STUDY QUESTION Do maternal serum levels of progesterone metabolites early in pregnancy correspond to an increased risk for very preterm delivery prior to 32 weeks? SUMMARY ANSWER Maternal serum levels of 11-deoxycorticosterone (DOC) measured during the late first trimester or early second trimester correlate with an increased risk for preterm delivery prior to 32 weeks, and the correlation becomes stronger when the ratio of DOC to 16-alpha-hydroxyprogesterone was measured. WHAT IS KNOWN ALREADY Progesterone is a pro-gestational steroid hormone that has been shown to decrease the risk of preterm birth in some pregnant women. Progesterone is metabolized by the body into various metabolites including members of the mineralocorticoid and glucocorticoid families. Our group has previously demonstrated that some progesterone metabolites enhance myometrial contractility in an ex vivo system, while others result in myometrial relaxation. The current exploratory study was designed to determine if pre-specified metabolites of progesterone measured early in pregnancy were associated with a woman’s risk for delivery prior to 32 weeks, which is referred to as a very preterm delivery. STUDY DESIGN, SIZE, DURATION The Building Blocks of Pregnancy Biobank (BBPB) is a biorepository at Indiana University (IU) that follows women prospectively through their pregnancy. A variety of biospecimens are collected at various time points during a woman’s pregnancy. Women participating in the IU BBPB who were enrolled after 8 weeks’ gestation with pregnancy outcome data were eligible for participation. PARTICIPANTS/MATERIALS, SETTING, METHODS Women delivering prior to 37 weeks (preterm) and at or after 37 weeks (term) who had blood samples collected during the late first trimester/early second trimester and/or during the early third trimester were identified. These samples were then processed for mass spectroscopy, and the amount of progesterone and progesterone metabolites in the samples were measured. Mean values of each measured steroid metabolite were calculated and compared among women delivering at less than 32 weeks, less than 37 weeks and greater than or equal to 37 weeks. Receiver operating characteristic (ROC) curves were constructed and threshold levels determined for each compound to identify a level above or below which best predicted a woman’s risk for delivery prior to 32 and prior to 37 weeks. Mann–Whitney U nonparametric testing with Holm–Bonferroni correction for multiple comparisons was utilized to identify steroid ratios that could differentiate women delivering spontaneously at less than 32 weeks from all other pregnancies. MAIN RESULTS AND THE ROLE OF CHANCE Steroid hormone levels and pregnancy outcome data were available for 93 women; 28 delivering prior to 32 weeks, 40 delivering between 32 0/7 and 36 6/7 weeks and 25 delivering at or greater than 37 weeks: the mean gestational age at delivery within the three groups was 27.0, 34.4 and 38.8 weeks, respectively. Among women delivering spontaneously at less than 37 weeks, maternal 11-deoxycorticosterone (DOC) levels drawn in the late first trimester/early second trimester were significantly associated with spontaneous preterm delivery prior to 32 weeks; a threshold level of 47.5 pg/ml had 78% sensitivity, 73% specificity and an AUC of 0.77 (P = 0.044). When DOC levels were analyzed as a ratio with other measured steroid hormones, the ratio of DOC to 16-alpha-hydroxyprogesterone among women delivering spontaneously prior to 37 weeks was able to significantly discriminate women delivering prior to 32 weeks from those delivering at or greater than 32 weeks, with a threshold value of 0.2 with 89% sensitivity, 91% specificity and an AUC of 0.92 (P = 0.002). When the entire study cohort population was considered, including women delivering at term and women having an iatrogenic preterm delivery, the ratio of DOC to 16-alpha-hydroxyprogesterone was able to discriminate women delivering spontaneously prior to 32 weeks from the rest of the population at a threshold of 0.18 and 89% sensitivity, 59% specificity and an AUC of 0.81 (P = 0.003). LIMITATIONS, REASONS FOR CAUTION This is a discovery study, and the findings have not been validated on an independent cohort. To mitigate issues with multiple comparisons, we limited our study to pre-specified metabolites that are most representative of the major metabolic pathways for progesterone, and adjustments for multiple comparisons were made. WIDER IMPLICATIONS OF THE FINDINGS Spontaneous preterm birth is increasingly being recognized to represent a common end pathway for a number of different disease phenotypes that include infection, inflammation, premature rupture of the membranes, uterine over distension, cervical insufficiency, placental dysfunction and genetic predisposition. In addition to these phenotypes, longitudinal changes in the maternal–fetal hypothalamic–pituitary–adrenal (HPA) axis also likely contribute to a significant proportion of the disease burden of spontaneous preterm birth. Here, we demonstrate that differential production of steroid metabolites is associated with very early preterm birth. The identified biomarkers may hint at a pathophysiologic mechanism and changes in the maternal–fetal dyad that result in preterm delivery. The early identification of abnormal changes in HPA axis metabolites may allow for targeted interventions that reverse the aberrant steroid metabolic profile to a more favorable one, thereby decreasing the risk for early delivery. Further research is therefore required to validate and extend the results presented here. STUDY FUNDING/COMPETING INTEREST(S) Funding for this study was provided from the Office of the Vice Chancellor for Research at IUPUI, ‘Funding Opportunities for Research Commercialization and Economic Success (FORCES) grant’. Both A.S.P. and C.A.G. are affiliated with Nixxi, a biotech startup. The remaining authors report no conflict of interest. TRIAL REGISTRATION NUMBER Not applicable.