Among binary prognostic factors, smoking (ever vs never), higher baseline cancer stage (IIIB+pleural effusion vs IIIA/IIIB), and type of chemotherapy regimen (single- vs double-agent regimen) were all independently associated with significantly higher mortality, while higher performance status (ECOG 2 vs 0–1) showed a trend toward worse prognosis ( P =0.086).
In addition, we have forced into the final Cox's PH model the performance status (ECOG 0–1 vs 2), even if it has a marginally nonsignificant effect in our analyses ( P =0.086), because it is an essential criterion for the treatment of unresectable NSCLC ( Pfister et al , 2004 ).
The third column of Table 2 shows that in the univariate Cox's PH models, all variables, with the exception of age and tumour pathology, had statistically significant ( P <0.05) or marginally significant (0.05< P <0.10) crude unadjusted associations with the hazard.
Both P -values are below 0.01, indicating that, after having accounted for their non-linear and time-dependent effects, both variables have highly significant associations with mortality.
In contrast, the adjusted effect of performance status did not reach statistical significance, possibly because of limited statistical power and/or misclassification, as physicians tend to underestimate the performance status ( Ando et al , 2001 ).