Non-invasive assessment of respiratory muscle activity during pressure support ventilation: accuracy of end-inspiration occlusion and least square fitting methods

• Published in: Journal of clinical monitoring and computing 2021-08, Vol.35 (4), p.913-921
• Main Authors: Natalini, Giuseppe, Buizza, Barbara, Granato, Anna, Aniballi, Eros, Pisani, Luigi, Ciabatti, Gianni, Lippolis, Valeria, Rosano, Antonio, Latronico, Nicola, Grasso, Salvatore, Antonelli, Massimo, Bernardini, Achille
• Format: Article
• Language: English
• Subjects: Accuracy; Anesthesiology; Bias; Critical Care Medicine; Health Sciences; Humans; Inspiration; Intensive; Least squares; Least-Squares Analysis; Medicine; Medicine & Public Health; Muscles; Occlusion; Original Research; Positive-Pressure Respiration; Respiration, Artificial; Respiratory Mechanics; Respiratory Muscles; Statistics for Life Sciences; Ventilation; Work of Breathing
• ISSN: 1387-1307; 1573-2614
• DOI: 10.1007/s10877-020-00552-5

Pressure support ventilation (PSV) should be titrated considering the pressure developed by the respiratory muscles (P musc ) to prevent under- and over-assistance. The esophageal pressure (P es ) is the clinical gold standard for P musc assessment, but its use is limited by alleged invasiveness and complexity. The least square fitting method and the end-inspiratory occlusion method have been proposed as non-invasive alternatives for P musc assessment. The aims of this study were: (1) to compare the accuracy of P musc estimation using the end-inspiration occlusion (P musc,index ) and the least square fitting (P musc,lsf ) against the reference method based on P es ; (2) to test the accuracy of P musc,lsf and of P musc,index to detect overassistance, defined as P musc  ≤ 1 cmH 2 O. We studied 18 patients at three different PSV levels. At each PSV level, P musc , P musc,lsf , P musc,index were calculated on the same breaths. Differences among P musc , P musc,lsf , P musc,index were analyzed with linear mixed effects models. Bias and agreement were assessed by Bland–Altman analysis for repeated measures. The ability of P musc,lsf and P musc,index to detect overassistance was assessed by the area under the receiver operating characteristics curve. Positive and negative predictive values were calculated using cutoff values that maximized the sum of sensitivity and specificity. At each PSV level, P musc,lsf was not different from P musc (p = 0.96), whereas P musc,index was significantly lower than P musc . The bias between P musc and P musc,lsf was zero, whereas P musc,index systematically underestimated P musc of 6 cmH 2 O. The limits of agreement between P musc and P musc,lsf and between P musc and P musc,index were ± 12 cmH 2 O across bias. Both P musc,lsf  ≤ 4 cmH 2 O and P musc,index  ≤ 1 cmH 2 O had excellent negative predictive value [0.98 (95% CI 0.94–1) and 0.96 (95% CI 0.91–0.99), respectively)] to identify over-assistance. The inspiratory effort during PSV could not be accurately estimated by the least square fitting or end-inspiratory occlusion method because the limits of agreement were far above the signal size. These non-invasive approaches, however, could be used to screen patients at risk for absent or minimal respiratory muscles activation to prevent the ventilator-induced diaphragmatic dysfunction.