Decision guarantee in tumour marker analysis: a cut-off independent assessment.

BACKGROUND

The analysis of tumour markers is based on the evaluation of data in relation to defined cut-off values. Changes in the method of determination or reference study group have led to different results. Generally, fuzzy logic-based classifiers for the evaluation of tumour-marker profiles have increased sensitivity and higher specificity. Drawing from experiences with cut-off-oriented evaluation, the suggestion that diagnosis be made using a non-cut-off-based approach is discussed.

PATIENTS AND METHODS

Two hundred and eighty-one consecutive patients with newly diagnosed, histologically confirmed lung cancer and a control group of 231 patients were examined. Histological classification of the tumour cases yielded 59 small cell carcinomas, 102 squamous cell carcinomas, 66 adenocarcinomas and 54 large cell carcinomas or mixed bronchial carcinomas without classification. The control group without tumours consisted of 23 healthy subjects, 125 patients with silicosis or asbestosis, 27 patients with chronic obstructive pulmonary diseases (COPD) and 56 persons suffered from inflammatory lung diseases.

RESULTS

CYFRA 21-1 was the most sensitive marker with a sensitivity of 57.3% and a specificity of 94.9%. The sensitivity increased when various tumour markers were evaluated. The loss of specificity through this type of multiparametric analysis could be compensated for by using a multiparametric evaluation such as a fuzzy logic classifier. A determination of the sensitivity-adapted confidence in diagnosis should be made. The decision guarantee and the cut-off-based evaluation are qualitatively equal because the decision threshold of 0.5 corresponds by definition to the cut-off value at 95% specificity. Every cut-off-based system of classification could be restated by using the decision guarantee instead of the measured values.

CONCLUSION

The evaluation based on decision guarantee has proven to be unaffected by changes in laboratory systems. The multiparametric system of classification based on decision guarantee of individual markers is self-adjusting and differences in measurement levels are eliminated if the ROC curves of the individual markers are in agreement.