The genetic relationship between ADHD and shortened telomeres points to neurodevelopmental mechanisms

Abstract

Attention-Deficit/Hyperactivity Disorder (ADHD) is often accompanied by other psychiatric and somatic conditions. Telomere shortening as an accelerated biological aging marker may underlie this aggregate burden. Telomere length (TL) has previously been associated with childhood inattention and hyperactivity, as well as other neuropsychiatric conditions. We hypothesize that TL and ADHD may share overlapping genetic influences, suggesting that some of the same genes could affect both cellular aging and ADHD traits. We investigated the pleiotropic and causal associations between these traits. Using the latest and largest publicly available genome-wide association study (GWAS) summary data for ADHD (38,691 cases and 186,843 controls) and TL (472,174 subjects), we tested causality, global and local genetic correlations, and joint genomic effects. Polygenic scores (PGS) were evaluated in an independent clinical sample of 665 ADHD cases and 995 controls, and TL was quantified via qPCR in a subsample of 370 subjects. The pleiotropy analysis showed global and local (a chromosome 17 locus) negative correlations. Shared genomic analysis showed loci enriched across several chromosomes, notably including the same chromosome 17 locus. This locus contains a diverse set of genes involved in many biological pathways impacting core cellular functions, tissue development, cell cycle, and apoptosis. Mendelian randomization indicated a unidirectional causal effect of ADHD on shorter TL across four methods. In our independent clinical sample, TL was shorter in individuals with ADHD. Lastly, higher PGS for longer TL was associated with reduced ADHD symptomatology. Our findings suggest that ADHD may contribute to TL, and genetic predisposition to shorter TL is associated with greater ADHD severity.

Introduction

Attention-Deficit/Hyperactivity Disorder (ADHD) is a neuropsychiatric condition associated with substantial morbidity [1, 2]. ADHD frequently coexists with various somatic conditions, including neurological, sleep-related, autoimmune, and metabolic disorders [3–5], as well as comorbid psychiatric disorders [6]. Both somatic and psychiatric conditions are linked to an elevated risk of mortality, potentially due to mechanisms that accelerate biological aging [7].

Telomeres are repetitive DNA sequences located at chromosome ends that preserve genomic stability and progressively shorten with each cell division, making telomere length (TL) a widely used marker of biological aging [8]. Accelerated telomere shortening reflects cumulative cellular stress, including oxidative damage and impaired DNA repair, ultimately leading to cellular senescence [9]. Beyond aging, TL is a highly heritable polygenic trait that has been associated with susceptibility to a range of somatic and neuropsychiatric conditions, highlighting its relevance as a marker of systemic biological vulnerability [7, 10].

ADHD-associated genes are enriched among those upregulated during early brain development [11], a period in which TL plays an important role in neurodevelopmental processes, including cellular proliferation and differentiation [12]. Moreover, altered TL has been consistently reported in psychiatric conditions that frequently co-occur with ADHD [13, 14]. Despite this, few studies have explored the relationship between ADHD and TL. Supporting TL’s potential influence on neurodevelopment, existing research has identified associations between shorter TL and both inattention [15] and hyperactivity symptoms in childhood [16]. Conversely, a study in a community-based sample of young adults did not find an association between current ADHD symptoms and TL but did report a link between past childhood hyperactivity symptoms and longer telomeres [17]. This conflicting evidence underscores the need for further research into the shared biology of ADHD and TL.

Since TL variability could act as a risk factor for the development of ADHD or reflect biological processes related to the disorder and its comorbidities, clarifying the direction and nature of this relationship remains essential. Based on prior evidence linking ADHD to increased somatic burden and biological aging, and considering the highly polygenic architecture of TL, we hypothesized that ADHD and this biological aging marker share a common genetic basis.

To test this hypothesis, we aimed to investigate the genomic architecture shared between ADHD and TL by evaluating pleiotropic and potential causal associations. Additionally, we examined whether both genetically predicted telomere length, derived from Polygenic Scores (PGS), and experimentally measured TL were associated with ADHD diagnosis, symptom dimensions, and related clinical features in adults. To our knowledge, this is the first study to integrate large-scale GWAS summary statistics with complementary post-GWAS approaches while jointly evaluating genetically predicted and directly measured TL in an independent adult ADHD clinical sample.

