Objective To assess nuclear morphometry as a predictor of prognosis in patients with renal cell carcinoma (RCC).
Patients and methods The study included 65 consecutive patients with RCC who underwent radical nephrectomy and were followed up for a median (range) of 80 (27–138) months. Nuclear morphometry was assessed using a computer-assisted image analysis system on histological sections and characterized by five nuclear variables (area, perimeter, major and minor diameter, and form factor). From the patients' records and pathology specimens, the clinicopathological prognostic variables (histological type, Fuhrman grade and pathological stage) were recorded. The proliferative activity was assessed using immunohistochemical staining with Ki-67 antibody.
Results Higher values of mean nuclear area, perimeter, and major and minor diameter were significantly related to higher nuclear grade, proliferative activity and advanced tumour stage. They were significant predictors of disease progression and survival, together with grade, stage, sarcomatoid histology and proliferative activity. Of all significant prognostic factors predicting progression-free survival, only stage was independent (T4 vs T1, hazard ratio 6.55, 95% CI 1.63–26.13, P =0.008).
Conclusion Although the significance of these preliminary results must not be overstated, nuclear morphometry might provide significant prognostic information in predicting survival and tumours at high risk of progression in RCC.
RCC is a highly unpredictable neoplasm with a tendency to recur or cause death many years after initial treatment [1–3]. Although surgery is the only definitive form of treatment for RCC, as new therapies for advanced disease are being developed, accurate methods are essential to identify patients who are at high risk for tumour progression.
Several morphological variables have been used to explain the unpredictable behaviour of RCC, but their prognostic value is controversial, mainly because of the subjectivity of morphological characterization and tumour heterogeneity . To date, stage and grade are the best predictors of disease progression and correlate well with long-term survival [1,4–6]. However, despite careful clinical staging, up to 30% of patients with RCC have disease progression after surgery and tumours, even at the same stage, may have different biological behaviour . In addition, the major criticism against nuclear grading is the subjectivity and difficulty in differentiating intermediate grades, which possibly contributes to the lack of uniformity in the use of nuclear grading [1,7]. Thus, improvements in predicting the risk of progression and in objective prognostic criteria are needed.
Distortion of nuclear shape and nuclear enlargement have been recognized as hallmarks of cancer and components of anaplasia. Many histological grading systems are also based on a subjective assessment of nuclear features. Therefore, nuclear morphometry might be useful for predicting prognosis for several human tumours. The quantitative assessment of nuclear morphometry is possible with computer imaging systems, providing a useful and reproducible method of predicting prognosis for many cancers [4,8,9]. However, whether it can yield additional and independent prognostic information in RCC remains uncertain. Thus, the purpose of the present study was to evaluate the prognostic significance of nuclear morphometry in predicting disease progression-free survival, and to correlate this variable with traditional clinicopathological prognostic factors in RCC.
Patients and methods
The study comprised 83 patients with RCC who underwent radical nephrectomy at our institution between 1989 and 1999; of these, 18 were excluded because the histological and archival material were inadequate. Files of the remaining 65 patients were reviewed; their median (range) age at the time of surgery was 60 (26–78) years and 45 were men (M : F 9 : 4). The follow-up was based on regular clinical and radiological examinations (abdominal ultrasonography, chest X-ray and abdominal CT) and blood chemistry (full blood count, ESR, and kidney and liver function tests).
The description of the gross specimen and of 5-µm thick histological sections (fixed in 10% neutral buffered formalin, paraffin-embedded after automated tissue processing, and stained with haematoxylin and eosin) from each patient were reviewed by one pathologist (K.Y.) with no previous knowledge of the patient's clinical course. The pathology data (histological type, nuclear grade and tumour stage) were recorded.
The histological classification was based mainly on the light microscopic appearance and included conventional, papillary, chromophobe and sarcomatoid types. Tumours were graded using the system proposed by Fuhrman et al., ranging from I to IV and exclusively based on nuclear features (nuclear diameter and form, and presence or absence of nucleoli). The 1997 TNM system was used for pathological staging .
