Hypotension during general anesthesia is associated with poor outcome. Continuous monitoring of mean blood pressure (MAP) during anesthesia is useful and needs to be reliable and minimally invasive. Conventional cuff measurements can lead to delays due to its discontinuous nature. It has been shown that there is a relationship between MAP and photoplethysmography (PPG) parameters like the dicrotic notch and perfusion index (PI). The objective of the study was to continuously estimate MAP from PPG. Pulse wave analysis based on PPG was implemented using either notch relative amplitude (MAPNRA), notch absolute amplitude (MAPNAA) or PI (MAPPI) to estimate MAP from PPG waveform features during general anesthesia. Estimated MAP values were compared to brachial cuff MAP (MAPcuff) and to radial invasive MAP (MAPinv). Forty-six patients were analyzed for a total of 235 h. Compared to MAPcuff, mean bias and limits of agreement were 1 mmHg (- 26 to +29), - 1 mmHg (- 10 to +8) and - 3 mmHg (- 21 to +13) for MAPNRA, MAPNAA and MAPPI respectively. Compared to MAPinv, mean absolute error (MAE) was 20 mmHg [10 to 39], 11 mmHg [5 to 18] and 16 mmHg [9 to 24] for MAP derived from MAPNRA, MAPNAA and MAPPI respectively. When calibrated every 5 min, MAPNAA showed a MAE of 6 mmHg [5 to 9]. MAPNAA provides the best estimates with respect to brachial cuff MAP and invasive MAP. Regular calibration allows to reduce drift over time. Beat to beat estimation of MAP during general anesthesia from the PPG appears possible with an acceptable average error.