Steroidopathies and hormonal imbalance in children and adolescents with autism spectrum disorder

Concetta de Giambattista : Conceptualization; methodology; writing—original draft; data curation; investigation. Patrizia Ventura : Methodology; investigation; writing—review and editing. Samuele Cortese : Supervision; writing—review and editing; conceptualization. Paolo Trerotoli : Methodology; validation; writing—review and editing. Alessandra Di Gioia : Investigation; writing—review and editing. Lucia Margari : Conceptualization; methodology; writing—original draft; writing—review and editing; funding acquisition; investigation.

The study was conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from all participants and/or their parents or legal guardians. The research protocol was approved by the Ethics Committee of the Azienda Ospedaliera‐Universitaria Consorziale Policlinico di Bari (Italy) on 16 February 2023 (approval number 7573).

In an attempt to create homogeneous groups, we consecutively recruited individuals referred to the Neuropsychiatry Unit of the Azienda Ospedaliero‐Universitaria Policlinico di Bari between January and October 2024. Inclusion criteria were an age range of 9–18 years and a diagnosis of neuropsychiatric disorders, including ASD, or other neurodevelopmental and psychiatric conditions such as attention deficit and hyperactivity disorder (ADHD), anxiety and depression, and conduct disorder. The Autism Group (AG) was recruited consecutively, including all individuals with an ASD diagnosis who met the inclusion criteria, until a total of 50 participants was reached. Recruitment ensured an equal sex distribution, with 25 males and 25 females enrolled in the order of their referral. The Neuropsychiatric Group (NPG) was then composed of the first 50 consecutive inpatients who met the inclusion criteria, had no ASD diagnosis, and concurrently matched the AG participants by age and sex. This procedure ensured demographic comparability between the two groups while preserving the consecutive recruitment within each group. Information on perinatal history, including mode of delivery (spontaneous vaginal delivery, elective caesarean section, or medically indicated caesarean section), was systematically collected for all participants. At the time of recruitment, information on current medication use was collected. Individuals receiving hormonal therapy were excluded from the study. Furthermore, in the analyzed sample, no statistically significant differences were observed between the AG and NPG groups with respect to the use of other medications. Detailed information about the purpose and the method of the study was provided to the participants and their families. Besides, written informed consent from parents was obtained before enrollment. All participants were diagnosed through a complete neuropsychiatric evaluation by Neuropsychiatrist specialists belonging to the research team, employing standard assessment (including Child Behavior Checklist, CBCL, and Youth Self Report Form, YSR) supplemented by Baron‐Cohen (BC) questionnaires (Baron‐Cohen et al.,  2001 , 2003 ; Baron‐Cohen & Wheelwright,  2004 ) and the semi‐structured questionnaire inspired by Testosterone‐related Medical Questionnaire (Ingudomnukul et al.,  2007 ) (Supporting Information  S1 : Appendix S1), from which we extracted the items most pertinent to our research objectives. The study was approved by the Ethics Committee of the Azienda Ospedaliero‐Universitaria Consorziale Policlinico di Bari (study code: NEUROSVI‐2022, protocol number: 0015608). Informed consent was obtained from the parents of all participants. Data on steroidopathies were collected through anamnestic interviews, supported by the semi‐structured questionnaire adapted from the Testosterone‐Related Medical Questionnaire. Participants were asked whether any first‐degree relatives had ever experienced hormonal or gynecological issues (e.g., PCOS, endometriosis, precocious or delayed puberty, breast, ovarian, uterine or prostate cancer, metabolic diseases such as gestational diabetes). Maternal medical history was also recorded, including the use of hormonal treatments, diabetes, preeclampsia, thyroid disorders, stress, smoking, medication use, and assisted reproductive technologies during pregnancy, as well as delivery complications. Data were also collected on maternal and paternal ages and breastfeeding duration. Moreover, females were asked whether they had hirsutism, severe acne, irregular menstrual cycles, painful periods, or excessive menstrual bleeding. Hormonal and Metabolic Blood Tests: Blood samples were collected from participants after fasting between 08:00 and 10:00 AM. The following hormone levels were measured: estradiol, prolactin, testosterone, folate, insulin, cortisol, and vitamin D. Electrochemiluminescence and chemiluminescence methods were used for the analysis. Quantitative variables (age, hormone levels, tests specific for autism, and tests specific for other co‐occurring conditions) were summarized as mean and standard deviation or median and interquartile ranges after assessing normality using the Shapiro–Wilk test. Independent group comparisons were conducted using t ‐tests (parametric) or Wilcoxon tests (nonparametric); Hedges' effect size was determined in both cases to evaluate the magnitude of the difference. Association between variables was assessed using Spearman's rank coefficient, as many variables did not sufficiently approximate the Gaussian distribution. Qualitative characteristics (sex, presence/absence of other diagnoses) were summarized as counts and percentages, and comparisons between independent groups were made using a chi‐square test or Fisher's Exact test, as appropriate. To evaluate the magnitude of the effect, Cramer's effect size was determined. For quantitative variables that showed significant differences between the AG and NPG, an analysis was also carried out to determine whether it was possible to associate the outcome of the characteristic with the presence of ASD. In the case of qualitative features, for this purpose, logistic regression and ROC curve analysis were used, with determination of the area under the curve (AUC) and associated confidence interval. To identify the characteristics that best predicted the risk of ASD, a logistic regression analysis was performed, with ASD status as the dependent variable. First, a univariable analysis was conducted including all variables as independent variables; subsequently, a multivariable model was analyzed, with independent variables chosen among those that were statistically significant in the univariable analysis (these were testosterone level, medication use during pregnancy, and presence of thyroid disease). Sex and age were included as covariates. The analyses were conducted with SAS 9.4 personal computer software. A p ‐value of < 0.05 was considered the threshold for defining statistical significance.

