Mass spectrometry-based quantitation of cervicovaginal progesterone and steroids in women at high risk of preterm birth.

Preterm birth (PTB), defined by the World Health Organization as a livebirth occurring before the 37th week of gestation, affects ∼10% of pregnancies worldwide, and remains the leading global cause of neonatal and infant mortality and morbidity ( Chawanpaiboon et al. , 2019 ). It has long-lasting social and economic consequences with survivors facing ongoing healthcare needs. Cervical cerclage and progesterone treatment are the main interventions for PTB prevention. Progesterone, typically administered vaginally, is widely recommended for women with short cervical length (≤ 25 mm) between 19 and 23 weeks of gestation or with a previous history of PTB ( Ramachandran et al. , 2024 ). However, evidence for its efficacy remains inconsistent ( Crowther et al. , 2017 ). The mechanistic rationale for progesterone supplementation is based upon the hypothesis that some preterm deliveries may result from localized subclinical progesterone deficiency, leading to phenotypic and morphological changes in the uterus and cervix ( Romero, 2011 ). This ‘progesterone insufficiency’ hypothesis is supported by the role of progesterone in mediating pro-labour and pro-parturition pathways in human gestational tissues ( Rokas et al. , 2020 ), including evidence that progesterone signalling inhibition results in the initiation of parturition and interruption of pregnancy ( Williams et al. , 2012 ; Nadeem et al. , 2016 ). However, quantitation of progesterone in blood, urine, or saliva, has failed to support an association between low progesterone levels during gestation and PTB ( Feng et al. , 2018 ). This may reflect tissue compartmentalization effects. For example, vaginal progesterone supplementation minimally alters circulating progesterone levels ( Meis et al. , 2003 ; Norman et al. , 2011 ), while markedly increases concentrations in endometrial tissue and uterine artery blood ( Cicinelli et al. , 2000 ; Ficicioglu et al. , 2004 ). Assessment of steroid concentrations in tissue or adjacent blood vessels is highly invasive, limiting their utility for studying local steroid contributions to PTB. Cervicovaginal fluid (CVF) provides a safe, minimally invasive alternative, which can be sampled via swabs, tampons, or soft-cup devices ( Hu et al. , 2006 ; Good et al. , 2007 ). CVF contains rich biochemical information that reflects key physiological events including microbe–host interactions, immunological, and hormonal responses ( Pruski et al. , 2021 ). Progesterone has also been detected in CVF using immuno-assays; however, such assays are limited by cross-reactivity between structurally similar compounds and their metabolites ( Krasowski et al. , 2014 ; Lisnawati and Wibowo, 2022 ). Improved selectivity, higher sensitivity, and quantitative accuracy of steroid hormones can be achieved using mass spectrometry (MS)-based approaches ( Stanczyk and Clarke, 2010 ). These have been largely developed for quantitation of endogenous steroids in serum, urine, and saliva samples using both gas-chromatography-mass spectrometry (GC-MS) ( Santen et al. , 2007 ), and liquid chromatography–tandem mass spectrometry (LC-MS/MS) ( Xu et al. , 2007 ). In this study, we report the adaptation of an LC-MS/MS assay for measuring progesterone and other related steroid hormones in CVF and use it to determine and compare steroid hormone concentrations in CVF during pregnancy in women who subsequently experience PTB with gestational age-matched term pregnancies.