Proliferative activity was assessed by immunohistochemical staining with Ki-67 antibody, scored semiquantitatively to divide the study group into three subgroups of Ki-67 labelling index (LI), as described previously . Each subgroup was determined by counting the number of immunopositive nuclei in 100 tumour cells in at least five representative high-power fields across the slide, i.e. group I <10%, group II 10–40% and group III >40%. Ki-67 immunostaining was undertaken in 60 patients, with the remainder excluded because immunostaining was unavailable for evaluating immunopositivity.
Nuclear morphometry was assessed by one author (E.Ö.) with no previous knowledge of the patients' clinical course, using a computer-assisted image-analyser system consisting of a microscope (Axiophot, Zeiss, Jena, Germany) equipped with a high-resolution video camera (JVC, Tokyo, Japan). The images were processed using an IBM-compatible personal computer, high-resolution video monitor and image analysis software (Samba 4000, Alcatel, Paris, France).
The highest-grade area of tumour on a representative slide from each case was outlined to ensure adequate and representative sampling of nuclei for morphometric evaluation. Identified areas on each slide were selected to avoid fixation artefacts, necrotic tumour, haemorrhage and inflammation, all of which might affect the morphometric analysis. For each patient's sample, 100 nuclei were digitized at an original magnification of × 400, as previously described, and standardized for reproducibility [8,9]. Twenty microscopic fields were selected for each sample. Briefly, the images were captured with the video camera, the nuclei viewed on the monitor (avoiding nuclei of stromal cells) and the five nuclei closest to the centre of each screen outlined by drawing, to minimize bias.
After each histological sample was completed, five morphometric descriptors were calculated for each case, i.e. the mean nuclear area (MNA), mean nuclear perimeter (MNP), mean nuclear major diameter (MNMjD), mean nuclear minor diameter (MNMnD) and mean nuclear form factor (MNFF). The latter represents a dimension-less size-invariant shape descriptor that yields a low value (1.00) for a perfect circle and increases if the nuclear contour deviates from a perfect circle. Nuclear area is the area enclosed inside the contour, the perimeter is the contour perimeter, and the major and minor diameters are the longest and shortest orthogonal projection, respectively.
The results were assessed using univariate analysis, the chi-square test, independent sample t-test, one-way anova, Spearman's correlation and the Kaplan-Meier method for survival analysis. The significance in the survival analysis was tested using the log-rank test. A Cox regression analysis was used to examine multivariate interactions and to determine the independent prognostic predictors. The analysis commenced by entering all significant factors from the univariate analysis into a backward stepwise selection. Removal testing was based on the probability of the Wald statistic (P=0.1). In addition, ROC analysis was used to evaluate the usefulness of morphological variables as a predictive marker of progression-free survival. Sensitivity and specificity were calculated for each level of morphological variable observed in the sample, by varying the level of the variable that signified a threshold. The prognostic significance of each factor was tested with the log-rank statistic. Kaplan-Meier plots were generated to describe progression-free risk groups according to this level. Sensitivity, specificity and likelihood ratios were also calculated for each variable. All calculated P values were two-sided and P<0.05 was considered to indicate statistical significance.
Table 1 lists the distribution of histological type, grade, stage and Ki-67 LI. None of the patients had either nodal or systemic metastases. The median (range) follow-up was 80 (27–138) months; overall, 31 patients (47%) died from recurrence (local relapse or metastasis) with a median survival of 64 months from the time of surgery. The progressive group included 32 patients (49%) who either had recurrence or died. Only one patient (2%) was still alive in the progressive group, with a follow-up of 85 months. The remaining 33 patients (51%) had no progression.
Table 1. Correlation of morphometric data with patient characteristics and histopathological findings
Mean (sd) variable
MNFF × 10 (%)
Measured in 60 patients; bold values show statistically significant differences.
