CK7 immunohistochemistry as a predictor of CIN1 progression: A retrospective study of patients from the quadrivalent HPV vaccine trials

Cervical high-grade squamous intraepithelial lesion (CIN2-3) is thought to arise from a distinct population of cells at the squamocolumnar junction (SCJ). Immunohistochemical (IHC) biomarkers that characterize the SCJ phenotype, including CK7, have been proposed as tools to separate the subset of low-grade squamous intraepithelial lesions (LSILs) (CIN1) that will progress to high-grade squamous intraepithelial lesion from the majority of cases, which will resolve without further intervention. We conducted a retrospective study of CK7 IHC on adjudicated CIN1 tissue from women in the placebo arm of the quadrivalent human papillomavirus (HPV) vaccine trials. Tissue sections were stained with CK7 IHC and scored as negative, patchy, gradation (ie, top-down), or full-thickness pattern. Results were assessed for the prediction of future diagnosis of CIN2-3/AIS (eg, CIN2+ progression) along with p16 IHC, antecedent high-grade cytology, and HPV16 status. A total of 517 patients with CIN1 biopsies and complete data were identified, 12% of whom showed CIN2+ progression on follow-up. Full-thickness CK7 staining showed the highest correlation with CIN2+ progression (odds ratio [OR] 2.8, P=0.021) relative to the other risk factors (HPV16: OR 2.0, P=0.035; antecedent high-grade cytology: OR 2.2, P=0.028; p16 IHC: OR 1.5, P=0.16). Inclusion of the gradation/"top-down" CK7 pattern resulted in a less robust association with progression (CIN2+: OR 2.0, P=0.028; CIN3+: OR 1.3, P=0.74). Interobserver variability ranged from slight to substantial and was not contingent on gynecologic pathology training experience (k=0.7078 for negative/patchy vs. gradation/full thickness; k=0.5672 for negative/patchy/gradation vs. full thickness). These data support the theory that SCJ-derived LSILs are precursors to a potentially aggressive subset of cervical SILs and that CK7 staining may inform risk stratification for LSIL (CIN1). However, clinical utility is significantly tempered by the relatively low amplitude of the risk increase, interpretative variability, and limitations of colposcopic sampling.