comorbid psychiatric diagnoses or psychotropic medication use, 85.7% (n=29,360) of exposed and 85.4% (n=146,262) of unexposed mothers remained, and the association between exposure and DIBD remained virtually unchanged (HR, 1.31; [95% CI, 1.22–1.41]). After excluding all mothers who lacked continuous enrolment or had restricted benefits during pregnancy, 33.3% (n=11,421) of exposed and 26.2% (n=44,822) of unexposed mothers remained, and the association between exposure and DIBD also remained virtually unchanged (HR, 1.28; [95% CI, 1.16–1.41]). The standardized differences for covariates from these analyses can be seen in eTables 8 and 9. In the analysis of mothers without psychiatric disorders, no standardized differences larger than 0.05 were observed, while in the analysis of mothers with continuous enrolment, no standardized differences larger than 0.2 were observed.

After matching on additional post-exposure variables, exposed and unexposed mothers were successfully matched (eTable 10, eFigures 3 and 4) and the primary outcome (HR, 1.29; [95% CI, 1.22–1.38]) and results from all secondary outcomes remained similar to those in the primary match where post-exposure variables were not included (Figure 4). Hazard of DIBD diagnosis was not found to differ based on preterm birth status (preterm [HR, 1.17; (95% CI, 1.03–1.34)] vs. non-preterm [HR, 1.33; (95% CI, 1.24–1.43)], p-value for exposure-subgroup interaction = 0.1).

Figure 4.

Hazard of Disruptive or Internalizing Behavioral Disorder Diagnosis and Secondary Outcomes after Prenatal Exposure to General Anesthesia Due to Maternal Appendectomy or Cholecystectomy During Pregnancy after also matching on Post-Exposure covariates

Notes: Exposed and unexposed mothers were exactly matched on pre-term and non-preterm birth. Subgroup analyses could therefore be performed on preterm vs. non-preterm children.

Discussion

In this national cohort of publicly insured mothers and children, prenatal exposure to general anaesthesia was associated with a 31% increase in risk of DIBD diagnoses in the children, with higher risk following second and third trimester exposures compared to first trimester exposure. Prenatal exposure was also associated with higher risk for a range of secondary neuropsychiatric childhood outcomes.

These results are consistent with the two studies evaluating prenatal exposures to general anaesthesia outside the labour and delivery period. The first evaluated 22 prenatally exposed children and reported significantly more externalizing behavioural problems,³⁶ which may occur at higher frequency in children with disruptive behavioural disorders such as ADHD, A second evaluated 129 pregnant mothers exposed to either general or regional anaesthesia and reported no differences in neurodevelopmental outcomes in their children.³⁷ However, in a subgroup analysis evaluating only the 111 children prenatally exposed to general anaesthesia, statistically significantly more problems with executive function were found.

Limitations

A number of limitations are present in this study. First, it is possible that factors such as infection, inflammation, fever and use of acetaminophen, physiological stressors on the mother due to the surgery, or other peri-operative factors contribute to the increased DIBD risk in prenatally exposed children.^38,39 Second, unmeasured confounders may still be present, and could potentially be responsible for these findings. However, such a confounder would need to be strongly associated with both prenatal exposure and DIBD risk to negate our results. From a causal perspective, it is also reassuring that the association between exposure and DIBD was seen in mothers with appendectomies, a non-elective procedure stemming from a nearly random event.^40,41 Third, for children born in the later years, there was limited follow-up which may have influenced their risk of receiving neuropsychiatric diagnoses. While it is possible that changes in anaesthetic practice between 1999 and 2013 may have contributed to the lower risk seen in children born in recent years, this difference more likely results from an underestimation of risk in those children due to limited follow-up, given that DIBD does not manifest fully in the first three years of life. Fourth, pregnancy loss following surgery may result in selection bias, but given that only 1% of foetuses are lost following appendectomy or cholecystectomy during pregnancy,²³ this is unlikely to bias our results. Fifth, while Medicaid offers health insurance to nearly half of pregnancies and children in the United States,^15,16 these women and children come from lower socioeconomic strata. Therefore, these results may not be generalizable to higher income populations.

A common criticism of studies of post-natal anaesthetic exposures is that children with conditions requiring surgery at a young age have coexisting medical problems that increase their risk of neuropsychiatric deficits. A key strength of our study is that prenatal exposure to anaesthesia results from medical conditions in the mother, so differences in baseline comorbidity in the exposed children and the resulting confounding by indication in the children should be attenuated.

Conclusions

Prenatal exposure to general anaesthesia was associated with a 31% increased risk of disruptive or internalizing behavioural disorders. Caution is advised, as many procedures in pregnant women may be necessary, and avoidance of necessary procedures can have detrimental effects on mothers and their children. However, these results may be considered when performing risk assessments for elective procedures in pregnant women, or when viable alternative treatments are available, such as antibiotics, which are now considered an acceptable first-line treatment for acute appendicitis.⁴²

Supplementary Material

Supplementary Appendix [1]

Supplemental Methods. Matching Algorithm

Supplemental Analysis. Sensitivity Analysis Evaluating a Potential Unobserved Covariate

eTable 1. ICD-9 and CPT Codes Used to Identify Maternal Appendectomy and Cholecystectomy

eTable 2. ICD-9 Diagnosis Codes Used to Identify Disruptive and Internalizing Behavioral Disorders

eTable 3. ICD-9 Diagnosis Codes for Medical Conditions in the Mothers and Children

eTable 4. Psychotropic Medications

eTable 5. ICD-9 and CPT Codes for Specific Post-exposure Variables

eTable 6. Number of Subjects and Percentage Contribution from Each Individual State

eTable 7. Full List of Covariates and Standardized Differences from the Nationwide Match

eTable 8. Full List of Covariates and Standardized Differences from the Nationwide Match Excluding Mothers with Comorbid Psychiatric Diagnoses or Psychotropic Medication Use

eTable 9. Full List of Covariates and Standardized Differences from the Nationwide Match Excluding Mothers Who Lacked Continuous Enrollment or Had Restricted Benefits During Pregnancy

eTable 10. Full List of Covariates and Standardized Differences from the Nationwide Match Including Matching of Post-exposure Covariates

eFigure 1. Standardized Differences in Pre-exposure Covariates Prior to Matching and After Matching from the Nationwide Match

Note: After matching, standardized differences for all matched pre-exposure variables were <0.05.

eFigure 2. Standardized Differences in Post-exposure Covariates Prior to Matching and After Matching from the Nationwide Match

Note: Post-exposure variables were not matched, but the majority still had standardized differences <0.05.

eFigure 3. Standardized Differences in Pre-exposure Covariates Prior to Matching and After Matching from the Nationwide Match Including Matching of Post-exposure Covariates

Note: After matching on pre-exposure variables as well as preterm birth, birth weight categories, and cesarean birth, standardized differences for all matched pre-exposure variables were <0.05.

eFigure 4. Standardized Differences in Post-exposure Covariates Prior to Matching and After Matching from the Nationwide Match Including Matching of Post-exposure Covariates

Note: After matching on pre-exposure variables as well as preterm birth, birth weight categories, and cesarean birth, standardized differences for all matched and unmatched post-exposure variables were <0.05.