Reproducibility and prognostic value of pattern of invasion scoring in low-stage oral squamous cell carcinoma.

AIMS: To evaluate and compare the prognostic value and reproducibility of different methods of pattern of invasion scoring in oral squamous cell carcinomas. The additional prognostic value to established histopathological prognostic factors was also analysed. METHODS AND RESULTS: The study group was confined to 211 previously untreated patients who underwent surgery for low-stage oral squamous cell carcinoma between 1997 and 2008. Median follow-up was 64 months (range 0-193 months). Pattern of invasion was scored using five previously described methods, at random and independently, by two observers. Pattern of invasion scoring showed moderate interobserver reproducibility (Cohen's kappa = 0.52-0.58). The predominant pattern of invasion and the summed predominant and worst pattern of invasion were independent prognosticators for locoregional recurrence-free survival (LRRFS) [hazard ratio (HR): 2.1, P = 0.033 and HR 2.2, P = 0.024, respectively] and disease-specific survival (DSS) (HR 2.3, P = 0.032 and HR 2.1, P = 0.044, respectively) in multivariate Cox regression analyses. The Harrell's C index for proven prognostic histopathological factors was 0.66 for LRRFS and 0.67 for DSS. This improved to 0.69 and 0.73 with the addition of pattern of invasion. CONCLUSIONS: Pattern of invasion is an independent prognostic factor in low-stage oral squamous cell carcinoma. However, it has a moderate reproducibility, and the contributory value next to other prognostic histopathological factors is minimal.