Obstetrics & Gynecology · Medical Literature
Identification of predictors of the ovarian response to clomiphene citrate in infertile women with polycystic ovary syndrome: A post-hoc analysis of a randomized controlled trial.
Abstract
Clomiphene citrate (CC) is the first-line medication for inducing ovulation in women with polycystic ovary syndrome (PCOS). However, approximately 20% of patients with PCOS are resistant to CC. This study aims to identify reliable baseline predictors of CC resistance in infertile women with PCOS. A post-hoc analysis of a large, multicenter randomized controlled trial (PCOSAct trial) conducted in China. The current analysis comprised the 471 participants who were randomized to the active CC arm and completed the requisite follow-up. To identify potential candidate variables, we employed multivariable logistic and LASSO regression analyses. Within the framework of a multivariable logistic regression model, we also estimated the independent associations between the identified candidate variables and resistance to CC. Additionally, we plotted the Receiver Operating Characteristic (ROC) curve and utilized the DeLong method to compare the statistical differences in the area under the curve (AUC). Finally, we constructed a restricted cubic spline (RCS) logistic regression model to illustrate the dose-response relationship between continuous predictor variables and CC resistance. CC resistance was identified in 32 (6.8%) participants. Body Mass Index (BMI), Total Testosterone (TT), and Anti-Müllerian Hormone (AMH) were useful predictors of ovarian response to CC. The "T+BMI" dual-factor model demonstrated high discriminative power (AUC = 0.801) and was statistically comparable to the three-factor model including AMH (AUC = 0.818; P = 0.347). TT was the strongest individual predictor (OR = 2.73 per 1-unit), while BMI was the most significant modifiable risk factor (OR = 2.49 per 1-SD). A simplified "T + BMI" assessment provides comparable prognostic utility without the need for AMH testing. For patients at high risk of CC resistance, we recommend upfront use of aromatase inhibitors or low-dose gonadotropins. This strategy avoids ineffective treatment cycles and enables personalized ovulation induction. The study was registered on ClinicalTrials.gov under the identification number NCT01573858 on July 6, 2012.
Clinical relevance
Among infertile women with PCOS receiving clomiphene citrate (CC), which baseline clinical and biochemical variables independently predict CC resistance, and can a simplified predictor model achieve clinically acceptable discriminative performance? __SNAPSHOT__ {"studyDesign":"Post-hoc analysis of a multicenter randomized controlled trial (PCOSAct trial)","setting":"27 hospitals in mainland China; multicenter academic setting; original trial conducted July 2012–November 2014","sampleSize":"n=471 (participants randomized to the active CC arm who completed follow-up, drawn from the 1,000-participant PCOSAct trial)","population":"Chinese women with PCOS-related anovulatory infertility diagnosed by modified Rotterdam criteria (oligomenorrhea or anovulation plus clinical/biochemical hyperandrogenism and/or polycystic ovaries); no concurrent insulin sensitizers or other PCOS medications at enrollment","intervention":"Clomiphene citrate 50 mg orally daily, cycle days 3–7, for up to four consecutive menstrual cycles; candidate predictors of resistance identified via multivariable logistic regression and LASSO regression applied to baseline demographic, anthropometric, hormonal, and metabolic variables","comparator":"Not applicable (single-arm post-hoc analysis of the CC-only group)","primaryOutcome":"CC resistance, defined as failure to achieve ovulation across all treatment cycles; discriminative performance assessed by ROC-AUC; dose-response relationships modeled via restricted cubic spline (RCS) logistic regression","followUp":"Four consecutive menstrual treatment cycles (duration in months not explicitly reported)","fundingSource":"Not reported in the abstract; the parent PCOSAct trial was a government/academic-sponsored trial registered at ClinicalTrials.gov (NCT01573858) and the Chinese Clinical Trial Registry (ChiCTR-TRC-12002081); no industry sponsorship reported"}
Key findings
CC resistance was identified in 32 of 471 participants (6.8%). This observed resistance rate is notably lower than the commonly cited population estimate of approximately 20%, a discrepancy that may reflect selection effects inherent in the parent trial population, the fixed 50 mg dose without escalation, or the four-cycle observation window. Three baseline variables emerged as independent predictors of CC resistance: Body Mass Index (BMI), Total Testosterone (TT), and Anti-Müllerian Hormone (AMH). TT was the strongest individual predictor (OR = 2.73 per 1-unit increase), while BMI was identified as the most significant modifiable risk factor (OR = 2.49 per 1 standard deviation increase). A dual-factor 'T+BMI' model demonstrated AUC = 0.801, which was not statistically significantly different from the three-factor model incorporating AMH (AUC = 0.818; P = 0.347 by DeLong method), indicating that AMH adds no statistically significant incremental discriminative value over TT and BMI alone in this dataset. Restricted cubic spline modeling was used to characterize the dose-response relationships between continuous predictor variables (BMI, TT, AMH) and CC resistance probability. Specific cut-off values, confidence intervals for individual ORs, sensitivity, specificity, and NNT/NNH were not reported in the abstract; these values are therefore not reported here.
