Lymphovascular invasion is an independent predictor of oncological outcomes in patients with lymph node-negative urothelial bladder cancer treated by radical cystectomy: a multicentre validation trial


  • C.B. and E.H contributed equally to this work

Christian Bolenz, Department of Urology, Mannheim Medical Center, University of Heidelberg, Theodor-Kutzer-Ufer 1-3, 68167 Mannheim, Germany.


Study Type – Prognosis (inception cohort)
Level of Evidence 1b


To validate the association of lymphovascular invasion (LVI) with disease recurrence and cancer-specific survival (CSS) in a multicentre cohort of patients treated with radical cystectomy (RC) for urothelial bladder cancer (UBC).


We collected pathological and clinical data on 1099 lymph node-negative patients treated with RC at six German institutions. LVI was defined as the presence of tumour cells within an unequivocal endothelium-lined space in haematoxylin and eosin-stained sections.


LVI was present in 295 (26.8%) patients; the presence of LVI correlated significantly with increasing tumour stage, i.e. pT1, 65 (29.4%); pT2, 88 (31.5%); pT3 110 (31.8%); and pT4 32 (38.1%) (P= 0.002) and grade (P < 0.001). In univariable analysis the presence of LVI was significantly associated with reduced recurrence-free survival (P= 0.008) and reduced CSS (P= 0.039). On multivariable Cox regression analysis tumour stage (P < 0.001), age (>75 vs ≥75 years; P= 0.018) and LVI (P < 0.001) were identified as independent predictors of CSS.


Our large multicentre study confirms the independent prognostic value of LVI in patients with node-negative UBC. LVI can be regarded as a surrogate variable for lymphatic micrometastasis in node-negative UBC. Assessment of LVI might improve the selection of patients who are likely to benefit from adjuvant therapy after RC. The identification of factors involved in the process of LVI could reveal new therapeutic targets for UBC.


urothelial bladder cancer


lymphovascular invasion


radical cystectomy




transurethral resection


carcinoma in situ


haematoxylin and eosin


cancer-specific survival


overall survival


recurrence-free survival


Urothelial bladder cancer (UBC) is among the most common cancers, and an estimated 70 980 new cases of UBC will be diagnosed in 2009 in the USA, causing ≈14 330 deaths [1]. Radical cystectomy (RC) with bilateral pelvic lymphadenectomy (LA) is the standard of care for muscle-invasive disease or recurrent high-grade non-muscle-invasive UBC despite transurethral resection (TUR) and intravesical therapy. Up to 40% of patients with organ-confined UBC at the time of RC develop disease recurrence, with most of these patients dying from the disease [2,3]. Accurate estimates of the clinical stage and prognosis are essential for adequate clinical decision-making and determination of follow-up schedules after radical surgery.

Numerous studies have been reported to identify clinical, pathological and molecular variables to stratify patients into risk groups [4–6]. Lymphovascular invasion (LVI) was found to be a strong prognostic variable for a poor oncological outcome in lymph node-negative patients after RC [5,7–11]. However, the role of LVI as a prognostic factor for UBC remained controversial due to the lack of standardized assessment of LVI [12] and insufficient validation in very large studies. In some studies LVI was not identified as a predictor of survival [7,13] and validation in a large and preferably prospective trial was proposed before LVI could be recommended to be included in the clinical staging of patients with UBC.

To date, valid prospective data on the prognostic role of LVI in UBC are lacking. Our goal was to first validate the importance of LVI as a prognostic marker in a large multicentre cohort of lymph node-negative patients who underwent RC.


