Clinical equivalence of two cytokeratin markers in mon-small cell lung cancer: a study of tissue polypeptide antigen and cytokeratin 19 fragments.

Study Objectives: We have longstanding experience with tissue polypeptide antigen (TPA), a tumor marker of the cytokeratin (CK) family. In the mid-1990s, a new CK marker, CK 19 fragments (CYFRA 21-1), became popular and widely accepted. This is the first study specifically designed to compare the two markers.
Design: Analysis of a single institution database over a 3-year period (ie, 1998 to 2000).
Setting: Community-based hospital and second referral level institution for a province of 500,000 people.
Patients: The study included 180 new consecutive patients (143 men) with pathologically documented non-small cell lung cancer (NSCLC), who were observed during and after treatment, and eventually were assessed for status.
Interventions: Anthropometric, clinical, and laboratory data, including TPA and CYFRA 21-1 serum levels, were recorded prospectively. Standard nonparametric tests, Kaplan-Meyer survival analyses, Cox proportional hazards models, receiver-operating characteristic (ROC) curves, and estimates were used for statistical analysis.
Measurements and Results: A total of 1,299 twin TPA and CYFRA 21-1 serum assays (180 performed at diagnosis and 1,119 performed during or after treatment) were obtained. Intermarker correlation tests revealed incredibly high Spearman rho indexes, ranging from 0.935 at diagnosis to 0.813 to 0.921 at the different follow-up times. The substantial equivalence of the two tests explained all the other results, as follows: their similar profile of correlation with the other variables (objective treatment response: TPA rho, 0.456; CYFRA 21-1 rho, 0.463; follow-up performance status: rho range, 0.424 to 0.435); their superimposable capability to predict important clinical situations (eg, recognizing a metastatic disease at diagnosis with areas under the ROC curve of 0.742 and 0.706, respectively); their nearly identical prognostic significance (the D statistic of the goodness-of-fit of a multivariate survival model: TPA, 851.0; CYFRA 21-1, 851.6).
Conclusions: In most of their traditional clinical applications the two serum tests are equivalent because of their virtual identity. We strongly recommend using a CK test in the evaluation of each NSCLC patient. The choice between TPA and CYFRA 21-1 can be based on nonclinical factors, such as the laboratory experience or preference, and the cost of the two kits.