Authors: Morgan R. Peltier, Michael J. Fassett, Nana A. Mensah, Nehaa Khadka, Meiyu Yeh, Vicki Y. Chiu, Yinka Oyelese, Darios Getahun
Categories: New Research, autism, pregnancy, perinatal depression, postpartum depression, race/ethnicity
Source: JAACAP Open
Authors: Morgan R. Peltier, Michael J. Fassett, Nana A. Mensah, Nehaa Khadka, Meiyu Yeh, Vicki Y. Chiu, Yinka Oyelese, Darios Getahun
Although maternal psychopathology has long been suggested to increase the risk of neurodevelopmental disorders, whether or not postpartum depression (PPD), a condition diagnosed after delivery, is associated with increased risk of diagnosis of autism spectrum disorders (ASDs) in the offspring, is unclear. Therefore, we tested the hypothesis that PPD diagnosis in the mother would increase the risk of ASD diagnosis in her offspring, and that the association would be independent of gestational age at birth, child’s sex, or race/ethnicity.
We conducted a retrospective cohort study among children born between 2010 and 2021 by examination of individual patient maternal–child linked electronic health records (EHRs) (N = 297,720) from Kaiser Permanente Southern California hospitals. International Classification of Diseases codes listed in the EHR were used to identify diagnosed PPD and ASD cases. Marginal Cox proportional hazard models were fit to evaluate the potential association between maternal PPD diagnosis and the diagnosis of ASD in the offspring. Results are reported as incidence rates and adjusted hazard ratios (HRs) with 95% CIs.
Children of mothers diagnosed with PPD had higher rates of ASD diagnosis than children of mothers without the diagnosis of PPD (9.11 vs 5.48 per 1000 person-years, HR =1.57, CI = 1.49, 1.65). PPD diagnosis in mothers was associated with ASD diagnosis in the offspring for both preterm and term-born children, boys as well as girls, and no strong racial/ethnic heterogeneity in the association was detected.
Postpartum depression in the mother is associated with an increased risk of ASD diagnosis in her child, independent of gestational age at birth, child sex, and race/ethnicity.
Postpartum depression (PPD) consists of symptoms of depression that usually occur within the first 6 weeks after delivery but that can occur at any time during the first year postpartum. Worldwide it is estimated to affect 10% to 15% of pregnancies^1^; however, this is likely an underestimate, due to a lack of universal screening. In a recent study of a large health care system, only about 66% of women were screened, with ethnic minority women being significantly less likely to be screened,^2^ despite the demonstrated effectiveness of screening instruments at detecting subclinical depression in peripartum women and reducing medical costs.3, 4, 5, 6, 7 Untreated PPD is associated with impaired maternal bonding,^8^^,^^9^ childhood developmental delays, reduced measures of intelligence,^10^ childhood behavior problems, and reduced vocabulary at 5 years of age,^11^ and, in severe cases, suicide,^12^^,^^13^ infanticide,^3^ and infanticide ideation.^14^
PPD is associated with a large number of risk factors that include the socioeconomic status,^15^^,^^16^ quality of marital relationship,^15^^,^^16^ immigration status,17, 18, 19, 20 history of domestic violence/abuse,17, 18, 19, 20, 21 smoking,^22^ cesarean or instrumented delivery,^23^ obstetric complications,24, 25, 26, 27 preeclampsia, previous history or family history of psychological illness,^20^^,^^28^^,^^29^ and adolescent pregnancy.^30^^,^^31^ The etiological factors behind PPD are elusive, but recent studies have suggested that aberrant inflammation during pregnancy may be a contributing factor.32, 33, 34, 35, 36 PPD is associated with high concentrations of proinflammatory cytokines such as interleukin (IL)–1β,^37^ IL-6,^32^^,^38, 39, 40, 41 and C-reactive protein.^42^^,^^43^
