Across millions of US siblings, researchers mapped 569 diagnoses to examine where birth-order patterns diverge and whether they persist within families.

Study: Birth order and disease risk across the human phenome. Image Credit: Phovoir / Shutterstock

In a recent study published in Nature Health, researchers assessed associations between birth order and recorded diagnoses at the phenome level. Analyzing more than 10 million individuals from approximately five million two-child families in the Merative MarketScan dataset, comprising inpatient, outpatient, and pharmacy claims, they found that first-born individuals had higher odds of claims-recorded neurodevelopmental and immune-allergic conditions.

In contrast, second-born siblings had higher odds of recorded diagnoses for substance abuse and gastrointestinal disorders. While the other/unspecified pervasive developmental disorder (PDD) code group, tics/Tourette syndrome, and autism showed the strongest first-born excesses, significant associations were also observed for attention-deficit/hyperactivity disorder (ADHD), food allergies, and allergic rhinitis. Second-born participants showed higher odds of conditions such as gastritis and duodenitis. The disease-level effect estimates were positively correlated between the within-family and between-family designs.

It is generally believed that younger siblings have a lower risk of allergic and immune-related conditions than older siblings. This may be because they typically receive greater microbial exposure from older siblings in early life, which may support the development of immune tolerance.

Previous studies have reported that later birth order is associated with a lower likelihood of asthma, hay fever, and eczema. The effects of birth order on disease risk have been studied for individual conditions.

Phenome-level assessments among large-scale sibling claims cohorts could systematically compare associations across organ systems and determine whether they persist across complementary study designs.

About the study

In the present study, researchers assessed phenome-scale disease risk by birth order. The study comprised 5.1 million two-child families used for within-family comparisons and 1.6 million matched pairs formed from individuals in different families for between-family comparisons. These individuals were identified in the 2003–2024 Merative MarketScan commercial claims database. The siblings were aged ≥12 years at their last observation and had at least 1 year of enrollment visibility.

To assess disease risk across families, the team paired first-born and second-born individuals from separate families, matching them on birth year, sex, urbanization status, census region, follow-up duration, parental age, and sibling age gap.

The study defined 569 diseases in the sibling cohorts, identified using the ninth and tenth versions of the International Classification of Diseases, Clinical Modification (ICD-CM) codes. Of these, 418 had at least 500 cases and entered the primary between-family analysis, while 541 had at least 100 sibling pairs in which only one sibling had the diagnosis and entered the within-family analysis.

Logistic regression estimated birth-order associations across families, while conditional logistic regression, stratified by family, compared first- and second-born siblings within the same family. The between-family models adjusted for measured demographic, parental-age, geographical, clinical, and observation-related factors. The within-family design helped reduce confounding from factors shared by siblings, such as parental genetics, geographical exposures, socioeconomic status, and family attitudes toward health and wellness.

The team developed a comprehensive disease atlas that provided information on the directionality of the findings. In the atlas, every disease tile was colored based on the extent and direction of related birth-order effects. The atlas comprised 75 diseases that met the stringent Bonferroni threshold in both designs and showed concordant directions of effect.

The researchers also performed sensitivity analyses, applying state fixed effects, full-sibling restrictions, and stricter clinically enriched rematching. To validate the analysis, they included positive and negative disease controls. Positive controls included diseases with expected birth-order associations based on previous literature, whereas negative controls included five diseases with primarily structural or genetic etiologies and two common acute conditions expected to show little or no association.

Results

The team found 150 associations that met the stringent Bonferroni threshold, accounting for testing multiple diseases among the 418 diseases analyzed across families. First-born individuals showed higher odds of recorded neurodevelopmental and immune-allergic diagnoses compared with second-born siblings.

All reported ORs compared second-borns with first-borns, meaning that an OR below 1 indicated first-born excess. In particular, first-born excess was observed for the other/unspecified PDD code group (OR, 0.569), tics/Tourette syndrome (OR, 0.693), autism (OR, 0.737), food allergy (OR, 0.797), allergic rhinitis (OR, 0.910), and ADHD (OR, 0.934).

In contrast, second-born individuals showed higher odds of recorded substance abuse (OR, 1.192) and gastrointestinal diagnoses such as gastritis or duodenitis (OR, 1.142). Herpes zoster had the strongest and most consistent association with second birth order (OR, 1.348). Several selected associations, including those for food allergy, allergic rhinitis, and substance abuse, weakened with wider sibling spacing. Anxiety and depression remained stable across spacing groups, while the autism association was strongest at gaps of four to six years.

In the disease atlas, the most pronounced first-born shift occurred in the neuropsychiatric domain, while dermatological conditions also showed predominantly first-born excess. Respiratory and endocrine/metabolic conditions showed bidirectional associations.

Contrastingly, second-born excess was enriched among digestive, circulatory, genitourinary, musculoskeletal, and infectious conditions, although several domain-level shifts weakened in the within-family analysis. Across the diseases analyzed, the between-family and within-family effect estimates were positively correlated (r = 0.66), with 74.2% directional concordance. Of the 150 Bonferroni-significant findings in the between-family analysis, 127 (84.7%) showed the same direction in the within-family analysis.

Models accounting for state-level differences, stricter clinical rematching, and full-sibling analyses produced highly concordant estimates. However, 110 of the 150 between-family associations moved toward the null in the within-family analysis.

The family-level analysis also examined associations between first-born diagnoses and whether families had a second eligible child. For instance, a first-born autism diagnosis was associated with 13% lower adjusted odds of having a second child (OR, 0.870). However, conditions such as ADHD showed negligible reproductive stoppage (OR, 1.011). The other/unspecified PDD code group was associated with slightly greater odds of having a second child (OR, 1.056), opposing the pattern expected under reproductive stoppage.

For several neurodevelopmental, dermatological, and immune-allergic conditions, the first-born excess was stronger in males, whereas the second-born excess for migraine and substance abuse was stronger in females.

Conclusion

The findings highlight differences in recorded diagnoses by birth order among first- and second-born individuals. Based on the study observations, first-born status was associated with greater odds of several neurodevelopmental and allergic diagnoses, whereas second-born status was associated with greater odds of several gastrointestinal and substance-related diagnoses.

These findings largely agreed between the within-family and between-family analyses and may identify birth order as a modestly informative risk marker relevant to family counseling and screening prioritization. Because claims data capture diagnoses recorded during healthcare encounters rather than true disease incidence, the findings do not establish that birth order causes these conditions. Residual healthcare-seeking differences, parental-age confounding, age-related differences, calendar-period and birth-cohort differences, and the restriction to two-child families also remain important limitations. 

In future studies, researchers should test whether the patterns replicate in larger sibships, Medicaid claims, independent datasets, and more socioeconomically, ethnically, and geographically diverse populations. They should also explore the influence of healthcare-seeking behavior and parental age on the effects of birth order before applying these findings to family counseling and screening prioritization.