Table 2.
Candidate (a priori) prognostic methylated tissue-based biomarker studies.
| Study [Ref] a | Approach (Discovery) | Cohort Size (Discovery) | Follow-Up (Discovery) [Years] | Approach (Validation) | Cohort Size (Validation) | Method (Validation) | Follow-Up (Validation) [Years] | Final Markers Identified (Validation) c,d,e,f | Key Results (Validation) g |
|---|---|---|---|---|---|---|---|---|---|
| Restriction-based methylation sequencing platforms h | |||||||||
| 1. Methylation sensitive arbitrarily primed PCR and methylated CpG island amplification | |||||||||
| Cottrell et al. [120] | Low GS (2–6 with no grade 4 or 5 patterns) vs. High GS (8–10) Early BCR (< 2 yrs post-RP) vs. no early BCR (> 4 yrs) |
n = 5 Low GS, n = 5 High GS, n = 5 no early BCR, n = 5 early BCR |
Range: 2–4 | Low GS vs. High GS Early BCR vs. no early BCR |
Cohort 1: n = 304 (130 low GS, 96 high GS; 88 no BCR, 63 BCR) Cohort 2: n = 233 (28 low GS, 27 high GS; 134 no BCR, 59 BCR) |
1. Custom methylation oligonucleotide microarray 2. MethyLight (qMSP) b |
Range: 2–4 | Low GS vs. High GS: Cohort 1—U: (1) ABDH9, (2) Chr3-EST, (3) GPR7, (4) NOTCH, (5) KBTBD6; M: not conducted Cohort 2—U: (1) ABDH9, (2) Chr3-EST; M: not conducted Early BCR vs. no early BCR: Cohort 1—U: (1) ABDH9, (2) ABDH9 (intermediate GS 6, 7 patients only) (3) Chr3-EST, (4) Chr3-EST (intermediate GS), (5) GPR7, (6) GPR7 (intermediate GS); M: not conducted Cohort 2—U: (1) ABDH9, (2) ABDH9 (intermediate GS), (3) Chr3-EST, (4) Chr3-EST (intermediate GS); M: (1) ABDH9 [ + GS + pathological T-stage + margin status], (2) Chr3-EST [ + GS + pathological T-stage + margin status] |
Low GS vs. High GS: Cohort 1—U: AUC (all Wilcoxon’s p < 0.001) (1) 0.71, (2) 0.70, (3) 0.72, (4) 0.71, (5) 0.71; M: NA Cohort 2—U: AUC (all Wilcoxon’s p < 0.001) (1) 0.77, (2) 0.79; M: NA Early BCR vs. no early BCR: Cohort 1—U: AUC (Wilcoxon’s) (1) 0.71 (p = 0.002), (2) 0.63 (p = 0.072), (3) 0.66 (p = 0.05), (4) 0.72 (p = 0.002), (5) 0.72 (p = 0.0002), (6) 0.70 (p = 0.005); M: NA Cohort 2—U: AUC (Wilcoxon’s) (1) 0.65 (p < 0.001), (2) 0.66 (p < 0.01), (3) 0.67 (p < 0.001), (4) 0.67 (p < 0.01); M: Logistic regression (1) p = 0.016; (2) p = 0.043, AUC = 0.81 & 0.79 (vs. 0.75 of GS + stage + margin status only) |
| 2. Enhanced Reduced Representation Bisulphite Sequencing | |||||||||
| Lin et al. [121] | Indolent (localised disease, no recurrence) vs. Advanced (aggressive CRPC) PCa |
n = 7 indolent, n = 6 advanced |
Indolent (range) 5–6 years |
Indolent vs. aggressive PCa |
n = 16 indolent, n = 8 advanced |
MassARRAY EpiTYPER | Indolent 3–7 years |
Panel of 13 CpG islands: KCNC2, ZDHHC1, TBX1, CAPG, RARRES2, GRASP, SAC3D1, TPM4, GSTP1, NKX2–1, FAM107A, SLC13A3, FILIP1L U: Panel M: Not conducted |