Autism spectrum disorders (ASDs) are a common neurodevelopmental problem that is characterized by social and communicative defects as well as repetitive/restrictive behaviors. Their prevalence has grown significantly over the past 50 years. This may be due to better diagnosis, screening, as well as recognizing that the degree to which individuals are affected varies widely. A recent global analysis suggested that the median prevalence in 71 studies was 1% but varied between 0.01 and 4.36%.^44^ Autism occurs about 4 times more frequently in male than in female individuals and is often comorbid with other neurodevelopmental disorders such as intellectual disability, epilepsy, and attention-deficit/hyperactivity disorder (reviewed in Zeidan et al.^44^). Although the condition has a strong genetic component (correlations of 0.98 for monozygotic twins, 0.53 for dizygotic twins,^45^ and a heritability that is as high as 90%^46^), epigenetic46, 47, 48 and environmental causes such as exposure to certain drugs,^49^ nutritional deficiencies,^50^ and obesity^51^ have also been suggested as potential risk factors.52, 53, 54 Maternal immune activation and inflammation during pregnancy have also long been associated with the development of autism in the offspring. Women who experienced infections during pregnancy were at significantly increased risk for having a child with autism,^55^ and inflammation-linked pregnancy complications increase the risk of ASD diagnosis in the offspring.56, 57, 58, 59 Experimental studies with mice have also demonstrated that exposure to viral infections during pregnancy results in behavioral, histologic, and neurological traits that are consistent with ASD that appear to be mediated through IL-6.^60^ IL-6 and other proinflammatory cytokines are known to interfere with the synthesis and metabolism of thyroid hormones61, 62, 63, 64 and to reduce the expression of biologically active tri-iodothyronine.^65^ Thyroid hormones increase the expression of neurotrophic factors, such as brain-derived neurotrophic factor (BDNF), that regulate the differentiation and myelination of neurons and the formation of synapses.^66^^,^^67^ Dysregulation of thyroid hormones in pregnancy68, 69, 70 or BDNF levels^71^^,^^72^ has been associated with ASD.
PPD has also been associated with thyroid dysfunction,73, 74, 75, 76, 77, 78, 79 BDNF deficiencies,^80^^,^^81^ inflammation,^34^^,^^82^ and obstetric conditions such as preterm birth^83^^,^^84^ or preeclampsia.^84^^,^^85^ Although PPD has risk biomarkers similar to those of ASD, it is unclear whether women who experience PPD are at increased risk for having children diagnosed with ASD, and how such a risk could be modified by other demographic or obstetric factors. A recent study performed in Taiwan reported that PPD diagnosis in the mother significantly increased the risk of ASD diagnosis in the offspring,^86^ but no such studies have been performed in the United States, where there is significant racial, ethnic, and socioeconomic diversity.
Kaiser Permanente Southern California uses universal screening protocols for neurodevelopmental disorders across all of its 15 medical centers and 236 medical offices. Through the use of standardized screening and diagnostic procedures, every child in this system is at equivalent baseline risk for being diagnosed with ASD by medical specialists trained to diagnose the condition. Therefore, we performed a study using electronic health records (EHRs) from this large health care system to examine whether PPD diagnosis is independently associated with ASD diagnosis in the offspring.
This study was conducted with the approval and oversight of the Kaiser Permanente Southern California Institutional Review Board with a waiver of informed consent.
We conducted a retrospective cohort study using individual EHRs of pregnant women and their children (N = 447,800) born between January 1, 2010, and December 31, 2021, at all Kaiser Permanente Southern California (KPSC) hospitals. Unique identifiers were used to link the child’s medical records with their biological mothers longitudinally. The linked data contained information on maternal sociodemographic and behavioral characteristics, maternal medical and obstetric history, along with complete child health care information. All data were collected during the provision of routine clinical care (ie, there was no enrollment or recruitment of patients). These individual-level data provided by the EHRs for every member of our patient population has many advantages over data collected for administrative databases or patient registries that are subject to loss of information by summarizing data, selection bias, and misclassification of exposures and outcomes. Furthermore, because this study is a longitudinal design, all maternal data were collected long before the child was diagnosed, eliminating the risk of recall bias.
