Study selection
A total of 1449 articles were collected from databases. After screening the full the articles, reasons for exclusion included manuscripts that focus on YKL-40 in the serum or plasma(n = 41), the study was systemic review(n = 1), insufficient data(n = 4), no survival data(n = 3). Finally, 30 articles that included 5160 participants were considered eligible in final meta-analysis. The flow chart of the study selection is shown in Fig. 1.
Characteristics And Quality Assessment Of The Included Studies
All cases of cancer were diagnosed via histopathology and included GBM, breast cancer, anal cancer, pancreatic cancer, renal cancer, epithelial ovarian carcinoma (EOC), etc. 28 studies included 4689 patients with OS data, and 11 studies included 2092 patients with DFS data. The study was carried out in Europe(21–34), America ༈18, 25, 35–38༉ and Asian༈39༉ ༈40–48༉ and its number was 14, 6 and 10, respectively. YKL-40 tissue expression was measured by immunohistochemistry in 23 studies ༈18, 22–24, 26–29, 31, 33–39, 41–46, 48༉and RT-PCR in 7 studies༈21, 25, 30, 32, 36, 47, 49༉, although the difference of cutoff values exist among these studies. 4 studies had a score of 9, 7 studies had a score of 8, 11 studies had a score of 7, and 7 studies had a score of 6. The details of assessment results and the relationship between YKL-40 and other clinical features were presented in the Table 1.
Table 1
Main characteristics of the eligible studies.
Study ID | Year | Country | Tumor | Number | Age | Sample | Method | YKL-40 associated with clinical features | Cutoff value | Percentage of positive or high expression | Survival | HR | NOS |
Arboix et al. (21) | 2008 | France | GBM | 42 | 62.5y | Frozen tissue | RT-PCR | NR | Mean | 50% | OS | R | 9 |
Pelloski et al. (18) | 2007 | USA | GBM | 509 | 55y | Tissue | IHC | NR | ≥ 10% stained cells | 60% | OS | R | 7 |
Salvati et al. (22) | 2012 | Italy | GBM | 105 | 58y | Tissue | IHC | NR | IRS ≥ 2 | 32% | OS | SC | 8 |
Batista1 et al. (23) | 2015 | Spain | GBM | 204 | 63y | Tissue | IHC | Subventricular | Median | 46% | OS | R | 6 |
Batista et al. (24) | 2016 | Spain | GBM | 152 | 65y | Tissue | IHC | Age and Karnofsky | Median | 41% | OS | R | 6 |
Steponaitis et al. (25) | 2016 | Lithuania | GBM | 98 | 50y | Frozen tissue | RT-PCR | Age and Pathological grade | Median | 25% | OS | R | 7 |
Castellano et al. (26) | 2009 | Italy | Anal cancer | 34 | NR | Tissue | IHC | Histological subtype | Median | 44% | OS, DFS | SC | 8 |
Mistrangelo et al. (27) | 2013 | Italy. | Anal cancer | 50 | 58.5y | Tissue | IHC | Sentinel lymph node and metastatic status | > 20% staining cell | 52% | OS, DFS | SC | 7 |
Yang et al. (39) | 2009 | China | EOC | 74 | 50.5y | Tissue | IHC | Clinical stage | IRS ≥ 4 | 58.3% | OS | SC | 8 |
Høgdall et al. (28) | 2009 | Denmark | EOC | 473 | 59y | Tissue | IHC | FIGO stage, histological type of cancer | ≥ 5% stained cells | 76% | OS | R | 8 |
Lawrenson et al. (36) | 2014 | USA | EOC | 105 | NR | Tissue | IHC | High tumor grade | Mean | 70% | OS | SC | 9 |
Chiang et al. (40) | 2015 | Taiwan | EOC | 180 | 53.8y | Frozen tissue | RT-PCR | Histological type and chemoresistance | Mean | 24% | OS | R | 6 |
Kim et al. (37) | 2007 | USA | BC | 109 | 52y | Tissue | IHC | Tumor size and differentiation | Median | 34% | DFS | SC | 6 |
Roslind et al. (29) | 2008 | Denmark | Primary BC | 630 | 47y | Tissue | IHC | High tumor differentiation | IRS ≥ 2 | 63% | OS, DFS | R | 8 |
Shao et al. (38) | 2010 | USA | BC | 79 | 56.5y | Tissue | IHC | Tumor grade | IRS ≥ 5 | 29% | OS | SC | 8 |
Kang et al. (41) | 2014 | Korea | BC | 425 | 45y | Tissue | IHC | Subtype, hormone receptor and molecular subtype | IRS ≥ 3 | 5% | OS,DFS | SC | 6 |
Vom Dorp et al. (30) | 2015 | Austria | Renal Cell Cancer | 101 | 65y | Frozen tissue | RT-PCR | Tumor stage | Median | 50% | OS | SC | 9 |
Table 1 (continued) |
Study ID | Year | Country | Tumor | Number | Age | Sample | Method | YKL-40 associated with clinical features | Cutoff value | Percentage of positive or high expression | Survival | HR | NOS |
Zhang et al. (42) | 2014 | China | Renal Cell Cancer | 73 | 53.9y | Tissue | IHC | Tumor size, TNM stage and metastasis | IRS > 3 | 13% | OS | SC | 6 |