U: AUC = 0.975 (Sensitivity = 95%; Specificity = 95%) M: NA |
| Capture-based methylation sequencing platforms | |||||||||
| MBD (methyl-CpG binding domain)-isolated genome sequencing (MiGS) | |||||||||
| Bhasin et al. [122] | Low GS (6) vs. High GS (8–10) |
n = 6 Low GS, n = 9 High GS |
NA | Low GS vs. High GS |
n = 46 Low GS, n = 203 High GS (TCGA) |
HM450K Microarray | NA | U: 101 DMRs including at CD14, PCDHGA11, EYA1, CCDC8, HOXC4; M: not conducted | U: LIMMA p = 2.81 × 1028–0.05 (range) M: NA |
| Microarray-based platforms | |||||||||
| 1. Agilent Human CpG Island Microarray | |||||||||
| Kron et al. [123] | Low GS (6 (3 + 3)) vs. High GS (8 (4 + 4)) |
n = 10 Low GS, n = 10 High GS |
NA | 1. Low GS vs. High GS | n = 20 low GS vs. n = 19 high GS (MethyLight) | MethyLight | NA | U: HOXD3 (detected in n = 2 GS6 vs. n = 6 GS8); M: not conducted |
Sample size too low for statistical U: 17.3% difference in methylation M: NA |
| 2. Kron et al. [124] GS ≤ 6 vs. GS7 BCR |
n = 232 (n = 101 GS ≤ 6, n = 107 GS 7, n = 147 no BCR, n = 85 BCR) |
MethyLight | Mean (range): 4.4 (0.2–9.5) | GS ≤ 6 vs. GS7-U: HOXD3; M: not conducted BCR-U: HOXD3; M: HOXD3 x pathological T-stage interaction term [+ GS + pathological T-stage + margin status] |
GS ≤ 6 vs. GS7-U: 10.1% difference in av. PMR values, Mann-Whitney U test p < 0.001; M: NA BCR-U: Log-rank p = 0.043; M: HOXD3 x pT3a [HR 3.78 (1.09–13.17), p = 0.037], HOXD3 x pT3b/pT4 [HR 5.23 (1.31–20.96), p = 0.019] |
||||
| 3. Liu et al. [125] GS≤6 vs. GS7 BCR |
n = 219 (n = 138 GS ≤6, n = 98 GS 7, n = ns BCR) (reduced cohort from Kron et al. [124]) |
MethyLight | ns | GS ≤ 6 vs. GS7-U: (1) APC, (2) TGFβ2; M: not conducted BCR-U: (1) APC, (2) HOXD3 + TGFβ2 + APC; M: (1) HOXD3 + TGFβ2 + APC [+ pathological T-stage + GS (7 and ≥8 groups)], (2) HOXD3 + TGFβ2 + APC [+ pathological T-stage + GS (7 (3 + 4), 7 (4 + 3) and ≥ 8 groups)] |
GS ≤ 6 vs. GS7-U: Mann-Whitney U test (1) p = 0.018, (2) p = 0.028; M: NA BCR-U: Log-rank (1) p = 0.028, (2) p < 0.001; M: (1) HR 2.01 (1.14–3.57), p = 0.017, (2) HR 2.068 (1.155–3.704), p = 0.014 |
||||
| 4. Jeyapala et al. [126] BCR |
Cohort 1: n = 435, n = 43 BCR |
Cohort 1: HM450K Microarray (TCGA) | Mean (range): Cohort 1: 1.9 (0–12.6) |
Cohort 1—U: GBX2; M: not conducted |
Cohort 1—U: Mann-Whitney Test cg09094860 [p = 0.003], cg00302494 [p = 0.01]; M: not conducted |
||||
| Cohort 2: n = 254 (n = 202, n = 52 BCR, n = 58 IDC/C-positive, n = 196 IDC/C-negative) |
Cohort 2: MethyLight | Cohort 2: 5.7 (0.1–12.3) | Cohort 2—U: GBX2 (and in IDC/C-negative patients only); M: (1) GBX2 [+ GS + Pathological T-stage + pre-op PSA], 2) GBX2 [+ pre-op PSA] |
Cohort 2—U: Mann-Whitney Test p = 0.0001, IDC/C-negative patients: Log-rank p = 0.002; M: (1) HR 1.02 (1.006–1.034), p = 0.004, (2) C-index 0.78 (vs. 0.71 for PSA alone) |