Records from all KPSC deliveries at 28 or more weeks’ gestation to women who had been members for at least 90 days were used as the starting point for this study. Records in which there was a neonatal death or missing outcome were excluded, as were pregnancies in which there was a multiple gestation or in which the mother was not a KPSC member. Records in which the child was disenrolled from the insurance program prior to 2 years of age were also excluded, as were patients who did not remain a KPSC member for more than 90 days after the child reached 2 years of age. Therefore, our final cohort (N = 297,720) comprised children who were born to KPSC members from a live singleton birth at 28^0/7^ to 42^6/7^ weeks’ gestation and who remained in the program for a minimum of 90 days when they were between the ages of 2 and 17 years (Figure 1). Validation of these methods for linking maternal–child data has been previously reported.^87^^,^^88^ These exclusion criteria were necessary because exposures, outcomes, and confounders cannot be adequately measured in patients who were members for less than 3 months. Furthermore, children under 2 years of age cannot be reliably diagnosed with ASD. Because these individuals are not at risk for being diagnosed with PPD or ASD, they were removed from the risk pool.Figure 1Flow Chart Documenting Our Initial Patient Population at Kaiser Permanente Southern California (KPSC) and the Exclusion Criteria Applied to Create the Final Study Cohort***Note:***PPD = postpartum depression.
International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) and International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnosis and procedural codes from inpatient and outpatient physician encounters, laboratory results, and pharmacy records were used to ascertain the exposure and outcome of interest.
The primary exposure of interest was the clinical diagnosis of PPD using ICD-9 and ICD-10 (since October 1, 2015) codes recorded in the patient’s EHR at the date of diagnosis that included the codes 300.4, 309.0, 311, F32.9, F33.0, F33.2, F33.3, F33.41, F33.9, F34.1, F43.21, and F53.0, or if they presented with depressive symptoms within 1 year after delivery and were prescribed antidepressants or other medications. In the KPSC setting, PPD is assessed using the Edinburgh Postnatal Depression Scale (EPDS). Patients were screened by social workers for PPD, and those who screened positive (10+) were further evaluated by a mental health specialist by whom they were formally diagnosed with PPD. The validity of the clinical diagnosis of PPD in the EHR was reported in previous publications to have a sensitivity of 98.3%, specificity of 83.3%, positive predictive value of 93.7%, and negative predictive value of 95.0% when compared to manual chart review.^89^
ICD-9 and ICD-10 (since October 1, 2015) codes at the date of diagnosis in the EHR were also used to identify cases of ASD that included the 299.0, 299.00, 299.01, 299.10, 299.11, 299.80, 299.81, 299.90, 299.91, 299.8, 299.9, F84.0, F84.5, F84.8, and F84.9. Diagnosis of ASD was made using the DSM-IV-TR for any of the following autistic disorder, childhood disintegrative disorder, Rett disorder, Asperger disorder, or pervasive developmental disorder–not otherwise specified (PDD-NOS). Children and adolescents 2 to 17 years of age with at least one documented DSM-IV-TR code for ASD on any 2 separate visits during the follow-up period were considered ASD cases. Clinical experts have validated the accuracy of these codes through medical record review (gold standard); ASD cases were identified with 100% sensitivity, 100% specificity, 100% negative predictive value, and 99.2% positive predictive value.^90^ Under the KPSC guidelines, the medical plan covers ASD services that require evaluation by professionals such as a child/adolescent psychiatrist, developmental/behavioral pediatrician, child psychologist, or neurologist, or contracted through a trained physician or psychologist. Almost all of the children with ASD (96%) were diagnosed by experts and received subsequent consultations by doctors within our system. The remaining 4% of ASD case patients were first diagnosed outside our health care system but whose ASD diagnosis was confirmed by specialists in our setting.^91^ All KPSC pediatricians are required to perform a series of behavioral and developmental surveillance at all well-child visits as early as 4 months of age. If there is a sufficient degree of suspicion, a modified version of the Checklist for Autism in Toddlers (M-CHAT) and developmental screening questionnaires for toddlers will be completed (as early as 18 months of age). Because all diagnoses were made by experts in diagnosing and treating ASD in a single large health care system with standardized protocols, all autistic children in the cohort are at equal probability of being diagnosed.^91^
Potential confounders/modifiers in this study include child’s race/ethnicity and sex, maternal age (<25, 25-34, ≥35 years), education (<12, 12, ≥13 years completed), median family household income based on census tract of residence, timing of prenatal care initiation, smoking during pregnancy (yes/no), gestational age at delivery based on clinical estimates, and maternal comorbidities including maternal diabetes and substance abuse. Maternal and paternal race/ethnicity defined the child’s race/ethnicity. A child was categorized as non-Hispanic White (White) if born to a non-Hispanic White mother and father. The same applied to non-Hispanic Black (Black), Hispanic, and Asian/Pacific Islander race/ethnicity groups. The Other/Multiple race/ethnicity category includes children born from interracial/interethnic relationships.
