Predicting natural conception leading to live birth for couples with infertility: a single-centre population-based cohort study of 7086 couples

Abstract STUDY QUESTION Can a model be developed to predict natural conception leading to live birth over a period of 12 months in couples with different causes of infertility? SUMMARY ANSWER We developed and validated a novel clinical prediction model that can be used to provide individualized estimates of the chance of a natural conception (leading to live birth) over a 12-month period for couples with different causes of infertility, including tubal factor, anovulation, male factor, unexplained, other (cervical, uterine, or sexual factor), and endometriosis. WHAT IS KNOWN ALREADY Existing prediction models for natural conception are primarily aimed at infertile couples with no identified cause (unexplained), mild male factor, or minimal endometriosis. Currently, there is no available model that is applicable across a wide range of fertility diagnoses at the point of initial assessment. STUDY DESIGN, SIZE, DURATION A population-based cohort study based on data from a single large tertiary UK fertility centre serving the entirety of the North of Scotland, including all 9757 couples who registered at the clinic for the first time from 1998 to 2015. PARTICIPANTS/MATERIALS, SETTING, METHODS Using a Cox proportional hazards survival model, we estimated the chance of conception within the first year from diagnosis of infertility leading to live birth. The predictive accuracy of the model was assessed using discrimination and calibration measures in an internal validation. The clinical utility of the model was assessed using decision curve analysis. MAIN RESULTS AND THE ROLE OF CHANCE After exclusions, 7086 couples with infertility were included, of whom 891 (13%) had a natural conception within 1 year of diagnosis leading to live birth. Longer duration of infertility [hazard ratio (HR): 0.64 (95% CI: 0.57–0.72)], increasing female age [0.62 (0.55–0.70)], and tubal infertility [0.62 (0.47–0.81)] had the strongest influence on reducing the chance of natural conception within 1 year leading to live birth. Conversely, previous pregnancy [1.30 (1.12–1.50)] and unexplained infertility [1.29 (1.00–1.67)] were most strongly associated with increased chances of live birth. For example, a 25-year-old woman with 1 year of secondary unexplained infertility has a 33.8% predicted chance of having a baby resulting from natural conception within a year of diagnosis. In contrast, a 35-year-old woman with tubal infertility and 1 year of primary infertility has an estimated 6.0% predicted chance of live birth over a similar time horizon. The model demonstrated acceptable discrimination (optimism-corrected C-statistic 0.645) and good calibration in internal validation. Decision curve analysis demonstrated that applying the model at a 20% probability threshold would correctly identify an additional 7.2 couples per 100 as being likely to benefit from fertility treatment when compared against a ‘treat-all’ policy, which incorrectly assumes that all couples would benefit from fertility treatment. A freely available online calculator was constructed from the model formula to generate individual estimates of the chance of live birth following natural conception. LIMITATIONS, REASONS FOR CAUTION The data used to inform this model were collected from a single centre until 2016, limiting generalizability. Any information not collected in this dataset, such as markers of ovarian reserve, severity of endometriosis, or subtype of ovulatory disorder, could not be included and may have improved the accuracy of the model predictions. WIDER IMPLICATIONS OF THE FINDINGS As we have demonstrated that couples diagnosed with infertility can conceive on their own, knowledge of their individual chances of having a baby can clarify the net value of active treatment, including assisted reproduction. Use of this model via the online calculator could improve the quality of decision-making around the nature and timing of initiating fertility treatment. FUNDING This work was funded in part by an NHS Grampian Charities Small Research Grant awarded to N.J.C, grant number SRG 24-30. We acknowledge the associated financial support of NHS Research Scotland, through NHS Grampian investment in the Grampian DaSH. DISCLOSURES N.J.C has no conflict of interest to declare. K.B. declares a previous role as Founder of the Fertility Alliance (charity) and receipt of speaker fees/honoraria from Merck and IBSA (paid to the Fertility Alliance). D.J.M. reports consulting fees from the Society for Assisted Reproductive Technology (SART) (paid to institution), honoraria from Merck for a fertility prediction modelling workshop, and financial support from ESHRE, Merck and IVIRMA to present at scientific conferences. S.B. reports royalties/licenses from Cambridge University Press, consulting fees from Merck, Ferring and Organon (paid to S.B. and to the institution), speaker fees/honoraria from Merck (paid to S.B. and to the institution), support for attending meetings/travel from Merck, and a leadership role as Board Member of the Fertility Alliance (charity). TRIAL REGISTRATION NUMBER N/A.