Methods

Overview of analytical strategy

To investigate the shared genetic architecture and potential causal relationship between ADHD and TL, we applied a two‑stage analytical framework combining large‑scale genomic analyses with experimental validation in an independent clinical cohort. First, we leveraged GWAS summary statistics to estimate global and local genetic correlations and identify shared genetic loci. These loci were submitted to enrichment analysis for suggested shared biological pathways. Further, summary statistics were used to test causal directionality using Mendelian randomization. Lastly, we evaluated the relevance of these genomic findings by examining associations between genetically predicted TL, experimentally measured TL, ADHD diagnosis, and symptom severity in an adult ADHD clinical sample. The supplementary material contains a detailed description of the methods. An overview of the bioinformatic tools employed in this study is provided in Supplementary Table 1.

Study samples

GWAS summary statistics

Publicly available GWAS summary statistics were used for all genomic analyses. ADHD summary data were obtained from the most recent meta‑analysis conducted by the Psychiatric Genomics Consortium and collaborators (38,691 cases and 186,843 controls [11]). TL summary statistics were derived from a large population‑based GWAS comprising 472,174 individuals [10]. All participants included in these GWAS were of European ancestry. Variants with ambiguous alleles, duplication, minor allele frequency < 1%, low imputation quality (INFO < 0.8), or located on sex chromosomes were excluded.

Adult ADHD porto alegre cohort

The experimental analyses were conducted in an independent adult clinical cohort from the Brazilian ADHD Research Network, based at the Hospital de Clínicas de Porto Alegre (HCPA). The sample comprised 665 adults with ADHD and 995 controls (Supplementary Table 2). ADHD diagnoses were established according to DSM-IV criteria (2001–2012) and DSM-5 criteria thereafter, using structured clinical interviews administered by trained clinicians.

To ensure diagnostic consistency across DSM versions, harmonized procedures were applied. Specifically, retrospective confirmation of symptom onset before 12 years of age was required for all participants, including those diagnosed during the DSM-IV period. In addition, the reduced adult symptom threshold introduced in DSM-5 had minimal impact in this cohort, as the vast majority of cases endorsed at least six symptoms in one dimension. Consequently, only a small number (n = 8) of participants met DSM-5 criteria without fulfilling DSM-IV thresholds. Controls screened negative for ADHD using standardized instruments. Exclusion criteria included neurological disorders affecting cognition and estimated IQ < 70. All participants were fully briefed on study procedures and provided signed consent, approved by the HCPA Institutional Review Board (IRB) (IRB 0000921) in accordance with the Declaration of Helsinki and Brazilian regulations.

Genomic analyses

Global genetic correlation

Global genetic correlation between ADHD and TL was estimated using linkage disequilibrium score regression (LDSC), which quantifies genome‑wide shared genetic liability between traits while accounting for linkage disequilibrium [18].

Local genetic correlation

Local genetic correlation was assessed using the Local Analysis of [co]Variant Association (LAVA) framework [19]. This approach partitions the genome into semi‑independent loci and tests for shared genetic effects within each locus. Only loci showing evidence of association with both traits were evaluated for local correlation, with correction for multiple testing. This approach identifies specific genomic locations that may contribute to the global correlations and, when combined with enrichment analysis, suggests genes and pathways that may be implicated in these correlations.

Shared genetic loci

To further identify genetic variants jointly associated with ADHD and TL, we applied the conjunctional false discovery rate (conjFDR) approach [20, 21]. This method leverages cross‑trait enrichment to detect shared loci beyond genome‑wide significance in individual GWAS.

Gene annotation and enrichment

Genes mapped to loci identified by local genetic correlation and conjFDR analyses were annotated and functionally characterized using publicly available databases. Pathway and gene‑set enrichment analyses were performed to identify biological processes over‑represented among the shared loci [18, 22–24].

Mendelian randomization

Bidirectional two‑sample Mendelian randomization (MR) analyses were conducted to evaluate potential causal relationships between ADHD liability and TL. Genetic instruments were selected based on genome‑wide significance and linkage disequilibrium clumping. Multiple MR estimators were applied, and causal inference was based on consistency across methods. Sensitivity analyses were performed to assess instrument validity and potential pleiotropy [25].

In house experimental analyses

TL measurement

TL was available in a subsample of 238 ADHD cases and 132 controls using quantitative PCR (Supplementary Table 3). TL values were log‑transformed and standardized prior to analysis. Group differences were assessed using linear regression models adjusted for age, sex, socioeconomic indicators, and relevant psychiatric comorbidities. Given the modest size of the TL subsample, analyses involving psychiatric comorbidity strata were treated as exploratory and primarily used to inform covariate selection rather than to draw disorder-specific conclusions.