Peng et al. (43) | 2010 | China | Endometrial Cancer | 68 | 52y | Tissue | IHC | NR | IRS > 3 | 38.2% | OS, DFS | SC | 7 |
Bi et al. (44) | 2009 | China | PGC | 172 | 58y | Tissue | IHC | Tumor invasion and tumor metastasis | IRS > 3 | 28.4% | OS | R | 9 |
Pelloski et al. (35) | 2005 | USA | GBM | 265 | 58y | Tissue | IHC | Resistance to radiation therapy | IRS ≥ 2 | 58% (Total section) 26%(subsection) | OS | SC | 7 |
Pan et al. (45) | 2013 | china | HC | 70 | NR | Tissue | IHC | Tumor size, invasion, stage and metastasis | IRS ≥ 3 | 61% | OS, DFS | R | 7 |
Harving et al. (31) | 2014 | Denmark | STS and LT | 49 | 58y | Tissue | IHC | Histological grade | > 20% staining cell | 43% | OS | SC | 7 |
Tschirdewahn et al. (32) | 2014 | Germany | UCB | 91 | NR | Frozen tissue | RT-PCR | YKL-40 concentration in serum | Median | 50% | OS, DFS | R | 7 |
Wang et al. (46) | 2014 | China | NSCLC | 95 | 60y | Tissue | IHC | Recurrence | IRS ≥ 4 | 56% | OS, DFS | SC | 6 |
Krogh et al. (33) | 2015 | Denmark | Melanoma | 204 | 51y | Tissue | IHC | Low Breslow thickness and Clark’s level | Mean | NR | OS | R | 7 |
Thorn et al. (34) | 2016 | Denmark | Osteosarcoma | 48 | 26y | Tissue | IHC | NR | > 50% staining cells | 17% | OS | SC | 6 |
Luo et al. (47) | 2016 | China | Thyroid carcinoma | 322 | 45y | Frozen tissue | RT-PCR | Tumor size, metastasis, invasion and stage | IRS ≥ 4 | 51.86% | DFS | R | 8 |
Chen et al. (48) | 2017 | China | Pancreatic Cancer | 234 | 59y | Tissue | IHC | Tumor size, stage, invasion, metastasis, and postoperative relapse | IRS > 3 | 63.7% | OS, PFS | R | 7 |
Cardona et al. (49) | 2019 | Colombia | Recurrent GBM | 59 | 43y | Frozen tissue | RT-PCR | CD133 mRNA | Mean | 11.9% | OS, PFS | SC | 7 |
Ykl-40 And Os
Overall survival data were reported or extrapolated from 28 of the studies(18, 21–36, 38–46, 48, 49), and 2 HRs were extracted from 1 study because of 2 cohorts in this study༈35༉. HR was merged by using a random-effects model and the forest plot was shown in Fig. 2. The pooled HR showed a clear association between high expression of YKL-40 and poor OS (pooled HR = 1.85 CI%=1.58–2.18;P < 0.001). Obvious heterogeneity was existed among these studies (p = 0.000; I2 = 74.8%).
Due to the high I2 values in the analysis, we then conducted the subgroup analyses following tumor type, tumor source, geographical location, detected method, number of the cases, HR acquisition method, and NOS score, the detailed data are summarized in Table 2. Of note, the results showed an obvious decrease in heterogeneity in the subgroup analysis of tumor type. The heterogeneity reduced to zero in patients with breast cancer [1.5(1.11–2.01);I2 = 0%; P = 0.39 ] and anal cancer [3.69(1.62–8.44)༛I2 = 0%; P = 0.985 ] (Fig. 3). In the two articles about renal cancer, the combined HR1.67 (95% CI: 0.62–4.48, P = 0.309) indicated that overexpression of YKL-40 would not precisely in predicting poor OS (Fig. 3). When coming to the tumor source subgroup analysis, the pooled HR 3.06(95%CI:2.10–4.44, P < 0.001)showed YKL-40 was a strong predictor of poor OS in patients with digestive system cancer. In the 3 articles about urogenital system cancer using OS to assess clinical outcome, the combined HR 1.31 (95% CI: 0.91–1.87, P = 0.144), implying the association between overexpression of YKL-40 and poor OS was not significant (Figures S1). Interestingly, the heterogeneity showed a remarkable change in the research region [European group: 1.36(1.17–1.57), I2 = 57.9%; P = 0.004; Asia group: 2.61(1.94–3.51), I2 = 41.2%; P = 0.093; America group: 2.15(1.74–2.66), I2 = 0%; P = 0.531] (Figures S2). However, the heterogeneity was not changed in patients with glioblastoma and epithelial ovarian carcinoma, or tumor source is nervous system and reproductive system (Table 2). A lower heterogeneity and prediction function of YKL-40 using RT-PCR method[1.37(1.02–1.848); I2 = 48.5%; P = 0.1] compare to immunohistochemistry (IHC) [1.85(1.58–2.18)༛I2 = 74.8%; P = 0.000 ]. Another interesting finding was lower heterogeneity in the higher NOS score group[1.53(1.31–1.79)༛I2 = 26.3%; P = 0.202] than the lower NOS score group (Table 2). The other subgroup analyses without statistically significant heterogeneity are according to HR acquisition method and number of cases.