Skip to main content
. 2021 Mar 12;2021(3):CD013316. doi: 10.1002/14651858.CD013316.pub2
    CpGs analysed (PCR‐based tests) Threshold for methylated D4: PF D4 (justification) D5: attrition D5 (justification) D6: other PFs D6 (justification) D7: sel. rep. D7 (justification)
Almuqate 2018 MS‐RE‐qPCR NR > 5% Unclear Cut‐off may have been based on performance. Unclear Insufficient information Low RoB No concerns Unclear Conference abstract – little information reported.
MS‐RE‐qPCR NR > 9% Low RoB No concerns Unclear Insufficient information Low RoB No concerns Unclear Conference abstract – little information reported.
Bady 2012 (E‐GBM) Bead array 31 and 83 > 0.358 Low RoB No concerns Low RoB No missing data N/A — Low RoB No concerns
Bead array 78–84 > 10% High RoB Threshold derived from outcome measurement. Low RoB No missing data N/A — Low RoB No concerns
PSQ 74–78 > 7.28% Low RoB Cut‐off does not seem to be determined by performance. Low RoB Only 3/50 missing. N/A — Low RoB No concerns
Bady 2012 (M‐GBM) Bead array 31 and 83 > 0.358 High RoB Threshold derived from outcome measurement. Low RoB No missing data Low RoB No concerns Low RoB No concerns
MSP 76–80 and 84–86 NR Low RoB No concerns Low RoB No missing data Low RoB No concerns Low RoB No concerns
Barault 2015 Methyl‐beaming 79–83 > 40.2% Low RoB No concerns Low RoB No missing data N/A — Unclear Unclear why there was no result for MSP for this cohort of people (MSP investigated in other cohorts of people in this study).
PSQ 76–81 > 29.6% Low RoB No concerns Unclear Missing data for 11/69. Unclear whether there were important differences between those included in the study and those who were not. N/A — Unclear Unclear why there is no result for MSP for this cohort of people (MSP investigated in other cohorts of people in this study).
Barbagallo 2014 MSP 76–80 and 84–87 Including weakly Low RoB No concerns Low RoB No missing data Low RoB No concerns Low RoB No concerns
MSP 76–80 and 84–87 Excluding weakly Low RoB No concerns Low RoB No missing data Low RoB No concerns Low RoB No concerns
PSQ NR > 9% Low RoB No concerns Low RoB No missing data Low RoB No concerns Low RoB No concerns
PSQ NR > 25% Low RoB No concerns Low RoB No missing data Low RoB No concerns Low RoB No concerns
Bell 2017 QF‐IHC (AQUA) N/A > median Unclear — Low RoB No concerns Low RoB No concerns Unclear Multiple HRs are reported for different overlapping subgroups of this population.
qMSP NR > 8 Unclear MSP was performed centrally in original RCT. Unclear Of the 833 people from the original RCT, 452 had available tissue and underwent the microarray of 22 proteins. From these, 320 had data for MSP. The study authors did compare OS between the 452 microarray samples and the remainder of the 833 from the RCT cohort that were not included in this secondary analysis – no significant difference. However, there does not appear to be any further examination of missing data. Low RoB No concerns Unclear Multiple HRs are reported for different overlapping subgroups of this population.
Brigliadori 2016 PSQ 74–83 > 9% Low RoB No concerns Low RoB No concerns N/A — Low RoB No concerns
PSQ 74–83 > 29% Low RoB No concerns Low RoB No concerns N/A — Low RoB No concerns
Chai 2018 (7‐site cohort) PSQ 72–78 > 12% High RoB Cut‐offs appeared to have been selected based on performance (ROC curve analysis, sensitivity, specificity). Low RoB people had to have information on methylation status and OS to be included, and therefore there is no missing data. The bias that selecting on this basis has already been covered in domain 1. N/A — Unclear The whole set of CpGs analysed was correlated with OS. In addition a subset that is part of a Qiagen kit was tested, and a third combination. Unclear why results for other combinations not presented.
