Survival of 2,991 patients with surgical lung cancer: the denominator effect in survival

To evaluate the current staging system of lung cancer, taking into account different selection criteria for the studied population.

A total of 2,991 consecutive patients with surgical lung cancer were prospectively compiled from 19 Spanish hospitals (Bronchogenic Carcinoma Cooperative Group of the Spanish Society of Pneumology and Thoracic Surgery) between 1993 and 1997.

The Kaplan-Meier method was used to calculate survival at 5 years (S5) for each pathologic stage, and the log-rank test was used for comparison purposes. These studies were performed in the total group (population 1, n = 2,972); excluding operative mortality and small cell lung cancer cases (population 2, n = 2,697); excluding cases with induction therapy (population 3, n = 2,542); excluding cases with exploratory thoracotomy (population 4, n = 2,304); and, lastly, excluding cases with incomplete resection (population 5, n = 2082) [70% of the initial population].

The global S5 was similar in populations 1, 2, and 3: 34% (95% confidence interval [CI] 32 to 36%), 37% (95% CI, 35 to 39%), and 38% (95% CI, 35 to 39%), but different from that of populations 4 and 5: 40% (95% CI, 39 to 43%) and 43% (41 to 45%), respectively. For pathologic stage I, pathologic stage II, and pathologic state IIIA (pIIIA), S5 was similar in the five reported populations. In pathologic stage IIIB (pIIIB), there were differences in S5 between populations 1, 2, and 3 (13 to 15%; 95% CI, 10 to 19%) and populations 4 and 5 (26 to 29%; 95% CI, 19 to 38%). In population 4, there was no significant prognostic difference between two specific stage groups, that is between pathologic stage IB (pIB) and pathologic state IIA (pIIA) [p = 0.70] and between pIIIA and pIIIB (p = 0.79); the pathologic T3N2M0 combination has a S5 (13%) lower than that for pIIIB (26%).

The definition of the population that constitutes the denominator for the analysis of survival in surgical lung cancer is important in pIIIB. The inclusion or exclusion of cases without resection is the most important factor for the selection of such population. This study detected that there are no prognostic differences between pIB and pIIA, and between pIIIA and pIIIB.