The association between ASD and elevated testosterone levels suggests a potential biomarker for early identification or risk stratification. Furthermore, the role of maternal hormonal imbalance and drug exposure during pregnancy as potential risk factors highlights the need for careful prenatal monitoring and risk evaluation.

The demographic characteristics (age and sex; Table  1 ) did not significantly differ between the AG and the NPG. Similarly, scores on the BC questionnaires (autistic quotient, AQ; empathy quotient, EQ; systemizing quotient, SQ; as well as on the Child Behavior Checklist (CBCL) and the Youth Self Report (YSR; Table  1 )), showed no significant differences between groups, with the exception of the CBCL subscales for Oppositional Defiant Disorder and Conduct Disorder, which were significantly higher in the NPG. Demographic characteristics and scores obtained by the two groups, AG and NPG, on the AQ, EQ, SQ, CBCL and YSR questionnaires. Abbreviations: AG, Autism Group; AQ, autistic quotient; CBCL, Child Behavior Checklist; EQ, empathizing quotient; IQR, InterQuartile Range; NPG, NeuroPsychiatric Group; SD, standard deviation; SQ, systematizing quotient; YSR, Youth Self Report. No significant differences were found between AG and NPG with respect to mode of delivery. The frequency distribution of spontaneous vaginal delivery (40.4% vs. 55.8%), scheduled caesarean section (25.5% vs. 20.9%), and emergency caesarean section (23.4% vs. 14.0%) was comparable across the two groups and was further confirmed by logistic regression, where the association between type of delivery and AG/NPG status was not statistically significant (Type III analysis of effects, Wald chi‐square = 2.38, df = 3, p  = 0.4965; Supporting Information  S1 : Table S2). Similarly, no statistically significant differences were observed between the two groups regarding the use of medications at the time of recruitment. In terms of familial steroidopathies, no significant differences were found in the prevalence of conditions like PCOS or endometriosis between the AG and NPG. Detailed data on familial, maternal, and paternal steroidopathies in the AG and Neuropsychiatric Group (NPG) are presented in Supporting Information  S1 : Table S1. However, there was a significantly higher prevalence of thyroid pathologies in the ASD group compared to NPG families (57.4% vs. 31.1%, Cramer's V  = 0.265; Fisher's Exact test, p  = 0.0115). Furthermore, the univariate analysis showed that thyroid pathologies in the family were associated with a statistically significant risk of ASD (OR = 2.98, 95% CI: 1.27–7.04, p  = 0.0122; Supporting Information  S1 : Table S2). Maternal hormonal treatment during pregnancy (with progesterone being the most frequently reported medication among those cited by the interviewed caregivers) was not associated with a statistically significant risk of belonging to the AG family (OR = 1.05, 95% CI: 0.46–2.41, p  = 0.9037; Supporting Information  S1 : Table S2). Although certain steroidopathies (hirsutism, severe acne, and irregular menstrual cycles) appeared more frequently in the AG compared to the NPG, these differences did not reach statistical significance (Table  2 ). Female ASD and NP participants' steroidopathies. Abbreviations: AG, Autism Group; ASD, autism spectrum