Background: Cardiopulmonary exercise testing (CPET) is used to assess perioperative risk in surgical patients. While previous studies have looked at short-term outcomes, this paper explores the ability of CPET to predict 5 yr survival after major surgery.
Methods: Over a period (1996-2009), 1725 patients referred for CPET subsequently underwent major surgery. Breath-by-breath data derived during each patient's CPET was processed using customized software to extract variables likely to impact on survival. Initial analysis examined the predictive power of single variables. Subsequently, Bayesian model averaging (BMA) was used to construct a multivariate model defining the association between CPET data and 5 yr survival.
Results: Six hundred and sixteen (36%) of the study patients died. Single variables were not significantly associated with 5 yr postoperative survival. BMA indicated the following major predictors of 5 yr survival: patient gender; type of surgery, and forced vital capacity. Four variables derived at the patient's anaerobic threshold were weaker predictors. These were end-tidal oxygen concentration, respiratory exchange ratio, oxygen consumption per unit body weight, and oxygen consumption per heart beat. The resulting model was then used to divide patients into low-, medium-, or high-risk categories, and 5 yr survival for each category was 87.8; 75.8, and 53.8% respectively. Survival was independent of patient age.
Conclusions: Multivariate analysis and model generation techniques can be applied to CPET data to predict 5 yr survival after major surgery more accurately than is possible with single variable analysis.