PSQ 74–78 > 12% High RoB Cut‐offs appeared to have been selected based on performance (ROC curve analysis, sensitivity, specificity). Low RoB People had to have information on methylation status and OS to be included, and therefore there were no missing data. The bias that selecting on this basis has already been covered in domain 1. N/A — Unclear The whole set of CpGs analysed was correlated with OS. In addition a subset that is part of a Qiagen kit was tested, and a third combination. Unclear why results for other combinations not presented.
PSQ 75–78 > 12% High RoB Cut‐offs appeared to have been selected based on performance (ROC curve analysis, sensitivity, specificity). Low RoB Participants had to have information on methylation status and OS to be included, and therefore there is no missing data. The bias that selecting on this basis has already been covered in domain 1. N/A — Unclear The whole set of CpGs analysed was correlated with OS. In addition a subset that is part of a Qiagen kit was tested, and a third combination. Unclear why results for other combinations not presented.
Chai 2018 (8‐site cohort) PSQ 75–78 > 13% High RoB Cut‐offs appeared to have been selected based on performance (ROC curve analysis, sensitivity, specificity). Low RoB People had to have information on methylation status and OS to be included, and therefore there were no missing data. The bias that selecting on this basis has already been covered in domain 1. N/A — Unclear The whole set of CpGs analysed was correlated with OS. In addition a subset that is part of a Qiagen kit was tested, and a third combination. Unclear why results for other combinations not presented.
PSQ 75–82 > 12% High RoB Cut‐offs appeared to have been selected based on performance (ROC curve analysis, sensitivity, specificity). Low RoB People had to have information on methylation status and OS to be included, and therefore there were no missing data. The bias that selecting on this basis has already been covered in domain 1. N/A — Unclear The whole set of CpGs analysed was correlated with OS. In addition a subset that is part of a Qiagen kit was tested, and a third combination. Unclear why results for other combinations not presented.
PSQ 76–79 > 11% High RoB Cut‐offs appeared to have been selected based on performance (ROC curve analysis, sensitivity, specificity). Low RoB People had to have information on methylation status and OS to be included, and therefore there were no missing data. The bias that selecting on this basis has already been covered in domain 1. N/A — Unclear The whole set of CpGs analysed was correlated with OS. In addition a subset that is part of a Qiagen kit was tested, and a third combination. Unclear why results for other combinations not presented.
Dahlrot 2018 (NS cohort) DIF N/A < 0.2 Low RoB Although no information was provided about blindness of analysis, all nuclear identification was performed automatically so there is reason to assume that the measurements ere objectives. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 9% Low RoB No concerns Low RoB No concerns N/A — Low RoB No concerns
Dahlrot 2018 (RSD cohort) DIF N/A < 0.2 Low RoB Although no information was provided about blindness of analysis, all nuclear identification was performed automatically so there is reason to assume that the measurements ere objectives. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 10% Low RoB No concerns Low RoB No concerns N/A — Low RoB No concerns
Dunn 2009 PSQ 72–83 > 9% Low RoB Cut‐off may or may not have been prespecified, but it was not data driven (i.e. not based on ROC curve analysis). Low RoB Missing data for 6/115 people treated with chemoradiation during the study period: four had histology elsewhere and two had inadequate tissue. Median OS for the complete cohort of 115 people was 12.8 months vs 12.4 months in the 109 included people. Low RoB No concerns Low RoB No concerns
PSQ 72–83 > 20% Unclear Methylated cases were ranked according to methylation and divided into 3 groups. This does not seem to have been a prespecified analysis, but not data driven. Low RoB Missing data for 6/115 people treated with chemoradiation during the study period: 4 had histology elsewhere and 2 had inadequate tissue. Median OS for the complete cohort of 115 people was 12.8 months vs 12.4 months in the 109 included people. Low RoB No concerns Low RoB No concerns