Polygenic score construction

Polygenic scores for TL were calculated using GWAS summary statistics as the discovery dataset. Scores were derived using a Bayesian continuous shrinkage (PRS-CS) approach that accounts for local linkage disequilibrium patterns [26]. PGS were standardized and tested for association with ADHD diagnosis, symptom severity, and psychiatric comorbidities.

Statistical analysis

Associations involving binary outcomes were tested using logistic regression, and continuous outcomes were evaluated using linear regression. All models included age, sex, and genetic ancestry principal components as covariates. Secondary models additionally adjusted for psychiatric comorbidities when appropriate, besides correction for multiple tests. A workflow of the analyses and samples used is shown in Supplementary Figure 1.

Results

Genetic Correlations (LDSC and LAVA)

LDSC analysis showed a negative global genetic correlation (r_g = -0.1074, P-value = 0.0004) between TL and ADHD (Supplementary Table 4). When examining it from a local perspective with LAVA, 189 loci passed for bivariate analysis (Supplementary Tables 5 and 6). After Bonferroni correction, only locus 2205 remained significant and presented a negative correlation (rho = -0.622488, p-value = 0.0082425546 - Supplementary Table 6). The area defined in LAVA as locus 2205 spans base pairs 40,233,546 to 42,348,003 on chromosome 17, including 110 genes listed in the GDV database (Fig. 1, Supplementary Table 7). Within the Reactome database, we identified 21 gene-sets that remained significant after multiple testing corrections (Supplementary Table 8). Included in these sets, the genes STAT5A, STAT5B, and STAT3 were involved in several pathways, especially in interleukin signaling pathways, such as Interleukin-2, IL-2 Family, IL-7, IL-9, IL-15, IL-20 Family, IL-21, as well as “Growth Hormone Receptor Signaling” (p-value < 0.05). In the GO Biological Process database, those three genes were also pointed out as part of “Growth Hormone Receptor Signaling Pathway (GO:0060396)” and the “Growth Hormone Receptor Signaling Pathway via JAK-STAT (GO:0060397)” (p-value < 0.05).

Fig. 1. Genes annotated within the locus identified by the LAVA analysis.

Genes located within the genomic interval spanning base pairs 40,233,546 to 42,348,003 on chromosome 17, as identified by the LAVA analysis.

ConjFDR

The conjFDR method revealed a significant shared genetic background between ADHD and TL, though TL has a more unique genetic architecture (Supplementary Figure 2). The results showed 14 independent significant hits jointly associated with both phenotypes (Fig. 2). These hits were distributed across 11 different chromosomes, including chromosome 17 (Fig. 2, Supplementary Table 9). The SNP in chromosome 17 was annotated to the NBR1 (NBR1 Autophagy Cargo Receptor 1), also known to be neighboring the BRCA1 gene. The SNPs found in other chromosomes were annotated to genes that were also hits in ADHD GWAS such as SORCS3, ACTBP13, and in TL GWAS such as MPHOSPH9. However, novel associations were also found in SEMA6D, GIGYF2, KPNA4, BAK1, and PHF2 genes. Reactome pathways and GO terms associated with the genes resulting from this annotation include axon guidance, autophagy, and interferon signaling (Supplementary Table 10). In addition, the terms that remained significant after multiple testing corrections (q-value < 0.05) revolved around molecular functions such as receptor binding and signal transduction, i.e. “Transmembrane Receptor Protein Kinase Activity GO:0019199”.

Fig. 2. Manhattan plot showing the genes associated in ADHD vs TL conjFDR GWAS.

The x-axis displays each chromosome, and the y-axis shows the -log10 of the conjFDR p value for each SNP tested. The dashed line indicates the significance threshold (-log10[conjFDR < 0.05]). The top 10 lead SNPs from significant loci are highlighted in red and labeled with their nearest gene.

