The predictive value of preoperative positive urine cytology for development of bladder cancer after nephroureterectomy in patients with upper urinary tract urothelial carcinoma: A retrospective multi-center case-control study and meta-analysis

Background: Upper urinary tract urothelial carcinoma (UUT-UC) is a rare and severe urinary malignancy. Urine cytology is a common clinical method for the early diagnosis of urologic neoplasms. Several studies have explored the relationship between preoperative urine cytology and intravesical recurrence in patients with UUT-UC. However, the results of these studies are controversial or even contradictory, and investigations with UUT-UC patients in Northeast China are rare. Methods: In this retrospective case-control study, we rst estimated the prognostic signicance of preoperative urine cytology in the outcomes of intravesical recurrence in 142 patients with UUT-UC after radical nephroureterectomy (RNU). A nomogram model was established and we evaluated its effectiveness. Subsequently, after searching several databases, including PubMed, Embase, and Ovid, we quantitatively combined our results with the published data in a meta-analysis. Results: Our case-control study demonstrated that patients with muscle-invasive tumors (T2-T4) were more prone to preoperative positive urine cytology than those with non-muscle-invasive tumors (Tis-T1), suggesting that preoperative positive urine cytology may be associated with the aggressive form of UUT-UC. Kaplan-Meier curves demonstrated that the patients with positive urine cytology had signicantly poorer intravesical recurrence-free survival (IV-RFS). Additionally, preoperative urine cytology (hazard ratio, HR=3.24) and lymph node status (HR=2.67) were validated as signicant independent predictors of intravesical recurrence by multivariate analysis. Nomogram, calibration plots, AUC values and the C-index demonstrated that the predictive accuracy was signicantly improved when preoperative urine cytology was combined. Subsequently, adding to our study, 11 eligible articles from 2010 to 2016 containing 2942 patients were sifted out for our meta-analysis. Overall analysis showed that preoperative positive urine cytology was associated with a 56% increased risk of intravesical recurrence (HR=1.56). In the subgroup analysis by region, study type, and sample size, the pooled HR was statistically signicant for the Japan subgroup (HR=1.45), China subgroup (HR=1.66), cohort study subgroup (HR=1.44), and case-control study subgroup (HR=1.77), the subgroup with sample size greater than or equal to 100 (HR=1.42), and the subgroup with sample size less than 100 (HR=2.95). Conclusions: Preoperative urine cytology signicantly correlated with intravesical recurrence in patients with UUT-UC after RNU, although these results need to be interpreted with caution. Large, prospective trials are required to further conrm its prognostic signicance.


Introduction
Upper urinary tract urothelial carcinoma (UUT-UC), which is widely acknowledged to be a urinary malignancy that originates from the urothelium, which runs from the renal pelvis and calyces to the distal ureter, is usually considered an infrequent urothelial tumor, accounting for 5-10% of urothelial carcinomas. Statistically, the average prevalence is 1-2 per 100,000 in the United States [1,2].UUT-UC is more commonly diagnosed in Asian countries, such as China (including mainland and Taiwan), and can even reach 20-30 cases per 100 persons in high-incidence areas [3,4]. Although the industry-recognized standard treatment for UUT-UC is radical nephroureterectomy (RNU) with bladder cuff excision, several authors have investigated whether subsequent bladder cancer occurrence is highly common after the management of UUT-UC. During follow-up, approximately 22-50% of UUT-UC patients who undergo RNU have bladder cancer recurrence; additionally, the recurrence rates at 5 years can reach more than 30% [5][6][7][8].Recurrence and rapid progression of bladder cancer after UUT-UC greatly increase the psychological anxiety and economic burden of patients and even lower their quality of life [9].
Previously, there were controversial predictive factors that resulted in poor prognosis for UUT-UC and intravesical recurrence, such as patient-speci c factors (age, sex, diabetes mellitus or history of bladder cancer), tumor-related factors (location, size, T stage or architecture) and environmental factors (smoking, long-term use of aristolochic acid, etc.) [9][10][11][12][13]. It is worth noting that the 2-year recurrence-free survival rate is less than 56% for stage ≥pT1 cancer, and the 5-year speci c survival rate is less than 50% for stage ≥pT2 [14]. Therefore, the exploration of the correlation between various predictors and intravesical recurrence in patients with UUT-UC after RNU plays a crucial role in clinical comprehensive prevention and treatment, which depend on many kinds of diagnostic techniques [15].
As one of the methods of bladder cancer diagnosis and follow-up, urine abscission cytology has been used for many years. In 1864, Samders discovered cancer cells in the urine of bladder cancer patients, opening the door to urine cytology as a diagnostic tool for bladder cancer [16,17]. Due to the characteristics of urine samples, it is always the goal of researchers to obtain high-quality diagnostic results. With the advancement of immunology and molecular biology, many bladder tumor markers have been found over time, which have clinical application value for the early diagnosis and treatment of bladder cancer [18]. Cystoscopy and urine cytology have already become standard for the diagnosis and follow-up of bladder cancer. Previous studies have shown that urine cytology correlates with the prognosis of bladder cancer [19][20][21].
Although there are studies about urine cytology in the prognosis of intravesical recurrence in UUT-UC patients who underwent RNU [22,23], the number of studies in mainland Chinese is limited [24,25]. Moreover, the current role of this modality lacks evidence. In order to draw a persuasive conclusion, we rst collected clinical data from UUT-UC patients who underwent RNU to retrospectively evaluate the prognostic value of urine cytology and intravesical recurrence in patients with UUT-UC in China. Then, we conducted a meta-analysis of published literature to systematically explore the role of urine cytology and its effects on postoperative intravesical recurrence.

