Although a wealth of literature exists on the discriminatory ability of a single serum β-hCG value on predicting pregnancy outcomes after IVF, significant heterogeneity and lack of local population data limit its application in our setting. This study aimed to evaluate the prognostic value of a single serum β-hCG cut off, 12 days after embryo transfer, on predicting pregnancy outcomes among Ugandan women.
In this population, a single serum β-hCG value had limited discriminatory ability to predict pregnancies that would yield a live birth. The cut off of 437.42mIU/ml is within range to that reported by [43](495 mIU/ml), [44] (377.8mIU/ml), and [45] (386 mIU/ml). However, the reported sensitivity varied between a comparable 75.9% [44] and a higher 80% [43]. In addition, the false positive rate, indicated similar suboptimal performance, ranged between a lower 23.4 [43] and a comparable 38.8% [44]. But for the most part, our population cut off was 2–4 fold higher than that reported by others such as 50mIU/ml [46], 76mIU/ml [33], 78mIU/ml [47], 85mIU/ml [45], 131 mIU/ml [48]; 104.5mIU/ml [27], 145mIU/ml [29], 199 mIU/ml [49], 222.86 mIU/ml [50], 253 mIU/ml [49], and 315.65mIU/ml [32].The associated false positive rates indicated variable predictive performance and ranged between 10% [47] and 50% [49].
The higher cut off in this study probably resulted from high proportion of multiple embryo transfers (2–4 embryos making up 98%) and subsequent high rate of multiple pregnancies at delivery (39%). Multiple pregnancies confer higher cut offs of serum β-hCG compared to singletons [45]. In addition, significant heterogeneity in the interval between embryo transfer and serum β-hCG reading, could account for the variation between this study’s cut off and that report by others. The interval varied by definition i.e, after oocyte retrieval [27, 44, 49] or after embryo transfer [43, 45, 48, 50, 51]. The interval also varied by duration i.e, 14–16 days after oocyte retrieval [27, 44, 47, 49] and 12–16 days after embryo transfer [43, 45, 48, 50–52]. Furthermore, majority of the studies used ROC (Receiver Operator Characteristic) analysis compared to our Youden index metric analysis to determined optimal cut offs. Importantly, these cut off values are not absolute and are to be interpreted with caution since, the lowest cut off, corresponding to the 5%centile, was 264.81 mIU/ml. This was echoed by Y Wu and H Liu [53], were optimal cut offs as low as 108.6 mIU/ml predicting live birth albeit with a false positive rate over > 50%.
The cut offs for clinical pregnancy and ongoing pregnancy had a modest performance with the Youden index metric with cut off values at 239.58 mIU/ml and 353.66 mIU/ml respectively. The cut off for clinical pregnancy was within range of that reported by others [50, 54]. In addition, cut off reported by N Singh, et al. [31] and MA Eskandar, et al. [32], for ongoing pregnancy were within range of our population. However, the cut off for clinical pregnancy was 2 fold higher than in other populations such as: 80.4 mIU/ml [27], 86.8mIU/ml [18], 100 mIU/ml [55], 123 mIU/ml [48], and 111mIU/ml [56] .Equally the cut off was approximately 2 fold higher that reported by others, 131mIU/ml [48], 137mIU/ml [57], for ongoing pregnancy. Others have reported higher cut off for ongoing pregnancy at 500mIU/ml [16] and 542.45 mIU/ml [54]. Furthermore, basing on the AUC (area under curve) characteristic and false positive rates, serum β-hCG cut offs performed better in predicting clinical pregnancy and ongoing pregnancy compared to livebirth, as reported by others [27, 48, 50]
Serum β-hCG had better discriminatory performance in women less than 30 years or greater than 40 years of age. This can be explained by the similar age of utilized oocytes since women above 40 years are likely to have utilized donor oocytes and oocyte donors are usually less than 30 years of age. The lack of association between age and serum β-hCG cut off values is in variance with that reported [51, 56], and could have resulted from the small sample size and high proportion of donor oocyte utilization in this population. Serum β-hCG had better discriminatory performance among cycles that utilized autologous oocytes compared to oocyte donor cycles. This probably because cycles using autologous oocytes had relatively more homogeneity in cycle characteristics compared to those using donor oocytes.
This study reports limited correlation with existing and studies and contributes to the ever-increasing variation in serum β-hCG cut off among different ART populations. Heterogeneity in population characteristics contributed to the disparity in cut off values between this study and others. Some of key characteristics that may have impacted the serum β-hCG cut off include: proportion of multiple gestations [45], embryo grade [56, 58], embryo developmental stage [47], fresh versus frozen embryo transfers [50, 56], maternal age [51, 56], and the interval between embryo transfer and serum β-hCG reading.