PheWas and enrichment

LAVA and conjFDR resulting genes were grouped into ten different categories: “Immunological features”, “Genome maintenance”, “Brain related features”, “MicroRNA”, “Transport”, “Signal Transduction”, “Cell development and differentiation”, “Cell cycle regulation/Apoptosis”, “Core intracellular function” and “Other”, the last four being the best represented. Considering PheWas results, the most prevalent domains were skeletal, immunological, metabolic, psychiatric, and reproductive. When selecting the most frequent traits across domains, we have found for skeletal “Heel bone mineral density”, “Height”, for immunological: “Mean corpuscular hemoglobin concentration”, “Reticulocyte fraction of red cells”, metabolic: “Body Mass Index”, “Impedance measures”, psychiatric: “Schizophrenia”, “Neuroticism”, and reproductive: “Age at menopause”, “Had menopause” (Supplementary Figures 3 and 4, Supplementary Table 10). In addition, STRING analysis resulted in one enriched association pathway between BRCA1, NBR1, and other related genes (Supplementary Figure 5). All genes leveraged by our genomic analysis are deeply characterized in Supplementary Table 11.

Mendelian randomization

MR analysis identified a unidirectional causal effect of ADHD on shorter TL, with consistent and significant results across four different models (Supplementary Tables 12 and 13, and Supplementary Figure 6). No causal effect of TL on ADHD was observed (Supplementary Figure 7). Sensitivity analyses confirmed the viability of instrumental variables (IVs) for both rounds, as the i² measure fell within the acceptable range of 0.6-0.9. MR-Egger regression detected no evidence of horizontal pleiotropy between the traits, and Steiger directionality tests ruled out reverse causality. However, heterogeneity tests indicated an excess of variation between IVs, which could introduce bias associated with high pleiotropy. Leave-one-out plots showed no evidence of outlier effects, while funnel plots revealed overdispersion of data in the ADHD-TL analyses (Supplementary Figures 8 to 10 and Supplementary Table 13).

In house experimental subset

Telomere length

TL was diminished in ADHD cases (mean = -0.1412, SD = 0.04559, N = 238) compared with controls (mean = 0.0054, SD = 0.04559, N = 132, p = 0.024) and in men (mean = -0.1734, SD = 0.05174, N = 175), compared to women (mean = -0.0131, SD = 0.04398, N = 195, p = 0.018). This sex effect seems to be carried by the differences observed in ADHD cases, since they were not observed in the analysis restricted to controls. As expected, shorter TL was significantly associated with older age (beta = -0.125, CI-95% = -0.015 - -0.002, p = 0.016). Of the psychiatric disorders tested for TL, only SUD was under the p-value threshold of 0.2 estimated for inclusion of covariates [27]. Therefore, we assembled a series of univariate linear regression models with ADHD as a predictor, TL as the outcome and including sex and age as covariates in the first model, adding SUD in a subsequent model. TL was nominally associated with ADHD diagnosis status across all models tested (Supplementary Table 14). No association was found between TL and ADHD symptom severity.

Polygenic Scores

Increased PGS for TL (longer TL) was associated with decreased ADHD symptomatology, especially in inattention symptoms, but not with case-control status (Fig. 3). When including only subjects of European descent we found a higher effect size and stronger associations, which is expected considering the ethnic component of the GWAS used for PGS calculation (Supplementary Table 15). We did not find any associations between TL PGS and ADHD comorbid profile (Supplementary Table 16). Experimentally measured TL was not associated with TL PGS (beta = -0.001, SD = 0.050, p = 0.991) or ADHD PGS (beta = 0.101, SD = 0.049, p = 0.132).

Fig. 3. Mean ADHD symptoms by TL PGS quintile.

The x-axis displays PGS for TL measures divided in 5 quintiles, and the y-axis displays mean ADHD symptom score. The graph shows that increase PGS for higher TL is associated with decrease of mean ADHD symptom score measured by ASRS.

Discussion

Our broad set of analyses provides novel insights into the relationship between ADHD and telomere biology, revealing both global and local negative genetic correlations between ADHD and TL, with several loci of interest and a specific effect identified on chromosome 17. MR analysis suggests a directional relationship in which genetic liability to ADHD is associated with genetically proxied TL. This aligns with our experimental data showing shorter telomeres in individuals with ADHD and with polygenic scores for shorter TL being associated with higher ADHD severity. Together, these findings reinforce previous studies indicating that ADHD may be linked to accelerated biological aging processes, potentially mediated by shared genetic pathways. Besides, our findings add substantial new evidence to the topic through various complementary genomic approaches, including LDSC, LAVA, conjFDR, MR, and PGS.

