High accuracy of common HIV-related oral disease diagnoses by non-oral health specialists in the AIDS Clinical Trial Group

Abstract

Objective: Many studies include oral HIV-related endpoints that may be diagnosed by non-oral-health specialists (non-OHS) like nurses or physicians. Our objective was to assess the accuracy of clinical diagnoses of HIV-related oral lesions made by non-OHS compared to diagnoses made by OHS. Methods: A5254, a cross-sectional study conducted by the Oral HIV/AIDS Research Alliance within the AIDS Clinical Trial Group, enrolled HIV-1-infected adults participants from six clinical trial units (CTU) in the US (San Francisco, New York, Chapel Hill, Cleveland, Atlanta) and Haiti. CTU examiners (non-OHS) received standardized training on how to perform an oral examination and make clinical diagnoses of specific oral disease endpoints. Diagnoses by calibrated non-OHS were compared to those made by calibrated OHS, and sensitivity and specificity computed. Results: Among 324 participants, the majority were black (73%), men (66%), and the median CD4+ cell count 138 cells/mm3. The overall frequency of oral mucosal disease diagnosed by OHS was 43% in US sites, and 90% in Haiti. Oral candidiasis (OC) was detected in 153 (47%) by OHS, with erythematous candidiasis (EC) the most common type (39%) followed by pseudomembranous candidiasis (PC; 26%). The highest prevalence of OC (79%) was among participants in Haiti, and among those with CD4+ cell count ≤ 200 cells/mm3 and HIV-1 RNA > 1000 copies/mL (71%). The sensitivity and specificity of OC diagnoses by non-OHS were 90% and 92% (for EC: 81% and 94%; PC: 82% and 95%). Sensitivity and specificity were also high for KS (87% and 94%, respectively), but sensitivity was < 60% for HL and oral warts in all sites combined. The Candida culture confirmation of OC clinical diagnoses (as defined by ≥ 1 colony forming unit per mL of oral/throat rinse) was ≥ 93% for both PC and EC. Conclusion: Trained non-OHS showed high accuracy of clinical diagnoses of OC in comparison with OHS, suggesting their usefulness in studies in resource-poor settings, but detection of less common lesions may require OHS.