Inter-clinic variation in the chances of natural conception of subfertile couples.

STUDY QUESTION

Are there differences between clinics in the chances of natural conception of couples?

SUMMARY ANSWER

We found significant differences between clinics in the couples' natural conception chances, which could not be explained by differences in characteristics of the patients or the clinics.

WHAT IS KNOWN ALREADY

In pooled data from multiple centers the synthesis prediction model for natural conception was found to be valid, yet the outcome of interest (i.e. natural conception) might differ between centers. Possible differences between clinics in natural conception rates, as well as the validity of the prediction model in each individual clinic are addressed in this paper.

STUDY DESIGN, SIZE AND DURATION: A secondary data-analysis of a prospective cohort study among 3020 subfertile couples recruited in 38 clinics in the Netherlands between January 2002 and December 2004. Clinics with less than 20 couples were excluded from the analyses, resulting in 21 clinics with 2916 couples.

PARTICIPANTS/MATERIALS, SETTING, METHODS: Inclusion of 2916 subfertile couples who underwent a basic fertility work-up. Couples were excluded who had a fertility disorder (one or two-sided tubal pathology, ovulation disorder, total motile sperm count <3 × 10(6)). Included couples were counseled for expectant management for at least six months or followed until the first day of treatment. Follow-up began at the completion of the fertility work-up. Couples lost to follow-up were censored at the last day of contact. Kaplan-Meier survival curves and a log-rank statistic were estimated. Crude and adjusted hazard ratios were determined, adjusted for patient characteristics and the type of clinic (university hospitals with an assisted conception unit (ACU), non-university hospitals with an ACU and non-university hospitals without an ACU). Hazard ratios were also ascertained with empirical Bayes (EB) estimates. Validation of the prediction model per clinic was performed through calibration.

MAIN RESULTS AND THE ROLE OF CHANCE

We found significant differences between clinics in the chance of ongoing pregnancy (P < 0.001); even after adjustment for female age, duration of subfertility, percentage of progressive motile sperm, primary/secondary subfertility and post-coital test (P < 0.001). Adjusted hazard ratios and EB estimates ranged from 0.50 to 2.21 and 0.58 to 1.53, respectively. Among the 21 clinics, there were 4 university hospitals, 10 non-university hospitals with an ACU and 7 non-university hospitals without an ACU. In the multivariable analysis, the type of clinic was not significant (P = 0.11). Calibration gave an average intercept of -0.25 (95% range: -1.04-0.53) and average slope of 0.81 (95% range: 0.03-1.60). Six clinics had a negative intercept that differed significantly from zero and three clinics had a negative or positive slope that differed significantly from one.

LIMITATIONS, REASONS FOR CAUTION: A more extensive model including more predictors could give less variation in the differences between the clinics. Variation in work-up protocol between clinics could also have played a role. In fertility prediction research the Cox proportional hazards regression is the most widely used statistical model, but as the underlying assumptions have rarely been evaluated, this model could lead to biased outcomes.

WIDER IMPLICATIONS OF THE FINDINGS

Our findings suggest that the synthesis model to predict natural conception is useful overall in clinical practice but in a minority of clinics the model is not well calibrated. Updating the synthesis model to include a center-specific baseline chance might improve the synthesis model for certain clinics.

STUDY FUNDING/COMPETING INTEREST(S): The study (on which this secondary data-analysis was based) was supported by grant 945/12/002 from ZonMW, the Netherlands Organization for Health Research and Development, The Hague, the Netherlands.