disorder; NPG, NeuroPsychiatric Group. Hormonal blood tests revealed significantly higher serum testosterone levels in the AG compared to the NPG (0.50 ng/mL vs. 0.275 ng/mL, Hedge's effect size = 0.28, p  = 0.0057), with no significant differences in estradiol, prolactin, or vitamin D levels (Table  3 ). Hormonal and metabolic values in AG and NPG. Abbreviations: AG, Autism Group; IQR, InterQuartile Range; NPG, NeuroPsychiatric Group. Variables that reached statistical significance in the univariate analysis (Supporting Information  S1 : Table S2) were subsequently included in the multivariate model (Table  4 ). Subsequently, results of the multivariate logistic regression analysis have shown that thyroid pathology (OR = 2.98; CI 95% 1.27–7.035; p  = 0.0122), medication use during pregnancy (OR = 3.56 CI 95%1.25–10.11; p  = 0.017), serum testosterone (OR = 1.45; CI 95% 1.06–1.98; p  = 0.0208) are independent risk factors for ASD regardless of sex, age in children (Table  4 ; Supporting Information  S1 : Table S2). Results of the multivariable logistic regression analysis, and corresponding 95% confidence interval. Abbreviations: CI, confidence interval; OR, odds ratio. To have a suggestion of a discriminant threshold for testosterone that could determine an increased risk of ASD, a ROC analysis was conducted (Figure  1 ), and it has resulted with a mild AUC 0.66 (95% CI 0.56–0.75), and the risk for ASD increases for a testosterone level > 0.42 ng/mL. ROC curves of testosterone. Association analysis between scores and hormone levels revealed a statistically significant correlation only between testosterone and SQ scores in males, in both the AG ( r  = 0.7, p  = 0.0019) and the NPG ( r  = 0.7, p  = 0.0123; Supporting Information  S1 : Table S3).

Emerging evidence suggests a link between steroid hormone dysregulation and ASD, yet current research is hindered by several methodological limitations. Female participants are often underrepresented, and few studies include first‐degree relatives, limiting insight into potential familial patterns. To address these gaps, the present study investigates the relationship between ASD, steroidopathies, and steroid hormone profiles in a well‐characterized, homogeneous cohort balanced for age and sex. By including first‐degree relatives, the study seeks to clarify familial contributions to hormonal dysregulation. Moreover, to determine whether these hormonal alterations are specific to ASD, the study compares the ASD group with a homogeneous group of individuals diagnosed with other neuropsychiatric conditions. Ultimately, this research seeks to advance the understanding of ASD pathophysiology and identify potential early biomarkers to support diagnosis and preventive strategies. Based on these objectives and prior evidence, we hypothesized that (i) first‐degree relatives of individuals with ASD would present a higher occurrence of endocrine disorders, (ii) individuals with ASD would show an increased prevalence of steroidopathies compared to individuals with other neuropsychiatric conditions, and (iii) individuals with ASD would display altered hormonal profiles, particularly elevated serum testosterone levels. Furthermore, given prior evidence of sex‐specific effects, we anticipated differences in hormonal alterations between males and females with ASD.

