Autofluorescence bronchoscopy (AF) has been shown to improve the detection of early central lung cancer. This study was designed to evaluate the D-light Autofluorescence System (Karl Storz Endoscopy America; Culver City, CA) in comparison to white light bronchoscopy (WL) in detecting class III (severe dysplasia, carcinoma in situ, early invasive cancer) in patients with known or suspected lung cancer. Patients with known or suspected lung cancer underwent diagnostic WL followed by AF. Findings were classified as follows: class I, normal; class II, normal, nonspecific; class III, abnormal/premalignant; and class IV, tumor. All class III lesions were sampled, along with two areas normal by AF and WL. Local and reference pathologists were blinded as to the location and bronchoscopic characteristics of the samples. The McNemar test was used to compare primary end points of sensitivity and specificity in AF vs WL. Exact χ2 was used to compare secondary end points, negative predictive value and positive predictive value, in AF vs WL. Three hundred patients were enrolled (180 men and 120 women, mean age, 65.4 years); 278 patients (93%) completed the protocol. Nine hundred one biopsy samples were obtained; 80 were excluded because of tumor or unsatisfactory. Eight hundred twenty-one biopsy samples from 293 patients (2.8 biopsy observations per patient) were analyzed. Under AF, sensitivity was 61.2%, compared to 10.6% under WL (p < 0.0001 by McNemar test). Specificity was 94.6% under WL and 75.3% under AF (p < 0.0001). Positive predictive value for AF was 22.2% vs 18.4% for WL (p = 0.49); negative predictive value for AF was 94.9% vs 90.2% for WL (p = <0.01). AF significantly improves the sensitivity of detecting class III mucosal lesions of the central airways. The negative predictive value of AF is superior to WL. Results with the D-light Autofluorescence System are similar to those reported with other laser-stimulated AF systems.