Comparison of Brock University, Mayo Clinic and Herder models for pretest probability of cancer in solid pulmonary nodules

Objective: Risk analysis models, which are used in the diagnostic algorithm of incidental pulmonary nodules, are based on patient data from developed countries. Mayo Clinic, Brock University and Herder are among the most known models. We aim to compare the reliability of these models in patients with indeterminate solid nodules and to investigate the contribution of the predictors used to the model.

Methods: We analysed 305 patients who performed transthoracic needle biopsy and positron emission tomography/computed tomography for solid nodules, retrospectively. For all three models, the malignancy risk probabilities of patients were calculated, and patients were classified as low (<5%), moderate (60%) and high (<60%) risk groups. Later, the malignancy rates of each model in three different risk groups were compared within each other and among the models.

Results: The malignancy rate is 73% in 305 patients. In the Mayo Clinic and Herder models, the difference in the low-, medium- and high-risk groups is significant (p < 0.001). In the medium-risk group, the rate of malignancy is 96.8% in the Brock model. In the high-risk group, the rate of malignancy in Herder is 88.3% and the rate of malignancy in Mayo Clinic is 28.8%. The optimal cutoff values for the Mayo Clinic, Brock University, and Herder were 29.6, 13.4 and 70 (AUC, respectively; 0.71, 0.67 and 0.73). Age, smoking, gender, size, emphysema and spiculation increase the likelihood of malignancy.

Conclusion: Close results were obtained in all three models. In the high-risk group, the Herder model has the highest reliability rate (odds ratio 3.3, confidence interval [1.1, 10.2]). Upper lobe predilection is not a reliable predictor.