Study population
Prior to this study, the information collection and experimental procedures of this study were approved by the The preoperative evaluation of urine cytology Unlike ureteroscopic evaluation with tumor biopsy or retrograde pyelography, preoperative urine cytology should be prioritized if it is necessary to conduct tumor cytology detection. According to the physician's judgment, urine samples for cytological examination were collected 1-3 times. In conventional urine cytology, three conditions could be de ned: (1) positive urine cytology; (2) suspicious urine cytology; and (3) both. Other cases were all deemed negative samples if they showed negative urine cytology and/or atypical cells. Positive cytology indicated malignancy, while negative cytology indicated benign or normal tissue [26]. The nal judgment was con rmed by two pathologists who were blinded to clinical outcomes.

Follow-up and surveillance regimen
In the rst year after the RNU operation, observations were made every 3 months. From the second year to the fth year, the frequency was slowed down to every 6 months, and once per year thereafter. The measurements of surveillance include serum chemistry studies, routine urine and blood examinations, cystoscopy and various imaging examinations, such as chest plain radiography/CT and radiographic examination of the urinary tract. In addition, the nonsurviving patients' responsible physicians identi ed the causes of death and then issued chart reviews or death certi cates.

Statistical analysis
The association between preoperative urine cytology and clinicopathological parameters was appraised using Fisher's exact test and the c 2 test. In order to directly re ect the degree of bladder recurrence after RNU, intravesical recurrence-free survival (IV-RFS), which was de ned as the death rate caused by recurrence of UUT-UC after RNU from a recurrent lesion located in the bladder, was used to statistically compare different groups by the log-rank test.
Taking various prognostic factors from previous statistics, univariate analysis was used to screen out the most correlated factors (p<0.05). Afterwards, the prognostic factors selected by univariate analysis were subjected to multivariate Cox proportional hazard regression models to compare with each other and lter out the signi cant factors (p<0.05). From the direct Kaplan-Meier survival curves drawn from the outcomes of the log-rank test, we compared the ratio of intravesical recurrence-free survival among all the patients and analyzed the survival time of certain cases visually. The aforementioned procedures were done in SPSS version 20.0 (SPSS Inc, Chicago, USA). To identify the independent risk factors for moderate/severe preoperative urine cytology, nomogram plot was programmed with R software (R Foundation for Statistical Computing, Vienna, Austria). The receiver operating characteristic (ROC) curves were plotted to compare the ability of estimate for patient mortality between nomogram and the criteria of score for UT-UUC set by American Joint Commission on Cancer (AJCC). In addition, calibration plot was also drawn in order to evaluate the potential clinical value of the nomogram. The concordance index (Cindex) was calculated by the means of bootstrapping to estimate the probability that the nomogram prognostic models are consistent with the actual observed results. P-value which is less than 0.05 corresponds with statistically signi cant.

