Table 6 Basic modelling information of papers included in systematic review
From: Prediction of resistance to chemotherapy in ovarian cancer: a systematic review
Study | Prediction | Prediction method | Predictive ability |
|---|---|---|---|
Jeong et al. [22] | Overall Survival | Student’s T test, Hierarchical clustering, Compound covariate predictor algorithm, Cox proportional hazards regression, Kaplan-Meier curves, Log-rank test, ROC analysis | ‘Taxane-based treatment significantly affected OS for patients in the YA subgroup (3 year rate: 74.4% with taxane vs. 37.9% without taxane, p=0.005 by log-rank test)’, ‘estimated hazard ratio for death after taxane-based treatment in the YA subgroup was 0.5 (95% CI=0.31−−0.82,p=0.005)’ |
Lisowska et al. [23] | Chemoresponse, Disease-Free Survival, Overall Survival | Support vector machines, Kaplan-Meier curves, Log-rank test | No genes found to be significant in the training set were significant in the test set, for chemoresponse, DFS or OS |
Roque et al. [24] | Overall Survival | Kaplan-Meier curves, Log-rank test, Student’s T test | ‘OS was predicted by increased class III β-tubulin staining by both tumor (HR3.66, 96%CI=1.11–12.1, p=0.03) and stroma (HR4.53, 95%CI=1.28–16.1, p=0.02)’ |
Li et al. [3] | Chemoresponse (chemoresistant vs. chemosensitive) | Correlation of p-CFL1 staining and chemoresponse | ‘immunostaining of p-CFL1 was positive in 77.3% of chemosensitive and in 95.9% of the chemoresistant’ (p=0.014, U=157.5) |
Schwede et al. [25] | Stem cell-like subtype, Disease-Free Survival, Overall Survival | ISIS unsupervised bipartitioning, Diagonal linear discriminant analysis, Gaussian mixture modelling, Kaplan-Meier curves, Log-rank test | OS (p values): Dressman =0.0354, Crijns =0.021, Tothill =4.4E−7 |
Verhaak et al. [26] | Poor Prognosis vs. Good Prognosis | Significance analysis of microarrays, Single sample gene set enrichment analysis, Kaplan-Meier curves, Log-rank test | Good or Poor prognosis, likelihood ratio =44.63 |
Obermayr et al. [27] | Disease-Free Survival, Overall Survival | Kaplan-Meier curves, Cox proportional hazards regression, χ2 test | ‘The presence of CTCs six months after completion of the adjuvant chemotherapy indicated relapse within the following six months with 41% sensitivity, and relapse within the entire observation period with 22% sensitivity (85% specificity)’ |
Han et al. [28] | Complete Response or Progressive Disease | Supervised principal component method | 349 gene signature: ROC AUC =0.702, p=0.022. 18 gene: ROC AUC =0.614, p=0.197. |
Hsu et al. [29] | Progression-Dree Survival | Semi-supervised hierarchical clustering | Good Response vs. Poor Response, p=0.021 |
Lui et al. [30] | Chemosensitivity, Overall Survival, Progression-Dree Survival | Predictive score using weighted voting algorithm, Kaplan-Meier curves, Log-rank Test, Cox proportional hazards regression | Response of 26 of 35 patients in an independent data set was correctly predicted, patients in the low-scoring group exhibited poorer PFS (HR=0.43, p=0.04), ROC AUC = 0.90(0.86–0.95) |
Kang et al. [31] | Overall Survival, Progression-Free Survival, Recurrence-Free Survival | Kaplan-Meier curves, Log-rank test, Cox proportional hazards regression, Pearson correlation coefficient | Berchuck dataset: HR=0.33, 95%CI=0.13–0.86, p=0.013; Tothill dataset: HR=0.61, 95%CI=0.36–0.99, p=0.044 |
Gillet et al. [32] | Overall Survival, Progression-Free Survival | Supervised principle components method, Cox proportional hazards regression, Kaplan-Meier curves, Log-rank test | ‘An 11-gene signature whose measured expression significantly improves the power of the covariates to predict poor survival’(p<0.003) |
