Sensitivity and Specificity of Malignant Melanoma, Squamous Cell Carcinoma, and Basal Cell Carcinoma in a General Dermatological Practice

Introduction. Incidence of melanoma and non-melanoma skin cancer is increasing worldwide. Melanoma is the sixth most common cancer in the United States, making skin cancer a significant public health issue. Background and goal. The goal of this study was to provide estimates for sensitivity (P(T+|D+)), specificity (P(T-|D-)), and likelihood ratios (P(T+|D+)/P(T+|D-)) for a positive test and (P(T-|D+)/P(T-|D-)) for negative test of clinical diagnosis compared with pathology reports for malignant melanoma (MM), squamous cell carcinoma (SCC) , basal cell carcinoma (BCC), and benign lesions. This retrospective cohort study collected data on 595 patients with 2,973 lesions in a Central Texas dermatology clinic, randomly selecting patients seen by the dermatology clinic between 1995 and 2011. The ascertation of disease was documented on the pathology report and served as the “gold standard.”
Hypotheses. Major hypotheses were that the percentage of agreement beyond that expected by chance between the clinicians’ diagnosis and the pathological gold standard were 0.10, 0.10, 0.30, and 0.40 for MM, SCC, BCC and benign lesions respectively. Results. For MM, the resulting estimates were: (a) 0.1739 (95% C.I. 0.0495, 0.3878), for sensitivity; (b) 0.9952 (95% C.I. 0.9920, 0.9974) for specificity; and (c) the likelihood ratios for a positive and negative test result were 36.23 and 0.83, respectively. For SCC, the resulting estimates were (a) 0.0833 (95% C.I. 0.0312, 0.1726) for sensitivity; (b) 0.9976 (95% C.I. 0.9950, 0.9990); and (c) the likelihood ratios for a positive and negative test result were 34.71 and 0.92, respectively. For BCC, the resulting estimates were: (a) 0.2178 (95% C.I. 0.1630, 0.2812) for sensitivity; (b) 0.9910 (95% C.I. 0.9867, 0.9941) for specificity; and (c) the likelihood ratios for a positive and negative test result were 24.20 and 0.79, respectively. For benign lesions, the resulting estimates were (a) 0.4942 (95% C.I. 0.4715, 0.5169) for sensitivity; (b) 0.9305 (95% C.I. 0.9135, 0.9450) for specificity; and (c) the likelihood ratios for a positive and negative test result were 7.11 and 0.54, respectively. Estimates for the kappa statistic (95% confidence intervals) were 0.1896 (0.0261, 0.3532), 0.1898 (0.0899, 0.2896), 0.3308 (0.2608, 0.3532), and 0.3585 (0.3319, 0.3850) for MM, SCC, BCC, and benign lesions, respectively. Conclusions. Over-biopsying lesions and fear of missing malignancy have a significant impact on the sensitivity and specificity of clinical diagnosis, leading to lowered accuracy. These results challenge clinicians to continue to work toward improving their diagnostic skills concerning MM, SCC, BCC, and benign lesions.