Abstract
Steroid hormones play a critical role in maintaining pregnancy and supporting fetal development. Accurate quantification of these hormones is essential for evaluating the endocrine status during pregnancy. While chemiluminescence immunoassay （CLIA） is widely used in clinical practice in China， it has inherent methodological limitations. In contrast， liquid chromatography-tandem mass spectrometry （LC-MS/MS） has been internationally recognized for its superior analytical performance but remains underutilized in Chinese obstetric practice. Moreover， systematic evaluations comparing these two methods in pregnant women are still lacking. This study aimed to develop and validate an isotope-dilution LC-MS/MS method for the simultaneous quantification of four key steroid hormones， including cortisol， dehydroepiandrosterone sulfate （DHEAS）， testosterone， and 17-hydroxyprogesterone （17-OHP）. In addition， a consistency analysis was conducted between LC-MS/MS and CLIA among pregnant women. Method validation followed international bioanalytical guidelines， evaluating linearity， limits of quantification （LOQs）， recovery， and precision. Serum samples from 94 pregnant women were collected at Guangdong Women and Children Hospital between January 2022 and December 2023， and were analyzed using both LC-MS/MS and CLIA. Method comparison was performed using paired t-tests， Pearson correlation， Bland-Altman plots， and Passing-Bablok regression to assess concentration differences， correlation， and systematic biases between the two methods. In this study， the established LC-MS/MS method exhibited excellent performance， with linear coefficients of determination >0.999， LOQs of 0.008-0.137 ng/mL， spiked recoveries of 92.1%-110.9%， and intra-assay relative standard deviations （RSDs） of 3.2%-9.0%， meeting the requirements for clinical detection. In a cohort of 94 pregnant women， paired t-tests revealed significant differences in hormone concentrations measured by LC-MS/MS and CLIA （all P<0.001）. Specifically， LC-MS/MS yielded significantly lower values for cortisol （64.31 vs. 120.60 ng/mL） and DHEAS （532.44 vs. 1 612.1 ng/mL）， but higher values for testosterone （1.10 vs. 0.88 ng/mL） and 17-OHP （2.00 vs. 1.07 ng/mL）. Bland-Altman analysis further revealed obvious negative proportional biases for cortisol and DHEAS， with bias values of -56.29 ng/mL and -1 079.68 ng/mL， respectively. Additionally， the biases increased with the increase of concentration. In contrast， testosterone and 17-OHP showed positive biases， with bias values of 0.22 ng/mL and 0.89 ng/mL， respectively. The two methods showed extremely strong positive correlations in the detection of the four hormones （P<0.001， r=0.819-0.974）， among which the correlation of testosterone detection results was the highest （r=0.974）. In the Passing-Bablok regression analysis， the consistency regression equation for cortisol was C(LC-MS/MS)=-3.58+0.57C(CLIA). The confidence interval （CI） of the slope did not include 1， indicating obvious proportional bias. The consistency regression equation for DHEAS was C(LC-MS/MS)=-59.77+0.38C(CLIA). The CIs of the intercept and slope did not include 0 and 1， respectively， suggesting the existence of fixed and proportional biases. The consistency regression equation for testosterone was C(LC-MS/MS)=-0.05+1.30C(CLIA). The CIs of the intercept and slope did not include 0 and 1， respectively， indicating the existence of fixed and proportional biases. In addition， the differences increased with the increase of concentration. For 17-OHP， the regression equation was C(LC-MS/MS)=0.10+1.75C(CLIA). Though the intercept was not statistically significant， the slope deviated significantly from 1， indicating a proportional bias. Additionally， LC-MS/MS values were significantly higher， and the difference increased with rising concentration. In conclusion， the established LC-MS/MS method offers high sensitivity， precision， and accuracy for the simultaneous quantification of multiple steroid hormones. The method reliably measured cortisol， DHEAS， testosterone， and 17-OHP levels in maternal serum. Additionally， our results revealed statistically significant differences between the two methods across all four hormones in pregnant women， with biases varying by concentration. These findings highlight the importance for clinicians to consider methodological differences when interpreting hormone test results during pregnancy， in order to improve the accuracy of endocrine assessment and reduce the risk of misdiagnosis due to analytical discrepancies. Future studies should incorporate large-scale maternal cohorts covering all stages of pregnancy and high-risk populations to further assess the clinical applicability of different methods and promote the standardization of LC-MS/MS for obstetric endocrine monitoring. In parallel， trimester-specific reference intervals should be established to support clinical interpretation and risk evaluation.