Effects of biographical variables on cervical fluorescence emission spectra.

Diagnostic algorithms can classify tissue samples as diseased or nondiseased based on fluorescence emission collected from the intact cervix. Such algorithms can distinguish high-grade squamous intraepithelial lesions from low-grade squamous intraepithelial lesions. An understanding of the effects of the values of biographical covariates, such as age, race, smoking, or menopausal status on the emission spectra for each patient could improve diagnostic efficiency. The analysis described was performed using data collected from two previously published clinical trials; one study measured spectra from 395 sites in 95 patients referred to a colposcopy clinic with abnormal Pap smears, and the second study measured spectra from 204 sites in 54 patients self-referred for screening and expected to have a normal Pap smear. For this analysis, data about age, race, menstrual cycle, and smoking were collected. The principal components from normalized data were compared. There are clear intensity differences observed with age and menopausal status; postmenopausal patients exhibit higher emission intensities. Differences associated with biographical variables need to be tested in larger studies, which stratify adequately for these variables. The addition of these biographical variables in the preprocessing of data could dramatically improve algorithm performance and applicability.