Although systemic levels of progesterone remain high during parturition, it has been hypothesized that local progesterone deficiency in the cervicovaginal niche could lead to premature cervical shortening, and PTB ( Romero, 2011 ). To examine this, we recruited 25 women from a specialist preterm birth prevention clinic, who provided CVF samples. Of these, 20 were sampled cross-sectionally and 5 were sampled longitudinally at two time points. All samples were collected between 12 and 24 weeks of gestational age (median gestational age at sampling = 17.6 weeks). Participants were retrospectively grouped into an uncomplicated outcome group (n = 14), comprising women who delivered at term without intervention, and an adverse outcome group (n = 11), which included women who experienced preterm birth (n = 9), miscarriage (n = 1), or required intervention to achieve term delivery (n = 1) ( Fig. 1B ). Of the 13 targeted steroids included in our UPLC-MS/MS assay, only progesterone, androstenedione, and cortisol were detected above the lower limits of quantitation (LLOQ) (0.005 ng/ml, 0.01 ng/ml, 0.05 ng/ml, respectively) ( Supplementary Table S3 ). Representative raw LC-MS MRM chromatograms are provided in Supplementary Fig. S2 . In CVF collected from term pregnancy samples (n = 19), the median (± SD) concentrations of the detected steroids were 6.61 ± 8.13 ng/g for progesterone, 4.16 ± 2.98 ng/g for cortisol, and 0.34 ± 0.21 ng/g for androstenedione ( Fig. 1C and Supplementary Fig. S3 ). One of the term patients required medically assisted labour. However, their levels of steroids were consistent with other term patients who did not require medical assisted labour (data not shown). Linear regression analyses showed no significant associations between progesterone, cortisol, or androstenedione concentrations and cervical length at sampling ( Fig. 2A ). Steroids levels were also found to be similar between women who subsequently delivered at term (n = 15) and those who delivered preterm (n = 9) (Wilcoxon–Mann–Whitney test: progesterone P  = 0.766; cortisol P  = 0.512; androstenedione P  = 0.18). A participant who miscarried was excluded from this comparison ( Fig. 2B ). Additionally, no significant relationships were observed between cervicovaginal steroid levels and gestational age at delivery ( Supplementary Fig. S4 ). These findings indicate that CVF progesterone levels in the second trimester do not differ meaningfully between women who deliver at term and those who experience spontaneous PTB. Consistent with these findings, multivariate analyses incorporating progesterone, cortisol, and androstenedione levels as well as maternal confounders did not identify distinct steroid profiles associated with adverse pregnancy outcomes ( Supplementary Fig. S5 ). Relationship between steroids concentration, cervical length, and preterm birth. ( A ) Steroid concentrations and cervical length at sampling. Linear regression analyses ( lm(steroid concentration ∼ cervical length) ) revealed no significant associations between cervical length at sampling and cervicovaginal fluid (CVF) concentrations of progesterone ( r 2 = 0.001, P  = 0.871), cortisol ( r 2 = 0.021, P  = 0.486), or androstenedione ( r 2 = 0.01, P  = 0.633). ( B ) Concentrations of progesterone, cortisol, and androstenedione in women who delivered at term (green, n = 15) versus preterm (red, n = 9). No statistically significant differences were observed between the two groups for any of the analytes (Wilcoxon–Mann–Whitney test: progesterone P  = 0.766, cortisol P  = 0.512, androstenedione P  = 0.18). The sampling window examined in this study was selected due to clinical guidelines recommending the commencement of preventative progesterone therapy occur between 16 and 24 weeks ( Jain et al. , 2020 ). A recent cross-sectional study of pregnant Indonesian women sampled late in the third trimester (28–34 weeks), reported that CVF progesterone levels measured by immuno-assay were marginally higher in women who subsequently delivered preterm ( Lisnawati and Wibowo, 2022 ). Comparisons of these findings with our own are challenging as clinical metadata were incomplete and progesterone therapy status was unknown. Further, reported concentrations for progesterone were uncorrected for CVF sample weight. Despite this, we observed comparable median CVF progesterone concentrations in term outcomes when recalculated using a similar approach (1.32 ± 1.63 ng/ml in our study vs 1.74 ± 1.19 ng/ml reported by Lisnawati and Wibowo (2022 ). These findings collectively indicate that the concentration of local cervicovaginal progesterone is unlikely to be a key