Altered serum vitamin profiles and a PCOS-specific vitamin A–DHEAS association: a three-group case-control study

Abstract Objectives Polycystic ovary syndrome (PCOS) is a common endocrine and metabolic disorder affecting reproductive-aged women. While individual micronutrients have been investigated, the integrated multi-vitamin landscape and its relationship with hyperandrogenism remain incompletely understood. This study aimed to characterize serum vitamin profiles via liquid chromatography-tandem mass spectrometry (LC-MS/MS) and explore their associations with androgen levels across a three-arm cohort. Methods Nine vitamins and three androgens were quantified via LC-MS/MS in 184 women: healthy controls (HC, n = 61), women with PCOS ( n = 66), and non-PCOS infertile women (NPI, n = 57). Between-group differences were adjusted for age and body mass index (BMI). The vitamin A–DHEAS association was formally tested using interaction regression and Fisher’s r -to- z tests. Diagnostic performance was evaluated using nested logistic regression with leave-one-out cross-validation. Results After adjustment, PCOS was associated with elevated vitamin A and lower vitamin E, B 1 , and 25(OH)D 2 relative to HC (all p < 0.001, FDR q < 0.05). Vitamin A correlated positively with DHEAS specifically in PCOS ( r = 0.394, p = 0.001), but not in HC ( r = − 0.007) or NPI ( r = 0.021); interaction regression ( β = 2.600, p = 0.005) and Fisher’s r -to- z tests ( p = 0.020 − 0.033) confirmed this group difference. Adding vitamin A to a testosterone-based model improved the AUC from 0.823 to 0.877 for distinguishing PCOS from non-PCOS (DeLong p = 0.035). Conclusions This LC-MS/MS study reveals distinct serum vitamin profiles in women with PCOS, including a statistically supported PCOS-specific vitamin A–DHEAS association. These findings indicate that altered vitamin status is part of the endocrine-metabolic phenotype of PCOS, and that vitamin A adds incremental diagnostic information beyond conventional androgens. However, the diagnostic models remain exploratory and require external validation in independent cohorts before clinical application.