|||||
| 5. Jeyapala et al. [127] BCR salvage RT/hormone therapy |
Cohort 1: n = 254 (n = 202 no BCR, n = 52 BCR, n = 205 no salvage RT, n = 42 salvage RT, n = 226 no hormone therapy, n = 21 hormone therapy) Cohort 2: n = 199 (n = 159 no BCR, n = 40 BCR, n = 180 no salvage RT, n = 19 salvage RT, n = 177 no hormone therapy, n = 22 hormone therapy) |
MethyLight | Median (range): Cohort 1: 6.7 (0.1–12.8); Cohort 2: 6.7 (0.2–18.6) |
4-G model: HOXD3, TGFβ2, CRIP3, APC (candidate) Cohort 1 BCR-U: 4-G model; M: Integrative model = 4-G model [+ CAPRA-S] Salvage RT/hormone therapy: U: 4-G model; M: Integrative model Cohort 2 BCR-U: 4-G model; M: Integrative model Salvage RT/hormone therapy: U: 4-G model; M: Integrative model |
Cohort 1 BCR-U: HR 2.72 (1.77–4.17), p < 0.001, Sensitivity = 90.9%, Specificity = 35.2%, AUC = 0.740; M: HR 1.49 (1.12–1.99), p = 0.006, Sensitivity = 92.9%, Specificity = 43.4%, AUC = 0.846 Salvage RT/hormone therapy: U: HR 2.20 (1.48–3.29), p < 0.001;M: HR 1.34 (1.03–1.75), p = 0.027 Cohort 2 BCR-U: HR 2.48 (1.59–3.86), p < 0.001, Sensitivity = 95.0%, Specificity = 27.5%, AUC = 0.670; M: HR 1.62 (1.17–2.24), p = 0.004, Sensitivity = 89.5%, Specificity = 37.3%, AUC = 0.726 (vs. 0.698 for CAPRA-S alone) Salvage RT/hormone therapy-U: HR 1.97 (1.21–3.21), p < 0.001, Sensitivity = 91.2%, Specificity = 27.4%, AUC = 0.636; M: HR 1.17 (0.79–1.72), p = 0.441, AUC = 0.731 (vs. 0.723 for CAPRA-S alone) |
||||
| 6. Savio et al. [128] BCR; Late BCR (5 and 7 yrs post-RP) salvage RT/hormone therapy [Biopsy specimens pre-RP] |
n = 86 (n = 61 no BCR, n = 25 BCR, n = 75 no salvage RT, n = 11 salvage RT, n = 70 no hormone therapy, n = 15 hormone therapy) |
MethyLight | Median (range): 5.1 (0.1–16) |
BCR-U: none; M: 4-G model [+ pre-op PSA] Late BCR-U: none; M: (1) 4-G model [+ pre-op PSA] (5 yrs), (2) 4-G model [+ pre-op PSA] (7 yrs) Salvage RT/hormone therapy: U: 4-G model; M: (1) 4-G model [+ pre-op PSA], (2) 4-G model [+ CAPRA] |
BCR: U: non-sig; M: Sensitivity = 78.6%, Specificity = 64.7%, AUC = 0.714 Late BCR: U: non-sig; M: (1) Sensitivity = 80%, Specificity = 60.5%, AUC = 0.705 (vs. 0.667 for PSA alone), (2) Sensitivity = 76.9%, Specificity = 62.9%, AUC = 0.688 (vs. 0.6 for PSA alone) Salvage RT/hormone therapy-U: Sensitivity = 66.7%, Specificity = 75%, AUC = 0.699; M: (1) Sensitivity = 75%, Specificity = 61.1%, AUC = 0.699, (2) Sensitivity = 76.9%, Specificity = 58.3%, AUC = 0.797 |
||||
| 2. Illumina GoldenGate Methylation Microarrays | |||||||||
| Goh et al. [129] | Low GS (6) vs. High GS (8–10) Overall survival |
n = 87 (n = 19 GS6, n = 48 GS8–10, n = ns death) |
Median (range): 4 (0–11.8) |
GS (6–9) BCR |
