A retrospective validation study of three models to estimate the probability of malignancy in patients with small pulmonary nodules from a tertiary oncology follow-up centre
Introduction
The management of patients with small pulmonary nodules (PNs) is clinically challenging. With the widespread availability of multidetector row computed tomography (CT) and the significant increase in the number of scans performed,1 this scenario is likely to increase in clinical practice, and clear evidence is required to guide management strategies.
Data from published case series in Europe,2, 3 Asia,4, 5 and America6, 7, 8 report the prevalence of PNs from 2% to 24% (mean 13%) with a mean lung cancer prevalence of 1.5%. The prevalence of PNs in a screening population is higher (mean 33%),9 but management of nodules in this scenario is predetermined for the majority of screening programmes.
The effective management of PNs requires knowledge of the “pre-test” probability of malignancy in each individual patient, prior to conducting further investigation.10 The British Thoracic Society Guidelines for the management of PNs recommends the use of prediction models when evaluating these patients.
Twenty-eight studies 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33 in the literature have evaluated the clinical and radiological characteristics of PNs in relation to the probability of malignancy. Four studies have developed composite prediction models that can be used in clinical practice and are easily accessible.
Swensen et al.11 (Mayo model) and Gould et al.12 (VA model) identified independent clinical and radiological predictors of malignancy and used this to create and validate models with an AUC of 0.8. The Mayo model was validated with the additional parameter of positron-emission tomography (PET) avidity by Herder et al.,13 increasing the AUC to 0.92 (the Herder model). McWilliams et al.14 developed a model based on a lung cancer screening population. The AUC for this model was 0.94 and specifically 0.91 for nodules ≤10mm.
A number of case series have attempted to address the likelihood of malignancy in patients with known extra-thoracic cancer34, 35; however due to their heterogeneity, and small patient numbers, these studies do not provide sufficient data to distinguish benign from malignant nodules or a metastasis from a primary lung cancer.
The aim of this study was to validate three existing composite prediction models in a non-screening clinically referred patient population with small, predominantly subcentimetre PNs.
Section snippets
Materials and methods
The Research and Development Department at Oxford University Hospitals NHS Foundation Trust approved the study and waived the requirement for informed consent.
Results
Seven hundred and two of the 1,209 patients met the inclusion criteria for the study. Baseline demographics of the population are presented in Table 1. Current or former smokers accounted for 457 (65%) patients and 306 (43.6%) patients had an active or previous history of lung or extra-thoracic cancer.
Of 702 nodules identified, 379 (54%) were benign, 303 (43%) were pulmonary metastases, and 20 (3%) were primary lung cancers. The overall prevalence of malignancy in this cohort was 46%.
Size of
Discussion
To the authors' knowledge, this is the largest study to validate predictive models for PNs in a non-screening clinically referred patient population. All three of the validated models in this study were not specifically designed for use in patients with a previous history of malignancy within 5 years of presentation with a PN; however, in clinical practice, it is often this group of patients that present a challenge to differentiate between primary lung cancer, pulmonary metastases, or benign
Acknowledgements
There was no direct funding for this work. Dr Talwar is funded by the Technology Strategy Board (now Innovate UK) for a research project in collaboration with Mirada Medical Ltd.
References (36)
- et al.
Frequency and significance of pulmonary nodules on thin-section CT in patients with extrapulmonary malignant neoplasms
Eur J Radiol
(2012) - et al.
Small pulmonary nodules on CT accompanying surgically resectable lung cancer: likelihood of malignancy
J Thorac Imaging
(2002) - et al.
Incidental pulmonary nodules on cardiac computed tomography: prognosis and use
Am J Med
(2008) - et al.
A clinical model to estimate the pretest probability of lung cancer in patients with solitary pulmonary nodules
Chest
(2007) - et al.
Clinical prediction model to characterize pulmonary nodules: validation and added value of 18F-fluorodeoxyglucose positron emission tomography
Chest
(2005) - et al.
Relationship between a history of antecedent cancer and the probability of malignancy for a solitary pulmonary nodule
Chest
(2004) - et al.
Histology and outcomes of newly detected lung lesions in melanoma patients
Ann Oncol
(2012) - et al.
Likelihood of malignancy in a solitary pulmonary nodule: comparison of Bayesian analysis and results of FDG-PET scan
Chest
(1997) - et al.
Development and validation of a clinical prediction model to estimate the probability of malignancy in solitary pulmonary nodules in Chinese people
Clin Lung Cancer
(2011) - et al.
Usefulness of circumference difference for estimating the likelihood of malignancy in small solitary pulmonary nodules on CT
Lung Cancer
(2007)
Solitary pulmonary nodules: clinical prediction model versus physicians
Mayo Clin Proc
Clinical and radiologic factors associated with pulmonary nodule etiology in organ transplant recipients
Am J Transplant
Risk of malignancy in pulmonary nodules: a validation study of four prediction models
Lung Cancer
Pulmonary nodule: evaluation with spiral volumetric CT
Radiology
Low dose CT in early lung cancer diagnosis: prevalence data
Radiolog Med
Pulmonary nodules detected by thoracic computed tomography scan after exposure to asbestos: diagnostic significance
Int J Tubercul Lung Dis
Significance of non-calcified pulmonary nodules in patients with extrapulmonary cancers
Thorax
Pulmonary nodules in patients with esophageal carcinoma
J Clin Gastroenterol
Cited by (12)
Differences in VA and Non-VA Pulmonary Nodules: All Evaluations Are not Created Equal
2023, Clinical Lung CancerPulmonary nodule malignancy probability: a diagnostic accuracy meta-analysis of the Mayo model
2022, Clinical RadiologyCitation Excerpt :The QUADAS-2 tool also implicated a higher concern of applicability in the same domain, as many studies calculated the index test incorporating patients with exclusion criteria or did not use the suggested cut-offs, according to the ACCP guidelines. Thirty-eight studies (46,422 number of patients) provided data for the AUC of the Mayo model for predicting the probability of malignancy.4,16–29,31–43,45–54 Overall, the pooled AUC was 0.730 (95% CI: 0.704-0.755, p<0.001, Fig 2).
Risk-Based lung cancer screening: A systematic review
2020, Lung CancerCitation Excerpt :However, the PanCan model was found to be less accurate than radiologists in data from the NLST [103]. The PanCan model was found to be the most accurate model in external validation studies that included six other risk prediction models for nodule malignancy [104–106]. In a comparative analysis of four models, Uthoff et al. found that the Mayo model and the PanCan model performed the best in a cohort of 317 individuals with pulmonary nodules (AUCs of 0.63 and 0.61, respectively) whereas the VA model and the PKUPH models had demonstrated lower risk discrimination (AUCs of 0.51 and 0.55, respectively) [107].
Evaluation of Four Predictive Models for Identifying Malignancy of Solitary Pulmonary Nodules in Health Check-up Population
2023, Cancer Research on Prevention and TreatmentPredictive model for the probability of malignancy in solitary pulmonary nodules: a meta-analysis
2022, Journal of Cardiothoracic Surgery