PSQ 72–83 > 29% High RoB ROC analysis used to separate cases into 2 prognostic groups. Low RoB Missing data for 6/115 people treated with chemoradiation during the study period: 4 had histology elsewhere and 2 had inadequate tissue. Median OS for the complete cohort of 115 people was 12.8 months vs 12.4 months in the 109 included people. Low RoB No concerns Low RoB No concerns
PSQ 72–83 > 35% Unclear Methylated cases were ranked according to methylation and divided into 3 groups. This does not seem to have been a prespecified analysis, but not data driven. Low RoB Missing data for 6/115 people treated with chemoradiation during the study period: 4 had histology elsewhere and 2 had inadequate tissue. Median OS for the complete cohort of 115 people was 12.8 months vs 12.4 months in the 109 included people. Low RoB No concerns Low RoB No concerns
PSQ 72–83 Cluster 1 vs 2 and 3 Unclear Methylated cases were ranked according to methylation and divided into 3 groups. This does not seem to have been a prespecified analysis, but not data driven. Low RoB Missing data for 6/115 people treated with chemoradiation during the study period: 4 had histology elsewhere and 2 had inadequate tissue. Median OS for the complete cohort of 115 people was 12.8 months vs 12.4 months in the 109 included people. Low RoB No concerns Low RoB No concerns
PSQ 72–83 Cluster 1 and 2 vs 3 Unclear Methylated cases were ranked according to methylation and divided into 3 groups. This does not seem to have been a prespecified analysis, but not data driven. Low RoB Missing data for 6/115 people treated with chemoradiation during the study period: 4 had histology elsewhere and 2 had inadequate tissue. Median OS for the complete cohort of 115 people was 12.8 months vs 12.4 months in the 109 included people. Low RoB No concerns Low RoB No concerns
Felsberg 2009 IHC N/A < 10% Low RoB No concerns Low RoB Missing data were due to issues with the IHC staining. N/A — Low RoB No concerns
MSP NR NR Low RoB No concerns Low RoB No concerns N/A — Low RoB No concerns
PCR‐mRNA N/A < 50% Unclear Insufficient information Unclear mRNA data not available for 64% of study cohort. Large amount of missing data but unclear differences between missing and included participants. N/A — Low RoB No concerns
Havik 2012 MSP 76–80 and 84–87 NR Low RoB No concerns Low RoB Missing data vs the population studied in Havik 2012 covered in Domain 1. N/A — Low RoB No concerns
PCR‐HRM 72–83 NR Unclear States that there is "no" threshold, although there must have been one. Unclear Missing data for methylation status for 11/48 participants. N/A — Low RoB No concerns
PSQ 74–78 > 2.68% Low RoB No concerns Low RoB Only participants treated with radiotherapy + TMZ included. N/A — Low RoB No concerns
PSQ 74–78 > 6% High RoB Not prespecified, data driven. "In order to compare the prognostic ability of the different methods, the optimal cut‐off value for PSQ needed to be identified. ROC curve analysis is the method of choice for predicting optimal cut‐off values (37, 38). The mean percentage methylation of the CpGs analyzed in the two PSQ assays were used in our ROC curve analysis, where methylation cut‐off scores (1–15%) were plotted to identify the optimum cut‐off value for the prediction of OS of 18 months or more after surgery. The AUROC results, including HR values, are listed in Table II. The highest values for AUROC were at a cut‐off of 7% for PSQ Therascreen and 7 and 8% for PSQ 96." Low RoB Missing data vs the population studied in Havik 2012 covered in Domain 1. N/A — Low RoB No concerns
PSQ 74–78 > 7% High RoB Not prespecified, data driven. "In order to compare the prognostic ability of the different methods, the optimal cut‐off value for PSQ needed to be identified. ROC curve analysis is the method of choice for predicting optimal cut‐off values (37, 38). The mean percentage methylation of the CpGs analyzed in the two PSQ assays were used in our ROC curve analysis, where methylation cut‐off scores (1–15%) were plotted to identify the optimum cut‐off value for the prediction of OS of 18 months or more after surgery. The AUROC results, including HR values, are listed in Table II. The highest values for AUROC were at a cut‐off of 7% for PSQ Therascreen and 7 and 8% for PSQ 96." Low RoB Missing data vs the population studied in Havik 2012 covered in Domain 1. N/A — Low RoB No concerns