Search Strategy
We assessed the pertinence of outcomes of preoperative urine cytology and the prognosis of UUT-UC after RNU by searching the PubMed, Medline, Embase, Cochrane Library and Scopus databases. The following retrieved words and medical subject headings that included all spellings were used: "Urine cytology", "Upper urinary tract", "Renal pelvis", "Ureter", "Urothelial cancer", "Urothelial carcinoma", "Bladder cancer", and "Prognosis". "Recurrence", "Intravesical". Additionally, we manually searched further eligible bibliographies and company reports to ensure the comprehensiveness of the experimental data from the literature. The publication country and time were not restricted.

Selection Criteria
The following criteria were used to measure whether or not valuable literature was collected. Inclusion criteria: (1) upper urothelial carcinoma was the histologic type; (2) selected samples had the corresponding data of preoperative urine cytology; and (3) the relationship between preoperative abscission urine cytology and postoperative IRFS was analyzed. The exclusion criteria were as follows: (1) studies that lacked original data, including reviews, letters to the editor, commentaries, and case reports; (2) studies containing data from earlier studies; (3) articles written by the same authors as another, chosen article; (4) studies without original data or no record of intravesical recurrence-free survival; and (5) studies with invalid data for calculating the hazard ratio (HR) and its standard error.

Data extraction
According to the aforementioned criteria, two investigators undertook a rigorous process to improve reliability and minimize bias. A third reviewer (Xishang Song) examined the extracted results and eliminated any discrepancies between the independent search results. If a study appeared appropriate, the full text was reviewed. The study was considered relevant if it met the inclusion criteria. Finally, 11 eligible articles [22,[24][25][26][27][28][29][30][31][32][33] from 2010 to 2016 was selected. The primary information the investigators recorded included author, publication year, region, study type, recruitment period, number of patients, baseline characteristics (mean age and sex, etc.), evaluation of positive urine cytology and so forth. The secondary information, based on the above-mentioned data, such as HRs and 95% con dence intervals (CIs) of IRFS, was also statistically analyzed.

Statistical analysis
The dichotomous outcomes were compared by assessment of HRs with 95% CIs. The c 2 test was used to qualitatively analyze heterogeneity. If the heterogeneity p-value was <0.05, a random-effect model rather than the xed-effect model was employed. The I 2 test was used to quantitatively delineate heterogeneity and measure the percentage of volatility. If I 2 > 50%, there was heterogeneity between studies. Subgroup analysis by Begg's test and Egger's test was visually re ected by forest and funnel plots to represent the estimated pooled effect using the Mantel-Haenszel random-effect model and to assess the publication bias. If the diamond crossed the vertical median line, statistical signi cance was con rmed. All statistical analyses were performed using Stata 12.0 (Stata Corporation, College Station, TX, USA).