Descriptive statistics regarding our patient population were obtained from the EHRs, and the association between maternal PPD diagnosis and risk of ASD diagnosis in the offspring was evaluated using survival methods. Follow-up of children started from the delivery date until the date of ASD diagnosis or censoring because of health plan disenrollment, 17th birthday, non-ASD–related death, or the end of the study. The hazard of the child being diagnosed with autism was examined using marginal Cox proportional hazard models. These account for the clustering of outcomes in women with 2 or more pregnancies in the cohort when estimating the 95% CIs. Results are presented as crude and adjusted HR with 95% CI, which reflect the relative instantaneous risk of a person being diagnosed with ASD at a given time point, to quantify the magnitude of the associations. Incidence rates of ASD are also presented as an index of absolute risk. Incidence rates reflect the number of cases observed in a population of individuals over a period of time. These adjust for the number of years that a person is followed and at risk for being diagnosed with ASD. Because the observation time varies from individual to individual, it is summed for the entire population and reported as events per person-years of observation.
Women whose children were diagnosed with ASD tended to be older, to be more highly educated, and to have incomes between USD 30,000 and 70,000 (Table 1). They were also more likely to be nulliparous and to have delivered preterm. No difference between the time of the initiation of prenatal care or maternal smoking was detected. Ethnic minority children were slightly more likely to be diagnosed with ASD in our cohort (3.70% vs 2.65% for ethnic minority and White children, respectively), and the sex distribution was significantly skewed toward male children (5.33% vs 1.68% for boys and girls respectively). ASD was also more frequently diagnosed in children whose mothers were diagnosed with type I/II (3.83% vs 3.54%, with and without type I/II diabetes) or for patients with and without gestational diabetes (4.24% vs 3.44%, respectively). No association between recreational drug use during pregnancy and rates of ASD diagnosis in the offspring was detected (Table 1).Table 1Baseline Characteristics by Autism Spectrum Disorder (ASD) StatusTotal N = 297,720 n (%)No ASD n = 287,153 n (%)ASD n = 10,567 n (%)pMaternal age, y**<.001** <208,081 (2.7)7,827 (2.7)254 (2.4) 20-29117,187 (39.4)113,217 (39.4)3,970 (37.6) 30-34100,908 (33.9)97,468 (33.9)3,440 (32.6) ≥3571,544 (24.0)68,641 (23.9)2,903 (27.5)Maternal education**<.001** Less than high school13,036 (4.4)12,587 (4.4)449 (4.2) High school graduate66,267 (22.3)63,827 (22.2)2,440 (23.1) Some college65,754 (22.1)63,279 (22.0)2,475 (23.4) Associate/Bachelor’s degree100,511 (33.8)96,914 (33.7)3,597 (34.0) Master's degree/above46,255 (15.5)44,864 (15.6)1,391 (13.2) Missing5,897 (2.0)5,682 (2.0)215 (2.0)Household income, USDa**<.001** <30,000-50,000-70,000-90,00071,124 (23.9)68,969 (24.0)2,155 (20.4) Missing245 (0.1)240 (0.1)5 (0.0)Parity**<.001** Multiparous169,637 (57.0)164,359 (57.2)5,278 (49.9) Nulliparous89,832 (30.2)86,043 (30.0)3,789 (35.9) Missing38,251 (12.8)36,751 (12.8)1,500 (14.2)Gestational age, wk**<.001** 20-281,293 (0.4)1,150 (0.4)143 (1.4) 29-322,390 (0.8)2,246 (0.8)144 (1.4) 33-3618,268 (6.1)17,383 (6.1)885 (8.4) Term birth275,769 (92.6)266,374 (92.8)9,395 (88.9)Smoking during pregnancy0.124 No290,477 (97.6)280,191 (97.6)10,286 (97.3) Yes7,243 (2.4)6,962 (2.4)281 (2.7)Initiation of prenatal care0.662 ≤3 months25,4859 (85.6)245,837 (85.6)9,022 (85.4) Late/no care38,563 (13.0)37,180 (12.9)1,383 (13.1) Missing4,298 (1.4)4,136 (1.4)162 (1.5)Child's race/ethnicity**<.001** Non-Hispanic White49,334 (16.6)48,027 (16.7)1,307 (12.4) Non-Hispanic Black15,673 (5.3)15,023 (5.2)650 (6.2) Hispanic122,463 (41.1)117,988 (41.1)4,475 (42.3) Asian/Pacific Islander28,601 (9.6)27,465 (9.6)1,136 (10.8) Other/multiple70,109 (23.5)67,609 (23.5)2,500 (23.7) Unknown11,540 (3.9)11,041 (3.8)499 (4.7)Child sex**<.001** Female145,233 (48.8)142,789 (49.7)2,444 (23.1) Male152,487 (51.2)144,364 (50.3)8,123 (76.9)Recreational drugs0.114No28,6951 (96.4)276,796 (96.4)10,155 (96.1)Yes10,769 (3.6)10,357 (3.6)412 (3.9)Type I/II diabetes**<.001No291,617 (98.0)28,1398 (98.0)10,219 (96.7)Yes6,103 (2.0)5,755 (2.0)348 (3.3)Gestational diabetes mellitus<.001**No257,119 (86.4)248,272 (86.5)8,847 (83.7)Yes40,601 (13.6)38,881 (13.5)1,720 (16.3)Note:aMedian household income based on census tract information with inflation adjustment. Statistically significant p-values are indicated in bold type.
There was a 1.66-fold higher incidence and 1.57-fold increased hazard of ASD diagnosis for children born to women whose pregnancies were complicated by PPD. This association remained significant and increased to 1.77-fold after adjustment for potential covariates (Table 2). Kaplan–Maier survival curves also indicated that the number of offspring diagnosed with ASD accumulated more rapidly for children of women who experienced PPD (Figure 2), suggesting that maternal PPD is independently associated with an increased risk of ASD diagnosis in the offspring. PPD was also associated with increased risk of ASD diagnosis among children of women who delivered at extreme (<28 weeks), severe (29-32 weeks), or moderate (33-36 weeks) preterm gestation by 2.27-, 1.83-, and 1.61-fold, respectively (Table 3). Tests for interaction between PPD diagnosis and gestational age at delivery on the association of ASD diagnosis in the offspring did not reach or attain statistical significance in either the crude or the adjusted analyses (p = .357 and p = .347, respectively). This suggests that the correlation of PPD diagnosis in the mother with ASD diagnosis in her child is not strongly modified by preterm delivery.Table 2Association Between Maternal Postpartum Depression (PPD) and Autism Spectrum Disorder (ASD) Incidence and RiskHR (95% CI)GroupnPerson-yearsIR (‰)CrudeAdjustedaNo PPD8,6421,578,1225.481.00 (Reference)1.00 (Reference)PPD1,925211,2429.11**1.57 (1.49, 1.65)*1.77 (1.60, 1.95)Note: Statistically significant associations are displayed in bold type. IR = incidence rates per 1000 person-years; HR = hazard ratio.aAdjustments were made for maternal age, maternal education, child’s race/ethnicity, gestational age at delivery, median household income, parity, prenatal care, smoking during pregnancy, child’s sex, drug use during pregnancy, type I/II diabetes, gestational diabetes mellitus, and an interaction term between PPD and child’s sex.Figure 2Kaplan–Maier Survival Curves for the Proportion of Children Diagnosed With Autism as a Function of Time When Born to Mothers With Postpartum Depression (PPD) and Without PPDNote:****Overall trends (A) and trends stratified by child sex (B) are shown.