Premium
Risk stratification in the investigation of pulmonary nodules in a high‐risk cohort: positron emission tomography/computed tomography outperforms clinical risk prediction algorithms
Author(s) -
Gibson Glenna,
Kumar Aravind Ravi,
Steinke Karin,
Bashirzadeh Farzad,
Roach Rebecca,
Windsor Morgan,
Ware Robert,
Fielding David
Publication year - 2017
Publication title -
internal medicine journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.596
H-Index - 70
eISSN - 1445-5994
pISSN - 1444-0903
DOI - 10.1111/imj.13576
Subject(s) - medicine , positron emission tomography , radiology , confidence interval , receiver operating characteristic , biopsy , malignancy , pet ct , nuclear medicine , tomography , area under the curve , retrospective cohort study , solitary pulmonary nodule , fluorodeoxyglucose , computed tomography , algorithm , surgery , computer science
Abstract Background Clinical prediction models and 18‐fluorine‐fluorodeoxyglucose positron emission tomography/computed tomography (18F‐FDG‐PET/CT) are used for the assessment of solitary pulmonary nodules (SPN); however, a biopsy is still required before treatment, which carries risk. Aim To determine the combined predictive benefit of one such model combined with modern PET/CT data to improve decision‐making about biopsy prior to treatment and possibly reduce costs. Methods Patients with a SPN undergoing 18F‐FDG‐PET/CT from January 2011 to December 2012 were retrospectively identified; 143 patients met inclusion criteria. PET/CT studies were rated (5‐point visual scale), and CT characteristics were determined. Tissue was obtained by endobronchial ultrasonography with guide sheath (EBUS‐GS), CT‐guided biopsy and/or surgery. EBUS‐transbronchial needle aspiration (TBNA) was used instead of nodule biopsy if there were PET‐positive sub‐centimetre lymph nodes. Results The prediction model yielded an area under the receiver operating characteristic curve (AUC‐ROC) of 64% (95% confidence interval (CI) 0.55–0.75). PET/CT increased this to 75% (95% CI 0.65–0.84). The 11% improvement is statistically significant. PET/CT score was the best single predictor for malignancy. A PET score of 1–2 had a specificity of 100% (CI 0.73–1.0), whereas a score of 4–5 had a sensitivity of only 76% (CI 0.68–0.84). No significant difference in clinical prediction scores between groups was noted. PET/CT showed the greatest benefit in true negatives and in detecting small mediastinal lymph nodes to allow EBUS‐TBNA with a higher diagnostic rate. Cost analysis did not support a policy of resection‐without‐tissue diagnosis. Conclusion PET/CT improves the clinical prediction of SPN, but its greatest use is in proving benignity. High PET scores had high false positive rates and did not add to clinical prediction. PET should be incorporated early in decision‐making to allow for more effective biopsy strategies.