n = 59 (n = 23 GS 6, n = 22 GS 7, n = 13 GS 8–10, n = 18 for BCR analysis) |
GoldenGate | No BCR-5 (1–13) |
“PHYMA” signature: 55 probes targeting CpG loci of 46 genes, including at ALOX12, PDGFRB which were selected for functional validation GS: U: PHYMA (GS 6–8) M: not conducted BCR: U: none M: not conducted |
GS: U: Logistic regression β-coefficient = 2.28, p = 0.2 (trend) M: NA BCR: U: non-sig M: NA |
| Angulo et al. 2016 Urol Int [130] | • BCR • PCa death |
n = 26 no BCR, n = 32 BCR |
Mean ± SD (range): 6.3 ± 3 (0.8–13.8) |
No validation | NA | NA | NA | Discovery only: BCR: U: (1) Gene hypermethylation profile of cluster 3 patients (including at GSTM2, GSTP1, RARB, ALOX12, APC, PDGFRB, SCGB3A1, CFTR, MT1A, PENK, NEU1, CCNA1, MET, KLK10, RARA, MFAP4, TERT, TBX1, TAL, MYCL2, (2) MT1A, (3) ALOX12, (4) GSTM2, (5) APC, (6) MYCL2, (7) RARB, (8) GSTM2 + MCLY2 M: (1) Gene hypermethylation profile [+ D’Amico classification]; (2) GSTM2 [+ ns]; (3) MCLY2 [+ ns] PCa Death: U: GSTM2 + MYCL2 M: not conducted |
Discovery only: BCR-U: (1) Log-rank p = 0.0054, Cluster 3 vs. 1 [HR 8.4 (1.86–38.46), 3 vs. 2 [HR 2.69 (1.13–5.95), 3 vs. 4 [HR 2.26 (0.89–5.72)]; (2) HR 2.14 (1.06–4.33), log-rank p = 0.029; (3) HR 2.21 (1.06–4.55), log-rank p = 0.025; (4) HR 4.59 (1.38–15.15), log-rank p = 0.0062; (5) HR 1.96 (0.97–3.97), log-rank p = 0.05; (6) HR 3.58 (1.6–8), log-rank p = 0.0009; (7) HR 2.5 (1.21–5.18), log-rank p = 0.01; (8) log-rank p = 0.0009; M: (1) Cox regression p = 0.064, Cluster 3 vs. 1 [HR 4.37 (0.94–20.41], 3 vs. 2 [HR 2.56 (1.11–5.88)], 3 vs. 4 [HR 2.26 (0.89–5.72)], C-index = 0.708 (vs. 0.679 for D’Amico alone); (2) HR 3.789 (1.11–12.83), p = 0.03;(3) HR 2.71 (1.21–6.09), p = 0.016 PCa death-U: HR 10.82 (1.96–59.67); log-rank p = 0.006; M: NA |
| 3. Illumina Infinium HumanMethylation 27K Microarray | |||||||||
| Kobayashi et al. [131] | BCR | n = 86 | Range: 0–5.5 |
No validation | NA | NA | NA | Discovery only: U: KCNK4, WDR86, OAS2, TMEM179 (FDR ≤ 1%; hypermethylated) M: not conducted |
Discovery only: U: ns M: |
| Mahapatra et al. [132] | Indolent vs. aggressive disease: No recurrence vs. recurrence BCR vs. clinical recurrence Local recurrence vs. metastatic relapse |
n = 75 no recurrence, n = 123 recurrence n = 43 BCR, n = 80 clinical recurrence n = 44 local recurrence, n = 36 metastatic relapse |
(Mean ± SD): No recurrence 6.2 ± 1.5 BCR: 5.9 ± 1.4 Local recurrence: 4.2 ± 1.7 Metastatic relapse: 4.4 ± 4.0 |
Indolent vs. aggressive disease: No recurrence vs. recurrence BCR vs. clinical recurrence Local recurrence vs. metastatic relapse |
n = 20 no recurrence, n = 20 BCR, n = 20 local recurrence, n = 20 metastatic relapse |
Pyro sequencing |
(Mean ± SD) No recurrence-6.5 ± 1.6 BCR-5.45 ± 1.3 Local recurrence-4.5 ± 1.9 Metastatic relapse-3.5 ± 1.9 |