The pathological and clinical data of 2011 patients with UBC were collected from six academic institutions in Germany. A multi-institutional database was created comprising all available information. The follow-up after RC was done according to protocols established at each institution, including hospital charts, physician records, and approved cancer registries. All patients included for analysis underwent LA, RC and urinary diversion between 1985 and 2008. Indications for RC included centre-based cystoscopic and biopsy findings, e.g. tumour invasion of the muscularis propria or prostatic stroma, presence of high-grade UBC, concomitant carcinoma in situ (CIS), recurrent multifocal non-muscle-invasive disease or CIS refractory to intravesical chemotherapy or immunotherapy, and recurrent multifocal non-muscle-invasive disease refractory to repeated TUR. Adjuvant chemotherapy was used in 64 patients (5.8%). We excluded patients with pathologically confirmed lymph node-positive disease (549), those with distant metastases (21), those who had RC for palliative reasons (39), those with benign disease of the urinary bladder and/or metastatic invasion by other primaries (including non-UBC, 12) and those with invalid and/or incomplete pathological data (on the presence or absence of LVI, 50) or follow-up information (227). We also excluded patients who received neoadjuvant chemotherapy (14). Thus, the final study population consisted of 1099 patients.

Physical examination, chest radiography, excretory urography, abdominal CT and bone scans were generally undertaken before RC to identify disease, according to institutional guidelines. Standard RC was performed at each institution, although the specific technique and extent of LA varied between surgeons and institutions.

All pathological specimens were processed according to standardized institutional protocols. All specimens were histologically confirmed as urothelial carcinoma, with a minority showing a partial differentiation other than urothelial carcinoma. RC specimens were staged according to the TNM classification, and tumours graded according to the 1973 WHO grading system. LVI was defined as the presence of tumour cells within an unequivocal endothelium-lined space with no underlying walls of smooth muscle cells, in standard haematoxylin and eosin (H&E)-stained sections.

Continuous normally distributed variables are presented as the mean (sd), and those not normally distributed as the median (interquartile range). The Kaplan-Meier method was used to derive the cumulative recurrence-free (RFS), cancer-specific (CSS) and overall (OS) survival, with the log-rank test used to compare curves of two or more groups. Univariable Cox regression analyses were used to identify differences within histopathological variables, and a multivariable Cox regression analysis to identify prognostic factors. All proportional hazard assumptions were verified using the Grambsch-Therneau test. All P values were two-sided and a value of <0.05 was considered to indicate significant differences between groups.


The clinical and pathological characteristics of the patients are shown in Table 1; the mean (sd, range) age of all patients at surgery was 64.9 (12.9, 29–92) years. Among the 1099 patients analysed, organ-confined tumour stages (pTa to ≤ pT2) were present in 621 (56.5%), with extravesical disease (pT3 and pT4) in 430 (39.1%) and high-grade tumours (G3) in 475 (43.8%).

Table 1.  The clinical and pathological characteristics of patients treated with RC for UBC
Variablen (%)5-year CSS, %
  • *

    Valid information was available in 1048 (95.4%) patients.

Male 906 (82.4)74.6
Female 193 (17.6)66.6
Age at surgery, years*  
 0–40  16 (1.6)66.7
 41–60 260 (24.8)82.3
 61–80 733 (69.9)69.2
 >80  39 (3.7)69.8
Histological subtype  
 Transitional1078 (98.1)73.1
 Partial squamous   11 (1.0)81.8
 Partial other  10 (0.9)60.0
No. of lymph nodes removed during LA  
 >15 22079.4
 ≤15 40272.6
Cancer-related death 213 (19.4) 
Death unrelated to cancer 177 (16.1) 
Disease recurrence, but alive  73 (6.6) 
No evidence of disease 636 (57.9) 

LVI was detected in 295 (26.8%) patients; the presence of LVI correlated significantly with increasing tumour stage (P= 0.002) and grade (P < 0.001; Table 2). Information on the number of removed lymph nodes during LA was available in 622 (56.6%) patients. The mean number of removed nodes was 14.66 (8.7, 1–65). On univariable analysis, removal of ≥15 nodes was significantly associated with a favourable RFS and CSS, compared with removal of <15 nodes (P < 0.001 and 0.044, respectively).

