In this study, the differences found in the nuclear measurements of the perimeter, circularity, solidity, and AR corroborate the occurrence of irregularities in the normal morphology of hepatocytes as a result of the malignant transformation process and demonstrate the excellent potential in combining this tool with the traditional histopathological analytic method.
We showed that morphometry can be used as a tool to discriminate tumoral and adjacent normal tissues. Our data complements the results of Hassan et al. [19] who performed imaging analysis of tumoral HCC nuclei and hepatocellular nuclei from surgical tumor-free safe margins in a cohort of patients with chronic hepatitis C; in that study, a significantly lower nuclear area was observed in tumor cells and in the surgical tumor-free margin hepatocytes than in patients without HCC. These data are indeed more significant when we consider that most patients included in our study also have a pre-transplantation history of hepatitis C.
The existence of nuclear morphometric changes has already been verified in studies with different types of tumors [10, 12, 20] including studies such as the one by Mendaçolli et al. [14] in which significant changes in morphometric and chromatin texture patterns were observed between basal cell carcinoma samples and the unaffected basal epithelium. Additionally in that study, the sclerodermiform type neoplasms presented larger nuclear area and diameter in relation to nodular and superficial types, suggesting that genomic or metabolic differences would also be determinant for independent biological behavior among basal cell carcinoma subtypes.
Regarding the chromatin texture, the FD in this study was lower in the HCC samples compared to the adjacent tissues. These findings differ from those observed in the study by Gheonea et al. [9] which obtained an increased FD value in HCC when compared to that observed in adjacent hepatic tissue. A possible explanation for the disagreement of FD values in tumor tissue between the studies may be related to inter-rater agreement in the measurement of this parameter: the non-significance of inter-rater agreement demonstrated in our study may have influenced the outcome, both in our study and in the study by Gheonea et al. [9]. In order to make stronger conclusions from the analysis of FD, it is necessary to improve the method for its measurement, increasing its reproducibility.
Quantitative analysis is a useful tool for developing new diagnostic methods [7]. Image softwares can make possible data checking between different researchers in all the samples measured, which allows for the exclusion of manual selections [21]. The correlation of inter-rater metrics was statistically significant in most parameters assessed, indicating that these findings might be reproducible in future studies and increasing the utility of our morphometric method for clinical practice.
Our study is the first to observe the association between changes in nuclear morphology and clinically relevant variables related to determination of the prognostic of post-LT patients. Our data showed an association of survival in a time period of less than five years with larger area and Feret diameter nuclei, and lower nuclear measurements of some parameters (Feret diameter and ncMGV) associated with higher MELD scores. This means that these parameters may be used to help predict outcomes of liver transplantation, providing a greater scientific basis for medical decisions making that directly affect medical practice and that broaden the scope of personalized medicine in HCC.
Lower AR values were found in individuals who had post-LT HCC recurrence. The risk assessment of post-LT HCC recurrence using pathological characteristics of the explant is an important finding as it can lead to refining of the prognostic assessment and in the future may help to delineate therapy and screening protocols [22].
A unique result of this study is the definition of cut-off values to differentiate malignant and healthy hepatocytes using the AR, which helps to establish more objective diagnostic criteria for cell differentiation. Values defined by the ROC curve related to AR are results not used in other studies. Sensitivity and specificity values for nuclear perimeter in our study are similar to the values found by Ambroise et al. [23] who showed that a cut-off level of 33.2 µm for nuclear perimeter could differentiate malignant and benign pleural effusions. However, despite the computer analysis by ImageJ following a similar methodology, they used analyses applied to effusion cytology, in addition to evaluating for each case only ten representative nuclei from ten different fields.
Our study does have some limitations. The first is the reduced number of samples, since many samples had to be excluded from the analysis due to the loss of tissue integrity caused by the prolonged storage time of the paraffin blocks, possibly hampering the power of this study. Secondly, it is known that cirrhosis, a subjacent abnormality in all cases, can affect the measurements obtained in the tissue adjacent to the tumor used in comparison with the HCC—although this does not limit the differentiation of cells from the same sample, it limits the applicability of the exact values to healthy liver parenchyma, and so further studies evaluating non-cirrhotic patients are necessary. Finally, we did not investigate the molecular events causing the observed morphometric differences in this study. Therefore, we propose that future studies incorporate the use of methods to evaluate these events such as chromatin immunoprecipitation associated with DNA sequencing for an in-depth elucidation of the mechanisms that trigger the morphometric changes observed here.