Key result
Current prolonged air leak risk models have low positive predictive values (0.13-0.19) but high negative predictive values (0.91-0.96), which can be useful for preoperative patient counseling.
Although prolonged air leak risk models lack the discriminatory capacity for clinical decision-making, their high negative predictive value makes them potentially useful for preoperative patient counseling.
In this issue of the European Journal of Cardio-Thoracic Surgery, the results of an interesting Delphi consensus on prevention and management of prolonged air leak (PAL) after lung resection are reported [1]. The authors, 35 well-known, experienced thoracic surgeons from around the world respond to a questionnaire consisting of 43 questions on PAL definition, risk assessment, intraoperative detection, classification and treatment and postoperative management of chest tubes and PAL. According to experts’ answers, the clinical practice was highly recommended on items where at least 75% of the agreement was reached. One of the points where experts reached the highest agreement rate (87%) is that current PAL risk models do not have enough discriminatory capacity and its use is not recommended in clinical practice. In the Discussion section, the authors comment that the high rate of agreement on this point confirms the results of a previous study authored by some PALAS participants [2] where the clinical and statistical performance of four of the existing PAL risk models is retrospectively re-evaluated on a cohort of almost 4000 patients who underwent video-assisted anatomical lung resection. The predictive performance of the four models is low, with a C-statistic no better than 0.65 and a high rate of false positives and low positive predictive values. The performance of other multi-institutional scores derived from databases is also low. The one from the Society of Thoracic Surgeons (STS) [3] reached a positive predictive value of 0.19, the European Society of Thoracic Surgeons (ESTS) Database [4] 0.13 (considering positive classes C and D of risk in this publication), and the French National Thoracic Database [5] 0.14 if predictive values are also calculated in the last considering a positive result of the test what the authors propose as the two higher classes of risk probability. But the negative predictive values are very high in the three reports: 0.91 in the STS [3], 0.92 in the ESTS [4] and 0.96 in the French National Thoracic Database [5]. That apparent discrepancy is understandable using basic clinical epidemiology knowledge since the prevalence of PAL is not higher than 0.13 in the recently reported analysis. PAL is recorded in 9.4% of cases after lobectomy and 12.5% after bilobectomy in the last report of the ESTS Database [6] and 10.4% after anatomical lung resection in the STS General Thoracic Surgery Database [3]. It is understandable that PAL scores are not found useful in clinical decision-making in the PALAS report [1] and that decisions on PAL prevention and management are rather guided by intraoperative findings. In fact, it is reported [7] that inter-hospital variability of PAL prevalence is mostly influenced by variations in real-life practice and local treatment patterns. If we look at the problem from the patient information and counselling point of view, we could see the relevance of PAL scores differently. PAL prevalence is not high enough that efficient predictive models can be built, but it is the most frequent complication after lobectomy and segmentectomy. Besides its frequency, PAL is associated with the occurrence of serious postoperative complications and hospital readmission [8] even in patients discharged home with chest tubes in place to avoid prolonged hospitalization [9]. Full information and counselling before hospitalization are strongly recommended for enhanced recovery after lung surgery [10]. Being PAL the most frequent postoperative complication after lung resection, potentially having severe consequences, the process of preoperative patient information and counselling must include accurate information on the risk of PAL and its consequences. In this context, preoperative information on the patient individual probability of not suffering the most frequent complication of lung resection could help counselling. PAL scores, despite not being useful to decide on surgical techniques to professionals, can facilitate informed decisions for patients since just understandable clinical variables and not knowledge of statistics are required.
No takes yet. Share an insight, caveat, or question.
Gonzálo Varela (2022) conducted an editorial in Prolonged air leak (PAL) after lung resection. Prolonged air leak (PAL) risk models was evaluated on Predictive performance of PAL risk models. Current prolonged air leak risk models have low positive predictive values (0.13-0.19) but high negative predictive values (0.91-0.96), which can be useful for preoperative patient counseling.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: