Use of lymph node ratio improves staging and selection criteria for adjuvant therapy of gastric cancer

4051 Background: Despite 1997 American Joint Commission on Cancer (AJCC) guidelines stipulating assessment of ≥15 lymph nodes (LN) for staging of gastric cancer, only one third of patients in the Surveillance, Epidemiology and End Results (SEER) database from 1998–2002 had ≥15 LN assessed (ASCO 2005 #4004), with resultant understaging and probable under-treatment. In series from Asia and Europe, Lymph Node Ratio (LNR), the ratio of positive to total LN assessed, has been shown to be more accurate for staging than number of positive LN. However, most of these excluded cases with <15 LN assessed. We examined the utility of LNR in a North American population. Methods: Using SEER data, we identified 9503 M0 resected gastric cancer cases from 1988–2002. LNR was categorized as 0%, 1–10%, 11–30%, 31–50% and >50%. For node negative cases (LNR = 0%, n = 3652), we stratified by number of LN assessed (A=1–4; B = 5–9; C = 10–14; D≥15). For each AJCC stage or LNR strata, the degree of understaging in patients with inadequate LN assessment was measured by survival difference on Kaplan-Meier curves. Cox proportional hazard ratio (HR) models determined the effect of stratifying node negative patients and the accuracy of LNR for prognostication. Results: 27% of patients had a LNR > 50%, a high proportion compared to Asian series. Fewer nodes assessed resulted in a higher likelihood of being node negative. In node negative cases, the HR of death increased for those with fewer LN assessed (vs. Group D, with 95% CI): A: HR=1.6 (1.5–1.8); B: HR = 1.3 (1.1–1.5); C: HR = 1.3 (1.1–1.5). Understaging was observed for patients with inadequate LN assessment when AJCC criteria were used (p < 0.0001); this effect significantly decreased by using LNR. LNR had superior prognostic accuracy in Cox models. Conclusions: This study examines LNR in the largest series of resected gastric cancer in the literature, and the only one in which the majority of cases were inadequately staged. LNR significantly decreases understaging and improves prognostic ability. Node negative patients, nearly one third of cases, should be risk stratified by number of LN assessed, and considered for adjuvant therapy on this basis. LNR should be used to stratify node positive patients in clinical trials, and to provide more accurate staging and prognostication. No significant financial relationships to disclose.