Table 2.  HRs for the probability of CSS within stage and grade groups when LVI is present
Variablen (%)LVI, %HR (95% CI)
  • *

    Valid information was available in 1084 (98.6%) patients;

  • no tumour in 43 (3.9%) patients; R, reference standard.

Tumour stage   
 pT0 48 (4.4)  0 
 PTa 51 (4.6)  0 
 PTis 70 (6.4)  0 
 pT1221 (20.1) 65 (29.4)R
 pT2279 (25.4) 88 (31.5)1.537 (0.620–3.809)
 pT3346 (31.5) 110 (31.8)2.411 (1.064–5.468)
 pT4 84 (7.6) 32 (38.1)5.311 (2.105–13.40)
Histological grade*   
 G1134 (12.4) 15 (11.2)R
 G2432 (39.9)132 (30.6)1.538 (0.796–2.971)
 G3475 (43.8)137 (28.8)2.542 (1.247–5.180)

The 5- and 10-year CSS of all patients was 73.2% and 68.8%, respectively, and the respective OS rates were 59.2% and 42.4%. In the univariable analysis the presence of LVI was significantly associated with earlier disease recurrence (5-year RFS 66.6% in LVI-positive and 73.9% in LVI-negative, P= 0.008; Fig. 1A) and a shorter CSS (5-year CSS 69.5% in LVI-positive and 74.5% in LVI-negative, P= 0.039; Fig. 1B). The most distinctive differences in RFS and CSS rates were in LVI-positive patients with advanced tumour stages (5-year CSS 82.9% in pT1, 74.1% in pT2, 64.5% in pT3 and 20.7% in pT4; P < 0.001) and high-grade tumours (74.2% in grade 1, 69.0% in grade 2, and 66.0% in grade 3; P= 0.03), in whom CSS was significantly shorter than in those who were LVI-negative. In the first subgroup (LVI-positive patients with advanced tumours), this difference was also highly significant for OS (P < 0.001), and for RFS (P < 0.001) and OS (P= 0.045) in the latter subgroup. In the subgroup of elderly patients (>75 years), the presence of LVI was associated with disease recurrence (P= 0.024), a shorter CSS (P= 0.028) and OS (P= 0.001). None of the other pathological and clinical variables (including gender, CIS, or number of nodes removed during LA) correlated with the LVI status.

Figure 1.

RFS (a) and CSS (b) estimates in 1099 patients treated with RC for UBC stratified by LVI status.

On multivariable Cox regression analyses including tumour stage, grade, number of nodes removed (≥15 vs <15) and age (>75 vs ≤75 years), LVI was identified as an independent predictor of RFS, with a hazard ratio of 3.502, and CSS (2.611) and OS (2.117) (all P < 0.001; Table 3).

Table 3.  Multivariable Cox regression analyses predicting RFS, CSS and OS in patients treated with RC for UBC
Prognostic variablesHR (95% CI), P
  1. R, reference standard.

Tumour stage   
 pT21.824 (1.226–2.712), 0.0031.968 (1.281–3.023, 0.0021.405 (1.055–1.872), 0.02
 pT32.996 (2.091–4.294), <0.0013.608 (2.466–5.279), <0.0012.158 (1.669–2.789), <0.001
 pT45.569 (3.491–8.885), <0.0016.870 (4.267–11.06), <0.0013.951 (2.781–5.613), <0.001
 G21.189 (0.801–1.766), 0.3911.340 (0.968–1.856), 0.0780.988 (0.693–1.444), 0.796
 G31.779 (0.589–4.282), 0.091.589 (0.927–2.112), 0.1321.885 (1.112–3.222), 0.06
LVI3.502 (2.184–5.617), <0.0012.611 (1.589–4.292), <0.0012.117 (1.449–3.093), <0.001
Nodes removed (>15)1.224 (0.626–3.212), 0.2341.268 (0.681–3.324), 0.4221.452 (1.086–1.942), 0.012
Age (>75 years)2.018 (1.252–3.252), 0.0041.796 (1.106–2.916), 0.0181.444 (1.002–2.081), 0.049


