The presence of lymphovascular invasion in radical cystectomy specimens from patients with urothelial carcinoma portends a poor clinical prognosis

Abstract

OBJECTIVES

To assess the prognostic significance of lymphovascular invasion (LVI) on clinical outcomes in patients with transitional cell carcinoma of the bladder treated with radical cystectomy (RC).

PATIENTS AND METHODS

We retrospectively evaluated a prospectively maintained and authorised cystectomy database; the presence or absence of LVI was determined by pathological examination of the RC specimen. Cox regression analysis and Kaplan-Meier tables were developed to evaluate the contribution of LVI to clinical outcomes.

RESULTS

In all, we analysed 356 patients treated with RC and urinary diversion between 1988 and 2006, with a mean follow-up of 45.6 months. Of these patients, 242 (68%) had no evidence of LVI in the RC specimen, whereas 114 (32%) had LVI. Patients with LVI tended to present with higher pathological stage; 84 (74%) had pT3 or pT4 disease. On univariable analysis the presence of LVI conferred a significant risk for decreased overall, cancer-specific and recurrence-free survival (P < 0.001); the mean values for LVI-negative patients were 96.8, 157.4, and 135.0 months, respectively, vs LVI-positive patients, whose survival times were 52.3, 82.7 and 75.2 months, respectively. The multivariable analysis showed significant independent risk for cancer-specific and overall survival for patients who were LVI-positive and had no lymph-node metastases. The hazard ratios (95% confidence interval) were 1.63 (1.06–2.51, P < 0.026) and 1.81 (1.06–3.08, P < 0.03) for overall and disease-specific survival, respectively.

CONCLUSIONS

The presence of LVI in the pathological RC specimen confers significant independent risk for reduced bladder cancer-specific and overall survival. This variable could be used to prospectively stratify patients who would benefit from adjuvant chemotherapy.

Abbreviations

LVI

lymphovascular invasion

RC

radical cystectomy

LNM

lymph node metastases

PLND

pelvic lymph node dissection

UC

urothelial carcinoma

OS

overall survival

DSS

disease-specific survival

RFS

recurrence-free survival.

INTRODUCTION

In the USA there will be ≈65 000 new diagnoses of bladder cancer, with ≈13 000 deaths from this disease each year [1]. By contrast with other cancers, the death rate from bladder cancer has not decreased substantially over the past two decades [2,3]. For patients with muscle-invasive bladder cancer or refractory superficial high-grade disease, the treatment that offers the best chance for durable cure is radical cystectomy (RC) with pelvic lymph node dissection (PLND) [4–6].

The disease-specific variables with confirmed prognostic value for cystectomy outcomes include tumour stage, tumour grade, and the presence of lymph-node metastases (LNM) [7]. Other investigators reported independent predictive value for the presence of hydronephrosis [8], serum tumour markers [9,10], and the expression of certain cell-cycle regulators [11].

The prognostic significance of lymphovascular invasion (LVI) in patients treated with RC was previously investigated by other groups [12–19]. There are conflicting data about the impact of LVI on patient prognosis and survival. Thus the objective of the present study was to evaluate the impact of LVI in patients with bladder cancer who were treated with RC, to determine if LVI provides independent predictive value for clinical survival.

PATIENTS AND METHODS

We retrospectively reviewed our prospectively maintained RC database at the University of Pennsylvania. This database consists of 440 consecutive patients who had a RC and urinary diversion by one surgeon (S.B.M.) at our institution from 1988 to 2006. Our analysis was limited to those patients with urothelial carcinoma (UC), thereby excluding 32 patients from the original cohort. Of the remaining 408 patients, 52 were excluded for incomplete data, including unavailable LVI information. The indications for RC in all cases were muscle-invasive bladder cancer diagnosed by transurethral resection, or high-grade Ta, T1 or carcinoma in situ refractory to repeat transurethral resection and intravesical immunotherapy or chemotherapy.