No recurrence vs. recurrence: U: (1) CRIP1; (2) RUNX3; (3) HS3ST2; (4) FLNC; (5) RASGRF2 M: Not conducted BCR vs. clinical recurrence: U: (1) PHILDA3; (2) TNFRSF10D; (3) RASGRF2 M: Not conducted Local vs. metastatic relapse: U: (1) BCL11B; (2) POU3F3; (3) RASGRF2 M: Not conducted |
No recurrence vs. recurrence: U: Sensitivity/Specificity, AUC, t-test (1) 65.0%/65.6%, 0.727, p = 0.0139; (2) 70.4%/75.3%, 0.788, p = 0.0018; (3) 65.0%/60.0%, 0.773, p = 0.0115; (4) 70.3%/60.4%, 0.660, p = 0.0835; (5) 75.7%/55.2%, 0.682, p = 0.0515 M: NA BCR vs. clinical recurrence: U: Sensitivity/Specificity, AUC, t-test (1) 65.6%/65.0%, 0.73, p = 0.0129; (2) 60.8%/75.6%, 0.692, p = 0.0373; (3) 75.4%/60.3%, 0.761, p = 0.0047 M: NA Local recurrence vs. metastatic relapse: U: Sensitivity/Specificity, AUC, t-test (1) 75.2%/60.0%, 0.741, p = 0.0091; (2) 65.5%/70.7%, 0.701, p = 0.0295; (3) 70.6%/75.4%, 0.748, p = 0.0071; M: NA |
| 4. Illumina Infinium HumanMethylation 450K (HM450K) Microarray | |||||||||
| Geybels et al. [133] | Low GS (≤6) vs. High GS (8–10) | n = 65 Low GS, n = 88 High GS (TCGA) | NA | Progression |
n = 323 no progression, n = 108 progression |
HM450K Microarray | Mean (SD): 8.0 (4.2) years | Signature consisting of 52 CpG sites (32 unique genes) including at RRM2, VWA3B, MFSD9, ANO7, GALNTL2, SEMA3F, ATXN7, SLC15A2, MME, USP17, KIAA0922, FOXI1, URGCP, PTPRN2, RP1, MRPS28, MKI67, CPT1A, KCNMB4, TMEM132D. U: (1) Signature; (2) Signature (GS7 only); (3) Signature (GS7 (3 + 4) only) M: [+ GS+ pathological T-stage + pre-op PSA] (1) Signature; (2) Signature (GS7 only); (3) Signature (GS7 (3 + 4) only) |
U: (1) HR 1.78 (per 25% increase) (1.48–2.16), p = 2.05 × 10−9; (2) HR 1.81 (1.42–2.31), p = 1.38 × 10−3; (3) HR 1.83 (1.36–2.45), p = 5.64 × 10−5 M: (1) HR 1.48 (1.21–1.81), p = 1.38 × 10–4, AUC = 0.78 (vs. 0.73 for GS + pathological T-stage + pre-op PSA only); (2) HR 1.59 (1.24–2.05), p = 1.38 × 10–6, AUC = 0.76 (vs. 0.64); (3) HR 1.65 (121–2.25), p = 1.54 × 10−3, AUC = 0.70 (vs. 0.62) |
| Zhao et al. [134] | Metastatic-lethal progression |
n = 304 no progression vs. n = 24 metastatic-lethal |
Mean: Metastatic relapse-8.1 Survival-12.2 |
1. Metastatic-lethal progression |
n = 41 no progression, n = 24 metastatic-lethal |
HM450K Microarray | Mean: 9 | U: (1) ALKBH5; (2) ATP11A; (3) FHAD1; (4) KLHL8; (5) PI15; (6) Intergenic region (chr1); (7) Intergenic (chr16); (8) Intergenic (chr17) M: [ + GS] (1) ALKBH5; (2) FHAD1; (3) KLHL8; (4) PI15 |
U: Mean β difference (t-test), AUC, pAUC (1)—5% (p = 0.037), 0.66 (p = 0.035), 0.001 (p = 0.566); (2)—6% (p = 0.049), 0.66 (p = 0.03), 0.009 (p = 0.022); (3)—6% (p = 0.007), 0.71 (p = 0.003), 0.004 (p = 0.159); (4)—10% (p = 0.002), 0.75 (p = 0.0004), 0.002 (p = 0.359); (5)—7% (p = 0.029), 0.68 (p = 0.014), 0.006 (p = 0.074); M: AUC, pAUC, p (from likelihood ratio test comparing with model with GS alone: AUC = 0.816, pAUC = 0.010) (1) 0.87, 0.024, p = 0.030; (2) 0.86, 0.013, p = 0.038; (3) 0.89, 0.008, p = 0.014; (4) 0.89, 0.006, p = 0.026 |
