A Novel Prognostic Assessment for Identifying Oral Pre-Malignant Lesions at High Risk for Progression to Cancer

Oral pre-malignant lesions (OPLs) are quite common, frequently asymptomatic, and generally detected during routine oral exams. While the transformation rate of pre-cancerous to cancerous lesions is less than 5% per year, most early stage cancers and OPLs are also asymptomatic, making identification more difficult. Most OPLs do not require aggressive treatment, however, preventing the transformation to malignancy is key to impacting oral cancer morbidity and mortality. Furthermore, the high mortality rate associated with oral cancer and the low transformation rate of OPLs creates a strong need for reliable assessments that more accurately identify lesions at high-risk of transformation, separating these lesions from those at lesser transformation risk. The standard of care for OPL risk assessment, dysplasia grading by histopathology, is impacted by intra- and inter-observer variation as well as significant overlap between grades, affecting its usefulness as a prognostic tool. The Straticyte^TM oral prognostic assessment has been developed to meet these needs.

OPL biopsy samples from 150 cases with a follow-up history of up to 12 years were used. Immunohistochemistry for the biomarker S100A7 on tissue biopsy slides and tissue microarrays was performed at Mount Sinai Hospital in Toronto, Ontario, Canada. The slides were then digitally scanned on a Hamamatsu Nanozoomer-XR slide scanner and images were visualized and analyzed using Visiopharm VIS (version 5; Horsholm, Denmark). This project was approved by the Mount Sinai Hospital Research Ethics Board.

All statistical analyses and model building were conducted using the R package (version 3). Stepwise Cox Regression was used to select the parameters. A multivariate Cox Regression model was fitted to selected parameters and the C-index was used to assess the model. Estimated Log Relative-Hazards from the Cox model were referred to as risk scores and used in the cut-off selection stage to classify all cases into three risk groups: low, intermediate, and high. The Nelson-Aalen-Breslow estimate, used to calculate the baseline cancer-free survival curve, is combined with the calculated risk score to produce the expected cancer-free survival probability for each case. The Aalen-Link-Tsiatis estimate, used to estimate the variance of expected cancer-free survival probability, provided the 95% confidence interval (CI) of the cancer-free survival curve.

From the 150 cases, the 95% CI of mild, moderate, and severe dysplasia grades based on histopathological assessment, overlapped extensively throughout the first 60 months, indicating ineffective differentiation. In contrast, the 95% CIs of the Straticyte classified groups had minimal overlaps at month 60, achieving better differentiation. The performance of Straticyte was evaluated by an internal validation study using the split-sample technique. Comparing the C-index (time-to-event response) and Area Under the Curve (AUC; binary response), Straticyte risk scores were more objective and discriminatory than histopathological dysplasia grading. Furthermore, Straticyte outperformed histopathological dysplasia grading in two clinical indices. The sensitivity between the low-risk vs. non-low-risk groups in Straticyte was 96% compared to mild vs. non-mild dysplasia gradings of 75%, with a negative predictive value of 80% and 59%, respectively.

Straticyte is intended to better categorize a patient’s 5-year risk for OPLs to progress to cancer and should be regarded as a complement to conventional histopathology. Straticyte can be easily incorporated into clinical practice as no additional tissue samples are needed for the assessment.

References:

1. Lingen, MW, et al: Critical evaluation of diagnostic aids for the detection of oral cancer. Oral Oncol 44(1): 10-22, 2008

2. Warnakulasuriya S. et al: Oral epithelial dysplasia classification systems: Predictive value, utility, weakness and scope for improvement. J Oral Pathol Med 37(3): 127-133, 2008