*Table 3Association Between Maternal Postpartum Depression (PPD) and Autism Spectrum Disorder (ASD) Incidence and Risk by Gestational AgeGestational Age, y, at deliveryPPD statusBirths, nASD, nPerson-yearsIR (‰)HR (95% CI)CrudeAdjusteda20-28No PPD9881016,04816.701.00 (Reference)1.00 (Reference)PPD305421,53727.331.55 (1.09, 2.21)****2.27 (1.24, 4.17)29-32No PPD1,97611912,2649.701.00 (Reference)1.00 (Reference)PPD414252,31110.821.06 (0.69, 1.63)1.83 (0.89, 3.75)33-36No PPD15,41870196,2647.281.00 (Reference)1.00 (Reference)PPD2,85018415,72411.701.51 (1.29, 1.78)****1.61 (1.16, 2.23)Term birthNo PPD240,1957,7211,463,5465.281.00 (Reference)1.00 (Reference)PPD35,5741,674191,6708.731.56 (1.48, 1.65)****1.76 (1.58, 1.96)**Note: Statistically significant associations are indicated in bold type. IR = incidence rates per 1000 person-years; HR = hazard ratio.aAdjustments were made for maternal age, maternal education, child’s race/ethnicity, median household income, parity, prenatal care, smoking during pregnancy, child’s sex, drug use during pregnancy, type I/II diabetes, gestational diabetes mellitus, and interaction term of PPD and child’s sex.
PPD was associated with increased hazard for ASD diagnosis for both boys and girls, but the difference in incidence rates between women whose pregnancies were complicated by PPD was 2.3 times higher for boys (13.48 − 8.35 = 5.13) than it was for girls (4.70 − 2.51 = 2.19). Furthermore, a significant PPD by child sex interaction was found in both the crude (p = .012) and adjusted (p = .010) analyses. This suggests that PPD may be more strongly associated with ASD in boys than in girls (Table 4). Analysis of data stratified by race/ethnicity also suggested that PPD was associated with diagnosis of ASD across all racial/ethnic groups studied to nearly an equivalent degree. PPD was associated with increasing the hazard of ASD diagnosis between 1.39- and 1.81-fold, and incidence rate differences (between PPD and not PPD) that ranged from 3.15 (for Black race) to 5.46 (unknown race). As with the previous analyses, the associations were largely unaffected by confounding variables (Table 5) but the hazards ratios for Black individuals and Asian/Pacific Islanders did not reach statistical significance after adjustment for confounders. No significant interaction between child race/ethnicity and maternal PPD rates of ASD diagnosis in the offspring was detected in either the crude (p = .288) or adjusted (p = .277) analyses. This suggests that the association of maternal PPD with risk of her child later being diagnosed with ASD was similar for all racial/ethnic groups studied.Table 4Association Between Maternal Postpartum Depression (PPD) and Autism Spectrum Disorder (ASD) Incidence and Risk by Child’s SexChild’s sexPPD statusBirths, nASD, nPerson-yearsIR (‰)HR (95% CI)CrudeAdjustedaFemaleNo PPD126,0261,951776,4772.511.00 (Reference)1.00 (Reference)PPD19,207493104,9894.70**1.76 (1.59, 1.94)****1.75 (1.58, 1.93)MaleNo PPD132,5516,691801,6458.351.00 (Reference)1.00 (Reference)PPD19,9361,432106,25313.481.52 (1.44, 1.61)****1.53 (1.44, 1.62)Note: Statistically significant associations are indicated in bold type. IR = incidence rates per 1000 person-years; HR = hazard ratio.aAdjustments were made for maternal age, maternal education, child’s race/ethnicity, gestational age at delivery, median household income, parity, prenatal care, smoking during pregnancy, drug during pregnancy, type I/II diabetes, and gestational diabetes mellitus.Table 5Association Between Maternal Postpartum Depression (PPD) and Autism Spectrum Disorder (ASD) Incidence and Risk by Child’s Race/EthnicityRace/ethnicityPPD statusBirths, nASD, nPerson-yearsIR (‰)HR (95% CI)CrudeAdjustedaNon-Hispanic WhiteNo PPD41,627991250,1363.961.00 (Reference)1.00 (Reference)PPD7,70731642,8487.371.81 (1.59, 2.05)****2.30 (1.78, 2.96)Non-Hispanic BlackNo PPD13,43753587,0936.141.00 (Reference)1.00 (Reference)PPD2,23611512,3769.291.39 (1.14, 1.70)1.24 (0.83, 1.86)HispanicNo PPD107,0603,699661,8405.591.00 (Reference)1.00 (Reference)PPD15,40377683,7989.261.56 (1.44, 1.68)****1.71 (1.47, 