determinant of preterm birth risk. However, it is important to acknowledge that patients who initiated progesterone therapy in early pregnancy, prior to our sampling window, were excluded from our study. This may have led to an underrepresentation of those experiencing progesterone-dependent adverse outcomes and should be further explored in future studies. To examine gestational dynamics of measured steroids, linear mixed-effects models incorporating participant ID as a random intercept were used to assess associations between steroid concentrations and gestational age at sampling. Progesterone levels showed a weak but significant positive association with increasing gestational age ( R 2 = 0.181, P  = 0.0215; Fig. 3A ). No significant associations between gestational age and levels of cortisol or androstenedione were observed (cortisol: R 2 = 0.002, P  = 0.830; androstenedione: R 2 = 0.008, P  = 0.590; Fig. 3A ). A paired analysis of samples collected from the same individuals (n = 5) at the beginning (12–16 weeks) and end of the second trimester (20–24 weeks) did not show statistically significant increases in progesterone, cortisol, or androstenedione levels with advancing gestation (Wilcoxon signed-rank test: progesterone P  = 0.313, cortisol P  = 0.813, androstenedione P  = 0.313; Fig. 3B ). This may be attributable to the small sample size, as paired samples were only available for term uncomplicated pregnancies. Our findings showing progesterone increases in the cervicovaginal niche during the second trimester, reflect those observed systemically ( Ku et al. , 2018 ). In contrast, local cortisol and androstenedione concentrations remain unchanged despite previous studies reporting systemic increase during the same period ( Makieva et al. , 2014 ; Vrijkotte et al. , 2023 ). Relationship between gestational age and steroid concentrations in cervicovaginal fluid (CVF). ( A ) Association between gestational age at sampling and cervicovaginal fluid steroid concentrations assessed using linear mixed-effects models, with participant ID included as a random intercept to account for repeated measurements. Progesterone showed a weak but significant positive association with gestational age (marginal R 2 = 0.176, P  = 0.0215). No significant associations were observed for cortisol (marginal R 2 = 0.001, P  = 0.829) or androstenedione (marginal R 2 = 0.001, P  = 0.589). ( B ) Paired analysis of longitudinal samples (n = 5) collected at early (12–16 weeks) and late (20–24 weeks) second trimester. ( C ) Boxplot representation of the longitudinal samples summarizing the distribution of progesterone, cortisol, and androstenedione concentrations at the early (A; 12–16 weeks) and late (B; 20–24 weeks) second-trimester timepoints. The box represents the interquartile range (25th–75th percentile), the horizontal line indicates the median, and the whiskers represent the minimum and maximum values. The absence of detectable levels of many of the steroids included in our assays, and the divergence between local and systemic levels, can be explained by several possible mechanisms. Tissue-specific expression of steroidogenic enzymes results in differential local metabolism and clearance compared with systemic circulation, with processes such as hydroxylation, reduction, and dehydrogenation playing key roles in steroid metabolism ( Wang and Abe, 2024 ). Steroid conjugation, particularly sulfation and glucuronidation, further regulates local availability. For example, sulfotransferases (SULTs) like SULT1E1, which is highly expressed in reproductive tissues, catalyse steroid sulfation modulating local oestrogen levels. SULT2A1 preferentially sulfates androgens in placental and foetal tissues ( Stanley et al. , 2001 ). In contrast, steroid sulfatase (STS) regenerates active steroids and is strongly expressed in the placenta and late pregnancy leukocytes ( Miyakawa et al. , 1994 ). Glucuronidation by UDP-glucuronosyltransferases (UGTs) also contributes to steroid homeostasis and clearance. Several UGT isoforms are expressed in reproductive tissues and implicated in regulating local oestrogen bioavailability during pregnancy ( Meech et al. , 2019 ). Pregnancy-related hormonal changes may thus promote steroid conjugation in CVF, rendering some steroids undetectable with the primary MS assay used in this study. Consistent with this, a pilot analysis of CVF using quadrupole time-of-flight (qTOF) mass spectrometry in both positive and negative ionization modes, indicates the presence of sulphated and glucuronidated derivatives of DHEAS, testosterone, and pregnenolone ( Supplementary Fig. S1 ). A limitation of our