A retrospective evaluation of prognosis and cost-effectiveness of IVF in poor responders according to the Bologna criteria

Study question: Do the Bologna criteria for poor responders successfully identify women with poor IVF outcome? Summary answer: The Bologna criteria effectively identify a population with a uniformly low chance of success. What is already known: Women undergoing IVF who respond poorly to ovarian hyper-stimulation have a low chance of success. Even if improving IVF outcome in this population represents a main priority, the lack of a unique definition of the condition has hampered research in this area. To overcome this impediment, a recent expert meeting in Bologna proposed a new definition of poor responders ('Bologna criteria'). However, data supporting the relevance of this definition in clinical practice are scanty. Study design, size, duration: Retrospective study of women undergoing IVF-ICSI between January 2010 and December 2012 in two independent infertility units. Women could be included if they fulfilled the definition of poor ovarian response (POR) according to Bologna criteria prior to initiation of the cycle. Women were included only for one cycle. The main outcome was the live birth rate per started cycle. The perspective of the cost analysis was the one of the health provider. Participants/materials, setting, methods: Three-hundred sixty-two women from two independent Infertility Units were selected. A binomial distribution model was used to calculate the 95% CI of the rate of success. Characteristics of women who did and did not obtain a live birth were compared. A logistic regression model was used to adjust for confounders. The economic analysis included costs for pharmacological compounds and for the IVF procedure. The benefits were estimated on quality-adjusted life years (QALY). To develop the model, we used the local life-expectancy tables, we applied a 3% discount of life years gained and we used a 0.07 improvement in quality of life associated with parenthood. Sensitivity analyses were performed varying the improvement of the quality of life and including/excluding the male partner. The reference values for cost-effectiveness were the Italian and the local (Lombardy) gross domestic product (GDP) pro capita per year in the studied period and the upper and lower limits suggested by NICE. Main results and the role of chance: Overall, 23 women had a live birth (6%, 95% CI: 4-9%), in line with the previous evidence. This proportion did not significantly differ in the different subgroups of poor responders. Positive predictive factors of success were previous deliveries (adjusted OR = 3.0, 95% CI: 1.1-8.7, P = 0.039) and previous chemotherapy (adjusted OR = 13.9, 95% CI: 2.5-77.2, P = 0.003). Age, serum AMH, serum FSH and antral follicle count were not significantly associated with live birth. The total cost per live birth was 87 748 Euros, corresponding to 49 919 Euros per QALY. This is above both the limits suggested by NICE for cost-effectiveness and the Italian and local GDP pro capita. Sensitivity analyses mainly support the robustness of the conclusion. Limitations, reasons for caution: We lack a control group and we cannot thus exclude that an alternative definition of poor responders may be equally if not more valid. Moreover, independent validations are warranted prior to concluding that IVF is not cost-effective. Women should thus not be denied treatment based on our findings. Noteworthy, there is also not yet a consensus on the most appropriate economic model to be used. Wider implications of the findings: We recommend the use of the Bologna criteria when designing future studies on poor responders. Large multi-centred international studies are now required to draw definite conclusions on the economic profile of IVF in this situation.