Autism Spectrum Disorder is a lifelong neurodevelopmental condition whose prevalence has significantly increased over the past few decades (Maenner et al.,  2023 ). It is characterized by difficulties in initiating and sustaining reciprocal interaction and communication, alongside restrictive, repetitive, and inflexible patterns of behaviors, interests, or activities. Diagnosis requires that these impairments be evident from early development and significantly affect daily functioning. However, symptoms may only become fully apparent as social demands and requirements exceed the individual's capacities or may be masked by learned strategies in later life (American Psychiatric Association,  2022 ). The etiology of ASD is complex and multifactorial, stemming from a tangled interplay of genetic as well as epigenetic and environmental factors (Fu et al.,  2022 ; Pugsley et al.,  2022 ; Yap et al.,  2023 ). Among these factors, the role of steroid hormones and associated clinical manifestations (referred to as steroidopathies) is gaining attention (Gillberg et al.,  2017 ). Studies have reported that fetal testosterone levels inversely correlate with visual attachment, language development, social skills and Empathy Quotient of the mother and infant (Auyeung et al.,  2010 ; Qiao et al.,  2025 ); and that postnatal testosterone levels (total serum concentration, its free fraction, and related precursor) are elevated in individuals with autistic traits and children with ASD (Artık et al.,  2023 ; He et al.,  2023 ; Wan et al.,  2024 ). Additionally, disruptions in estrogen receptor signaling (Hwang et al.,  2021 ) and in cholesterol metabolism (Puljko et al.,  2025 ) as well as dysregulation of the hypothalamic‐pituitary‐adrenal (HPA) axis (including altered circadian rhythms, abnormal cortisol stress responses, and low plasma vitamin D levels), have been observed in ASD individuals (Wang et al.,  2022 ; Ward et al.,  2023 ). Furthermore, conditions such as PCOS, hirsutism, obesity, menstrual irregularities, severe acne, and an elevated risk of reproductive system cancers have been noted in both ASD females and mothers of ASD individuals and symptoms of hyperandrogenism were associated with autistic traits (Gasser et al.,  2020 ; Jiang et al.,  2022 ; Ward et al.,  2023 ).

This study provides evidence of an association between ASD and altered steroid hormone levels, particularly elevated serum testosterone. Familial steroidopathies, especially thyroid pathologies, were more prevalent in families of individuals with ASD. These findings suggest that steroidopathies and hormonal imbalances could play a role in the etiopathogenesis of ASD.

