Impact of lymphovascular invasion in bladder cancer ~Comparison of transurethral resection and total cystectomy: a retrospective cohort study

Background: This study aimed to evaluate the associations of lymphovascular invasion (LVI) at rst transurethral resection of bladder (TURBT) and radical cystectomy (RC) with survival outcomes, and to evaluate the concordance between LVI at rst TURBT and RC. Methods: We reviewed 216 patients who underwent rst TURBT and 64 patients who underwent RC at our hospital. Medical records were reviewed uniformly, collecting the following data: age, sex, clinical and pathological T stage, neoadjuvant chemotherapy, grade, metastasis, urinary cytology, carcinoma in situ and LVI. Results: LVI was identied in 22.7% of patients who underwent rst TURBT, and 32.8% of patients who underwent RC. Univariate analysis identied ≥ cT3, metastasis and LVI at rst TURBT as factors signicantly associated with overall survival (OS) and cancer-specic survival (CSS). Multivariate analysis identied metastasis (hazard ratio (HR) 6.560, p=0.009) and LVI at rst TURBT (HR 9.205, p=0.003) as signicant predictors of CSS. On the other hand, in patients who underwent RC, ≥ pT3, inclusion of G3 and LVI were signicantly associated with OS and CSS in univariate analysis. Multivariate analysis identied inclusion of G3 as a signicant predictor of OS and CSS. The concordance rate between LVI at rst TURBT and RC was 48.0%. Patients with positive results for LVI at rst TURBT and RC displayed poorer prognosis than other patients (p<0.05). Conclusions: We found that the combination of LVI at rst TURBT and RC was likely to provide a more signicant prognostic factor.

Further, LVI has been described as a risk factor in the guidelines of the American Urological Association.
Those guidelines de ne high-grade and T1 tumors, recurrent high-grade and Ta tumors, high-grade Ta and large (> 3 cm) tumor, multifocal high-grade Ta tumor, any carcinoma in situ (CIS), any Bacille de Calmette et Guérin failure in high-grade cases, any variant history, any high-grade prostatic urethral involvement, and any LVI as high-risk factors [18]. However, other guidelines such as those of the European Association of Urology do not de ne LVI as a high-risk factor [19]. Although T stage and grade are common high-risk factors, LVI has also been reported as independently associated with recurrence and progression, even when limited to pT1 high-grade patients [20]. We thus consider that LVI is also likely to represent an important high-risk factor in patients with other high-risk factors, such as T1 and high-grade bladder cancers. High-risk factors are important because the frequency of progression from non-MIBC (NMIBC) to MIBC is higher for high-risk patients than for low-risk patients. NMIBC progresses to MIBC in about 15% of patients, resulting in poor prognosis [21]. Although the standard treatment for MIBC is RC, up to 50% of patient treated in this manner experience disease recurrence and mortality [22]. For NMIBC patients, assessment of LVI at TURBT is very important for risk strati cation and decision-making regarding further treatment. In comparison, for patients after RC, assessment of LVI at RC is suggested to help in determining adjuvant therapy.
As noted above, LVI has been studied in case reports of small numbers of cases, in larger international collaborative groups, and in meta-analyses. In particular, an extensive body of literature exists regarding LVI at RC, which is strongly suggested to be associated with poor clinical outcomes. On the other hand, LVI at TURBT has also been increasingly reported recently, and is suggested to be associated with poor prognosis. TURBT is the rst-line treatment for bladder cancer, and we considered rst-time ( rst) TURBT as the most important for diagnosis and treatment. Although many investigations have examined TURBT and RC individually, few have looked into LVI at TURBT and RC.
The present study aimed to evaluate the relationship between LVI on both TURBT and RC, as well as overall survival (OS) and cancer-speci c survival (CSS). The present study also aimed to assess the concordance rate between TURBT and RC specimens with regard to the presence of LVI. Furthermore, we analyzed differences in prognosis due to the combination of LVI at TURBT and RC.

