The predictive value for in vitro fertility delivery rates is greatly impacted by the method used to select the threshold between normal and elevated basal follicle-stimulating hormone.

OBJECTIVE: To evaluate the predictive accuracy of different methodologies for selecting a basal FSH threshold level that prognosticates live birth after IVF.
DESIGN: Retrospective.
SETTING: Academic private practice. PATIENT(S): Eight thousand nineteen patients who had their basal FSH levels determined by the program's endocrinology laboratory. INTERVENTION(S): Thresholds between normal and elevated basal FSH levels were calculated by using six different methodologies. MAIN OUTCOME MEASURE(S): Live birth rate per initiated IVF cycle. RESULT(S): The thresholds selected by using the manufacturer's normal range or using 95% confidence intervals of a fertile population, the infertile population, or distinct age groups within the infertile population all proved unsatisfactory. The live birth rates for patients in whom there had been a previously elevated FSH level were <or=43.7%. The efficiency curves that were created for each of the five age groups were the most useful. With appropriately selected thresholds, the predictive value of an abnormal result approaches 100%. CONCLUSION(S): The methodology used to select thresholds for basal FSH levels has enormous impact on the utility of the screening test. The use of anything other than clinical outcome-based methods to select the threshold greatly decreases the utility of the test and may lead to false conclusions or inaccurate patient counseling.