Ferriss et al. [33] | Overall Survival | COXEN coefficient, Mann-Whitney U test, ROC analysis, Unsupervised Hierarchical Clustering | Carboplatin: sensitivity = 0.906, specificity = 0.174, PPV = 60%, NPV = 57% (UVA-55 validation set) |
Brun et al. [34] | 2-year Disease-Free Survival | Student’s T test, Principal component analysis, Concordance index, Kaplen-Meier curves, Log-rank test | No genes were found to have prognostic value |
Skirnisdottir and Seidal [35] | Recurrence, Disease-Free Survival | χ2 test, Kaplan-Meier curves, Log-rank test, Logistic regression, Cox proportional hazards regression | p53-status (OR=4.123, p=0.009; HR=2.447, p=0.019) was a significant and independent factor for tumor recurrence and DFS. |
Brenne et al. [36] | OC or MM, Progression-Free Survival, Overall Survival | Mann-Whitney U test, Kaplan-Meier curves, Log-rank test, Cox proportional hazards regression | Cox Multivariate Analysis: EHF mRNA expression in pre-chemotherapy effusions was an independent predictor of PFS (p=0.033, relative risk=4.528) |
Sabatier et al. [37] | Progression-Free Survival, Overall Survival | Cox proportional hazards regression, Pearson’s coefficient correlation score | Favourable vs. Unfavourable: ‘sensitivity = 61.6%, specificity = 62.4%, OR=2.7, 95%CI=1.7–4.2; p=6.1×10−06, Fisher’s exact test’ |
Gillet et al. [38] | Overall Survival, Progression-Free Survival, Treatment Response | Linear regression, Hierarchical clustering, Kaplan-Meier curves, Log-rank test | ‘6 gene signature alone can effectively predict the progression-free survival of women with ovarian serous carcinoma (log-rank p=0.002)’ |
Chao et al. [39] | Chemoresistance | Interaction and expression networks for pathway identification, pathway intersections, betweenness and degree centrality, Student’s T test | No statistical measure available. Many genes identified have previously been found experimentally |
Schlumbrecht et al. [40] | Overall Survival, Recurrence-Free Survival | Linear regression, Logistic regression, Cox proportional hazards regression, Kaplan-Meier curves, Unsupervised cluster analysis, Log-rank test, Mann-Whitney U test, χ2 test | ‘Greater EIG121 expression was associated with shorter time to recurrence (HR=1.13 (CI=1.02–1.26), p=0.021)’, ‘Increased expression of EIG121 demonstrated a statistically significant association with worse OS (HR=1.21 (CI1.09–1.35), p<0.001)’ |
Glaysher et al. [41] | Chemosensitivity | AIC gene selection, Multiple linear regression | Cisplatin: \(R^{2}_{\textit {adj}} = 0.836\), p<0.001 |
Yan et al. [42] | Chemosensitivity | ANOVA, Student’s T test, Mann-Whitney U test | ‘Immunostaining scores [Annexin A3] are significantly higher in platinum-resistant tumors (p=0.035)’ |
Yoshihara et al. [43] | Progression-Free Survival | Cox proportional hazards regression, Ridge regression, Prognostic index, ROC analysis, Kaplan-Meier curves, Log-rank test | ‘Prognostic index was an independent prognostic factor for PFS time (HR=1.64, p=0.0001)’, sensitivity = 64.4%, specificity = 69.2% |
Williams et al. [44] | Overall Survival | COXEN score, Kaplan-Meier curves, Student’s T test, ROC analysis, Spearman’s rank correlation coefficient, Logistic regression, Log-rank test | Carboplatin and Taxol: sensitivity = 77%, specificity = 56%, PPV=71%, NPV=78% |
Denkert et al. [45] | Overall Survival | Semi-supervised analysis via Cox scoring, Principal components analysis, Kaplan-Meier curves, Log-rank test, Cox proportional hazards regression | Duke et al.: ‘clinical outcome is significantly different depending on the OPI (p=0.021), with an HR of 1.7 (CI 1.1–2.6)’ |