Methods
This was a retrospective study approved by the institutional review board (approval number: S17108). We identi ed 423 patients with bladder cancer treated by TURBT at Toho University Sakura Medical Center between January 2012 and December 2016. Moreover, a total of 291 patients underwent rst TURBT for bladder cancer. Of those, 75 patients were excluded from the present study due to the following reasons: tumor of any other origin (n = 1); concurrent or recurrent bladder cancer with upper urinary tract urothelial carcinoma (UC) (n = 16); bladder Tis lesions only (n = 8); and missing values (n = 50). A nal total of 216 patients were analyzed for prognostic factors of rst TURBT, and to determine the utility of LVI (Supplementary table 1).
Furthermore, we analyzed prognostic factors for RC, and considered the relationship between LVI in samples from RC and LVI in samples from rst TURBT. A total of 75 patients who underwent RC for bladder cancer were seen at our institution between 2010 and 2017. Of those, 11 patients were excluded for the following reasons: the pathology was not UC (n = 2); and missing values (n = 9). Eventually, a total of 64 patients were analyzed regarding prognostic factors for RC (Supplementary table 2). Twenty patients (31.3%) received neoadjuvant chemotherapy (NAC). Generally, we administered two cycles of gemcitabine-cisplatin (GC). In addition, 50 patients underwent both rst TURBT and RC at our institution, and we analyzed the concordance rate of each LVI. More details can be seen in the ow chart in Fig. 1.
Medical records were surveyed retrospectively, and the following data were collected from medical charts: age at surgery, sex, clinical T stage, presence or absence of NAC, pathological T stage, grade (G1, G2, and G3), presence of metastasis, urinary cytology, CIS, and LVI. LVI was de ned as the presence of tumor cells within an arterial, venous, or lymphatic lumen. LVI was de ned as the presence of tumor cells within an arterial, venous, or lymphatic lumen. Presence of LVI in specimens was assessed using conventional hematoxylin and eosin (HE) staining. If we encountered any di culty in evaluations using only HE staining, we added Elastica van Gieson staining. Positive urinary cytology was de ned as ≥ class IV (malignancy suspected) according to Papanicolaou's classi cation. Lymph node metastasis and distant metastasis were collectively de ned as metastasis. Pathological evaluation was performed by two of three pathologists as a council system in Toho University Sakura Medical Center using the tumor, node, metastasis

Statistical analysis
Results are presented as median and range or mean ± standard deviation, as appropriate. Continuous parametric variables were compared using t-tests. Non-parametric variables were compared using Mann-Whitney U tests. The Kaplan-Meier method was used to estimate OS and CSS following TURBT and RC, respectively. The log-rank test was used to compare statistical signi cances in each curve. After variables were selected in univariate analyses, regression analyses of Cox proportional hazards were performed to determine factors signi cantly associated with OS and CSS.
These statistical analyses were carried out using JMP Pro version 13 (SAS Institute, Cary, NC). All statistical tests were two-sided, and values of p < 0.05 were considered signi cant.

LVI at rst TURBT
The clinical and pathological characteristics of patients who underwent rst TURBT are shown in Supplementary table 1 (n = 216). Mean patient age was 71.3 ±9.3 years, and males represented 79.6% of the study population. G3 was present in 43.5% of patients, 22.7% of patients had LVI, and only 9.7% had CIS. We compared the LVI-positive and -negative groups ( Table 1). The LVI-positive group showed higher pathological T stage (p < 0.001) and greater frequency of G3 (p < 0.001) compared to the LVI-negative group. Subsequently, we analyzed prognostic factors from rst TURBT ( Table 2). Univariate analyses identi ed cT stage, metastasis and LVI as signi cantly associated with OS (p < 0.001, p < 0.001, p = 0.003, respectively). Multivariate analysis identi ed metastasis (hazard ratio [HR] 4.717, 95% con dence interval (CI) 1.336-15.755; p = 0.017) as signi cant predictors of OS after rst TURBT. According to univariate analyses, ≥cT3 (p < 0.001), metastasis (p < 0.001) and LVI (p < 0.001) were signi cantly associated with CSS. Patients in the LVI-negative group at rst TURBT (n = 167) displayed signi cantly better CSS than those in the LVI-positive group at rst TURBT (n = 49; p < 0.001) ( Supplementary Fig. 1

