M27 Systematic review of methods to predict postoperative gas transfer following lung cancer resection

Background

Prediction of postoperative spirometry and gas transfer is a key part of the risk stratification process for patients with early stage non small cell lung cancer for whom resection with curative intent may be possible. Generally accepted cut off values for proceeding with resection are postoperative values of between 30% and 40% predicted for the patient demographics. A range of different prediction techniques have been reported but their performance has not been formally compared before, patients could be denied access to efficacious treatment or exposed to higher risk than expected if predictions are inaccurate.

Methods

A systematic review of was performed with the primary outcomes being accuracy (mean difference) and precision (standard error of the mean) of different techniques to predict postoperative lung function. Predicted values for gas transfer were compared to measured postoperative values. The review was registered with PROSPERO before commencement [CRD42017058955]. Risk of bias of included studies was assessed using a pre-publication version of the PROBAST tool.

Results

A total of 3817 potentially eligible studies were retrieved with the search strategy; 23 studies that reported gas transfer met inclusion criteria but only 2 of these had low risk of bias. One paper reported prediction with segment counting and found a mean difference of 2 percentage points with a standard error of 1.5 percentage points. The other study reported prediction with perfusion scintigraphy and found a mean difference of 11 percentage points with a standard error of 1.7 percentage points.

Conclusion

The available literature suggests that prediction using segment counting may be the most accurate and precise technique. However, prediction of postoperative gas transfer requires direct comparison of different techniques in the relevant patient population. Given the gravity of decisions that are based upon the prediction result clinicians need to know which is the most accurate and precise technique to use to aid their recommendations.