P-433 Comparative analysis of serum obtained in early pregnancy using Fourier transform-infrared spectroscopy to predict tubal ectopic pregnancy and miscarriage

Abstract

Abstract Study question Can Fourier transform-infrared (FT-IR) spectroscopy be used to discriminate adverse early pregnancy outcomes such as tubal ectopic pregnancy and miscarriage from live normally-sited pregnancies? Summary answer Biochemical information acquired from serum samples using FT-IR spectroscopy does not discriminate adverse early pregnancy outcomes from live normally-sited pregnancies. What is known already Ectopic pregnancy and miscarriage are associated with maternal morbidity and mortality. Current management of pregnancy of unknown location and viability may involve serial blood human chorionic gonadotrophin-beta measurements and several ultrasound scans. Multiple interventions and length of time to diagnosis are costly and have a profound psychological impact on women. Vibrational spectroscopic methods such as Fourier transform-infrared (FT-IR) spectroscopy have been increasingly shown to discriminate disease states such as cancer from controls utilising the ‘metabolomic signature’ of molecules within a biological sample. It is hypothesised that this signature will differ between live normally-sited pregnancies, and tubal ectopic pregnancy and miscarriage. Study design, size, duration A total of 380 women were recruited to the prospective cohort study, Ectopic Pregnancy Diagnosis sTudy (ExPeDITe) between November 2018 and November 2021. Inclusion criteria comprised pregnant women ≥18 years presenting with pain and/or bleeding at < 10 weeks of gestation. A sample of blood and demographic data were collected at the time of consent, and a full history was taken. Early pregnancy outcomes were collected and classified according to ESHRE guidance. Participants/materials, setting, methods Serum samples prepared from collected blood (n = 340) were diluted 1:4 to prevent oversaturation of spectra and quality controls were created from the pooled diluted serum. Samples were transferred to silicon plates and dried at 37.5 °C. Samples were subjected to FT-IR spectroscopy and spectral data were submitted to baseline correction, smoothing, and vector normalisation prior to multivariate statistical analysis including Principal Component Analysis (PCA) and Partial Least Squares – Discriminant Analysis (PLS-DA). Main results and the role of chance Infrared spectral data collected from live normally-sited pregnancies, tubal ectopic pregnancies, pregnancies of unknown location, and miscarriages displayed similar infrared signatures, exhibiting bands associated to vibrational modes related to the main molecular constituents commonly found in most biological samples such as proteins, lipids, nucleic acids, and carbohydrates. Spectral data were used as inputs to PCA and PLS-DA models as clustering algorithms in order to classify the different pregnancy outcomes. The output obtained by the classification model generated with PLS-DA shows two distributions; a ‘real’ and ‘null/random’ model. These models displayed poor correct classification rates, indicating no differences in the infrared signatures between the different pregnancy outcomes. Similar findings were also observed in the PCA results as no clustering pattern indicating correct classification between the experimental groups was observed in the scores plot. Although infrared spectroscopy has been well demonstrated as an analytical tool to discriminate several disorders from healthy/control samples including cancer, diabetes and many others, these findings indicate that the information about the overall biochemistry probed by infrared spectroscopy acquired from serum samples currently lacks accuracy/sensitivity to discriminate different pregnancy outcomes. Limitations, reasons for caution Proteins were not removed from samples prior to analysis, which may mask discreet changes in small molecule metabolites. As FT-IR spectra multivariate in nature power calculations are not possible. If this study had shown case-control separation then these findings would have been validated with a second set of samples. Wider implications of the findings The findings of this study suggest that FT-IR spectroscopy of serum may not be discriminatory to identify important early pregnancy outcomes, but further analytical techniques should be utilised to confirm if metabolomic analysis of blood or other biofluids would provide a novel diagnostic test. Trial registration number REC 17/NW/0646