Case-control study
Relationship between examination of urine cytology and the clinicopathological characteristics of UUT-UC patients In order to investigate the clinical relationship between the outcomes of urine cytology and intravesical recurrence after radical nephroureterectomy in UUT-UC patients, we redivided our patients into two groups: the preoperative positive urine cytology group and the preoperative negative urine cytology group. The descriptive clinicopathological characteristics of the 142 patients (sex, age, tumor side, tumor location, tumor focality, pathological stage, histological grade, lymph node status, distant metastasis, and type of operation) are listed in Table 1. In total, there were 68 patients (47.9%) with positive urine cytology. The sex composition was 86 males and 70 females. There was no signi cant difference between groups in the outcome of urine cytology or in clinicopathological factors, including tumor side, histological grade, lymph node status, type of operation, etc. The primary tumors were located in the calix or pelvis in 57.0% (n=81) of patients, and multifocal tumors were present in 18 patients (12.7%). Positive lymph nodes were present in 11 patients (7.8%), and distant metastases were present in 2 patients (1.4%).
Pathological stage was the only variable positively correlated with preoperative positive urine cytology (p=0.001). By Spearman's correlation analysis, the correlation coe cient was 0.289 (p<0.001) between preoperative positive urine cytology and pathological stage, indicating that preoperative positive urine cytology may be associated with the aggressive form of UUT-UC.  16 32 Predictors and intravesical recurrence-free survival in UUT-UC patients To determine whether preoperative urine cytology is a signi cant predictor of bladder cancer recurrence after radical nephroureterectomy, Kaplan-Meier curves were calculated to determine the relationships between positive or negative preoperative urine cytology and bladder cancer recurrence (Fig. 1). The intravesical recurrence-free survival of patients with preoperative positive urine cytology was signi cantly lower than that of patients with preoperative negative urine cytology (p<0.001  Nomogram of prognostic prediction based on the preoperative urine cytology Using R language, we combined preoperative urine cytology, age, gender, tumor side, tumor location, tumor focality, histological grade, pathologic stage, lymph node status, distant metastasis, and surgery type to establish a novel nomogram, as shown in Fig. 2. The clinician can locate the data on the nomogram, sum the score up and get a total prognostic point. The 1-, 3-and 5-year OS rates of UT-UUC patients can be converted accordingly. The calibration plots was demonstrated in Fig. 3. We have observed that, in comparison with the prediction of 5-year OS, the actual curves made by using the nomogram in the prediction of 1-and 3-year OS have a better coincidence with the ideal curves. The AUCs of ROC curves of 1-, 3-and 5-year nomograms which are respectively 0.67 (95% CI=0.65-0.83), 0.76 (95% CI=0.66-0.84) and 0.74 (95% CI=0.65-0.83) also revealed better discrimination by the nomogram than AJCC score (Fig. 4) . The model C-index and the internally veri ed C-index respectively equal to 0.75 and 0.68. Therefore, we could speculate that the aforementioned nomogram prognostic model has a signi cant potential to predict 1-, 3-and 5-year OS.
Meta-analysis of the correlation between preoperative positive urine cytology and the risk of intravesical recurrence in UUT-UC patients

Method of systematic selection of relevant publications
The detailed process of selection is shown in Fig. 5. A total of 339 initial studies were identi ed from electronic database and literature searches, and additional records were found from other approaches (n=3). After the duplicated publications were removed (n=114) and we manually screened the titles and abstracts (n=87), 27 studies were left. With the careful assessment of the remaining studies based on the information from the title, abstract, and full text, those that had HRs or 95% CIs that could not be extracted, those that were irrelevant to our analysis (such as papers on extravesicular recurrence, lower urothelial neoplasm or nonurinary carcinoma, etc.), reviews, commentaries, and animal studies were excluded (n=16). Eleven candidate studies remained. We have accepted no different studies from the same group. Adding these to our cohort study (n=1), 12 independent articles including quantitative synthesis were nally included in our meta-analysis. The major characteristics of the twelve included studies are summarized in Table 3. They were published from 2010 to 2016. One of the studies assessed patients from within Korea [22], 3 from China [24,25], and 8 from Japan [26][27][28][29][30][31][32][33]. Only one study was an RCT [30], 4 were casecontrol [22,27,31,32] studies, and the others were retrospective cohort searches [24-26, 28, 29, 33]. The size of the studies varied from 36 to 754, with a total number of 2942 patients. The mean or median follow-up duration varied from 8.4 to 69.3 years. Only one publication did not provide the recruitment period, and demographics of only one publication were missing. Six studies showed the assessment of positive urine cytology.

Overall analysis
Forest plots are depicted in Fig. 6, which presents the relative risk estimates from 12 studies for the relationship between preoperative urine cytology and intravesical recurrence. A random-effect model was selected because signi cant heterogeneity was detected (I 2 =45.3%, p=0.038). Overall analysis showed that the combined HR was 1.56 (95% CI=1.26-1.93, p<0.001), suggesting that preoperative positive urine cytology was associated with a 56% increased risk of intravesical recurrence. Visually, the shape of Begg's funnel plots (Fig. 8)  In the subgroup analysis by sample size, when the sample size was 100 or more, the risk of intravesical recurrence was 1.42 times higher in patients with positive urine cytology than in those with negative cytology (95% CI=1.16-1.73, p=0.001). When the sample size was less than 100, the same trend was seen (HR=2.95, 95% CI=1.68-5.17, p<0.001). To explore whether the assessment of urine cytology in uenced preoperative urine cytology and intravesical recurrence, subgroup analysis was done under a random-effect model.
The results showed that preoperative positive urine cytology was associated with an increase in intravesical recurrence, not only in the group with assessment of urine cytology (OR=1.96; 95% CI=1.23-3.13; p=0.005) but also in the group without assessment of urine cytology (OR=1.44; 95% CI=1.14-1.83; p=0.002). Fig. 7 has shown all the forest plots.