Practice implications
This post-hoc analysis of a large, well-characterized Chinese PCOS cohort establishes that higher baseline BMI, total testosterone, and AMH are independently associated with CC resistance, and that a simplified 'T+BMI' two-variable screening model achieves good discrimination (AUC ~0.80) without requiring AMH measurement. This finding has potential clinical utility: if externally validated, pre-treatment assessment of BMI and total testosterone may allow clinicians to identify patients at elevated risk for CC failure and either counsel weight reduction before pharmacological treatment or proceed more directly to alternative agents such as letrozole or low-dose gonadotropins. However, several important limitations constrain immediate translation. This analysis is post-hoc and exploratory by design; the predictors identified require prospective validation in independent cohorts before they can be recommended for clinical decision-making. The study population is exclusively Chinese women diagnosed by modified Rotterdam criteria, and results may not generalize to other ethnicities, healthcare settings, or women diagnosed by NIH or ESHRE criteria. The observed CC resistance rate of 6.8%—substantially lower than the widely reported 20–25%—raises questions about dose adequacy (fixed 50 mg only, no escalation to 100–150 mg as used in many clinical protocols), patient selection, or the definition of resistance applied, and merits scrutiny before the model's performance is assumed to apply to clinical practice. The study also does not report live birth rates, clinical pregnancy rates, or patient-important reproductive outcomes as endpoints; ovulatory failure is a surrogate. Availability of this analysis only in abstract form at time of summary limits full critical appraisal. __LIMITATIONS__ Post-hoc, exploratory design with attendant risk of overfitting and multiple comparisons; predictors require prospective validation.,Fixed CC dose of 50 mg only (no dose escalation to 100–150 mg), which is inconsistent with standard clinical practice and likely contributes to the unexpectedly low resistance rate of 6.8%, potentially inflating the apparent predictive value of baseline variables.,Exclusively Chinese population limits generalizability to other ethnicities and healthcare systems.,Parent RCT enrolled participants between 2012 and 2014; temporal generalizability to contemporary clinical cohorts uncertain.,No patient-important outcome (live birth, clinical pregnancy) modeled as the dependent variable; ovulatory failure is a surrogate endpoint.,Specific OR confidence intervals, cut-off values, sensitivity, and specificity for the individual predictors and composite models not reported in abstract; full critical appraisal requires access to complete manuscript.,No insulin sensitizers were permitted in the parent trial, reducing confounding but limiting applicability to real-world patients commonly co-managed with metformin.,Sample size of n=471 with only 32 resistance events (6.8%) raises concerns about model stability and adequate power for multivariable analysis including LASSO and RCS components.,Funding source for the current post-hoc analysis is not reported; the original PCOSAct trial's sponsorship structure should be reviewed for potential conflicts.,The DeLong method p-value of 0.347 confirms non-inferiority of the simpler T+BMI model but does not constitute formal equivalence testing; the clinical threshold for AMH cost-benefit decisions requires further evaluation. __BOTTOM_LINE_CATEGORY__ promising_but_preliminary __BOTTOM_LINE__ In a post-hoc analysis of 471 Chinese women with PCOS treated with CC, baseline BMI and total testosterone together predicted CC resistance with an AUC of 0.801, statistically equivalent to a three-variable model also including AMH, suggesting that expensive AMH testing may not be required for pre-treatment risk stratification. These findings are hypothesis-generating and require prospective validation before clinical adoption, particularly given the unusually low observed resistance rate (6.8%) that likely reflects the fixed low-dose CC protocol. Clinicians managing PCOS-related anovulatory infertility should note that elevated BMI and hyperandrogenism remain the most actionable, modifiable pre-treatment risk factors for CC failure.