The identification of patients at high risk of disease recurrence and cancer-specific death despite radical surgery remains an important challenge in the treatment of UBC. This is especially true for patients who have favourable pathological characteristics according to TNM staging, e.g. only microscopic involvement of perivesical fatty tissue and negative lymph nodes. Adjuvant chemotherapy is rarely used in this group of patients even though cancer-specific death occurs in up to 25% of them [2,3]. An optimized risk stratification might help to select candidates for more aggressive treatment approaches among patients assumed to have favourable outcomes after RC.

In our multicentre study, we identified the presence of LVI in the surgical specimen as an independent predictor of survival in patients with node-negative UBC treated with RC and pelvic LA. We identified LVI in 26.8% of patients with node-negative disease, which is in line with the proportion reported in a previous study [5]. LVI occurred more frequently in tumours with higher stage and grade, but also ≈30% of pT1 tumours showed LVI, indicating metastatic potential despite a low stage. We confirmed the independent prognostic value of LVI for the clinical outcome. The present study is amongst the largest investigating the prognostic value of LVI in UBC to date, contributing to the further validation of LVI as an important prognostic variable [5,8–11,13,14].

The process of LVI remains poorly understood but there is consensus that it represents an early step in the systemic spread of malignant cells [15,16]. LVI can be regarded as a surrogate marker for the presence of lymphatic micrometastases at the time of RC. In a large multicentre study, Lotan et al.[5] previously evaluated 151 LVI-positive patients with node-negative UBC after RC and showed that LVI is a predictor of recurrence, CSS and OS. In their study, the prevalence of LVI increased with higher pathological stage and grade. The presence of LVI retained independent prognostic value in competing-risks regression models in which other-cause mortality was considered as a competing risk. Given these findings and some confirmatory retrospective studies [5,8–11,13,14], LVI has been suggested to be included in clinical staging models of UBC. However, insufficient validation and problems with the pathological assessment of LVI yet precluded an adoption into TNM classification. To improve accuracy in determination of LVI and reproducibility, Lapham et al.[17] advised using immunohistochemistry for assessing LVI in RC specimens. However, immunohistochemical methods for detecting LVI are still not part of the routine pathological evaluation. There is a lack of comparative studies on the detection rate of LVI in H&E- and specifically stained sections using lymph-endothelial markers (e.g. podoplanin). The frequent occurrence of retraction artefacts mimicking LVI in H&E-stained sections of invasive UBC might lead to false-positive findings and staging errors. Concerns also exist in terms of the experience level of different pathologists evaluating LVI, potentially creating a bias due to the misinterpretation of LVI [12]. The ability of immunohistochemistry to reduce these biases should be investigated in UBC.

Prospective validation trials with central pathological review by experienced genitourinary pathologists should be conducted to give sufficient evidence for inclusion of LVI into routine clinical staging of UBC, comparable to the staging system of testicular germ cell tumours. In this entity, the presence of vascular invasion in an organ-confined tumour is classified as pT2 rather than pT1 disease. Similarly, patients with pT1-4 pN0 UBC could be upstaged accordingly.

One goal of an optimized risk-stratification of patients treated with RC is to select appropriate candidates for adjuvant treatment strategies. The results of studies on adjuvant chemotherapy after RC have been inconsistent. Several trials showed no benefit, whereas other studies showed some survival benefits [18–23]. Most of these studies have been criticized for major deficiencies in terms of sample size, early stopping of patient entry, and inappropriate statistical analyses. Inadequate patient selection which was based on TNM criteria might be the reason for the variation in outcomes of published adjuvant chemotherapy trials for UBC. The chemotherapy regimens applied in the present study were heterogeneous and not always cisplatinum-based. Only 7.8% of LVI-positive patients received adjuvant chemotherapeutic agents and no benefit was found when compared to LVI-positive patients who did not receive adjuvant chemotherapy. The value of adjuvant chemotherapy might have been underestimated because there were relatively few patients in this subgroup analysis. LVI is one potential pathological feature contributing to better patient selection and homogeneity in the group of patients who will receive adjuvant chemotherapy.

Sufficient evidence for improved CSS has only been reported with the administration of neoadjuvant chemotherapy in conjunction with RC [24]. Interestingly, the presence of LVI has also been identified as an important prognostic factor in TUR specimens of UBC [25,26]. Kunju et al.[26] showed that the presence of LVI at the time of TUR predicted LVI in RC specimens in 65% of patients, and the presence of pelvic lymph node metastases in 41%. Thus, LVI might be used to identify high-risk patients who could benefit from neoadjuvant chemotherapy or more extended LA, as the probability of aggressive UBC is higher when LVI is detected at TUR. Support for this hypothesis is given by Millikan et al.[27], who found that LVI was associated with a higher rate of pathological upstaging after RC, from clinically organ-confined to more advanced tumours and/or node-positive disease.

Our study had some limitations; it was retrospective and various physicians and pathologists were involved in generating the data. The study period exceeded two decades and the change in certain practice patterns, including surgical techniques of LA and pathological evaluation criteria might have created heterogeneity in the patient cohort. Similar to previous reports on the role of LVI in UBC [5,14], there was no re-evaluation of the pathological slides. However, to reduce bias, only patients for whom the pathology reports contained explicit information on the presence or absence of tumour cells within a clear endothelial lining were included. Regarding the prognostic value of LVI, we confirmed that it was an independent predictor. Our data are limited in providing information on a gain in predictive accuracy when compared to other pathological variables, e.g. tumour stage. Given the heterogeneity of UBC, the best way to improve the characterization and prognostication in an individual patient might be an assessment of molecular changes in UBC specimens and/or patient urine and serum. Recent publications have shown that biomarkers can provide additional prognostic value [28,29]. We consider our study as confirmatory for LVI as an aggressive characteristic of UBC, representing an impetus for further investigations into the mechanisms of LVI on a molecular level. Elucidation of these processes will potentially identify new molecular therapeutic targets.

In conclusion, we confirmed the independent prognostic value of LVI in patients with node-negative UBC. The presence of LVI predicted CSS in patients who had RC, indicating a high-risk group. Assessment of LVI might be useful as an additional staging tool and could improve the selection of patients who are likely to benefit from adjuvant therapy after RC. The identification of factors involved in the process of LVI might reveal new therapeutic targets for UBC.


None declared.



LVI is an essential and important step in the systemic dissemination of cancer cells. Several small studies reported an association of LVI with adverse pathological features and poor clinical outcomes in patients with UBC. However, despite the increasing number of published studies that have added to the general knowledge about the prognostic role of LVI in UBC, LVI is not part of the TNM staging system and it is not used in treatment guidelines for bladder cancer. This is largely due to the lack of external validation in large, multicentre studies. While encouraging, single-institutional studies are not adequate for definitive evaluation of any marker’s predictive ability. Independent, large multi-institutional studies are needed to fully evaluate prognostic markers, and decrease the chance of overestimation of a marker’s predictive power. In the current study, Bolenz et al. aimed to externally validate the value of LVI as a prognostic marker for UBC in a large multi-institutional cohort of patients treated with RC and bilateral LA for UCB.

The authors confirmed that LVI is a feature of biologically and clinically aggressive UBC. They also confirmed that LVI is independently associated with clinical outcomes after RC. However, the question of whether a biomarker such as LVI can improve the ability of established predictors of cancer outcome requires more than the conventional univariable and multivariable analyses with associated hazard rates and P values. For biomarkers to be clinically useful, they must add unique predictive information, thus improving the performance of a predictive model constructed without the new biomarker by a significant margin [1–4]. Many prognostic factors that are ‘independent predictors’ in multivariable analysis add minimal information to what can be obtained by optimally combining a set of standard factors.

