Criteria for the selection of referential groups in tumor marker statistical evaluation on the basis of a retrospective study.

The authors of this study are concerned with the analysis of optimal criteria for the selection of referential groups in the statistical evaluation of tumor markers for early detection of recurrent disease. Although criteria for the selection of optimal referential groups have already been published on a number of occasions (EGTM recommendation), these criteria are not followed in daily routine, which leads to a false interpretation of results and the impossibility of comparing individual studies. The commonest problem is an incorrect determination of cut-off, caused by not following the recommended specificities at 95%, which results in an incorrect assessment of tumor marker sensitivities. Other faulty interpretations happen in consequence of inaccurate and not clearly defined referential groups, which differ from each other by, for example, stage of the disease, length of the follow-up and so on. Comparing tumor marker results still remains a problem, since they are assessed with diagnostic kits from different manufacturers which may misrepresent the final value of the results, and thus imitate remission or progression of the tumor disease. Similarly, mutual comparison of results from prospective and retrospective studies without standardization of clinical conditions leads to an unreliable interpretation. The authors show, through concrete examples, the possibility of a completely different interpretation of the results in identical referential groups in consequence of their inaccurate characteristics.