| 2. Zhao et al. [135] Metastatic-lethal progression |
Training dataset (from Zhao et al. [134]): n = 344 no recurrence, n = 48 metastatic-lethal; Testing dataset: n = 11 no recurrence, 23 metastatic-lethal |
Pyro sequencing |
Training (Mean (minimum))—8 (5) Testing At least 5 years |
Methylation score: ALKBH5 + ATP11A + FHAD1 + KLHL8 + PI15 [+ GS] Training-U: not conducted; M: Methylation score Testing-U: not conducted; M: Methylation score |
Training-U: NA; M: Logistic regression β-coefficient-ALKBH5 [−0.75], ATP11A [−0.7], FHAD1 [−9.72], KLHL8 [−0.33], PI15 [0.70], GS [1.13] Testing-U: NA; M: Mean difference = 2.49 (p = 6.8 × 10–6), OR 4.0 (1.8–14.3), p = 0.006, AUC = 0.91 (vs. 0.87 for GS alone), pAUC = 0.037 (vs. 0.025 for GS alone), sensitivity at 95% specificity= 74% (vs. 53% for GS alone) |
||||
| Mundbjerg et al. [136] | Aggressiveness (individual PCa foci vs. matched lymph node metastasis) | n = 14 (n = 92 samples: multiple tumour foci, adjacent normal tissue, lymph node metastases and normal lymph nodes) | NA | Aggressive (lymph node metastases and pathological stage T3 tumours) vs. non-aggressive | n = 351 (TCGA) | HM450K Microarray | Mean: 3.2 years | Aggressiveness classifier: 25 probes, including in NXPH2, NCAPH, TRIB1, PCDHA1-PCDHA8, C3orf37, C9orf3, CPN1, TCF7L2, ROBO1, GFPT2, FBXQ47, SKI, HDAC9, CARS, SLC6A17, BCAT1, GAS1, RAI1 U: Aggressiveness classifier M: Not conducted |
U: Specificity = 97.4%, Sensitivity = 96.2%, Negative predictive value = 76%, Positive predictive value = 99.7%, Lymph node metastases [Fishers exact test p = 9.2 × 10−5], pathological stage T3 [Fishers exact test p = 2.2 × 10−7] M: NA |
| Toth et al. [137] | Good prognosis vs. Poor prognosis |
n = 35 good prognosis, n = 35 poor prognosis (80% training set, 20% testing set) |
Range: 3–5 | Good prognosis vs. poor prognosis (Cohort 1, 2) BCR (Cohort 1, 2, 3) |
Cohort 1: n = 222 (n = 63 for prognosis analysis, n = ns BCR) (ICGC cohort [138]) Cohort 2: n = 477 (n = 27 good prognosis, 57 poor prognosis, n = ns BCR) (TCGA) Cohort 3: n = 12,581 (n = 3612 BCR) (for ZIC2 immunostaining analysis only) |
HM450K Microarray | Cohort 1 & 2 At least 5 years Cohort 3 (Median (range)) 4 (0.08–20.08) |
Signature consisting of 598 CpG sites. Top 20: CCT8L2, NOP56, FCRL1, OR5W2, ZFP36L2, PRMT8, SLC1A6, DOK5, CCT8L2, ZFP36L2, MMP16, ESR1, ZIC2, GPR137B, NANOS1, LCE3A, C11orf87, PEG3, ZIM2, CTSC Good prognosis vs. good prognosis: Cohort 1—U: (1) Signature, M: not conducted Cohort 2—U: Signature; M: not conducted BCR: Cohort 1—U: Signature, M: not conducted Cohort 2—U: Signature; M: Signature [+ GS] Cohort 3—U: ZIC2 protein; M: ZIC2 protein [+ GS + pathological T-stage + nodal stage + PSA] |
