Abstract
BACKGROUND: High-quality randomized controlled trials have demonstrated the benefit of radiotherapy (RT) in patients with radical resected gastric cancer (GC), however, utilization rates of postoperative RT remain remarkably low. Patterns, incidences, and time of recurrence provide biological bases for clinical monitoring of GC patients and guiding potential complementary therapies. Thus, the aim of this study is to understand the location of locoregional recurrence which may allow individualized RT strategies and minimize radiation-related toxic effects. METHODS: A relatively large sample of GC patients in a single institution who had undergone curative D2 resection was retrospectively reviewed and the relevant recurrence patterns were illustrated. Independent recurrence-related risk factors were analyzed by logistic regression analysis. New logistic regression models were further developed to predict the probability of recurrence. RESULTS: Overall, among 776 GC patients who had continuous and complete follow-up data, 300 cases relapsed after curative resection. Lymphovascular invasion, lymph node metastases, and tumor stage were indicators for early recurrence. Peritoneal, regional, local, and distant recurrence initially occurred in 51 (6.6%), 151 (19.4%), 56 (7.2%), and 164 (21.1%) patients, respectively. Among patients with regional recurrence, the most common sites were lymph node stations 16a2, 8, 12, 16b1, and 9. Remnant stomach recurrence was not so prominent that it seemed reasonable to be excluded from an irradiation field for patients with negative surgical/pathologic margins. CONCLUSIONS: For GC patients who underwent radical D2 resection, distant and regional recurrences were still common. Besides, optimizing regional control of lymph nodes outside the D2 dissected area was crucial for rational design of the RT field. Furthermore, the new logistic regression models might act as useful tools to evaluate recurrence risk and determine which patients should receive postoperative chemoradiotherapy.