Skin examinations by primary care clinicians were highly specific but had low sensitivity for the detection of skin cancer
ACP J Club. 1997 Nov-Dec;127:77. doi:10.7326/ACPJC-1997-127-3-077
Whited JD, Hall RP, Simel DL, Horner RD. Primary care clinicians' performance for detecting actinic keratoses and skin cancer. Arch Intern Med. 1997 May 12;157:985-90.
To compare the accuracy of examinations done by primary care clinicians with those done by dermatologists for the diagnosis of skin cancer.
A blinded comparison of examinations for the detection of skin cancer done by primary care clinicians and dermatologists.
General internal medicine and dermatology outpatient clinics in a Veterans Affairs medical center in North Carolina, USA.
190 men who were ≥ 40 years of age (mean age 67 y, 100% white) and who presented to the general medicine clinic or dermatology outpatient clinic.
Description of test and diagnostic standard
Complete skin examinations above the waistline were done independently by 1 primary care clinician and 1 dermatologist on the same day. Patients were examined by a primary care clinician who pointed out skin lesions believed to be malignant (those the clinician thought required biopsy) and actinic keratoses. A second examination was done by a dermatologist (the diagnostic standard) who was unaware of the findings of the primary care clinician. Patients judged to have malignant lesions by primary clinicians but not by dermatologists were reexamined. If the dermatologist still judged the lesions to be nonmalignant, the patient was followed for 6 months.
Main outcome measures
Sensitivity and specificity for the diagnoses of malignancies by primary care clinicians.
Biopsies were done on 104 of 109 lesions thought to be malignant by dermatologists, and 64 were tissue-proved malignancies. Sensitivity, specificity, and likelihood ratios for the presence of disease if the test result was positive or negative as determined by a primary care clinician are listed in the Table for both the patient and lesion as the unit of analysis.
Skin examinations done by primary care clinicians were specific but had low sensitivity for the detection of skin cancer.
Source of funding: Not stated.
For article reprint: Dr. J.D. Whited, Center for Health Services Research in Primary Care, Durham Veterans Affairs Medical Center (152), 508 Fulton Street, Durham, NC 27705, USA. FAX 919-416-5836.
Table. Test features for the diagnosis of skin cancer*
|Unit of analysis||Sensitivity, % (CI)||Specificity, % (CI)||+LR||-LR|
|Patient||57 (44 to 68)||88 (81 to 93)||4.9||0.48|
|Lesions||38 (29 to 47)||95 (93 to 96)||7.1||0.66|
*+LR = likelihood ratio for presence of disease if the test is positive; -LR = likelihood ratio if the test is negative. CI defined in Glossary.
The study by Whited and colleagues shows, like others, that examinations for detecting skin cancer done by dermatologists have greater sensitivity than those done by nonspecialists (1). The additional training required to equalize the diagnostic performance of nondermatologists and dermatologists is measured in years rather than weeks (2).
The study also confirms that even the predictive value of a clinical diagnosis of skin cancer made by an experienced dermatologist is usually between 50% and 60% (3).
Although the use of patients (rather than pictures) is an advantage of this study, it also causes problems. Because the authors attempted to generate a high prevalence of skin cancer, 60% of the recruited patients had previously had skin cancer or solar keratosis. Therefore, the group is unlike the general population and patients may have been recognized by the examining dermatologists who had treated them previously for skin cancer. In addition, the study involved only non-melanoma tumors.
Readers may also wonder whether the primary care clinicians involved (primarily internal medicine attending physicians) are similar to primary care physicians elsewhere. Restricting screening for skin cancer to dermatologists, however, cannot be strongly recommended on the basis of this limited evidence. Important uncertainty persists about the overall benefit of early tumor detection, and arguably, a need remains for experimental evaluation of screening (4, 5).
Ian Harvey, MB, PhD
University of BristolBristol, England, UK