Here is an article from www.InfertilitySpecialists.com that lists some statistics on ovarian reserve:
1. High age is limiting even with normal FSH. Original reports demonstrated lower pregnancy rates in women past 40 years of age, no matter their basal FSH level . Even with today’s treatments, successful pregnancies past age 42 are uncommon, and past 45 are rare.
2. High FSH is limiting even with normal age. The original reports saw a declining pregnancy rate as FSH rose above 20, and no ongoing pregnancies beyond an FSH of 25 IU/L [5,6]. While the assay has since changed and altered these cutoffs, there still tends to be a threshold above which declining performance (egg production and pregnancy rate) is detected, and a higher threshold above which egg production is quite limited, and almost no pregnancies have occurred.
3. Cutoffs for FSH depend on the lab test employed. Up through the early 1990’s, most commercial assays reported FSH levels about twice as high as those now in wide use. Thus, whereas the cutoff of normal FSH was 20 IU/L in early reports, it is now more commonly about 10 IU/L. And whereas markedly abnormal ovarian reserve was formerly seen only above 25 IU/L, now that threshold occurs above about 15 IU/L. It is still best if clinics develop their own thresholds to define then end of the normal range, and the entry into the very abnormal range, for FSH and estradiol assays commonly employed for their patients.
4. The highest-ever FSH is the one most likely to be true. Several early reports demonstrated the futility of delaying treatments until a cycle with a normal FSH occurs. More recent studies have continued to affirm this effect [7,8]. Once an FSH elevation is observed, egg production capacity will be limited thereafter. This is to be expected, given the on-again, off-again nature of basal FSH elevations once egg numbers become critically short (as illustrated in Fig. 2).
5. Prediction of ovarian reserve is easier than predicting pregnancy. Basal FSH levels are better able to predict outcomes more closely related to ovarian function, such as cancellation (R2=77%), follicles aspirated (R2=35%), and oocytes retrieved(R2=21%) than more distal events such as pregnancy rate (R2=4%)[5,9].
You can read the whole article here.