Histopathologic prognosis of thymomas: another example of medical surrogacy.

Particularly in anatomic pathology, the past 2 decades have witnessed a proliferation of tests that are surrogates for selected analytes of predictive or prognostic interest. Those targets include such moieties as functional hormone receptor proteins and the composition of genes that govern responses to selected growth factors. Common respective examples of those groups are estrogen and progesterone receptor proteins 1 and the HER-2/neu, c-kit, and epidermal growth factor receptor (EGFR) genes. 2-8 Thousands of tests are done each year, predominantly through immunohistochemical analysis, to assess the status of such targets in malignant human neoplasms. And yet, immunohistochemical analysis represents an imperfect substitutive method in reference to the goal just mentioned. Immunohistology does not necessarily detect functional hormonal receptors (HRs), 9 and it is suboptimally effective in separating HR-" positive " from HR-" negative " tumors, when compared with the performance of traditional biochemical HR assays. 10 General assumptions also have been made (and accepted) that hold that the increased cellular expression of the proteins related to the neu, c-kit, and EGFR genes—as visualized by immunohistochemical analysis—is an invariable reflection of gene amplification or an activating mutation. 11 Recent publications have demonstrated the fallacy of those suppositions. Increasingly, the published literature shows that fundamental, " first-hand " laboratory tests are incontrovertibly superior to immunohistochemical analysis as medical devices. 8,12 Pertinent examples are the dextran-coated charcoal assay (DCCA) for HRs 13 and in situ hybridization or polymerase chain reaction–based blotting techniques for the assessment of gene structure. This situation speaks to a common tendency that many physicians have, that is, a propensity to exchange scientific rigor—because it demands scrupulous attention to detail— for expediency. If one assessment of prognosis is quicker, cheaper, and less technologically intensive than another, the first will be chosen by practitioners nearly every time. Moreover, that paradigm applies even if the selected test is statistically inferior to others, providing that seemingly convincing rationalizations can be made for its use. Returning to an assertion made in the previous paragraph, it has become clear that the old-fashioned DCCA, if properly performed, gives a more accurate depiction of the likelihood that breast carcinomas will respond to hormonal modulation than does immunohistochemical analysis. 10,14,15 However, it is cumbersome to obtain aliquots of frozen tissue in each case of breast cancer and equally problematic to transmit them under controlled conditions to the laboratory. In addition, frozen-section examination of the tissue aliquots is ideally …