2.00)Asian/Pacific IslanderNo PPD26,9341,040161,6226.431.00 (Reference)1.00 (Reference)PPD1,667968,19211.721.64 (1.33, 2.02)1.46 (0.91, 2.33)Other/multipleNo PPD59,9612,003360,1265.561.00 (Reference)1.00 (Reference)PPD10,14849753,6039.271.57 (1.42, 1.74)****1.76 (1.44, 2.14)UnknownNo PPD9,55837457,3056.531.00 (Reference)1.00 (Reference)PPD1,98212510,42511.991.72 (1.40, 2.10)****1.87 (1.27, 2.79)**Note: IR = incidence rates per 1000 person-years; HR = hazard ratio.aAdjustments were made for maternal age, maternal education, gestational age, median household income, parity, prenatal care, smoking during pregnancy, child’s sex, drug during pregnancy, type I/II diabetes, gestational diabetes mellitus, and the interaction term of PPD and child’s sex. Statistically significant associations are displayed in bold type.
Electronic health records, coupled with large health care systems that use consistent diagnostic and treatment protocols across their network, have greatly advanced the field of epidemiology. These large databases with structured data now make it possible to perform longitudinal studies to identify risk factors for disorders for conditions that emerge years after exposure. This use of individual data collected under standardized conditions has many advantages over administrative data and registries that summarize outcomes at institutions, and are less subject to selection bias and patient misclassification. Furthermore, such large databases make it possible to study conditions such as PPD and ASD where there are large racial/ethnic disparities in prevalence that could be partially due to inequities in health care access as well as social determinants of health. By understanding the role of race/ethnicity and other social determinants of health through large population-based studies, it may be possible to reduce the disparities in underserved populations.
Approximately 13.2% of the women in our cohort were diagnosed with PPD, and this is consistent with the yearly incidence rate of 10% to 20% reported for most studies.^92^ Our overall ASD incidence rate of 5.9 cases per 1,000 person-years was slightly higher than a recently published longitudinal study in the United Kingdom of more than 7 million children 429.1 per 100,000 person-years.^93^ Our rates may be higher because we relied on specialist-diagnosed cases in EHR data rather than what was recorded in school databases, where there could have been underreporting of milder cases.
Our findings suggest that PPD diagnosis in the mother is a risk factor for ASD diagnosis in the offspring, and that the association occurs for both boys and girls, term and preterm infants, and is largely unaffected by child race/ethnicity. These findings are consistent with a recent study performed in Taiwan that found that PPD was associated with the development of ASD in the offspring.^86^ To our knowledge (and a PubMed search conducted on November 21, 2023), this is the first study performed in the United States to address this question and to evaluate potential risk modifiers. As expected, we found differences in the risk of ASD due to race/ethnicity and gestational age at delivery, but these factors did not appear to be strong risk modifiers for the observed association between maternal PPD and increased risk of ASD diagnosis in the offspring. Although the hazard ratios were similar for both boys and girls, the incidence rate difference between PPD and non-PPD mothers for ASD was much larger for boys than it was for girls. This suggests that male sex may be associated with greater risk of ASD diagnosis in children whose mothers were diagnosed with PPD than for female sex.
