Objective To determine the relative prognostic importance of microvascular invasion in apparently localized renal cell carcinoma (RCC).
Patients and methods A retrospective clinical and pathological review was conducted of 176 consecutive patients identified from pathology records who had a nephrectomy for RCC with a median follow-up of 44 months. Vascular invasion was recorded and categorized by the level of microvascular invasion (MVI), renal vein invasion (RVI) and inferior vena cava invasion (IVCI). Tumour type, grade and size were also assessed. These variables were assessed by univariate and multivariate analysis to determine their effect on disease-free survival.
Results In the univariate analysis tumour size, grade, vascular invasion and young age each predicted reduced disease-free survival. On multivariate analysis for all 176 patients, grade, vascular invasion and young age were the significant independent predictors of reduced disease-free survival. In a subgroup of 149 patients from whom those with very high risk determinants were excluded (those with grade 4 tumours and/or IVCI) most of the risk of metastasis could be accounted for by vascular invasion and young age alone (MVI vs no vascular invasion, hazard ratio 3.18, 95% confidence interval 1.29–7.84; RVI vs no vascular invasion 2.41, 0.989–5.89; and age per year 0.963, 0.94–0.992).
Conclusions Grade, vascular invasion and young age are the main independent predictors of relapse in clinically localized RCC after nephrectomy. For most patients, who do not have very high risk indicators, the main adverse predictors are vascular invasion and young age. These findings are important when selecting patients for trials of adjuvant therapy and have implications for pathological staging.
RCC has an incidence of ≈ 5 per 100 000 and accounts for 1–2% of all cancer deaths. At presentation, about two-thirds of tumours are localized and these patients will usually undergo nephrectomy. Although apparently cured by surgery, 40% will subsequently develop metastases and die from these. Predicting which patients will relapse is notoriously difficult [1–3] and presently available prognostic indicators do not satisfactorily allow the selection of high-risk patients for trials of adjuvant therapy.
The characteristics of localized RCC currently accepted as prognostic indicators are the histological type, grade, size, perinephric invasion and the presence of renal vein and/or inferior vena cava invasion (IVCI). Venous invasion, size, and perinephric invasion are combined as a tumour stage in the UICC TNM  system.
A recent report by van Poppel et al. suggested that the presence of microscopic vascular invasion (MVI) is the most relevant prognostic factor in nonmetastatic RCC. However, this report does not compare the significance of MVI with that of renal vein invasion (RVI) and thus leaves an important question unanswered.
We analysed retrospectively, by pathological and clinical review, 176 consecutive patients with localized RCC treated in two hospitals between 1991 and 1995, to determine the effect of different stages of vascular invasion.
Patients and methods
Patients who had undergone surgery for a primary kidney tumour were identified by a computer search of the histopathology records of two hospitals between 1 January 1991 and 31 December 1996. Of the 235 cases found the following were excluded: 19 TCCs, 15 Wilms' tumours, nine oncocytomas, two cystic nephromas and one clear cell sarcoma. The remaining 189 patients had RCC, and of these 11 had lymph node metastasis at the time of surgery and two had concurrent pulmonary metastasis, leaving 176 patients in the study group. All had undergone a radical nephrectomy but with no formal sampling of lymph nodes.
Histology reports and slides were available in all these cases, and all included at least two blocks of the edge of the tumour and a block of the renal vessels. For each tumour the slides were reviewed by a pathologist (D.F.R.G. or A.V.) with no knowledge of the clinical outcome, and assessed for histological type by the Heidelberg classification , Fuhrman nuclear grade  and the extent of any vascular invasion seen by microscopy. Vascular invasion was classified into three categories, i.e. (i) IVCI, (ii) invasion into the major hilar vessels, designated RVI and (iii) vascular invasion seen microscopically but not in the IVC or major hilar vessels, designated MVI. Cases in which there was doubt about vascular invasion were examined by both pathologists for a consensus opinion. The size of the tumour was taken from the macroscopic description in the pathology report.
Most patients had been reviewed annually as outpatients for 3–7 years; the following information was recorded from the patients' notes: date of birth, sex, date of surgery, date last seen, date of death, cause of death and the date on which recurrent or metastatic disease was first identified. In those cases where the hospital notes were not complete, information was sought from the patient's GP and from the Wales Cancer Information and Surveillance Unit (cancer registry). Two patients were lost to follow-up at 25 and 29 months. In all other survivors the last recorded clinical contact was after 1 January 1998, giving a median (range) follow-up of 44 (25–99) months. In 14 patients in whom the cause of death was recorded as RCC the date of first recurrence was unavailable; in these patients the date of death was considered the endpoint for disease-free survival.
