The use of biomarkers in the prediction of survival in patients with pulmonary carcinoma.

Data on ten variables and 16 biomarkers were obtained on 119 patients with newly diagnosed pulmonary cancer. The prognostic value of 16 biomarkers (alpha-1-antitrypsin [AAT], adrenocorticotropic hormone [ACTH], alpha-fetoprotein [AFP], carcinoembryonic antigen [CEA], human chorionic gonadotropin [HCG], immune complexes, immunoglobulins, N-terminal peptide of proopiomelanocortin [NTERM], and tumor-associated antibody [TAA]) was tested by adding these to the model of age, gender, stage, morphology, Feinstein's classification of symptoms, Karnofsky scale, leukocyte count, recent weight loss, and liver enzymes. Using Cox's regression method and a forward stepwise procedure, seven biomarkers (ACTH, AAT, AFP, calcitonin, HCG, TAA, and prolactin) entered the model. Elevated levels of cortisol and TAA were associated with longer survival. The selection of biomarkers by stepwise regression needs to be interpreted with caution, especially since the Z scores were found to be dependent on the particular variables included in the model. Furthermore, when dichotomized on maximum of the normal laboratory values, HCG and AFP were infrequently (2%) elevated. The lack of correlation among the biomarkers supports the hypothesis of random derepression of the genome of cancer cells. Further studies in improved modeling and the formulation of a biomarker index could enhance our understanding of the biology of cancer.