Objectives: A crucial point in the work-up of a solitary pulmonary nodule (SPN) is to accurately characterise the lesion on the basis of imaging and clinical data available. We introduce a new Bayesian calculator as a tool to assess and grade SPN risk of malignancy.
Methods: A set of 343 consecutive biopsy or interval proven SPNs was used to develop a calculator to predict SPN probability of malignancy. The model was validated on the study population in a "round-robin" fashion and compared with results obtained from current models described in literature.
Results: In our case series, receiver operating characteristic (ROC) analysis showed an area under the curve (AUC) of 0.893 for the proposed model and 0.795 for its best competitor, which was the Gurney calculator. Using observational thresholds of 5% and 10% our model returned fewer false-negative results, while showing constant superiority in avoiding false-positive results for each surgical threshold tested. The main downside of the proposed calculator was a slightly higher proportion of indeterminate SPNs.
Conclusions: We believe the proposed model to be an important update of current Bayesian analysis of SPNs, and to allow for better discrimination between malignancies and benign entities on the basis of clinical and imaging data.
Key points: • Bayesian analysis can help characterise solitary pulmonary nodules • Volume doubling time (VDT) is a good predictor of malignancy • A VDT of between 25 and 400 days is highly suggestive of malignancy • Nodule size, enhancement, morphology and VDT are the best predictors of malignancy.