Sensitivity and specificity of manual versus automated methods of anaerobic threshold detection in patients undergoing colorectal surgery: implications for clinical outcome?

Background: Cardiopulmonary exercise testing (CPX) is used to determine cardiorespiratory fitness in patients prior to major surgery given its association with post-operative survival. Typically an automated anaerobic threshold (AT) value of <11.0 ml O2.kg-1.min-1 (Older et al., 1999) has been employed as an objective biomarker of increased perioperative risk. In the present study, we compared to what extent differences between automated versus manual (gold-standard) methods of AT detection have the theoretical potential to influence surgical risk stratification. 
Methods: A randomised sample of 213 patients scheduled for elective colorectal surgery who underwent CPX testing were retrospectively examined. Manual AT results were calculated using the gold standard ‘V-slope’ method (Beaver et al., 1986) and confirmed by two independent clinicians. Automated AT results were compiled using default settings in Breeze software (Medgraphics, UK). Ventilatory equivalent for CO2 (VE/VCO2) slope and respiratory exchange ratio (RER) were also recorded at both Manual and Automated ATs. Following confirmation of distribution normality (Shapiro W Wilks tests), data were analysed using a combination of paired samples t -tests and Chi-Squared tests. Data are expressed as mean ± SD and significance established at p < 0.05. 
Results: Pulmonary oxygen uptake ( V O2) at the AT was 11.0 ± 3.0 versus 12.5 ± 3.8 ml.kg-1.min-1 for the Manual and Automated methods respectively ( p < 0.05). One hundred and twelve ATs <11.0 ml O2.kg-1.min-1 were reported for the Manual versus 70 for the Automated method ( p < 0.05). Fifty two false negatives were reported for the Automated method (sensitivity 55%, specificity 91%). 
Conclusions: Automated detection of the AT overestimates V O2 by 13% and is associated with a high rate of type II errors (false negatives). This could result in some patients transcending risk stratification boundaries thus leading to incorrect decision making and inappropriate surgical risk stratification. Despite the ease of use of automated software based AT predictions, clinicians should be encouraged to use the manual and gold standard V-slope method for a more accurate assessment of patient cardiorespiratory fitness. 
Reference 1: Beaver et al. (1986). J Appl Physiol 60(6), 2020-2027 Reference 2: Older et al. (1999). Chest 116, 355-362 (No Image Selected)