ABSTRACT
Despite clearly defined histologic criteria, the prediction of tumor behavior for patients with gastrointestinal stromal tumors (GIST) still poses a challenge to pathologists. Therefore, searching for alternative markers that allow for better prognostic evaluation is an important task. To determine the practicability of immunohistochemical staining for p16 in clinical cases, we examined p16 protein expression in a group of 284 GISTs, a subset of which had long-term follow-up (median, 45 months; range, 1-204 months). P16 protein expression was ascertained on tissue microarrays as well as on standard sections. Survival analyses were carried out in 157 patients. P16 loss was found in 50% of GISTs, there being no correlation with age, sex, histologic subtype, signs of necrosis, or metastases. Patients having p16-negative tumors had a worse prognosis than those with p16-positive tumors (P = 0.012) with a 2.3-fold relative increased risk of dying of disease. P16 loss identified a subgroup of gastric tumors with a worse prognosis (P = 0.03). The multivariate configural frequency analysis identified two "antitypes," whose observed frequency was found to be significantly lower than the expected frequency [i.e., marker combinations: p16 positive, no metastases, and death of disease and p16 loss, metastases, and still alive]. The "type" whose observed frequency was significantly higher than the expected frequency consisted of the following marker pattern: p16 loss, necrosis, and death of disease (P < 0.001). In the multivariate Cox regression analysis, p16 loss, necrosis, and metastases each had independent prognostic value. P16 loss is a common molecular abnormality in GISTs and might be used in routine diagnosis to identify patients with high-risk tumors.