Thus the question is whether adding LVI status improves the accuracy of standard histopathological features for prediction of both disease recurrence and cancer-specific death, especially in patients with no lymph node metastasis. If this is confirmed in large, well-designed studies, the level of evidence would be sufficient to suggest the assessment and reporting of LVI in all urothelial carcinoma specimens. In addition, LVI should be included in predictive tools and should be used for clinical decision-making to help select patients with node-negative UBC who could benefit from adjuvant therapy [4]. Similarly, after further confirmation, patients with LVI on staging biopsy specimens might be offered early aggressive therapy, e.g. early RC and/or neoadjuvant chemotherapy. Finally, LVI could be considered for inclusion in the TNM staging system, as upstaging tumours on the basis of the presence of LVI might improve prognostication. In other malignancies such as hepatic [5] and testicular cancer [6], LVI has been added to the TNM staging system, allowing for improved cancer staging and treatment decision-making. In testicular germ cell neoplasia, the presence of vascular invasion in organ-confined testicular tumour is classified as pT2 rather than pT1 disease. Patients with stage I non-seminomatous germ cell tumour and LVI are recommended to undergo adjuvant chemotherapy, whereas patients without LVI are recommended to undergo surveillance [7]. Similarly, in UBC, improved prognostication might be provided by upstaging tumours on the basis of the presence of LVI.

However, before integrating LVI into daily clinical decision-making, it is of utmost importance that strict morphological criteria are established to standardize and ensure reproducibility and reliability [8]. It is inherently difficult to determine the presence of LVI at the morphologic level and there are significant differences between local pathologists and central pathology review [9,10]. For example, retraction artefacts of the surrounding stromal tissue can mimic vascular invasion [11,12]. Therefore, experts have recommended reporting LVI only in unequivocal cases and to make use of immunohistochemistry if necessary [11,12]. However, the use of immunohistochemical staining to identify the vessels remains controversial (e.g. heterogeneity in the expression of immunohistochemical markers by different capillary structures) and not practical for everyday clinical use [13,14]. In addition, the differential value of LVI location (i.e. peritumoral or outside cancer perimeter-positive invasion vs intratumoral invasion) [15–17] and the differential value of lymphatic vs blood vasculature [18,19] on outcomes needs to be investigated.

Shahrokh F. Shariat
Urology, University of Texas Southwestern Medical Center, TX, USA



RC with urinary diversion remains the standard treatment for patients with invasive bladder cancer. Unfortunately, a significant percentage of patients with organ-confined disease at the time of RC will ultimately recur and eventually die from their disease ([2] in the paper). This is a humbling statistic for any surgeon who treats bladder cancer. The question is how can we improve outcomes for such patients? Certainly improved risk stratification is needed to better identify patients who might benefit most from adjuvant therapies after RC.

LVI has been shown in several retrospective studies to be a potential surrogate for disease recurrence. This multi-institutional study by Bolenz et al. adds to the growing body of evidence supporting LVI as an adverse pathological feature. The authors reported a significantly worse RFS (hazard ratio 3.5; P < 0.002) on multivariable analysis for patients with LVI on final pathology after RC. Urologists should consider LVI when counselling patients about their risk of disease recurrence. Unfortunately, it is unlikely that one pathological variable will provide significantly robust prognostic information to further risk-stratify patients for adjuvant therapy after RC. In this regard, predictive, prospectively validated nomograms based not only on pathology but also on molecular markers and genomic expression are sorely needed.

Thomas J. Guzzo
Assistant Professor of Urology, The Hospital of the University of Pennsylvania, Philadelphia, PA, USA