Discussion
The ratio of intravesical recurrence has ranged from 22% to 50% in patients with UUT-UC who underwent RNU in most studies [34]. In our study, intravesical recurrence was seen in 24.3% of samples and usually occurred within 33.1 months after surgery. Therefore, we think that the postoperative follow-up within 2 years should be strengthened. In our univariate analysis, we found that preoperative urine cytology and lymph node status were signi cant factors for recurrence of bladder cancer. In the multivariate analysis, they were still independent risk factors for bladder cancer recurrence in the future. Pathological stage was the only variable positively correlated with preoperative positive urine cytology (p=0.001). In fact, as various risk factors correlated with intravesical recurrence have been found and analyzed, studies differ signi cantly in the identi cation of multifactorial prognosis. For patients with UUT-UC, a history of bladder cancer, tumor location in the ureter or tumor multifocality have been validated as prognostic factors in the recurrence of non-muscle-invasive bladder cancer [35][36][37]. The number of patients with multiple upper urothelial carcinomas who had recurrent urothelial carcinomas was signi cantly higher than that of patients with single urothelial carcinomas [38]. In addition, tumor size, preoperative hydronephrosis and ureterorenoscopy are more likely to lead to postoperative recurrence in the bladder [39][40][41]. Some of those conclusions are consistent with ours. The differences among studies explain the diversity of characteristics of urothelial carcinoma and have led to uncertainties in clinical follow-up and treatment, which should be studied further in long-term, systematic, large-sample research.
In our meta-analysis of 11 published studies [22,[24][25][26][27][28][29][30][31][32][33] and our case-control study, preoperative positive urine cytology was associated with a 56% increased risk of intravesical recurrence (HR=1.56, 95% CI=1.26-1.93), not only in the subgroup of cohort studies (HR=1.44, 95% CI=1.15-1.80) but also in the subgroup of case-control studies (HR=1.77, 95% CI=1.06-2.93). Interestingly, there is an attractive hypothesis that can explain the mechanism of urinary abscission cytology in the prediction of recurrence in the bladder after upper urothelial neoplasms. In a previous study, the dissemination and oating of cancer cells from UUT-UC in the bladder occurred postoperatively [42,43]. During radical nephroureterectomy, the bladder is cut open, and a urethral catheter is placed, resulting in an increased degree of bladder exposure. Injured urothelium is more likely to provide a site for adherence compared with complete urothelium. Along with continuously monitoring the removal of postoperative remnants, urinary bladder suspensions can be used to determine whether tumor cells have adhered to the damaged urothelium and proliferated [44,45].
In fact, tumors can occur in one or more sites simultaneously or at different times. Recurrent bladder cancer after upper urinary tract urothelial carcinoma is a kind of multiple-organ carcinoma of the urinary tract epithelium [46,47].
It is universally acknowledged that the polycentricity of lesions plays an important role in the biological characteristics of urothelial carcinoma [43,45,47]. Based on our ndings and implantation theory, it seems that urothelial multiorgan carcinoma originates from the same cell clone. According to this view, urothelial carcinoma originates from a single cell clone, and urothelial multiorgan carcinoma is caused by tumor cell implanting along the direction of urine ow and migrating intraepithelially to form multiple tumor lesions [43,48]. Many conditions can be explained by implantation theory, such as unilateral upper urinary tract urothelial carcinoma, tumors formed in the same direction as urine ow and consistent tumor histologic patterns in multiple organs.
From the perspective of tumorigenesis, another theory is the multifocal origin theory. The polycentricity of urothelial carcinoma is caused by chromosome deletion and gene mutation of normal cells under the action of local carcinogenic factors, forming independent cell clones. These clones proliferate to produce multiple tumor lesions at the same or different times. This theory can explain the bilateral upper urinary tract tumors appearing at the same or different times and the long-term neoplasms that occur in the same direction. Some scholars still emphasize that intravesical recurrence, even a multiorgan carcinoma in the reverse direction from urinary ow, is related to tumor cell implantation. In fact, studies have reported that multiple or recurrent urothelial carcinomas probably arise from the same tumor cell implantation rather than from multicentric tumors [43,47]. From clinical observations, the incidence of upper urinary tract urothelial carcinoma after bladder cancer is lower than that of upper urinary tract urothelial carcinoma followed by bladder cancer. It was further con rmed that the mechanism of recurrence of bladder tumors in patients with UUT-UC is related to cystoureteral re ux. In summary, tumor cell implantation may be the main cause of recurrent bladder cancer after urothelial carcinoma and is worthy of further study [46,49].
Although we did a case-control study and meta-analysis of the data in the published literature to overcome the limitation of our single-center sample source to some extent and to further con rm the accuracy of our research results, our study has some potential limitations. First, our sample was enrolled from only one a liated hospital of a medical university, restricting the universality of our retrospective ndings. Additionally, the different surgeons used different surgical procedures, which could lead to certain errors. Compared with multicenter data, the sample size of single-center data is limited, which causes certain limitations in the statistical analysis of predictors. The metaanalysis results also have some defects. Although we searched for literature from multiple sources, even including unpublished trials and abstracts that can cause higher risks of null results, we could not refrain from the potential bias. Publication bias might have been caused by several things: (1) In these articles, we selected the hazard ratios from multivariate survival analysis. Univariate data will be adopted when the available results emerge. (2) If HRs and their variance were not reported, we performed a calculation from the survival comparison statistics and its variance, which has less reliability than direct data [50][51][52]. In addition, because of the differences in the diagnostic criteria for urinary cytology among different countries and regions, the quality of this study would have been improved by standardized protocols.