RC was performed with en bloc excision of the bladder, prostate and seminal vesicles in men and en bloc excision of the bladder, uterus, ovaries and anterior vagina in women. All patients had a bilateral PLND to the bifurcation of the common iliac vessels cranially, the genitofemoral nerve laterally and the node of Cloquet distally. LND also included the obturator lymph-node packets bilaterally and a single presacral lymph- node packet. Lymph-node packets were sent separately for pathological analysis. Pathology specimens were examined according to our institution’s protocol under the direction of a genitourinary pathologist (J.E.T.). Tumours were staged according to the 1997 American Joint Committee on Cancer staging manual and graded according to the 1973 WHO classification system. The presence or absence of LVI was determined by microscopic analysis of the dominant tumour in the RC specimen.

The follow-up complied with our institution’s protocol; patients were seen every 4 months for the first 2 years, every 6 months until the fifth year, and then annually. Patients were monitored with a physical examination, routine general laboratory tests, chest X-ray and urine cytology at each visit. The upper tract was monitored yearly with either IVU or a loopogram, depending on the patients’ serum creatinine level. Abdominal and pelvic cross-sectional imaging was also done biannually. Abnormal surveillance imaging or a positive cytology prompted a further evaluation, as necessary.

Univariable and multivariable Cox regression analysis were used to determine the significance of the clinical and pathological variables for overall survival (OS), disease-specific survival (DSS), and recurrence-free survival (RFS). Multivariable Cox regression analysis was used to examine the relationship between LVI and OS, DSS and RFS while controlling for age at surgery, sex, pathological stage and nodal status. Survival probabilities were estimated using Kaplan-Meier curves, with the OS and DSS calculated from the date of RC to the date of death or last known follow-up, and RFS calculated from date of RC to the date of detectable recurrent disease or last known follow-up. The log-rank statistic was used to compare OS, DSS and RFS by the presence or absence of LVI. In all tests, P < 0.05 was considered to indicate statistical significance.

RESULTS

The study comprised 356 patients who had RC for bladder TCC from 1988 to 2006 (mean age 65.5 years, sd 10.0; male-to-female ratio 4:1). The mean follow-up was 46.4 months; 242 (68%) of the patients had no evidence of LVI in their RC specimen, while 114 (32%) had LVI. Of the 114 patients with LVI, 84 (74%) had a pathological stage of T3 or T4. Table 1 shows the pathological stage distribution of patients according to the presence or absence of LVI. Of all patients, 99 (28%) had LNM at the time of RC; 57 (58%) of these had LVI while 42 (42%) had no evidence of LVI in their pathological specimen.

Table 1.  The distribution of patients by pathological stage and LVI status

pT stage	Total	LVI −ve	LVI +ve, n (%)
pT0	23	23	0
pTIS	53	50	3 (6)
pTa	4	4	0
pT1	29	23	6 (21)
pT2	85	64	21 (25)
pT3	112	55	57 (51)
pT4	50	23	27 (54)

The OS, DSS and RFS are shown in Fig. 1; patients with no LVI had a statistically significant survival advantage for all three variables when compared with those with LVI. The mean OS, DSS and RFS for LVI-negative patients were 96.8, 157.4, and 135.0 months, respectively, vs LVI-positive patients, whose survival times were 52.3, 82.7 and 75.2 months (P < 0.001). Further stratification of the two groups by pathological stage indicated that LVI was most significant in patients with pT3 disease. Interestingly, patients with pT2 or pT4 disease had no statistical difference in their clinical outcome when stratified by LVI. The mean OS, DSS and RFS for patients with no LVI and pT3 disease was 82.2, 114.5 and 106.7 months, respectively; the respective values for those with LVI and pT3 disease were 35.0, 41.3 and 38.3 months (P = 0.005, 0.001, and 0.001, respectively).

Figure 1. Kaplan-Meier survival estimates by LVI status.

We then analysed survival using LVI and pathological stage (Fig. 2); as for all patients, the mean OS, DSS and RFS were statistically significantly longer in patients with pT3 disease, no LVI or LNM, with mean times of 97.7, 130.6 and 135.5 months for patients with no LVI, vs 41.0, 50.7 and 41.3 months, respectively (P = 0.044, 0.028 and 0.007).

Figure 2. Kaplan-Meier estimates for overall survival, disease-specific survival, and recurrence-free survival by LVI status, pT stage.

