Limited value of logistic regression analysis in solid solitary pulmonary nodules characterization: A single-center experience on 288 consecutive cases

• Published in: Journal of surgical oncology Vol. 110; no. 7; pp. 883 - 887
• Main Authors: Perandini, S., Soardi, G.A., Motton, M., Dallaserra, C., Montemezzi, S.
• Format: Journal Article
• Language: English
• Published: United States Blackwell Publishing Ltd 01.12.2014; Wiley Subscription Services, Inc
• Subjects: Adult; Aged; Aged, 80 and over; Female; Follow-Up Studies; Humans; Logistic Models; lung cancer; Lung Neoplasms - pathology; Male; Middle Aged; Neoplasm Staging; preoperative risk assessment; Prognosis; Retrospective Studies; ROC Curve; solitary pulmonary nodule; Solitary Pulmonary Nodule - pathology; Tomography, X-Ray Computed; lung cancer; preoperative risk assessment; solitary pulmonary nodule
• ISSN: 0022-4790; 1096-9098
• DOI: 10.1002/jso.23730

Background and Objectives Preoperative characterization of the solitary pulmonary nodule is a delicate task faced by surgeons, radiologists, and clinicians. Mathematical models have been developed to overcome subjectivity. The Mayo Clinic model was suggested in the latest ACCP evidence‐based clinical practice guidelines for the preoperative risk assessment of solitary pulmonary nodule malignancy. The aim of the study is to assess the validity of the Mayo Clinic model in a current continuous case series of biopsy‐proven nodules. Methods The Mayo Clinic model was applied to estimate probability of malignancy in 288 consecutive cases in this single‐center retrospective study. Results ROC curve analysis returned an AUC of 0.767, while analysis performed on 158 malignant nodules showed a mean predicted risk value of 38.15%. In our clinical setting, using a risk observational threshold set at 5% and a risk surgical threshold set at 60%, there would have been 4 cases of unnecessary surgery (false positives) at the cost of 13 cases of cancer progression (false negatives), while 68.75% of all nodules would have received non‐decisional values. Conclusions Surgeons should be aware that current data shows how the Mayo Clinic model is of little use in preoperative nodule characterization. J. Surg. Oncol. 2014 110:883–887. © 2014 Wiley Periodicals, Inc.