Limitations of the statistical approach must be discussed, because they are widely unrecognized. The multivariate analysis involving many factors that have been suggested to be of prognostic value and the large number of patients are strengths. Nevertheless, this is an exploratory analysis, in which a cohort is analyzed and whatever falls out as statistically significant is reported. This approach is associated with a fairly high chance of a false-positive finding (ie, identification of a factor that will not hold up to validation studies). Until validation of these factors is demonstrated by an independent cohort, we must remain somewhat skeptical of the results. The fact that the significant factors for stage I and II are not the same should raise suspicion that we are not uncovering an important aspect of tumor biology. Although other studies have suggested that most of the factors studied may be prognostic, the results have been inconsistent. In fact, in the large International Association for the Study of Lung Cancer database, visceral pleural invasion could not be shown to be prognostically important. In other words, when leaping into undefined territory, we must first prove that we can land on our feet consistently.