PSQ 74–78 > 8% High RoB Not prespecified, data driven. But this was not the optimal cut‐off. "In order to compare the prognostic ability of the different methods, the optimal cut‐off value for PSQ needed to be identified. ROC curve analysis is the method of choice for predicting optimal cut‐off values (37, 38). The mean percentage methylation of the CpGs analyzed in the two PSQ assays were used in our ROC curve analysis, where methylation cut‐off scores (1–15%) were plotted to identify the optimum cut‐off value for the prediction of OS of 18 months or more after surgery. The AUROC results, including HR values, are listed in Table II. The highest values for AUROC were at a cut‐off of 7% for PSQ Therascreen and 7 and 8% for PSQ 96." Low RoB Missing data vs the population studied in Havik 2012 covered in Domain 1. N/A — Low RoB No concerns
PSQ 74–78 > 9% High RoB Not prespecified, data driven. But this was not the optimal cut‐off. "In order to compare the prognostic ability of the different methods, the optimal cut‐off value for PSQ needed to be identified. ROC curve analysis is the method of choice for predicting optimal cut‐off values (37, 38). The mean percentage methylation of the CpGs analyzed in the two PSQ assays were used in our ROC curve analysis, where methylation cut‐off scores (1–15%) were plotted to identify the optimum cut‐off value for the prediction of OS of 18 months or more after surgery. The AUROC results, including HR values, are listed in Table II. The highest values for AUROC were at a cut‐off of 7% for PSQ Therascreen and 7 and 8% for PSQ 96." Low RoB Missing data vs the population studied in Havik 2012 covered in Domain 1. N/A — Low RoB No concerns
PSQ 76–79 > 6% High RoB Not prespecified, data driven. But this was not the optimal cut‐off. "In order to compare the prognostic ability of the different methods, the optimal cut‐off value for PSQ needed to be identified. ROC curve analysis is the method of choice for predicting optimal cut‐off values (37, 38). The mean percentage methylation of the CpGs analyzed in the two PSQ assays were used in our ROC curve analysis, where methylation cut‐off scores (1–15%) were plotted to identify the optimum cut‐off value for the prediction of OS of 18 months or more after surgery. The AUROC results, including HR values, are listed in Table II. The highest values for AUROC were at a cut‐off of 7% for PSQ Therascreen and 7 and 8% for PSQ 96." Low RoB Missing data vs the population studied in Havik 2012 covered in Domain 1. N/A — Low RoB No concerns
PSQ 76–79 > 7% High RoB Not prespecified, data driven. "In order to compare the prognostic ability of the different methods, the optimal cut‐off value for PSQ needed to be identified. ROC curve analysis is the method of choice for predicting optimal cut‐off values (37, 38). The mean percentage methylation of the CpGs analyzed in the two PSQ assays were used in our ROC curve analysis, where methylation cut‐off scores (1–15%) were plotted to identify the optimum cut‐off value for the prediction of OS of 18 months or more after surgery. The AUROC results, including HR values, are listed in Table II. The highest values for AUROC were at a cut‐off of 7% for PSQ Therascreen and 7 and 8% for PSQ 96." Low RoB Missing data vs the population studied in Havik 2012 covered in Domain 1. N/A — Low RoB No concerns
PSQ 76–79 > 8% High RoB Not prespecified, data driven. But this was not the optimal cut‐off. "In order to compare the prognostic ability of the different methods, the optimal cut‐off value for PSQ needed to be identified. ROC curve analysis is the method of choice for predicting optimal cut‐off values (37, 38). The mean percentage methylation of the CpGs analyzed in the two PSQ assays were used in our ROC curve analysis, where methylation cut‐off scores (1–15%) were plotted to identify the optimum cut‐off value for the prediction of OS of 18 months or more after surgery. The AUROC results, including HR values, are listed in Table II. The highest values for AUROC were at a cut‐off of 7% for PSQ Therascreen and 7 and 8% for PSQ 96." Low RoB Missing data vs the population studied in Havik 2012 covered in Domain 1. N/A — Low RoB No concerns
