Sensitivity and specificity of the diagnosis of periocular lesions by oculoplastic surgeons.

[BACKGROUND] A few studies have assessed the sensitivity of ophthalmologists in diagnosing periocular lesions. However, no study has assessed their diagnostic specificity or the degree to which they overdiagnose various lesions. We performed a study to determine the sensitivity and specificity of oculoplastic surgeons' preoperative diagnoses of common and uncommon periocular lesions.

[METHODS] Using the patient records of four oculoplastic ophthalmologists, we reviewed all charts documenting periocular surgery in which lesions were excised and sent for biopsy. For each of 358 cases we recorded the preoperative diagnosis as listed by the surgeon and the postoperative diagnosis as reported by the pathologist. Each preoperative diagnosis was then scored as correct or incorrect. The relative number of each type of lesion was determined, and the ophthalmologists' sensitivity and specificity were calculated.

[RESULTS] The sensitivity and specificity for the eight most frequently lesions were (in order of frequency) basal cell carcinoma 92.1% and 81.6% respectively, papilloma 81.6% and 66.0%, cyst 66.7% and 69.7%, nevus 53.6% and 75.0%, seborrheic keratosis 27.8% and 71.4%, chalazion 93.3% and 100.0%, squamous cell carcinoma 33.3% and 55.6%, and xanthelasma 100.0% and 76.9%. The values for melanoma were 50.0% and 28.6%, for sebaceous gland carcinoma 0.0% and 100.0%, and for uncommon lesions (nonspecific inflammation, actinic keratosis, granuloma, cavernous hemangioma, folliculitis, benign sweat gland tumour, eccrine hidrocystoma and collagenous tissue) 27.8% and 57.1%. The overall accuracy in lesion identification was 70.0%.

[INTERPRETATION] Although ophthalmologists have excellent diagnostic sensitivity and specificity for some types of lesion, such as basal cell carcinoma and chalazion, the identification of other lesion types, such as cyst, squamous cell carcinoma and possibly melanoma and sebaceous gland carcinoma, is less optimal.