Multivariable analyses were then used to determine if LVI was an independent predictor of worse clinical outcome. For all patients, the presence of LVI was a statistically significant independent predictor of poor OS and DSS but not RFS (P < 0.01, 0.007 and 0.1). Table 2 shows the effect of LVI on OS, DSS and RFS for all patients. We then repeated the multivariable analysis dividing patients into those with or with no LNM. The prognostic significance of LVI was lost in patients with lymph-node-positive disease at the time of RC, but was maintained in patients with node-negative disease. Table 2 also shows the relevance of LVI in patients with no LNM. As in all patients, OS and DSS were statistically significantly longer (P <0.026 and 0.03) but RFS was not (P <0.09).

Table 2.  Multivariable analysis of all patients and node-negative patients

Group, LVI status	Hazard ratio (95% CI), P
	OS	DSS	RFS
* Reference category.
All patients
Age	1.05 (1.03, 1.06), <0.001	1.03 (1.01, 1.05), 0.006	1.02 (1.00, 1.04), 0.040
LVI, +vs vs −ve*	1.53 (1.11, 2.11), 0.010	1.74 (1.16, 2.59),0.007	1.38 (0.93, 2.03), 0.108
Pathological stage	<0.001	<0.001	<0.001
pT1 vs pT0, pTIS, pTa*	0.93 (0.44, 1.95), 0.850	0.23 (0.03, 1.77), 0.158	0.41 (0.12, 1.40), 0.156
pT2 vs pT0, pTIS, pTa*	1.12 (0.70, 1.80), 0.643	1.47 (0.73, 2.95), 0.276	1.25 (0.69, 2.27), 0.466
pT3 vs pT0, pTIS, pTa*	1.95 (1.26, 3.02), 0.003	3.58 (1.90, 6.73), <0.001	2.64 (1.52, 4.60), 0.001
pT4 vs pT0, pTIS, pTa*	4.53 (2.75, 7.47), <0.001	6.98 (3.47, 14.03), <0.001	3.75 (1.95, 7.22), <0.001
Nodal status, +ve vs −ve*	1.54 (1.11, 2.13), 0.009	1.56 (1.05, 2.34), 0.029	2.05 (1.38, 3.05), <0.001
Node-negative patients
Age	1.06 (1.04, 1.08), <0.001	1.03 (1.01, 1.06), 0.009	1.03 (1.00, 1.05), 0.036
LVI, +ve vs −ve*	1.63 (1.06, 2.51), 0.026	1.81 (1.06, 3.08), 0.030	1.59 (0.93, 2.71), 0.093
Pathological stage	<0.001	<0.001	<0.001
pT1 vs pT0, pTIS, pTa*	0.72 (0.30, 1.73), 0.458	0.29 (0.04, 2.24), 0.235	0.18 (0.02, 1.39), 0.101
pT2 vs pT0, pTIS, pTa*	1.21 (0.73, 2.00), 0.465	1.59 (0.75, 3.40), 0.229	1.24 (0.64, 2.42), 0.521
pT3 vs pT0, pTIS, pTa*	1.79 (1.07, 2.98), 0.027	3.47 (1.66, 7.25), 0.001	2.48 (1.28, 4.79), 0.007
pT4 vs pT0, pTIS, pTa*	5.29 (2.85, 9.83), <0.001	9.81 (4.32, 22.26), <0.001	6.29 (2.89, 13.68), <0.001

The survival effect of chemotherapy on patients with LVI but no LNM was also examined. Of the 63 patients in this subset, 11 received chemotherapy while the other 52 did not. There was no advantage in those patients receiving adjuvant chemotherapy for OS, DSS and RFS (P = 0.8, 0.2 and 0.75).

DISCUSSION

The presence of LVI at the time of RC is an independent predictor of a poor prognosis in patients with UC. To our knowledge, this analysis represents the second largest single-institution experience on this subject. In the present study this pathological variable was the strongest predictor in patients with lymph node-negative disease. Our analyses indicate that patients with lymph-node negative disease, the absence of LVI, and pT3 disease had the greatest survival advantage over patients with LVI.

