The definition of malignancy classically requires that there be uncontrolled growth of tissue. There are histological correlates of this behavior, and this correlation is absolute in late-stage tumors. If it looks like a malignancy on histological examination, the lesion will exhibit uncontrolled growth. However, at the earliest stages of malignancy, the correlation between histological appearances and malignant behavior may not be so clear. In very early hepatocellular carcinoma (HCC), for example, although there are features suggesting malignancy, such as stromal invasion, the malignant nature of these lesions (i.e., uncontrolled growth) has not been demonstrated. Furthermore, at the boundary between premalignant and malignant states, it is difficult to identify morphological correlates of malignant behavior, because dysplasia and early neoplasia share many common histological features. Although dysplasia is a precursor to malignancy, not all dysplastic lesions ultimately develop malignancy. Some, admittedly a minority, may even disappear. Others remain unchanged in size for years before ultimately developing malignancy. By the strict definition of malignancy, whatever radiological or histological characteristics a lesion may have, if it does not grow, it is not malignant, and therefore may not need treatment.
The adoption of ultrasound screening for early diagnosis of HCC has allowed us to identify smaller and smaller lesions in the liver. Obviously, the smaller a lesion, the closer it lies to the dysplasia/neoplasia boundary, and consequently, the more difficult it becomes to be certain on histological analysis whether the lesion is cancer. Ideally, if it is cancer, we do not want to wait for it to declare itself by growing, because the more advanced the lesion, the lower the likelihood that therapy will be curative.
A number of tools have been developed to help determine whether a liver nodule is malignant. A new classification of premalignant and early malignant changes have been described by an international group of pathologists.1 New staining techniques are intended to distinguish early malignancy from dysplasia. Several authors have described glypican-3, glutamine synthetase, and heat shock protein-70 as early histological markers of HCC.2-4 Di Tommaso et al.5 described that positivity for at least two of three of these markers is highly specific for HCC, when applied to biopsies of small nodular lesions in a cirrhotic liver. However, the combination is not highly sensitive. In this issue of Hepatology, the same group6 describe yet another marker that might be useful to distinguish early malignancy from premalignancy, namely staining for clathryn heavy chain (CHC). In their study, they looked at liver cancers that were confirmed by histology and stained the tissue with the aforementioned three markers and with CHC. They found that positivity of any two markers correlated 100% with HCC, i.e., specificity and the positive predictive value were both 100%. Addition of CHC improved the sensitivity in HCCs smaller than 2 cm to approximately 64% from 47%, a clinically useful improvement.
This study highlights some of the difficulties in identifying biomarkers, whether histological or serological, that can improve the diagnostic accuracy of small hepatocellular nodules, and also illustrates why some caution is required in applying these findings to clinical practice. The current study6 included controls, namely samples from nontumor tissue, cirrhosis, and focal nodular hyperplasia, and of course, these controls are necessary. However, all these studies probably overestimate the true performance characteristics of the biomarker in question. It is one thing to do a study in which the characteristics of the target tissue or of the patients is known, but another to evaluate the performance characteristics of these biomarkers in tissue of patients in whom the diagnosis is suspected but not confirmed. This study was performed in patients with known HCC. In other words, the diagnosis of HCC had already been made on other grounds, presumably morphological. In essence, if the diagnosis can be made on morphological grounds alone, what need is there for additional markers? Therefore, these results are not useful when patients have HCC that can be confirmed by radiology or morphology alone. However, the clinical problem lies in distinguishing HCC from dysplasia in cases where morphology or radiology are not sufficiently specific, e.g., in the earliest HCC lesions. Whether staining for CHC or any of the other markers is useful under these circumstances remains unknown. Unfortunately, the studies required to prove the utility of these markers under these circumstances will likely never be done. This would require finding an uncertain diagnosis on biopsy of a small lesion, not intervening, and allowing it to declare itself as malignant over time, and then analyzing the biomarker positivity in the biopsy to see whether a positive stain or positive panel of stains predicted the development of malignancy. There might be ethical objections to such a study, and in any case, recruitment would be difficult when a minor procedure like radiofrequency ablation would remove all risk of traces of the lesion in the vast majority of cases.
The authors tried to overcome the bias that studying established tumors brings by separately analyzing small HCCs from larger HCCs. Small HCC was defined as tumors less than 2 cm in size. This approximates the size of a lesion that might be detected on screening, and that might pose diagnostic difficulty. They showed that specificity was just as good as with larger lesions, although sensitivity (68%) was reduced compared to larger lesions (92%). However, this still does not remove the issue that these were HCCs that were all diagnosed initially on morphological grounds.
The authors also discuss the use of biomarkers in situations where pathologists are perhaps not as expert in distinguishing early HCC from nonmalignant tissue. Once again, the use of these markers has not been tested in this situation. One can envisage the possibility that the use of makers provides a false confidence. An inexperienced pathologist might misdiagnose marker-negative tissue as benign. Thus, the consequences of a false negative might be significant.
The same considerations apply to evaluation of serological biomarkers for use in HCC screening. There are many studies that describe the performance characteristics of alpha-fetoprotein, des-gamma carboxyprothrombin, and other markers in patients who have established HCC.7-10 In one recent example, Marerro et al.7 studied patients with HCC, and despite finding a sensitivity that was only 66% for alpha-fetoprotein, concluded that this was a good screening test. They defined early HCC as being within Milan criteria. However, most patients will not receive a liver transplant for their HCC, because there are not enough livers to satisfy the demand. The point of screening is to reduce mortality from the disease. Diagnosing tumors that are within Milan criteria means diagnosing tumors that are already beyond the early stage. Treating these cancers might not be associated with a reduction in mortality. Absent the availability of a liver transplant, a 5 cm lesion or three lesions less than 3 cm will have a lower cure rate than lesions 1-2 cm in size. Radiofrequency ablation or surgery will cure 1- to 2-cm lesions in more than 90% of cases. Therefore, to determine whether alpha-fetoprotein, des-gamma carboxyprothrombin, or other biomarkers are really useful requires at a minimum that they be evaluated in patients with lesions smaller than 2 cm. Furthermore, ideally they should be evaluated in screened population, not a population with known HCC.
Thus, although the findings reported by Di Tommaso et al.6 are promising, they are not yet ready for widespread application. First, the study needs to be confirmed in a larger sample. Second, there should be some real-life experience using the markers in expert centers. Nonetheless, the recent revision of the American Association for the Study of Liver Diseases guidelines cautiously endorses the use of histological markers to evaluate screening-detected lesions suspected to be HCC.11
Because the ideal study will likely never be performed, we have to decide whether to accept the current findings, if confirmed, and apply them more generally, or whether to adopt a completely different approach. Possible approaches might be to repeat the biopsy, to ablate the lesion whatever the results of histology, or observe the lesion at short intervals for growth. These approaches would require formal testing.
In summary, histological markers of HCC may be useful to diagnose the very earliest detectable stage of HCC. However, the results have to be interpreted with caution.