Conclusion And Outlook
With the development of molecular biology approaches, the processes of urothelial carcinoma pathogenesis have been widely and deeply explored. We have classi ed bladder cancer by gene sequencing and comprehensive pro ling of RNA expression and have used DNA sequencing results from biopsy to understand survival distributions and prognostic factors [53][54][55]. This new understanding of heterogeneous diseases such as UT-UUC and bladder cancer can lead to improved outcomes and quality of life through better-customized treatments. For positive tumor cells and moderately or severely abnormal cells, signi cant differences in genes arise as the disease occurs and progresses. Therefore, we predict that the genetic analysis of exfoliated cells in urine may be helpful for making new discoveries on the evolution of urinary tract tumors. In addition, in cancer pathological states, abnormal activation of export machinery, such as exosomes (products of vesicular transport), results in excessive discharge of many key proteins and misexpression of miRNAs, which are molecules with great differences in size [56,57]. mRNA expression analysis is an effective tumor phenotyping tool that has been used to de ne the molecular subtypes of almost every major tumor type [58]. It is worth focusing on the key molecular processes and analyzing gene and protein expression systematically. We believe that successful exploration of cancer therapeutic targets requires an interdisciplinary understanding of these cellular activating or inhibiting mechanisms and how secreted biomolecules The nomogram for 1-, 3-and 5-years OS prediction of UT-UUC and its predictive performance. The independent prognostic factors (including age, gender, tumor side, tumor location, tumor focality, histological grade, pathologic stage, lymph node status, distant metastasis, and surgery type and preoperative urine cytology) were combined with the point designations to get total prognostic score   Forest plot showing the risk of intravesical recurrence in UUT-UC patients with preoperative positive urine cytology in total analysis. Squares in each trial are on behalf of hazard ratios; 95% con dence interval(CI) was represented by the horizontal line which traverse the square. the estimated combined effect was represented by the diamonds.
According to heterogeneity analysis, the Mantel-Haenszel random-effect model was selected Forest plot showing the risk of intravesical recurrence in UUT-UC patients with preoperative positive urine cytology by subgroup analysis of (A) region (B) study types (C) sample size (D) assessment of urine cytology. Squares in each trial are on behalf of hazard ratios; 95% con dence interval(CI) was represented by the horizontal line which traverse the square. the estimated combined effect was represented by the diamonds. According to heterogeneity analysis, the Mantel-Haenszel random-effect model was selected