The invasion of bladder cancer into lymphatic and vascular vessels could be construed as a marker for gross or micrometastatic disease. This supposition could clearly explain why LVI has no independent prognostic value in patients with lymph node-positive disease. However, when examining patients with lymph node-negative disease, LVI has independent prognostic value, which is driving its statistical significance for the whole cohort. Further stratification examining a stage-for-stage effect showed prognostic significance only for patients with pT3 disease. The same effect would be expected in patients with pT2 and pT4 disease; the explanation for this disparity could be that pT2 disease is an organ-confined process, thus the presence of metastatic disease is too low to discern a difference when stratifying with LVI. Thus, the ability to detect a significant difference would require a large group of patients.

Patients with pT4 disease have significantly advanced cancer, often with bulky local involvement or invasion into the prostate, both of which are markers for a poor outcome. The few overall patients (50) in this subset make it difficult to draw any definitive conclusions, but it is likely that the significance of LVI is lost in these patients due to the advanced stage of their disease.

There have been many published reports of both lower [12–17] and upper [18,19] urinary tract UC and the implications of LVI or perineural invasion or vascular/blood vessel invasion. The largest single-institution series investigating LVI was published by Quek et al.[12] in 2005. They showed, in 702 patients with UC, that the presence of LVI was an independent predictor of poor outcome, using a multivariable analysis. Lotan et al.[13] also published their results of 750 patients from a multi-institutional RC database in 2005. They found that LVI was significant specifically in patients with no lymph node disease, in terms of local and distant recurrence, OS, DSS and RFS. Finally, Leissner et al.[14] reported their analysis of 283 patients, specifically analysing the importance of lymphatic invasion, perineural invasion, and/or blood vessel invasion. They found that only the presence of blood vessel invasion in 37 of the 283 patients was statistically significant on multivariable analysis.

There have been negative studies of LVI and its independent prognostic significance, most notably by Bassi et al.[15]. In their series of 369 patients, they found that vascular, lymphatic or perineural invasion provided no statistical significance on multivariable analysis. Cheng et al.[16] and Bell et al.[17] also reported no independent significance with their series, but both reports only assessed vascular or blood vessel invasion, and both contained fewer than 75 patients. It seems relatively convincing that the present findings taken in concert with the published reports strongly suggest that the prognostic significance of LVI in the final pathological specimen. In >2000 patients it was shown that the presence of LVI compromises survival.

Clearly, there are limitations to the present study; being retrospective limits the conclusions that can be drawn. Furthermore, despite our suppositions, it is unclear why patients with pT4 disease but no LVI had no statistically significant survival advantage, as seen in the pT3 group. These subset analyses are under-powered to draw any definitive conclusion. Finally, the data relating to RFS are less robust, as there is no clinical protocol by which these patients are being followed. Enrolling these patients in a clinical trial would ensure a uniform follow-up programme.

A recent article describing the ‘Will Rogers phenomenon’ in relation to urological cancers noted that there has been a stage migration and change in survival in bladder cancer from the 1970s to the 1990s, mainly due to improved surgical techniques, i.e. the implementation of routine PLND [20]. However, since the 1990s, the stage migration seen with other urological cancers has not been apparent in bladder cancer. Furthermore, a recent review of practice patterns associated with the use of perioperative chemotherapy for stage III bladder cancer showed a statistically significant increase in the use of chemotherapy from 1998 to 2003; however, in real terms, this increase only amounted to 62 more patients receiving chemotherapy [21], i.e. of ≈7000 patients, 204 vs 142 received perioperative chemotherapy. Finally, a recent study by Herr et al.[22] showed that early RC resulted in improved survival in patients with high-grade T1 disease, vs delayed RC. All these findings underscore the need for a more aggressive treatment paradigm for bladder cancer. The difficulty with bladder cancer is identifying those patients with non-metastatic disease who are at greater risk of recurrence, and thus should be treated more aggressively with chemotherapy. LVI is one such variable that can be used to stratify the risk in patients who might benefit from adjuvant chemotherapy.

In conclusion, the presence of LVI confers an independent risk of a poorer clinical outcome in patients treated for UC with RC. LVI is more common in patients with a higher pathological stage and its presence has the most independent predictive value in patients with node-negative disease. This pathological variable might provide further prognostic value in risk-stratifying patients for more aggressive treatments.

CONFLICT OF INTEREST

None declared.