For disease-free survival, deaths from causes other than RCC were considered to be censored at the date of death. Univariate analysis of disease-free survival was assessed by the Kaplan-Meier method using the Breslow test. Multivariate analysis by Cox regression was carried out using S-Plus 2000 (MathSoft Inc, Seattle). In the case of grade and vascular invasion a linear contrast was used such that they were compared with grade 1 and the absence of vascular invasion, respectively. Initially, all variables were entered into the model but tumour type was omitted from the final model as the P values for the coefficient for the factors for each tumour type were > 0.5. The significance of departures from proportional hazards was assessed using the scaled Schoenfeld residuals (using the S-Plus function cox.zph), and in the case of all covariates, there was no significant departure from proportionality. Where appropriate, all tests of significance were two-tailed.
The median (range) age of the 176 patients was 65 (34–88) years; 120 were men and 56 women; 119 of the tumours were conventional (clear cell) carcinomas, 22 papillary, five chromophobic, and three collecting duct, with 27 unclassified by conventional histology (four were entirely sarcomatoid, the others consisted of apparent mixtures of other morphological subtypes or their pattern of differentiation made confident morphological identification according to the Heidleberg classification difficult). MVI was identified in 24 patients who had no other evidence of vascular invasion; this was usually seen in the fatty connective tissue around the pelvicalyceal system (the renal sinus) although some was seen in the parenchyma adjacent to the tumour. The analysis of grade and vascular invasion is shown in Table 1.
Table 1. Cross-tabulation of grade and the level of vascular invasion
Of the 24 cases with MVI the vascular invasion was seen in one vascular cross-section only in 13, in two in nine and in three or more in the remaining two cases. In the 31 cases with RVI, MVI in other sites was seen in 21; tumour in the renal vein was the only evidence of vascular invasion in the other 10 cases.
Of the 176 cases 69 had died during the follow-up, 53 from disease-specific causes and 16 from other causes; five patients were alive with metastatic disease. For those who developed metastasis the median (range) time from surgery to the identification of metastasis was 11 (0.5–72) months. The median (range) time between identification of metastasis and death was 4 (0.1–43) months.
Kaplan-Meier survival plots of disease-free survival are shown in Fig. 1. In a univariate analysis, tumour size, nuclear grade and vascular invasion were all strongly associated with both disease-free and absolute survival (P < 0.005 in all cases). Young age was associated with reduced disease-free survival (Fig. 1d) so that the 5-year disease-free survival in patients aged 60 years was significantly less than in those aged > 60 years (56% and 71%, respectively, P=0.04). Analysis of tumour type showed that patients with papillary carcinomas had a longer and those with unclassified RCCs a shorter disease-free survival than those with conventional tumours (Fig. 1e).
When the data from all 176 patients were entered into a Cox regression model the independent predictors of disease-free survival were grade, vascular invasion level and age (Table 2). In contrast to the results of the univariate analysis, tumour size and tumour type were not independent predictors, probably because of their correlation with vascular invasion. The mean (sd) size of tumours having no vascular invasion, at 6.21 (2.67) cm, was significantly less than those in which there was invasion, at 8.57 (3.02) cm. For tumour type, a lower proportion of papillary carcinomas showed vascular invasion (two of 22) than those of conventional type (48 of 119), and unclassified tumours were of higher mean grade and showed an excess of vascular invasion (17 of 27). Analysis of the categories within each variable showed that the significance of grade was largely caused by the presence of grade 4 tumours. In the case of vascular invasion, the presence of MVI, RVI and IVCI all predicted a significantly greater hazard than the absence of vascular invasion, although the significance of the presence of MVI was marginal (P=0.057). When treated as a continuous variable, and in agreement with the univariate analysis, age was a significant independent predictor (Table 2) and the hazard ratio was < 1, so that increasing age was associated with a reduced hazard. Specifically, an increase in age by 25 years yielded an estimated hazard ratio of 0.52 (95% CI 0.28–0.95).
Table 2. Multivariate disease-free survival analysis for predictors of relapse
Hazard ratio (95% CI)
Hazard ratio is for each increase in age by one year;
hazard ratio is for each increase in size by 1 cm.
Clearly, grade 4 tumours and those with IVCI have a particularly poor prognosis. Further multivariate analysis was therefore carried out on a subset in which these high-risk cases were excluded. In this subset (149 patients) the only significant independent predictors were vascular invasion level and age (Table 2). Compared with the absence of vascular invasion, both MVI and RVI were significantly associated with an increased hazard. Analysis of the subgroup of 149 patients was repeated with MVI and RVI cases combined as ‘vascular invasion’. This gave a hazard ratio of 2.73 (95% CI 1.23–6.07; P=0.013) for vascular invasion compared with no vascular invasion, without materially affecting the hazard ratios or CIs for the other covariates.