The shared biology between ADHD and TL, as revealed by our genomic analyses, appears to involve broad mechanisms related to cellular proliferation and survival. Beyond the global correlation observed, we identified one specific shared locus on chromosome 17 (chr17:4023354–42348003). Gene annotation of this region highlighted several candidates with plausible roles in early embryonic development and genome maintenance, including BRCA1, involved in DNA repair; RAB5C, a component of the Ras signaling pathway; and STAT3, STAT5A, and STAT5B, key members of the JAK-STAT molecular network. The Ras pathway is well established in regulating cell growth, proliferation, and apoptosis [28], while the JAK-STAT pathway integrates growth hormone signaling with immune and inflammatory processes, playing a central role in cytokine-mediated immune responses and neuroimmune communication [29], both contributing to essential cellular and systemic functions. Gene enrichment analyses indicated biological processes related to inflammation and growth hormone (GH) signaling.

Consistently, the conjFDR analysis identified 14 independent loci across 11 chromosomes, including a locus on chromosome 17 annotated to NBR1, which encodes a selective autophagy receptor involved in cellular stress responses and genomic maintenance. Among the shared loci, SEMA6D stands out as a gene with prior neurodevelopmental relevance, playing a central role in axon guidance and early neural circuit formation [30], and was recently identified among probable causal genes in a GWAS of ADHD symptoms in children [31]. In contrast, other identified genes -including GIGYF2, KPNA4, PHF2, MPHOSPH9, and BAK1 - have not been consistently linked to ADHD but participate in fundamental cellular processes such as transcriptional regulation, nucleocytoplasmic transport, apoptosis, and growth factor signaling. Notably, BAK1 is a key regulator of apoptosis, a mechanism directly connected to cellular senescence and telomere shortening [32]. The presence of these loci suggests that the shared genetic architecture between ADHD and TL extends beyond classical neurodevelopmental pathways to include broader mechanisms of cellular stress regulation and aging biology.

Based on functional enrichment analyses, we categorized the identified genes into distinct biological domains. The most prominent categories were: “Core Cellular Function,” “Cell Cycle Regulation / Apoptosis,” and “Cell Development and Differentiation.” Notably, these genes also interact within a BRCA1-related pathway, which may be closely associated with cellular aging [33]. Among all functional pathways identified, immunological features emerged alongside domains related to genome maintenance, brain-related processes, and core intracellular functions. This pattern is consistent with growing evidence supporting interactions between immune-related biological processes and neurodevelopmental and psychiatric conditions, including ADHD [34]. Brain–immune crosstalk has been implicated in neural development, synaptic regulation, and behavioral phenotypes relevant to ADHD [35] and immune-related traits identified in the PheWAS analysis further reinforce this cross-system biological overlap. In addition, immune activation and chronic low-grade inflammation have been associated with telomere dynamics, providing a plausible biological context linking immune-related pathways and TL [36]. While these observations are exploratory and do not imply causality, they suggest that immune-mediated mechanisms may contribute to the broader biological landscape underlying the association between adult ADHD and telomere biology. Of note, telomere shortening occurs most rapidly during the early years of life [37], a critical period for neurodevelopment and when these shared mechanisms with ADHD may be particularly active.

MR results suggest that genetically predicted ADHD affects TL and not the other way around. This improves our understanding of causality between the two phenotypes, suggesting that genetic liability to ADHD may intersect with pathways involved in cell cycle regulation and biological aging. The influence of ADHD-related genetics on shorter TL may reflect shared pathways active during early development, which compromise neural differentiation and growth, likely driven by overall genomic integrity maintenance. A similar genomic effect has been observed in the association between other neurodevelopmental disorders, such as autism spectrum disorder [38], suggesting that genomic regulation of brain development could affect these disorders through mechanisms that involve biological aging acceleration.

Our findings could also be interpreted under alternative explanations since the length of telomeres is impacted by several factors related to psychiatric illness like somatic and mental comorbid disorders, mostly substance and alcohol use [39]. Indeed, the commonalities between ADHD and TL could also be explained by the fact that telomere maintenance is affected by inflammation and oxidative stress mediators [40], which are also frequently associated with psychiatric disorders [40], as well as cell cycle regulation components. We have addressed this alternative perspective by exploring the relationship across experimentally measured TL and ADHD, including psychiatric comorbidities in the tested models, and found that the association remained stable across all models.

