Autism spectrum disorder (ASD) affects more than 1% of children in the USA. The male-to-female prevalence ratio of roughly 4:1 in ASD is a well-recognized but poorly understood phenomenon. An explicit focus on potential etiologic pathways consistent with this sex difference, such as those involving prenatal androgen exposure, may help elucidate causes of ASD. Furthermore, the multi-threshold liability model suggests that the genetic mechanisms in females with ASD may be distinct and may modulate ASD risk in families with female ASD in the pedigree. We examined umbilical cord blood from 137 children in the Early Autism Risk Longitudinal Investigation (EARLI) cohort. EARLI is an ASD-enriched risk cohort with all children having an older sibling already diagnosed with ASD. Fetal testosterone (T), androstenedione (A4), and dehyroepiandrosterone (DHEA) levels were measured in cord blood using liquid chromatography-tandem mass spectrometry (LC-MS/MS). Robust linear regression models were used to determine associations between cord blood androgen levels and 12-month Autism Observation Scales for Infants (AOSI) scores and 36-month Social Responsiveness Scale (SRS) scores adjusting for potential confounders. Increasing androgens were not associated with increasing 12-month AOSI score or 36-month total SRS score in either boys or girls. However, the association between T and autistic traits among subjects with a female older affected sibling was greater at 12 months (test of interaction, P = 0.008) and deficits in reciprocal social behavior at 36 months were also greater (test of interaction, P = 0.006) than in subjects whose older affected sibling was male. While increased prenatal testosterone levels were not associated with autistic traits at 12 or 36 months, our findings of a positive association in infants whose older ASD-affected siblings were female suggests an androgen-related mechanism that may be dependent on, or related to, genetic liability factors present more often in families containing female ASD cases. However, this initial finding, based on a small subgroup of our sample, should be interpreted with considerable caution.