study was that we did not investigate the potential role of the vaginal microbiome to mediate steroid metabolism and availability. While in the gut, activity and function of steroids is mediated via microbial modification ( McCurry et al. , 2024 ), our previous work indicates that vaginal progesterone does not change microbiota composition ( Kindinger et al. , 2017 ). Potential microbial mediation of local steroid metabolism and activity warrants further investigation. Our study also included a restricted sampling window and population, including exclusion of some women who had already initiated progesterone therapy before sampling. The limited gestational age window (15–24 weeks) may have restricted detection of biomarker changes occurring outside this period. Longitudinal sampling in high-risk women is not feasible because these patients routinely receive progesterone or other standard treatments, and withholding therapy to observe natural steroid trajectories would be inappropriate. Although beyond the scope of the current study, prospective longitudinal profiling is achievable in low-risk populations and could determine whether repeated local steroid measurements provide clinically meaningful information. Vaginal progesterone treatment also poses analytical challenges, including carryover and system saturation, underscoring the need for further customization of the assay in future work. In summary, we present a robust UPLC-MS/MS method for detecting progesterone, androstenedione, and cortisol in CVF. While our data show that CVF progesterone levels increase with gestational age, they do not support a link between CVF progesterone deficiency and preterm birth. The presented method enables larger studies to assess the clinical relevance of local steroid dynamics, particularly early in pregnancy, where local measurements may help distinguish normal from altered patterns of cervical remodelling. More broadly, the capacity to quantify local steroid dynamics opens avenues for applications across women’s health and reproductive medicine, including studies of endometriosis, IVF-related physiology, and other conditions in which local hormonal activity is likely to play an important biological role.

Detailed methods and analytical workflows are provided in the Supplementary Materials and Methods . Briefly, CVF samples were obtained from pregnant women enrolled in the Vaginal Microbiome and Metabolome in Pregnancy 2 (VMET2) study who provided written, informed consent. The study was conducted with approval of the National Research Ethics Service Committee London–Stanmore of the National Health Service (REC 14/LO/0328). Participants with singleton pregnancies and clinical risk factors for preterm birth were recruited from specialist clinics in London, UK. Women receiving progesterone before sampling or meeting predefined exclusion criteria were omitted. CVF was collected using a menstrual disc, diluted in 5 ml of phosphate-buffered saline, aliquoted, and stored at −80°C. Cohort characteristics are summarized in Supplementary Table S1 . Steroids were extracted from CVF using sequential organic solvent extraction with isotopically labelled internal standards, vacuum dried, and reconstituted prior to analysis. Targeted quantification of 13 steroid hormones was performed using UPLC-MS/MS with polarity switching and multiple reaction monitoring ( Fig. 1A ). Chromatographic conditions, instrument parameters, analytes, and internal standards are detailed in the Supplementary Materials and Methods and Supplementary Table S2 . Additional high-resolution qTOF experiments were performed to investigate the presence of conjugated steroids ( Supplementary Fig. S1 ). Raw data were processed in TargetLynx using linear calibration with isotopic normalization within MassLynx (v4.1) software (Waters Corporation, Milford, MA, USA). Quantification limits and calibration procedures are described in the Supplementary Materials and methods . Statistical analyses were conducted in R using linear models to assess associations with gestational age and Wilcoxon–Mann–Whitney tests for group comparisons. Overview of sample processing, study design, and steroid concentrations in cervicovaginal fluid (CVF). ( A ) Schematic workflow of the sequential solvent extraction process for CVF sample preparation. (Created in BioRender. MacIntyre, D. (2026) https://BioRender.com/a181syj ) ( B ) Schematic illustrating participant gestational age at sampling and classification into uncomplicated outcome and adverse outcome groups. ( C ) Measured concentrations of progesterone, androstenedione, and cortisol in CVF from term pregnancy samples.