P. Foucher et al., COMPARATIVE-EVALUATION OF RELATIVE SURVIV AL VERSUS CLASSICAL METHODSAPPLIED TO PATIENTS WITH NON-SMALL-CELL LUNG-CANCER, Bulletin du cancer, 81(10), 1994, pp. 857-865
The aim of the study is to compare the usefulness of a recent relative
survival model versus more classical methods for univariate and multi
variate survival analysis, applied to a population of patients with su
rgically cured non small cell lung cancer, in determination of prognos
tic factors and appreciation of the exact role of age on survival. We
studied 156 patients surgically treated between 1975 and 1988. Both un
ivariate and multivariate analysis were performed, using the actuarial
method and the Cox model for crude survival and the additive Hakuline
n model (1985) for relative survival (total risk of death equal to nat
ural risk of death in general population plus disease specific risk of
death) which is an age-adjusted survival corrected for normal life ex
pectancy. In addition, the loss in life expectancy was also calculated
. Our 156 patients (including six females), whose age ranged from 30 t
o 78 (mean age 59) were almost all current or former smokers (97%) and
63% had clinical trouble. Squamous cell carcinoma was the most common
histology (76%) before adenocarcinomas (20%). Pneumonectomy and lobec
tomy were equally performed. Post surgical TNM staging was stage I = 7
8 (50.3%), II = 23 (14.8%), IIIa = 44 (28.4%), IIIb = 10 (6.5%). By 31
December 1990, 116 patients had died, 24 were alive and 16 lost to fo
llow-up. In univariate analysis, overall survival is (crude/relative):
1 year (75.8%/77.5%), 2 years (53.8%/56.0%), 5 years (28.7%/32.5%), 1
0 years (14.4%/18.9%). Univariate prognostic factors are histopatholog
y, surgical procedure and post operative TNM staging. The overall loss
in life expectancy is 71.4% (5.5 years of life expectancy vs 19.21).
The loss is higher for the younger patients than for the older ones (7
3% for the 30-49 year old group - 59.2% for the more than 70 year old
group). In multivariate analysis, prognostic factors are: Cox model: p
ost-surgical TNM staging, histopathology and age (RR = 2.18 (1.13-4.23
) for patients over 65); Hakulinen model: TNM staging. In this model,
age is no longer a significant prognostic factor. In conclusion, this
study confirmed the poor prognosis of NSCLC, even if a curative surgic
al procedure has been possible, with a 5-year survival of 48% for stag
e I tumours but only 6% for stage III tumours. The most significant pr
ognostic factor is the post-surgical TNM staging. The relative surviva
l model of Hakulinen dismissed age as a significant prognostic factor.
Our study underlined the usefulness of relative survival methods whic
h should be more frequently sued to allow comparisons between series o
f different origin because of its ability to eliminate natural life ex
pectancy as a confounding factor.