Matsumura et al. [46] | Taxane sensitivity, Overall Survival | Hierarchical clustering, Kaplan-Meier curves, Log-rank test | ‘Patients in the YY1-High cluster who were treated with paclitaxel showed improved survival compared with the other groups (p=0.010)’ |
Crijns et al. [47] | Overall Survival | Supervised principal components method, Cox proportional hazards regression, Kaplan-Meier curves, Log-rank test, χ2 test | OSP: (High-risk vs. low-risk) HR=1.940, CI=1.190–3.163, p=0.008 |
Mendiola et al. [48] | Progression-Free Survival, Overall Survival | Kaplan-Meier curves, Log-rank test, AIC-based model selection, ROC curves, Cox proportional hazards regression | OS: sensitivity = 87.2%, specificity = 86.4% |
Gevaert et al. [49] | Platin Resistance/Sensitivity, Stage | Principal component analysis, Least squares support vector machines | Platin-Resistance/Sensitivity: sensitivity = 67%, specificity = 40%, accuracy = 51.11% |
Bachvarov et al. [50] | Chemoresistance | Hierarchical Clustering, Support vector machines | No prediction metric applied |
Netinatsunthorn et al. [51] | Overall Survival, Recurrence-Free Survival | Kaplan-Meier curves, Cox proportional hazards regression | OS: HR=1.98, 95%CI=1.28–3.79, p=0.0138 ; RFS: HR=3.36, 95%CI=1.60–7.03, p=0.0017 |
De Smet et al. [52] | Stage I vs. Advanced stage, Platin-sensistive vs. Platin-resistant | Principal component analysis, Least squares support vector machines | Estimated Classification Accuracy: Stage I vs Advanced Stage =100%, Platin-sensitive vs. Platin-resistant =76.9% |
Helleman et al. [53] | Chemoresponse (responder vs. non-responder) | Class prediction, Hierarchical clustering, Principal component analysis | Test set: PPV=24%, NPV=97%, sensitivity =89%, specificity =59% |
Spentzos et al. [54] | Chemoresponse (pathological-CR or PD), Disease-Free survival, Overall Survival | Class prediction analysis, Compound covariate algorithm, Average linkage hierarchical clustering, Kaplan-Meier curves, Log-rank test, Cox proportional hazards regression | Cox PH (resistant vs. sensitive): Recurrence HR=2.7 (95%CI=1.2–6.1), Death HR=3.9 (95%CI=3.1–11.4) |
Jazaeri et al. [55] | Clinical response | Class prediction | 9 most significantly differentially expressed genes, primary chemoresistant vs. primary chemosensitive: accuracy =77.8% |
Raspollini et al. [56] | Overall Survival (high vs. low) | Univariate logistic regression, χ2 test | COX-2: OR=0.23, 95%CI=0.06–0.77, p=0.017; MDR1: OR=0.01, 95%CI=0.002–0.09, p=<0.0005 |
Hartmann et al. [57] | Time To Relapse (early vs.late) | Support vector machine, Kaplan-Meier curves, Log-rank test, average linkage clustering | Accuracy =86%, PPV=95%, NPV=67% |
Spentzos et al. [58] | Disease-Free Survival, Overall Survival | Supervised pattern recognition/class prediction, Kaplan-Meier curves, Log-rank test, Cox proportional hazards regression | Unfavourable vs. Favourable OS : (CPH) HR=4.6, 95%CI=2.0–10.7, p=0.0001 |
Selvanayagam et al. [59] | Chemoresistance (chemoresistant vs. chemosensitive) | Supervised voice-pattern recognition algorithm (clustering) | PPV=1, NPV=1 |
Iba et al. [60] | Chemoresponse, Overall Survival | Kaplan-Meier curves, Log-rank test, Cox propotionate hazards regression, ROC analysis, χ2 test, Student’s T test, Mann-Whitney U test | ‘Patients with c-myc expression of over 200 showed a significantly better 5-year survival rate (69.8% vs. 43.5%)’, p<0.05 |
Kamazawa et al. [61] | Chemoresponse (CR or PR vs. NC or PD) | Defined threshold expressionto divide responders and non-responders | MDR-1 (all samples): specificity =95%, sensitivity = 100%, predictive value =96% |
Vogt et al. [62] | Chemoresistance | Correlation of AUC from in-vitro ATP-CVA and gene expression | All p values for correlation of drugs and genes were >0.05 |