A revised prediction model for natural conception

Highlights

• The Hunault model is the standard prediction model for natural conception.

• This model can be revised by including additional predictors.

• The revised model is applicable to a broader population.

• The revised model needs to be externally validated.

Abstract

One of the aims in reproductive medicine is to differentiate between couples that have favourable chances of conceiving naturally and those that do not. Since the development of the prediction model of Hunault, characteristics of the subfertile population have changed. The objective of this analysis was to assess whether additional predictors can refine the Hunault model and extend its applicability.

Consecutive subfertile couples with unexplained and mild male subfertility presenting in fertility clinics were asked to participate in a prospective cohort study. We constructed a multivariable prediction model with the predictors from the Hunault model and new potential predictors. The primary outcome, natural conception leading to an ongoing pregnancy, was observed in 1053 women of the 5184 included couples (20%). All predictors of the Hunault model were selected into the revised model plus an additional seven (woman's body mass index, cycle length, basal FSH levels, tubal status,history of previous pregnancies in the current relationship (ongoing pregnancies after natural conception, fertility treatment or miscarriages), semen volume, and semen morphology. Predictions from the revised model seem to concur better with observed pregnancy rates compared with the Hunault model; c-statistic of 0.71 (95% CI 0.69 to 0.73) compared with 0.59 (95% CI 0.57 to 0.61).

Introduction

One of the aims in reproductive medicine is to differentiate between couples with favourable chances and those with unfavourable chances of natural conception, as treating couples with favourable chances implies overtreatment, causing unnecessary risks and costs. The tools to do this are prediction models.

In the past decade, several prediction models for natural conception have been developed (Leushuis et al., 2009). Currently, the established standard prediction model for natural conception is the Hunault model, also known as the synthesis model, which was based on the original data collected in three cohorts of subfertile couples between 1974 and 1995 (Collins et al, 1995, Eimers et al, 1994, Snick et al, 1997). Of all the prediction models for natural conception, it is the one with the best calibration, and it has been externally validated (van der Steeg et al., 2007b). The synthesis Hunault model encompasses five predictors: female age, duration of subfertility, whether female subfertility is primary or secondary, sperm motility, and referral status. Data used for construction of the model was restricted to a subfertile population in whom female age was below 40 years, and in whom any tubal pathology and azoospermia had been excluded (Hunault et al., 2004).

Since the development and external validation of the synthesis Hunault model, characteristics of the subfertile population have changed. Obesity is increasing rapidly worldwide, and also affects women in their reproductive age (Haslam, James, 2005, Talmor, Dunphy, 2015). Also, women tend to postpone their child wish, because of their career or other reasons (Mills et al., 2011). Consequently, ovarian reserve tests have been introduced as part of the fertility work-up. In addition, tests already incorporated into routine fertility work-up, such as semen analysis and tubal patency tests, are now reported in more detail than during the time the Hunault model was developed (Leushuis et al, 2010, van der Steeg et al, 2011, Verhoeve et al, 2011).

It is, therefore, currently unclear whether the natural conception chances predicted by the Hunault model are reliable, either because possible predictors are not taken into account by the model, or because characteristics of a couple may differ substantially from the data that the model was developed upon. The aim of this study was to assess whether we could build an improved prediction model with additional predictors for the contemporary subfertile population, i.e. including couples with one-sided tubal pathology, severe male subfertility and maximum female age of 45 years.