Accordingly, our findings are reinforced by the fact that our MR indicated a unidirectional genomic effect of ADHD on TL despite the substantial exposure to environmental hazards during the lifespan of ADHD subjects that could impact TL. In a relatively small experimental sample, we did not find an association between ADHD PGS and TL (N = 370), although TL PGS was associated with ADHD symptoms (N = 1660). We argue that our sample size with experimentally measured TL may be limited to detect the association, especially accounting for differences in ethnic background; nevertheless, in an augmented sample these effects from polygenic scores are observed. Sample size, ethnic background and environmental factors such as comorbidity profile can also address the lack of association between TL PGS and experimentally measured TL.

This study should be interpreted in light of several limitations. First, although we combined multiple complementary genomic approaches to strengthen inference, analyses based on genetically predicted liability are inherently limited to the genetic component of ADHD and cannot fully capture environmental or developmental influences. Second, experimentally measured telomere length was available in a subset of the clinical cohort, which reduced statistical power for subgroup analyses, particularly those stratified by psychiatric comorbidities; accordingly, these analyses were treated as exploratory and primarily used to inform covariate selection. Third, although we identified biologically plausible genes within shared loci, many of these genes are highly pleiotropic and participate in multiple cellular pathways. Thus, the present findings should be interpreted as evidence of shared genetic architecture rather than support for specific molecular mechanisms, which will require targeted functional studies. Finally, longitudinal studies with repeated telomere length measurements beginning in childhood will be needed to more directly capture telomere dynamics across the life course and their relationship with ADHD.

In conclusion, our findings suggest that genes involved in cell proliferation, immune response and genome maintenance, which play a crucial role during early brain development and contribute to ADHD pathophysiology [11], may also be linked to shorter TL in adult ADHD patients. This evidence deepens our understanding of the cellular mechanisms associated with ADHD and suggests that telomeres could serve as potential biomarkers for further research and clarification of the disorder.

Author contributions

MEAT conceived the study, curated and analyzed the data, performed statistical analyses, interpreted the results and wrote the original draft of the manuscript. JBS contributed to the conception and design of the study, acquisition of resources, and funding support. IFT, ESV, IJS, NPC, VFO, and CEB contributed to data analyses, interpretation of results, and drafting of the corresponding sections of the manuscript. EHG, LAR, BSS, DLR, and CHDB contributed to study supervision, project administration, acquisition of resources, and critical revision of the manuscript. CHDB supervised the work throughout its development. All authors contributed to the interpretation of the findings, critically reviewed the manuscript, and approved the final version.

Funding

This work was supported by the National Council for Scientific and Technological Development (CNPq; grant numbers 405434/2023-5, 444628/2024-0, 404121/2024-1, and 421559/2025-0), the Coordination for the Improvement of Higher Education Personnel (CAPES; Finance Code 001), the Fundo de Incentivo à Pesquisa e Eventos do Hospital de Clínicas de Porto Alegre (FIPE-HCPA; grant numbers 16-0600 and 21-0254), the Research Foundation of the State of Rio Grande do Sul (FAPERGS; grant numbers 25/2551-0001880-3 and 25/2551-0002733-0), and the São Paulo Research Foundation (FAPESP; grant number 2020/05652-0). The funding sources had no role in the design of the study; the collection, analysis, or interpretation of data; the writing of the manuscript; or the decision to submit the article for publication.

Data availability

The data that support the findings of this study are openly available in PhysioGenLab github at https://github.com/PhysioGenLab.

Code availability

All scripts are available in https://github.com/PhysioGenLab/Tavares-et-al.-2025.

Competing interests

LAR has received grant or research support from, served as a consultant to, and served on the speakers’ bureaus of Abdi Ibrahim, Abbott, Aché, Adium, Apsen, Bial, Knight Therapeutics, Medice, Novartis/Sandoz, Pfizer/ Upjohn/Viatris, and Shire/Takeda in the last 3 years. The ADHD and Juvenile Bipolar Disorder Outpatient Programs chaired by LAR have received unrestricted educational and research support from the following pharmaceutical companies in the last 3 years: Novartis/Sandoz and Shire/Takeda. LAR has received authorship royalties from Oxford Press and ArtMed. E.H.G. reported receiving personal fees from Takeda and EMS for lectures outside the submitted work during the 3 years prior to submission. All other authors report no biomedical financial interests or potential conflicts of interest.

Declaration of Generative AI and AI-assisted technologies in the writing process

To improve the clarity and fluency of the manuscript, we used ChatGPT (OpenAI) for assistance with English grammar, punctuation, and style. The content was entirely generated by the authors, and ChatGPT was not used for data analysis, interpretation, or writing of scientific content.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04180-6.