Management of Pulmonary Nodules by Community Pulmonologists

BACKGROUND Pulmonary nodules (PNs) are a common reason for referral to pulmonologists. The majority of data for the evaluation and management of PNs is derived from studies performed in academic medical centers. Little is known about the prevalence and diagnosis of PNs, the use of diagnostic testing, or the management of PNs by community pulmonologists. METHODS This multicenter observational record review evaluated 377 patients aged 40 to 89 years referred to 18 geographically diverse community pulmonary practices for intermediate PNs (8-20 mm). Study measures included the prevalence of malignancy, procedure/test use, and nodule pretest probability of malignancy as calculated by two previously validated models. The relationship between calculated pretest probability and management decisions was evaluated. RESULTS The prevalence of malignancy was 25% (n = 94). Nearly one-half of the patients (46%, n=175) had surveillance alone. Biopsy was performed on 125 patients (33.2%). A total of 77 patients (20.4%) underwent surgery, of whom 35% (n = 27) had benign disease. PET scan was used in 141 patients (37%). The false-positive rate for PET scan was 39% (95% CI, 27.1%-52.1%). Pretest probability of malignancy calculations showed that 9.5% (n = 36) were at a low risk, 79.6% (n = 300) were at a moderate risk, and 10.8% (n = 41) were at a high risk of malignancy. The rate of surgical resection was similar among the three groups (17%, 21%, 17%, respectively; P = .69). CONCLUSIONS A substantial fraction of intermediate-sized nodules referred to pulmonologists ultimately prove to be lung cancer. Despite advances in imaging and nonsurgical biopsy techniques, invasive sampling of low-risk nodules and surgical resection of benign nodules remain common, suggesting a lack of adherence to guidelines for the management of PNs. Materials and Methods This was a multicenter, community-based, retrospective observational study of patients with PNs, ranging from 8 to 20 mm in diameter, presenting to 18 geographically representative outpatient pulmonary clinics across the United States. The study was approved at 15 sites by a central institutional review board and at three sites by local institutional review board approval. Site Selection Four hundred forty sites were identified based on investigator databases and claims data from a large insurance carrier whose coverage population was representative of the overall US population. Of these, 77 sites expressed interest in participating, and 48 sites went on to sign confidentiality agreements. Of these, 17 did not request additional information, leaving 31 sites undergoing qualification review. Eighteen outpatient pulmonary clinics were chosen to participate based on the following criteria: (1) management of patients with PNs, (2) availability of medical records, and (3) ability to perform data abstraction. In addition, investigators targeted enrollment of geographically diverse patients to limit the potential bias associated with differences in practice patterns and to account for variation in disease prevalence (eg, endemic mycoses) that could alter management decisions. Patient Selection Patients were identified by querying databases (eg, billing and scheduling systems) using five International Classification of Diseases, Ninth Revision, Clinical Modification codes for PN (793.1, 786.6, 518.89, 519.8, 519.9) to ensure homogeneity in patient identification and inclusion. 17 Manual chart abstraction was then used to identify those who met the criteria. To minimize selection bias, the sites were not permitted to use additional codes during database query to identify patients. To ensure a systematic sample, patient eligibility was determined by examining consecutively referred patients to the site. Inclusion criteria included age ≥ 40 years and ≤ 89 years at the time of nodule finding, presentation to a pulmonologist, nodule size 8 to 20 mm, and definitive diagnosis ascertained by tissue diagnosis or radiographic follow-up for 2 years. Exclusion criteria included chest CT scan performed > 60 days prior to the initial visit, prior diagnosis of any cancer within 2 years of nodule detection, or incomplete chart data. Patients were categorized into three groups by the most invasive procedure performed during management, as follows: surveillance (serial imaging), biopsy (CT scan-guided transthoracic needle aspiration [TTNA] or bronchoscopy), or surgery (including mediastinoscopy, video-assisted thorascopic surgery, and/or thoracotomy). Data Collection Clinical data were abstracted retrospectively by designated study staff into an electronic data capture system from initial consultation through establishment of a definitive diagnosis (ie, pathology results) or a minimum 2-year follow-up. Data included patient demographic and clinical characteristics, PN characteristics, imaging tests, invasive testing, and surgery. PET scan reports were reviewed and abstracted where available for a subset of patients. To adjudicate a PET scan report as positive or negative, an algorithm was developed that prioritized the following components of the report from highest to lowest: final radiology impression, description of findings, and standard uptake values (SUVs) (e-Fig 1). PET scanning was defined as negative if the report included any of the following statements: no evidence of malignancy, no 18F-fluorodeoxyglucose uptake or hypermetabolic activity, or an SUV of > 2.5. A positive PET scan was defined as a report that included any of the following statements: concern or suggestion of malignancy, findings that described increased 18F-fluorodeoxyglucose uptake or hypermetabolic activity, or an SUV ≥ 2.5. In the 27 cases in which the findings were discordant with the final impression, adjudication was performed by two independent pulmonologists and agreement was reached in all cases. Data quality was ensured through ongoing site monitoring. Programmed edit checks were built into the electronic data capture system and at the conclusion of chart abstraction, each site provided access to a random 10% sample of deidentified patient records for review. Pretest Probability of Malignancy Two previously developed and validated models 9,10 were used to estimate the pCA in each patient. Model accuracy was determined by comparing the pCA with the final diagnosis. Receiver operating characteristic curves and the area under the curve were generated with 95% CIs. The pCA was calculated for each patient and categorized into three groups ( Data Analysis χ 2 or analysis of variance tests were used to compare subgroups, and P values > .05 were considered significant. A nodule was classified as benign based on confirmed benign pathology or the absence of radiographic change as determined by the managing physician during surveillance for at least 2 years. Multivariate logistic regression was performed to identify factors associated with the use of an invasive diagnostic procedure. All statistical analyses were performed using SAS/STAT, version 9.3 (SAS Institute inc).