Comparison of LVI between rst TURBT and RC
We analyzed the concordance rate between LVI at rst TURBT and LVI at RC (Table 3). We analyzed data from the 50 patients who underwent RC, excluding 14 patients who did not undergo rst TURBT in our hospital or for whom data were missing. Of the 31 patients with LVI at rst TURBT, 10 patients (32.3%) also showed LVI at RC. Of the 19 patients without LVI at rst TURBT, 14 patients (73.7%) were also found to have no LVI at RC (Table 3).
We con rmed the effect of NAC on LVI in RC specimens. The clinical and pathological characteristics of patients according to use of NAC are shown in Supplementary table 3. In the group receiving NAC (NAC + group) at rst TURBT, cT stage was higher (p = 0.006), and LVI was more frequently evident (p = 0.003) compared to the group not receiving NAC (NAC-group). On the other hand, patients in the NAC + group at RC displayed signi cantly higher pT stage than those in the NAC-group (p = 0.044). No signi cant difference in the presence or absence of LVI was evident between NAC + and NAC-groups at RC (p = 0.312). In patients showing LVI (LVI+) at rst TURBT, but not at RC, pT stage was pT0 in 9 patients (42.9%). Supplementary table 4 shows changes in LVI according to the use of NAC in LVI + cases at rst TURBT. LVI in RC specimens showed no signi cant difference according to use of NAC. No signi cant difference in LVI + or LVI-status was seen between the NAC + group (n = 14) and NAC-group (n = 17) in LVI + cases at rst TURBT (p = 0.252).
Finally, survival analysis was conducted across all combinations of LVI at rst TURBT and RC. Patients with no evidence of LVI at both rst TURBT and RC (TURBT-/RC-) had the best prognosis, and patients with evidence of LVI at both rst TURBT and RC (TURBT+/RC+) had a worst prognosis (p = 0.016) (Fig. 2). Five-year CSS rates for each combination of LVI at rst TURBT and RC were determined. Fiveyear CSS rate was 92.9% for patients with TURBT-/RC-, 80.0% for those with LVI at RC but not rst TURBT (TURBT-/RC+), 77.0% for those with LVI at rst TURBT but not at RC (TURBT+/RC-), and 50.0% for those with evidence of LVI at both rst TURBT and RC (TURBT+/RC+).

