Feasibility and accuracy of noninvasive continuous hemoglobin monitoring using transesophageal photoplethysmography

Background Continuous and noninvasive hemoglobin (Hb) monitoring during surgery is essential for anesthesiologists to make transfusions decisions. The aim of this study was to investigate the feasibility and accuracy of noninvasive and continuous Hb monitoring using transesophageal descending aortic photoplethysmography (dPPG) based on near-infrared spectroscopy (NIRS) technology. Methods Nineteen landrace pigs, aged 3 to 5 months and weighing 30 to 50 kg, were enrolled in this study. A customized oximetry sensor, including red (660nm) and infrared (940nm) lights, was placed in the esophagus for dPPG signal detection to pair with the corresponding reference Hb values (Hbi−STAT) measured by blood gas analysis. The decrease and increase changes in Hb concentration were achieved by hemodilution and transfusion. Metrics, including alternating current (AC), direct current (DC), and AC/DC for both red and infrared light were extracted from the dPPG signal. A receiver operating characteristic (ROC) curve was built to evaluate the performance of dPPG metrics in predicting Hb level. Agreement and trending ability between Hb measured by dPPG (HbdPPG) and by blood gas analysis were analyzed by Bland-Altman method and polar plot graph. Error grid analysis was also performed to evaluate clinical significance of HbdPPG measurement. Results A total of 376 pairs of dPPG signal and Hbi−STAT were acquired. ACred/DCred and ACinf/DCinf had moderate correlations with Hbi−STAT, and the correlation coefficients were 0.790 and 0.782, respectively. The areas under the ROC curve for ACred/DCred and ACinf/DCinf in predicting Hbi−STAT < 60 g/L were 0.85 and 0.75, in predicting Hbi−STAT > 100 g/L were 0.90 and 0.83, respectively. Bland-Altman analysis and polar plot showed a small bias (1.69 g/L) but a wide limit of agreement (-26.02–29.40 g/L) and a poor trend ability between HbdPPG and Hbi−STAT. Clinical significance analysis showed that 82% of the data lay within the Zone A, 18% within the Zone B, and 0% within the Zone C. Conclusion It is feasible to establish a noninvasive and continuous Hb monitoring by transesophageal dPPG signal. The AC/DC extracted from the dPPG signal provided a sensitive prediction for Hb < 60 g/L and Hb > 100 g/L. The Hb concentration measured by dPPG signal has a moderate correlation with that measured by blood gas analysis.