Good vs. poor prognosis: Cohort 1—U: AUC = 0.997; M: NA Cohort 2—U: AUC = 0.775; M: NA BCR: Cohort 1—U: Log-rank p < 0.0001, M: NA Cohort 2—U: Log-rank p < 0.0001; M: Cox regression p = 0.011 Cohort 3—U: Log-rank p < 0.0001; M: ns |
Abbreviations: AUC = area under the curve; av. = average, CAPRA-S = Cancer of the Prostate Risk Assessment Score; CRPC = Castration-Resistant Prostate Cancer; GS = Gleason Score; HR = hazard ratio; IDC/C = Intraductal Carcinoma and Cribriform Architecture; M = multivariate analysis; NA = not applicable; non-sig = nonsignificant; ns = not specified; OR = odds ratio; pAUC = partial area under the cure; PCa = prostate cancer; PMR = percent methylated ratio; PSA = prostate- specific antigen; RT = radiotherapy; U = univariate analysis. Definitions: BCR: Biochemical recurrence: PSA elevations ≥ 0.2n g/mL post-RP, except [130] > 0.4 ng/mL and [131] > 0.07 ng/mL; clinical recurrence = local recurrence or metastatic relapse; good prognosis: organ-confined disease (pT2) and lack of BCR for at least 5 years; local recurrence: cancer observed on prostatic bed, confirmed by histological analysis of biopsies; metastatic relapse: metastatic deposits (visceral, bony metastasis) confirmed by positive biopsies or cT/bone scans; metastatic-lethal progression = metastatic relapse or PCa death; pathological T-stage: tumour staging based on pathological examination of surgically removed prostate tissue; PCa death: prostate cancer-specific death; poor prognosis: systemic presence of metastatic disease, indicated by recurrence within 3 years and no response to local radiation therapy; progression: either of BCR, metastatic relapse or PCa death; recurrence: either of BCR, local recurrence or metastatic relapse. a All studies are on prostate cancer tissues from radical prostatectomy, unless specified. b MethylLight is a quantitative methylation-specific PCR (qMSP) platform. c Univariate (U) or Multivariate (M) analyses. d Plus (+) sign indicates variables in the same multivariate model or methylation score together. e Square bracket ([]) indicate the clinicopathological factors adjusted for in each multivariate model f The use of bracketed numbers; e.g., (1), (2), indicates different genes, sets of genes or multivariate models validated in the respective study. g Number in brackets following HR or OR indicate the 95% confidence interval. h Bolded headings within the table subdivide the different types of genome-wide platforms (restriction-based, capture-based or microarray-based) used in these studies.