|
Author | Country | Sample type | SCC stage | Targeted/untargeted method | Analytical platform | Statistical analysis/prediction model | STARD score | QUADAS | CAWG-MSI metabolite ID level | CAWG-MSI score | Sn (%) | Sp (%) | AUC |
Risk of bias | Applicability |
|
Studies of oesophageal squamous cell carcinoma |
Liu 2013 | China | Plasma | Late: 17 Control: 53 | Untargeted | UPLC-ESI-TOF-MS | PCA, hierarchical cluster analysis | 37 | Low | Low | 2 | 13 | — | — | — |
Wang L 2013 | China | Tissue | Early: 28 Late: 71 Control: 30 | Untargeted | 1H-NMR | OPLS-DA | 35 | Low | High | 2 | 7 | — | — | — |
Jin 2014 | China | Plasma | Early: 49 Late: 31 Control: 30 | Untargeted | GC-MS | Model of 3 compounds based on OPLS-DA model | 33 | Low | Low | 2 | 16 | 90 | 96.67 | 0.964 |
Ma 2014 | China | Plasma | Early: 51 Control: 60 | Targeted | HPLC | Student -test, PLS-DA | 32 | Low | Low | 2 | 10 | — | — | — |
Wang J 2016 | China | Plasma | Early: 28 Late: 30 Control: 105 | Untargeted | UHPLC-QTOF/MS | Model of 16 compounds based on random forest model | 37 | Low | Low | 1 | 19 | 85 | 90.5 | 0.929 |
Xu 2016 | China | Urine | Late: 40 Control: 62 | Untargeted | LC-MS/MS | Model of 7 compounds based on binary logistic regression and ROC curve | 25 | Unclear | Low | 2 | 19 | 90.2 | 96.0 | 0.961 |
Cheng 2017 | China | Plasma | Patient: 40 Control: 27 | Targeted | LC-MS/MS | Model of 4 compounds based on fivefold cross-validation test | 32 | Low | Low | 2 | 15 | 77.5 | 85.33 | 0.798 |
Zhang 2017 | China | Plasma | Early: 17 Late: 23 Control: 40 | Untargeted | 1H-NMR, UHPLC | Model of 9 compounds based on binary logistic regression and ROC curves | 33 | Low | Low | 2 | 14 | 97.4 | 95 | 0.988 |
|
Studies of lung squamous cell carcinoma |
Song 2010 | China | Breath | Early: 20 Late: 33 Control: 41 | Untargeted | SPME GC-MS | Wilcoxon rank sum test, ROC | 29 | Low | Low | 2 | 11 | — | — | — |
De Castro 2014 | Spain | Plasma | Patient: 30 Control: 35 | Targeted | GC-MS | Model of 1 compound based on ROC curves | 30 | Unclear | Low | 1 | 13 | 77 | 66 | 0.7 |
Handa 2014 | Germany | Breath | Early: 19 late: 31 Normal: 39 | Untargeted | IMS | Model of 11 compounds based on decision tree algorithm | 27 | Unclear | Low | 3 | 6 | 97.4 | 60 | — |
Rocha 2015 | Portugal | Tissue | Patient: 19 Control: 37 | Untargeted | 1H-NMR | PLS-DA, Wilcoxon rank sum test | 25 | Unclear | Low | 2 | 7 | — | — | — |
Sanchez-Rodriguez 2015 | Spain | Plasma | Late: 18 Control: 50 | Targeted | GC-MS | Model of 1 compound based on ROC curves | 31 | Low | Low | 1 | 17 | 69 | 68 | 0.68 |
|
Studies of head and neck squamous cell carcinoma |
Mizukawa 1998 | Japan | Saliva | Patient: 18 Control: 18 | Targeted | HPLC | Nil–peak detection only | 21 | Low | High | 1 | 7 | — | — | — |
Somashekar 2011 | USA | Tissue | Patient: 22 Control: 22 | Untargeted | HR-magic angle spinning proton NMR spectroscopy | PCA | 23 | Low | Low | 1 | 8 | — | — | — |