Further studies need to be performed to confirm these findings and to identify the potential mechanism(s) or reasons why maternal PPD is associated with increased risk of ASD diagnosis in the offspring. Previous studies have demonstrated that autistic women are more likely to suffer from PPD than non-autistic women.^94^ Women who display a broader autistic phenotype in Japan were also reported to be at increased risk for PPD.^95^ Therefore, women with PPD may be more likely to have autistic traits that are inherited to some extent by their children, which would explain our observed associations. It is also possible that PPD and ASD may be different consequences of exposure to inflammation and/or dysregulation of cytokines and hormones during pregnancy. If that is the case, then it would be expected that these conditions would cooccur more frequently than what would be predicted by random chance. Previous studies have demonstrated that dysregulation of biomarkers for inflammation,96, 97, 98, 99 hypothyroidism,^100^^,^^101^ and neurodevelopment^71^^,^^101^^,^^102^ are associated with increased risk of both PPD and ASD symptoms or diagnosis. Recent studies have also suggested that low concentrations of allopregnanolone (AP), a neurosteroid that has a key role in the development of PPD103, 104, 105 and is now approved by the US Food and Drug Administration for treatment of the condition^106^ may also be part of the pathophysiology of ASD. A combination of clinical-, animal-, and laboratory-based studies are needed to explore these possible biomarkers and their potential role in the observed association between PPD diagnosis in the mother and ASD diagnosis in her offspring.
Strengths of our study include a large number of both PPD and ASD cases from a well-defined cohort of women and their children using individual-level data collected during the provision of routine care. Because data were collected longitudinally, no subjects needed to be enrolled, and because the exposure was documented prior to diagnosis of the outcome, there was no risk for recall bias. Furthermore, by implementing consistent screening and diagnostic procedures, all children were at equal risk for being diagnosed due to medical surveillance. This is far more reliable than parent- or teacher-report for screening, and with all cases being formally diagnosed by specialists greatly reducing the risk for patient misclassification. Our findings that PPD is a potential risk factor for ASD diagnosis in the offspring held up after sensitivity analyses in which we considered only those cases diagnosed by psychiatrists and psychologists (HR = 1.63, 95% CI = 1.49,1.65; adjHR = 1.59, 95% CI = 1.51, 1.67) and to women who had only one pregnancy during their time in the cohort (HR = 1.57, 95% CI = 1.48,1.66; adjHR = 1.58, 95% CI = 1.50, 1.67). This suggests that our findings are robust. Furthermore, the annual incidence rate of PPD and the incident rate for ASD were consistent in our population with previous studies, suggesting that our findings are likely generalizable to other patient populations.
Our findings are not without limitations, however. First, we did not have ready access to the results of screening instruments that could have been useful for identifying dose-dependent relationships between maternal PPD and the severity of symptoms or the presence of co-morbidities (eg, intellectual disability) in the offspring diagnosed with ASD. We were also unable to identify the mechanism(s) by which a diagnosis of PPD could be associated with ASD diagnosis in the offspring. This, however, is very difficult to do with a retrospective cohort design because the impact that unmeasured confounders can have on the observed associations is difficult to assess. For example, antidepressants such as fluoxetine are among the most frequently prescribed medications for pregnant women.^107^ The possibility that antidepressants increase the risk of ASD diagnosis in the offspring has been proposed, but it has been difficult to prove because of the challenge of separating the impact of medications from the underlying maternal depression.^108^^,^^109^ We are similarly limited by our ability to discern PPD from the conditions in pregnancy that preceded it, such as antenatal depression or thyroid disorders that could be completely responsible for the observed association of PPD diagnosis in the mother with ASD diagnosis in her child.
In conclusion, we found that diagnosis of PPD in the mother is associated with increased risk of ASD diagnosis in the offspring, and that this association is consistent for both boys and girls, regardless of preterm birth status or race/ethnicity.
Morgan R. Peltier: Writing – review & editing, Writing – original draft, Visualization, Investigation. Michael J. Fassett: Writing – original draft, Resources. Nana A. Mensah: Methodology, Investigation. Nehaa Khadka: Methodology, Investigation. Meiyu Yeh: Methodology, Investigation, Formal analysis, Data curation. Vicki Y. Chiu: Software, Investigation, Formal analysis. Yinka Oyelese: Writing – review & editing, Investigation. Darios Getahun: Writing – review & editing, Writing – original draft, Supervision, Resources, Project administration, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.