In this work, we show that 1H NMR spectra of spent culture media of day 3 embryos that implant (100 % implantation) are not consistently different from the spectra of embryos that fail to implant (0 % implantation). Specifically, the discrimination offered by NMR-based metabolomics could not increase the efficiency of embryo selection and thus IVF success rate.
Our study contrast several publications that showed a correlation between specific metabolic profiles of embryo spent culture media and successful implantation [18
]; however, our findings are consistent with the results from two recent randomized controlled trials [11
] showing no improvement in pregnancy outcome when media profiling was added to the conventional embryo selection protocol. In particular, our findings are contrary to the study of Seli et al, [26
] and to the study of Pudakalakatti et al [18
] where a positive correlation was found between implantation and NMR profiling. The reason for the difference is unclear. However, it is important to note that our study had a much larger sample size. In our study we included 108 patients: more specifically we included 106 embryos (from 54 patients) in the 100 % implantation and 140 embryos (from 54 patients) in the zero implantation group. The Seli study only included 18 patients: 17 embryos (from 10 patients) implanted and led to delivery (100 % sustained implantation), whereas 17 embryos (from 8 patients) did not implant (zero percent implantation). The Pudakalakatti study included 48 patients (and the profile of 43 implanted embryos and 57 not implanted embryos) and specifically analyzed only few metabolites: lactate, pyruvate and alanine. In addition, the authors used a “per patient” definition of successful implantation: because multiple embryos were transferred in the same patient, a group of embryo was defined as implanted if at least one embryo in its cohort was implanted successfully. The particular embryo that resulted in successful implantation could not be determined. However, for patients resulting in pregnancy (successful implantation), it was found that all embryos had a similar metabolic signatures in the NMR spectrum. This point is important and we believe it represent a critical limitation of the Pudakalakatti study, since our data analysis showed that even among sibling embryos that resulted in implantation the NMR signature can be very different.
Most importantly, none of the reported studies performed validation of their prediction model in an independent data set. Our study was specifically designed to validate results of the initial experiments in independent data sets. NMR data were acquired for five independent sets of media samples in order to make a most comprehensive attempt to validate the findings for the individual sets. While Sets 1, 2 and 4 (Table ) were conducted on embryos cultured on the same type of media (G1.3 and HSA), set 3 were performed on embryos cultured on G1.5+HSA medium and set 5 on G1.5+SSS medium. Overall, it was not possible to jointly subject NMR profiles from different sets to multivariate data analysis since the differences of the culture media composition and their supplements clearly dominate class separation. The inability to compare metabolomic profiles between media containing different components is a key finding, since culture media manufacturers routinely change and improve culture media. Furthermore, different IVF laboratories might use different culture media; here we show that changes of media composition profoundly affect the metabolomic profile, even when a very rigid standard operating procedure in a single laboratory was adopted. A robust embryo predictor technology for routine use in a clinical setting should, however, not be influenced by these differences.
In addition, even when limiting analysis to embryos cultured in same media but in different sets (Sets 1, 2, 4), prediction model could not be confirmed in independent data sets (Tables and ).
Several studies have analyzed the differences between metabolites present in media from implantable versus non-implantable embryos [5
]. For example, Conaghan showed that pyruvate uptake was lower, on average, in embryos that implant compared to embryos that did not implant [5
]. However, pyruvate uptake in individual embryos differed widely, indicating that pyruvate uptake is a rather poor marker for embryo selection [5
]. Pudakalakatti et al also analyzed the concentration of pyruvate in spent culture media; in addition, lactate and alanine were assessed in 57 embryos that implanted and 43 embryos that did not implant using NMR [18
]. They found that embryos transferred on day 3 (after 72 h in vitro culture) with successful implantation exhibited significantly higher pyruvate uptake (not lower, as in the Conaghan study) and lower pyruvate/alanine ratios compared to those that failed to implant. Lactate levels in media were similar for all embryos. On the contrary, Seli et al, also using NMR, did not find any difference in pyruvate or pyruvate/alanine ratio but found that implanted embryos had higher glutamate and lower alanine levels, and a decrease in the alanine to lactate ratio [26
]. In a more recent study Gardner et al analyzed the daily glucose consumption of post compaction embryo in spent culture media using microfluorimetry [9
]. The author found that glucose consumption in embryo that implanted was significantly higher on both Day 4 and Day 5 than that by embryos which failed to develop post-transfer (p
0.01) and that glucose uptake was independent of embryo grade. In summary the above mentioned studies indicate that the mean (average) concentrations of several metabolites can be different in the two groups; however each study showed a large intra-group variability that does not support robust individual predictions. Importantly, none of the research group attempted to validate their findings in independent data sets. More concerning, the metabolites found to be altered in spent culture media of implanted embryo in different NMR studies are not consistent, indicating lack of reproducibility between different studies.
A potential limitation of our study is inclusion of single as well as multiple embryo transfer cases. Our experimental design was however modeled to reflect the more common practice of multiple embryo transfer prevalent in US [10
]; moreover the studies published using NMR technology used a similar multiple embryo transfer design [18
One additional consideration to ponder while analyzing our study and embryo selection and implantation studies in general, is the fact that the endometrium could play a role in allowing an embryo to implant. However, it is important to note that all patients in our study had evidence of normal endometrium defined as 1) lining of more set the than 7 mm and 2) had to have a normal saline sonogram. A potential confounding factor is that patients whose embryos implanted showed some evidence of better clinical prognosis (had higher estrogen levels on day of hCG, a thicker endometrial lining, more 2PN embryos and had less number of embryos transferred). However, other important clinical markers (age, fragmentation score or number of embryonic cells) were not different among the groups.
In conclusion, we have found no significant difference between 1H NMR profiles of spent culture media of implanted versus non-implanted embryos while following a rigorous protocols and using one of the most sensitive technologies available today to study metabolite profiling. Although it might be possible that future advancements of NMR hardware may improve class separation and thus predictive accuracy of NMR-based profiling, our findings suggest that variations of embryo-dependent metabolite concentrations in media are too small to be of predictive value for embryo selection.