The initial search yielded 998 references, and 539 studies were excluded because of duplication. After title and/or abstracts screening, 169 articles remained for full-text assessment, and 290 articles were excluded, including reviews, letters, meeting abstracts, and other articles irrelevant to our study. In accordance with the study inclusion criteria, 138 articles were excluded for repeated crowds or without enough extratable data. Finally, thirty-one studies were retrospective design were included in this meta-analysis. A flow diagram about the literature search and study selection process is presented in Figure 1.
Features of included studies
Summary characteristics of these studies are shown in Table 1. A total of 14,653 UTUC patients (ranged from 81-1363) were included. The median or mean age of patients ranged from 62 to 71 years. The 31included articles were published from 2009 to 2019. Geographically, 20 studies were conducted in Asia, 7 in Muti-centers, 2 in USA, 1 in France and 1 in Serbia. All patients had received RNU as primary treatment for UTUC. Of this studies, 12 studies reported OS, 24 studies reported CSS, 14 studies reported RFS, 5 studies reported CSM and 6 studies reported recurrence. The characteristics of tumor features and pathologic outcomes are summarized in Table 2. LVI was detected in (3,635/14,653) 24.8% in pathological specimens of the included patients. According to the NOS, we assessed the quality of the 31 eligible studies[10-40]. The quality scores of the studies varied from 7 to 9, with a mean of 8.7, which means that all the studies were of high quality. (Supplementary Table S1)
The pooled result indicated that the presence of LVI in UTUC specimen was associated with poor CSS (RE model, HR=1.62, 95 % CI: 1.49-1.76, p < 0.001; I2 = 69.9%; Figure 2), OS (RE model, HR=1.55, 95 % CI: 1.41-1.71, p < 0.001; I2= 73.2%; Figure 3A), RFS ( RE model, HR=1.46, 95 % CI: 1.32-1.61, p < 0.001; I2= 78.6%; Figure 3B), CSM (RE model, HR=1.25, 95 % CI: 1.00-1.56, p = 0.047; p < 0.001; I2= 91.6%; Figure 3C) and recurrence (RE model, HR=1.23, 95 % CI: 1.03-1.48, p = 0.026; I2= 89%; Figure 3D). To explore the heterogeneity for CSS, OS, RFS, the prognostic value of LVI was evaluated further via subgroup analysis under the geographical region (Asia vs. non-Asian), year of publication (≥ 2014 vs. < 2014), TNM stage (T3+T4 %) (≥ 40 vs. < 40), tumor grade (G2+G3 %) (≥ 60 vs. < 60), No. of patients (≥ 500 vs. < 500) and median follow-up (≥40 months vs. < 40 months) (Table 3). Because of too few cohorts in the CSM and recurrence group, no subgroup analysis was conducted. The results in subgroup analyses were consistent with the primary findings, which suggested LVI still as a prognostic factor despite heterogeneity among some groups. Although no significant changes for the inter-study heterogeneity were detected, the observed heterogeneity dropped significantly in some subgroup models, such as No. of patients < 500, Grade (G3+G4 %) ≥ 60.
The risk estimate with pooled ORs were used to assess the associations between the LVI and clinicopathological parameters in UTUC patients. As shown in Table 4, LVI was significantly related to TNM stage (III/IV vs. I/II: OR = 7.63, 95% CI: 5.60–10.39, p < 0.001) (Supplementary Figure 1A), higher tumor grade (3 vs. 1/2: OR = 5.61, 95% CI: 3.71–8.48, p < 0.001) (Supplementary Figure 1B), lymph node metastasis(LNM) (yes vs. no: OR = 4.95, 95% CI: 3.66–6.71, p < 0.001) (Supplementary Figure 1C), CIS (yes vs. no: OR = 1.92, 95% CI: 1.36–2.70, p < 0.001) (Supplementary Figure 1D) and positive surgical margin(PSM) (yes vs. no: OR = 4.38, 95% CI: 2.71–7.07, p < 0.001) (Supplementary Figure 1E), but not related to gender (male vs. female: OR = 0.98, 95% CI: 0.80–1.19, p = 0.825) (Supplementary Figure 2A) and multifocality (multifocal vs. unifocal: OR = 1.10, 95% CI: 0.82–1.47, p = 0.539) (Supplementary Figure 2B) .No significant heterogeneity were observed in those groups.
In sensitivity analyses omitting enrolled studies in turn, the results showed that the pooled HRs did not alter significantly which suggested that the findings were reliable and robust. (Supplementary Figure 3).
We conducted the publication bias assessment of the studies by funnel plots and Egger’s test. As shown in Figure 4, No obvious asymmetry was founded in all groups. The P-values of the Egger’s test were all greater than 0.05 in CSS (p-Egger = 0.701, Figure 4A), OS (p-Egger = 0.330, Figure 4B), RFS (p-Egger = 0.811, Figure 4C), CSM (p-Egger = 0.984, Figure 4D) and recurrence (p-Egger = 0.843, Figure 4E).