The aim of this study was to assess prognostic indicators in patients with non-metastatic RCC; this study differs from others in using a recent histological classification and by classifying vascular invasion by the vessels it has reached. Thus we were able to assess these factors independently. There was also an almost complete follow-up in this consecutive series. The main findings are that for all tumours the independent variables affecting disease-free survival are grade, vascular invasion and age. Excluding a clear group with a poor prognosis (grade 4 tumours or those with IVCI) the significant independent variables are vascular invasion and age.
Stage was not used as a variable, as in the TNM system the T (tumour) stage is assessed as a combination of size, perirenal invasion and vascular invasion. Furthermore, in the TNM system the criteria for T staging has changed significantly in each of the revisions. We would have preferred to include perirenal invasion but in many cases it was difficult to decide histologically if such invasion was present. RCCs commonly have a prominent pseudocapsule; some pathologists require tumour cell infiltration of surrounding tissues to indicate extension outside the kidney, as recently recommended by Eble . Others have accepted significant intrusion of the tumour pseudocapsule into the perinephric fat; this issue is unresolved. Van Poppel et al. found no significant difference in prognosis between pT2 and pT3a cases (i.e. those with and with no perirenal invasion).
The most surprising finding was that age is an independent variable; younger patients have a higher risk of recurrence irrespective of grade or level of vascular invasion. Age was not significantly associated with grade, vascular invasion or size. We are unaware of any other reports of young age being a significant adverse prognostic indicator. However, previous series have had variable follow-up and imprecise criteria for entry; this may have selected younger survivors, a selection that could have masked the effect of this variable. The present study is a consecutive series in which the follow-up is almost complete and so is unlikely to be subject to selection bias. The hazard ratio given by the regression analysis suggests a significant clinical effect so this clearly requires further study. The reason for the association is unclear; one possibility is that more tumours in older patients are incidental (as older patients are more often investigated by scanning methods). Two recent reports [9,10] of the characteristics of incidentally discovered tumours confirm an older mean age at diagnosis, but in those patients incidental tumours were associated with lower stage and grade. If the explanation of the age effect is that there are fewer incidental tumours in older patients, then the present data would suggest that the better prognosis of these tumours is independent of grade and vascular invasion. Alternatively, tumours in younger patients may be intrinsically more aggressive, perhaps reflecting a different genetic contribution to their aetiology.
The present findings on grade and size are consistent with many other series and need no further comment. Although our definition of vascular invasion differs in detail from the definitions of others, the finding that it is a strong and independent prognostic indicator is broadly similar [5,11,12]. We have also shown that RVI and MVI have a similar prognostic effect.
The identification of MVI depends on histological sampling; in each of the present patients with MVI very few invaded veins were found. The sampling in this retrospective series, although consistent with UK practice, was limited compared with that recommended by Bonsib et al., who found that complete sampling of the interface between the tumour and the renal sinus identified vascular invasion in six of 16 RCCs with no RVI, more than twice the incidence found in the present study. A more detailed histological examination of the resection specimen would be expected to identify more cases with MVI; histopathologists will need to respond to this challenge.
The present findings have important implications for pathological staging systems; such systems are only of use if they indicate prognosis or assist with management. In the cases of localized RCC, where further local treatment is rarely needed, the main value of a local staging system is to assess metastatic risk, and for this it should be constructed from validated and preferably independent variables. As MVI has a similar predictive value to RVI then there is no good reason for pathological staging to separate cases with these features. The present results and those of van Poppel et al. also suggest that if grade and vascular invasion are considered then tumour size adds no further significant prognostic information. If it can be confirmed that age is an independent prognostic variable then it will be necessary to determine how age can be added to the equation when assessing metastatic risk. The issue of whether invasion outside the kidney (i.e. distinguishing pT2 from pT3a) remains significant once grade, vascular invasion and age are considered still needs to be resolved.
We thank Dr R.G. Newcombe, Department of Medical Statistics, UWCM for statistical advice; Mrs Sarah Bennett, Department of Urology, University Hospital of Wales, for assistance in drawing and abstracting notes; and Mrs Joy McCrae, Wales Cancer Intelligence and Surveillance Unit, for assisting in obtaining causes of death.
D.F.R Griffiths, Department of Pathology, University of Wales College of Medicine, Heath Park, Cardiff, CF14 4XN, UK. e-mail: firstname.lastname@example.org