Table 1 also summarizes the statistical analysis for nuclear morphometry and the clinicopathological prognostic factors. MNA, MNP, MNMjD and MNMnD were significantly related to histological type, tumour stage, nuclear grade and Ki-67 LI. Higher values of these variables were significantly related to sarcomatoid histology, advanced tumour stage, higher nuclear grade and proliferative activity. Omitting the sarcomatoid histology from the entire group did not change the significance of morphometry. In contrast, MNFF showed no significant relation to any variable except sarcomatoid histology.
Higher values of MNA, MNP, MNMjD and MNMnD were significant predictors of progression-free survival (Tables 1 and 2). Univariate analysis also showed that prognostic factors such as histology (predominantly sarcomatoid type), advanced tumour stage, higher nuclear grade and proliferative activity were also significant predictors of progression (Table 2). Using these morphometric variables and prognostic factors, the multivariate analysis (Table 3) indicated that only tumour stage was an independent predictor (T4 vs T1).
Table 2. Univariate analysis of morphometric variables and clinicopathological prognostic factors as predictors of progression
The optimal thresholds for the sensitivity and specificity values and likelihood ratios for the significant morphometric variables (MNA, MNP, MNMjD and MNMnD), from the ROC analysis, were 150 µm2, 55 µm, 17 µm and 13 µm, respectively. All of the variables were diagnostically useful, because the likelihood ratios were >1 (Table 4). Figure 1 shows the probability of progression-free survival according to the thresholds in the study group.
Table 4. ROC analysis of morphometric variables as predictors of progression-free survival
Accurate methods for identifying patients at high risk of tumour progression are clinically important. Tumour-related prognostic factors include several gross, microscopic and molecular findings. Of these, tumour stage and grade are well-established [1,4], but they provide limited information about the individual patient. Tumours even in the same stage may show different behavioural patterns . Although the histological subtype, including sarcomatoid histology and labelling indices of proliferation markers such as Ki-67 antigen, correlate with survival, their prognostic significance in separating the risk groups for recurrence is limited [6,13,14].
The subjectivity of nuclear grading, particularly in the intermediate grades, has fostered quantitative morphometric approaches to evaluating nuclear features . There is increasing evidence of the value of nuclear morphometry as a marker of tumour behaviour [16,17]. First introduced by Diamond et al. in 1982 and later standardized by Mohler et al.[8,9], the method is known to be accurate, reproducible and efficient, and has now been used to predict the prognosis in bladder and prostate cancer. Nuclear morphometry has been used to predict tumour behaviour in RCC [4,19–23] and with stereology .
To improve the predictive ability of the well-established prognostic variables (stage, grade, histological subtype and proliferative activity) we assessed the correlation between nuclear morphometry and these factors. Higher values of MNA, MNP, MNMnD and MNMjD were significantly related to the clinicopathological factors. This correlation might not be surprising, because these nuclear variables are key quantitative features of nuclear enlargement in malignant cells, and criteria in most grading systems. Similar results were reported by Eskelinen et al., who found that higher grade renal tumours had larger nuclei and were also proliferating rapidly. Similarly, in the present study, a higher MNFF was related only to sarcomatoid histology, which might be explained by the presence of characteristic bizarre nuclei in this histological subtype. However, omitting this histological type from the analysis did not change the significance of histology as a predictor.
In the univariate analysis, all the clinicopathological and nuclear variables were significant predictors of progression-free survival, but only stage was an independent predictor. Incorporating nuclear morphometry with tumour-related factors might enhance their prognostic ability. In addition, from the ROC analysis, the threshold of each significant nuclear variable provided excellent separation of progressive and non-progressive cases; in practice, tumours with a MNA of >150 µm2 are at high risk of progression.
In the present study there were many patients (65) and a long median follow-up (80 months); the results show that morphometric variables can be used to identify patients at greater risk of progression. This preliminary report confirms that a prognostic model incorporating stage and other significant factors, including nuclear morphometry, might provide significant prognostic information in RCC on the clinical course, and for stratifying patients at risk in the adjuvant setting. Further validation of this conclusion is required.
Z. Kirkali, Department of Urology, Dokuz Eylül University School of Medicine, Inciralti, 35340 Izmir, Turkey. e-mail: email@example.com