Discussion
This study detected LVI in 22.7% of rst TURBT specimens, and 32.8% of RC specimens. These rates were comparable to those described in the literature. The rate of LVI at TURBT has been reported as 6-70%, and the rate of LVI at RC has been reported to range from 30-50% [1]. Clinical stage and pathological T stage were higher in the LVI-positive group than in the LVI-negative group at rst TURBT. Moreover, we found that patients with LVI often showed evidence of G3 in the resected specimen. Patients with LVI were thus suggested to be more likely to be at high risk. In clinical practice, choosing the next treatment for patients with high-grade T1 is di cult. This is because high-risk patients are more likely to progress to MIBC, and RC is also a treatment option. Mathieu et al. showed that high-grade pT1 bladder cancer and LVI detected in biopsy specimens were poor prognostic factors for disease recurrence and progression [23]. RC is thus suggested to be a treatment option before progression for high-grade T1 patients with LVI.
At rst TURBT, ≥cT3, presence of metastasis, and LVI were found to be signi cantly associated with OS and CSS in univariate analyses. Furthermore, LVI was found to be a signi cant predictor of CSS in multivariate analysis. Various studies have reported that LVI was associated with CSS, but not with OS. A meta-analysis of LVI at rst TURBT has shown that the pooled hazard ratio was signi cant for CSS (HR 1.35, 95%CI 1.01-1.81; p = 0.04), but not for OS (HR 1.55, 95%CI 0.90-2.67; p = 0.11) [11]. Moreover, Streeper et al. showed that patients with LVI of clinical stage I or II showed lower survival than those without LVI (HR 2.68, 95%CI 1.55-4.64; p = 0.049) [3]. Resnick et al. also associated LVI at TURBT with an increasing likelihood of node-positive disease (48.3% vs. 25.0%; p < 0.001) [24]. In addition, the association between LVI in TURBT specimens and RFS has also been reported, and LVI at TURBT is a poor prognostic factor [6,20,25].
Subsequently, at RC, factors of ≥pT3, presence of G3 and presence of LVI were found to be signi cantly associated with OS and CSS in univariate analyses. In multivariate analysis, inclusion of G3 was identi ed as a signi cant predictor of OS and CSS. High tumor grade is included as a high-risk factor in all guidelines, and is also included as a factor in the nomogram that predicts risk of recurrence after RC [26]. On the other hand, LVI was shown to be signi cantly associated with OS and CSS in univariate analysis. Although LVI was not identi ed as a signi cant factor in multivariate analysis in the present study, several studies have reported LVI as associated with OS, CSS, and RFS even in multivariate analysis [16,[27][28][29][30][31][32]. Mari et al. summarized the relationship between LVI and clinical outcomes after RC [15]. They found that patients with LVI at RC exhibited higher risks of disease recurrence (HR 1.57, 95%CI 1.45-1.70) and cancer-speci c mortality (HR 1.59, 95%CI 1.48-1.73). In particular, LVI was associated with recurrence and cancer-speci c mortality with lymph node-negative bladder cancer. We suggest that lymph node-negative patients with LVI were at increased risk of lymph node metastasis, and LVI was likely to be a risk factor for recurrence and cancer-speci c mortality. In this study, 5 patients who underwent RC had lymph node metastasis, and positive LVI was seen at TURBT in all these patients. Lymph node metastasis was not a signi cant prognostic factor after RC, probably because the sample size was too small. However, metastasis should be considered a strong prognostic factor in general.
Next, we considered the concordance between LVI in rst TURBT and RC specimens. A total of 10 patients (32.2%) with LVI at rst TURBT showed the presence of LVI at RC. On the other hand, a total of 14 patients (73.7%) without LVI at rst TURBT had negative LVI at RC. In other words, the concordance rate between LVI at TURBT and at RC was 48.0%. The concordance rate was about 60-70% in some other reports, but was lower in this study [3,24,33]. The reasons for this were as follows. First, although many pathologists assess and diagnose specimens in daily clinical practice, a central pathological review of TURBT specimens was not performed in this study. Therefore, some variation in pathological evaluations may have been present. Second, some patients underwent NAC before RC at our hospital.
Furthermore, the determination of which patients should undergo NAC was made at the discretion of each attending physician, and no clear, standardized criteria for selection were established. This might have affected the concordance rate. In fact, patients with LVI at rst TURBT but not at RC (TURBT+, RC-) included patients who had undergone NAC and those whose pathological stage from the RC specimen was pT0. Moreover, we analyzed differences in prognosis due to the combination of LVI at rst TURBT and RC, and found an interesting result: all patients with evidence of LVI at both rst TURBT and RC displayed very poor prognosis and died within 2 years. LVI does not show a very high concordance rate, but may offer very useful information when both LVI at TURBT and RC are positive. Consequently, in patients with LVI at both rst TURBT and RC, the results may be useful for considering adjuvant therapy after RC, such as adjuvant chemotherapy, to improve prognosis.

Conclusions
LVI at rst TURBT and RC were associated with OS and CSS. Speci cally, LVI at rst TURBT was suggested to represent an independent prognostic factor for CSS. Furthermore, we found that the combination of LVI at rst TURBT and RC was likely to offer a more signi cant prognostic factor than either factor individually. Ethics board approval number is S17108. Additional informed consent to participate was not required due to the retrospective nature of this study. However, we publish the details of this clinical study on our website, and we provide an opportunity for patients to reject (opt out).

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.