This study aimed to investigate the association between ASD, steroidopathies, and steroid hormone profiles, with a particular focus on familial endocrine conditions and maternal factors. Our findings indicate that steroid hormone dysregulation in ASD persists after controlling for sex distribution, familial history, and comparison with other neuropsychiatric conditions. Specifically, elevated serum testosterone, increased prevalence of familial thyroid pathology, and maternal hormonal treatment during pregnancy emerged as independent risk factors for ASD. These results provide new evidence linking hormonal alterations with ASD risk and pathophysiology and suggest potential pathways for identifying early biomarkers. Steroid hormones have been investigated in recent years for their potential contribution to the etiopathogenesis of ASD, from BC's studies on the Extreme Male Brain (Baron‐Cohen,  2002 ), to Gillberg's hypotheses regarding the impact of cholesterol metabolism alterations on neurological development (Gillberg et al.,  2017 ), to Pohl's research on the presence of steroidopathies in autistic females (Pohl et al.,  2024 ), and Lai's analysis of the influence of prenatal androgens on the social brain (Lombardo et al.,  2020 ). These hormones are essential for neurodevelopment, influencing neural proliferation, neurotransmission, oxidative stress regulation, and immune function. They may also affect cerebrovascular development, act as epigenetic fetal programming factors shaping the early brain environment and interact with thyroid and glucocorticoid hormones—via the placenta as a mediator—thereby influencing brain development (Banik et al.,  2017 ; Collignon et al.,  2024 ; Ward et al.,  2023 ). This study supports the notion that steroid hormonal abnormalities, particularly elevated serum testosterone levels, may represent risk factors for ASD. Previous studies, comparing ASD individuals with neurotypical children, have indicated that maternal conditions characterized by a hyper‐androgenic status, such as PCOS, are linked to a higher probability of offspring being diagnosed with autism (Cesta et al.,  2020 ; Katsigianni et al.,  2019 ; Kosidou et al.,  2016 ). Kosidou et al. ( 2016 ), in a large population‐based study conducted in Sweden, reported that maternal PCOS increased the odds of ASD in offspring by 59%, with no differences observed between sexes (Kosidou et al.,  2016 ). Katsigianni et al. ( 2019 ), through a systematic review and meta‐analysis, concluded that women with PCOS show increased odds of having a child with ASD, based on a large sample and high‐quality studies (Katsigianni et al.,  2019 ). Expanding on these findings, Cesta et al. (2020) analyzed data from over one million births in the Swedish national registry and confirmed that children ‐especially females‐of mothers with PCOS may have a higher likelihood of developing neuropsychiatric disorders, particularly ASD, ADHD, and Tourette Syndrome (Cesta et al.,  2020 ). Although the difference in the prevalence of familial steroidopathies between individuals with ASD and those with other neuropsychiatric disorders was not statistically significant, this study found a significantly higher prevalence of familial thyroid pathologies in the ASD group, independent of sex and age. This finding confirms previous studies suggesting that thyroid dysfunction could be a risk factor for ASD. Thyroid hormones are essential for neuronal differentiation, synaptogenesis, and myelination in the fetal brain. Variations in maternal thyroid hormone levels (also associated with thyroid autoimmunity) particularly during the first trimester, may exert a significant impact on fetal brain development, increasing the risk for ASD and other neurodevelopmental disorders (Rotem et al.,  2020 ; Ward et al.,  2023 ). Regarding maternal medication use, our statistical analysis indicates that the use of at least one medication during pregnancy is a borderline significant risk factor. The most commonly used medications by our participants' mothers during pregnancy, apart from those for thyroid conditions, are anti‐abortive drugs, primarily progesterone. Our results are consistent with recent scientific literature, including both animal models and population‐based case‐control studies (Davidovitch et al.,  2018 ; Zou et al.,  2017 ), which have shown that progesterone exposure during fetal life may increase the risk of ASD, potentially through epigenetic modification. Davidovitch et al. (2018) found that while infertility treatments themselves were not directly linked to a significant increase in ASD risk, the use of progesterone during early pregnancy may contribute to a higher likelihood of ASD development (Davidovitch et al.,  2018 ). Similarly, Zou ( 2017 ) investigated the relationship between prenatal exposure to synthetic progestins, the downregulation of estrogen receptors, and the emergence of cognitive deficits and autism‐like behaviors in both animal models and national registry datasets (Zou et al.,  2017 ). In the female AG, a high prevalence of steroidopathy‐related symptoms was observed, including hirsutism (29.2%), acne (16.7%), and menstrual irregularities (25.0%). Although these rates were not statistically different from those observed in the NPG, they represent clinically relevant findings within