Wei 2011 | China | Saliva | Early: 21 Late: 16 Control: 66 | Untargeted | UPLC-QTOF-MS | Model of 5 compounds based on ROC curves | 31 | Low | Low | 3 | 13 | 86.5 | 82.4 | 0.89 |
Yonezawa 2013 | Japan | Tissue, plasma | Early: 7 Late: 10 Control: 22 | Untargeted | GC-MS | Student’s -test, Bartlett’s test, Wilcoxon rank sum test | 27 | Low | Low | 2 | 17 | — | — | — |
Gruber 2014 | Israel | Breath | Early: 9 Late: 11 Control: 40 | Untargeted | GC-MS, sensors | Model of 3 compounds based on discriminant factor analysis | 30 | Low | Low | 3 | 10 | 77 | 90 | 0.83 |
Wang Q (Clinica Chimica Acta) 2014 | China | Saliva | Early: 13 Late: 17 Control: 0 | Targeted | UPLC-MS | Model of 4 compounds based on ROC curves | 30 | Unclear | Low | 1 | 24 | 92.3 | 91.7 | — |
Wang Q (Scientific Reports) 2014 | China | Saliva | Early: 13 Late: 17 Control: 30 | Untargeted | RPLC-MS, HILIC-MS | Model of 5 compounds based on ROC curve | 24 | Unclear | Low | 1 | 16 | 100 | 96.7 | 0.997 |
Wang Q (Talanta) 2014 | China | Saliva | Early: 13 Late: 17 Control: 60 | Targeted | UPLC-ESI-MS | Model of 2 compounds based on logistic regression model | 25 | Low | Unclear | 1 | 25 | 92.3 | 91.7 | 0.871 |
Gupta 2015 | India | Plasma | Early: 28 Late: 72 Control: 175 | Untargeted | H-NMR | Model of 2 compounds based on OPLS-DA | 33 | Unclear | Low | 2 | 10 | 90 | 94 | 0.979 |
Szabo 2015 | Hungary | Breath | Cancer: 14 Control: 11 | Targeted | OralChroma and GC-MS | Nil–peak detection only | 22 | Unclear | Low | 1 | 8 | — | — | — |
Kekatpure 2016 | India | Urine | Early: 14 Late: 64 Control: 94 | Untargeted | LC-triple quadrupole-MS/MS | Kruskal-Wallis, Fisher exact test, Cox proportional hazards model | 23 | Low | High | 2 | 13 | — | — | — |
Mukherjee 2016 | USA | Tissue, saliva | Early: 2 Late: 5 Control: 7 | Untargeted | LC-MS, LC-MS/MS, GC-MS | Kruskal-Wallis with adjustment for multiple testing | 36 | Low | Low | 3 | 15 | — | — | — |
Shoffel-Havakuk 2016 | Israel | Saliva | Cancer: 6 Control: 4 | Untargeted | GC-MS | Mann–Whitney , Fisher exact test | 24 | Low | Low | 2 | 11 | — | — | — |
Bouza 2017 | Spain | Breath | Early: 11 Late: 15 Control: 26 | Untargeted | SPME, GC-MS | Kruskal-Wallis, Mann–Whitney, PLS-DA, SIMCA prediction | 25 | Unclear | Low | 2 | 10 | — | — | — |
Hartwig 2017 | Germany | Breath | Early: 5 Late: 5 Control: 4 | Untargeted | GC-MS | Jackknife/leave-one-out cross-validation | 34 | Unclear | Low | 3 | 6 | — | — | — |
Kamarajan 2017 | USA | Tissue, saliva, plasma | Early: 17 Late: 30 Control: 19 | Untargeted | UPLC-MS/MS, GC-MS | Anova, -test, random forest classification, PCA | 31 | Low | Low | 2 | 20 | — | — | — |
Ohshima 2017 | Japan | Saliva | Early: 14 Late: 8 Control: 21 | Untargeted | CE-TOF-MS | Hierarchical cluster analysis, Wilcoxon rank sum test | 37 | Low | Low | 3 | 9 | — | — | — |
|