[18F]Fluorodeoxyglucose positron emission tomography-computerized tomography and lung cancer: a significant referral bias exists

Objective: [ 18 F]Fluorodeoxyglucose positron emission tomography-computerized tomography (PET-CT) scan is a tool widely used in the diagnosis and staging of lung cancer. Referral bias is present when the results of a diagnostic study affect the decision to proceed with definitive testing. This bias artificially increases the sensitivity and decreases the specificity, and may inappropriately alter the decision to undergo definitive testing. The accuracy of PET-CTscan in suspected lung cancer and the role of referral bias were investigated. Methods: From January 2005 through June 2007, 584 consecutive patients undergoing PET-CTscan for suspected lung cancer were studied. Endpoints measured included qualitative and quantitative results of PET-CT scans and pathologic results from patients, who underwent invasive procedures for diagnosis, staging and/or therapy. A positive PET scan was defined as one in which the standard uptake value (SUV) was greater than 2.5. A standard mathematical model, based on overall results of PET-CTscan in all patients, was used to create adjustments to account for the effect of referral bias. Results: A total of 414 (71%) of PET-CTscans were reported as positive, while 170 (29%) were negative. Attempt at tissue diagnosis occurred in 417 patients (71%); surgical intervention was performedin 246(42%). Whereas86% (355/414) of patientswith a positive PET-CTscan underwent tissue sampling, only 36% (62/170) with a negative PET-CTscan had an attempt ( p < .001). In patients with a positive study, a higher SUV was associated with an attempt at tissue diagnosis (p < .001). Before adjustment, the sensitivity and specificity of the PET-CTscan for any cancer were 87% and 43%. After adjustment, sensitivity fell to 75%, while specificity rose to 64%. The unadjusted sensitivity of PET-CTscan for detecting mediastinal disease was only 79%. Conclusions: The estimate of the accuracy of PET-CT scan is significantly influenced by referral bias, and a strong reliance is placed on its results. Furthermore, patients with a positive PET-CTscan are more likely to undergo tissue sampling as the SUV increases. Given the relatively low sensitivity of the PET-CTscan in detecting disease and its poor performance in the mediastinum, the PET-CT scan may have too high an impact on the decision to undergo further invasive diagnostic procedures. Patients should not be deferred from tissue sampling based solely on a negative PET-CT scan. # 2010 European Association for Cardio-Thoracic Surgery. Published by Elsevier B.V. All rights reserved.