the sample. It is also worth noting that the study population ranged from 9 to 18 years of age, which may precede the typical age of onset for some of these conditions, potentially leading to an underestimation of their actual prevalence. To better contextualize these findings, it is important to compare them with prevalence estimates from the general population, where hirsutism affects approximately 4%–10% of women worldwide (Spritzer et al.,  2022 ), acne affects approximately 9.4% (Tan & Bhate,  2015 ), and menstrual irregularities range from 14% to 25% (American College of Obstetricians and Gynecol ogists (ACOG) Committee Opinion No. 760,  2018 ). The presence of alterations in steroid hormones—particularly in females with ASD‐ and the related expression of clinical conditions known as steroidopathies‐have been extensively discussed in recent literature (Ward et al.,  2023 ). Ingudomnukul, Pohl, and Ward (2007, 2014, 2023) have explored the increased prevalence of testosterone related disorders and steroidopathic symptoms in women with ASD compared to neurotypical controls (Ingudomnukul et al.,  2007 ; Pohl et al.,  2024 ; Ward et al.,  2023 ). From BC's early studies to more recent research, attention has increasingly focused on the possibility of measuring these steroid hormonal alterations by assessing prenatal and postnatal steroid hormone levels in amniotic fluid and in other bodily fluids (saliva, urine, blood) (Artık et al.,  2023 ; Auyeung et al.,  2010 ; He et al.,  2023 ; Katsigianni et al.,  2019 ; Qiao et al.,  2025 ; Wan et al.,  2024 ). Baron‐Cohen et al. analyzed amniotic fluid samples, searching for androgens in 2015 (Baron‐Cohen et al.,  2015 ) and estrogens in 2020 (Baron‐Cohen et al.,  2020 ), finding an association between hormonal levels and likelihood of autism. Gasser et al. ( 2020 ) found, in a small group of autistic female adolescents, elevated androgen levels compared to neurotypical controls (Gasser et al.,  2020 ). Artik et al. ( 2023 ) found that serum levels of testosterone in prepubertal drug‐naive boys with ADHD were associated with more pronounced autistic traits (Artık et al.,  2023 ). He et al. ( 2023 ) investigated salivary steroid hormones, variations in children and adolescents with ASD versus neurotypical controls, finding that subjects with ASD had significantly higher salivary testosterone precursors levels (dehydroepiandrosterone and pregnenolone), than their neurotypical peers (He et al.,  2023 ). Wang et al. ( 2024 ) provided a systematic review and meta‐analysis on androgen levels in blood, urine, saliva finding significantly higher serum free testosterone levels, in individuals with ASD compared to control groups and total testosterone and DHEA in urine (Wan et al.,  2024 ). In the present study, we analyzed serum hormone levels in children and adolescents with ASD and other neuropsychiatric conditions, finding that serum testosterone is an independent risk factor in autistic children regardless of sex or age. Moreover, our results showed that serum testosterone levels are associated with hyper‐systemizing traits, measured with the SQ, exclusively in males ‐both in those diagnosed with autism and those diagnosed with other neuropsychiatric disorders (Supporting Information  S1 : Table S2). These results are consistent with previous literature, which stems from BC's Extreme Male Brain theory demonstrating that elevated testosterone levels are associated with higher scores of autistic traits measured by BC's scales (Baron‐Cohen,  2002 ; Baron‐Cohen et al.,  2020 ). Although the differences between these scores in ASD and non‐ASD groups were not statistically significant, our results suggest that the role of testosterone may be significant for autistic traits across diagnostic categories, rather than solely in cases of confirmed autism diagnosis ‐ as a trans‐nosographic feature. In addition, when comparing the ASD group with the NP group, a higher number of individuals with Oppositional Defiant Disorder—a condition notably associated with elevated testosterone levels (Shenk et al.,  2012 )—was observed in the NP group. However, in our ASD group, testosterone values were still higher, suggesting the significant role and the specificity of testosterone levels associated with autism, regardless of the presence of other disorders. The main limitation of this study is the absence of a control group of healthy subjects. Additionally, the sample size may not be sufficient to allow for generalization of the findings to the broader population. Another limitation concerns potential bias due to the retrospective nature of the study, which required reliance on available clinical data. Further studies with larger sample sizes and an appropriate healthy control group are needed to confirm and expand upon these results.

Prof. S.C. has declared reimbursement for travel and accommodation expenses from the Association for Child and Adolescent Central Health (ACAMH) in relation to lectures delivered for ACAMH, the Canadian AADHD Alliance Resource, the British Association of Psychopharmacology, Healthcare Convention and CCM Group team for educational activity on ADHD, and has received honoraria from Medice. The remaining authors have declared that they have no competing or potential conflicts of interest.

Supporting Information S1