PSQ 76–79 > 9% High RoB Not prespecified, data driven. But this was not the optimal cut‐off. "In order to compare the prognostic ability of the different methods, the optimal cut‐off value for PSQ needed to be identified. ROC curve analysis is the method of choice for predicting optimal cut‐off values (37, 38). The mean percentage methylation of the CpGs analyzed in the two PSQ assays were used in our ROC curve analysis, where methylation cut‐off scores (1–15%) were plotted to identify the optimum cut‐off value for the prediction of OS of 18 months or more after surgery. The AUROC results, including HR values, are listed in Table II. The highest values for AUROC were at a cut‐off of 7% for PSQ Therascreen and 7 and 8% for PSQ 96." Low RoB Missing data vs the population studied in Havik 2012 covered in Domain 1. N/A — Low RoB No concerns
qMSP 71–73 and 75–86 NR Low RoB No concerns Low RoB Missing data vs the population studied in Havik 2012 covered in Domain 1. N/A — Low RoB No concerns
qMSP 71–86 > 0% Low RoB No concerns Low RoB Only people treated with radiation therapy + TMZ included. N/A — Low RoB No concerns
Hsu 2017 (see Hsu 2015) IHC N/A < 10% Low RoB No concerns Low RoB No missing data Low RoB No concerns Low RoB No concerns
MSP 76–80 and 84–87 NR Low RoB No concerns Low RoB No missing data Low RoB No concerns Low RoB No concerns
PSQ 76–79 > 5% Low RoB No concerns Low RoB No missing data Low RoB No concerns Low RoB No concerns
qMSP 77–80 and 84–87 > 0.04% Low RoB No concerns Low RoB No missing data Low RoB No concerns Low RoB No concerns
qMSP 77–80 and 84–87 > 0.1% Low RoB No concerns Low RoB No missing data Low RoB No concerns Low RoB No concerns
Karayan‐Tapon 2010 IHC N/A < 15.5% Low RoB No concerns Low RoB Missing data for 3/81 people. N/A — Low RoB No concerns
MSP 76–80 and 84–87 NR Low RoB Likely to be based on presence/absence of bands on a gel but threshold not reported. Low RoB No missing data N/A — Low RoB No concerns
PCR‐mRNA N/A < 0.39 Low RoB No concerns Low RoB Missing data for 1/81 people. N/A — Low RoB No concerns
PSQ 74 > 5.5% Low RoB No concerns Low RoB Missing data for 2/81 people. N/A — Low RoB No concerns
PSQ 74–78 > 8.0% Low RoB No concerns Low RoB Missing data for 2/81 people. N/A — Low RoB No concerns
PSQ 75 > 8.7% Low RoB No concerns Low RoB Missing data for 2/81 people. N/A — Low RoB No concerns
PSQ 76 > 8.0% Low RoB No concerns Low RoB Missing data for 2/81 people. N/A — Low RoB No concerns
PSQ 77 > 7.85% Low RoB No concerns Low RoB Missing methylation data for 2/81 people. N/A — Low RoB No concerns
PSQ 78 > 7.8% Low RoB No concerns Low RoB Missing data for 2/81 people. N/A — Low RoB No concerns
SQ‐MSP 76–80 and 84–87 > 35 Low RoB No concerns Low RoB No missing data N/A — Low RoB No concerns
Kim 2016 MSP 76–80 and 84–87 NR Low RoB Likely to be based on presence/absence of bands on a gel but threshold not reported. Low RoB Data for all people treated with TMZ. N/A — Low RoB No concerns
PSQ 74–78 > 9% High RoB Data driven – based on the results of ROC analysis (although for the whole cohort, not just those treated with TMZ). Low RoB Data for all people treated with TMZ. N/A — Low RoB No concerns
Kristensen 2016 IHC N/A at 0% Unclear It is unclear whether the investigators analysing the results were blinded to clinical outcomes. Low RoB Low proportion with missing data. N/A — Low RoB No concerns
PSQ NR > 10% Low RoB No concerns Unclear 12% missing data. N/A — Low RoB No concerns
qMSP‐PSQ NR > 0.1% Low RoB No concerns Low RoB No missing data N/A — Low RoB No concerns
qMSP‐PSQ NR > 5% Low RoB No concerns Low RoB No missing data N/A — Low RoB No concerns
qMSP‐PSQ NR > 20% Low RoB No concerns Low RoB No missing data N/A — Low RoB No concerns
Lalezari 2013 IHC N/A < 30% Low RoB No concerns Unclear Missing data on MGMT status for about 15% of the study population. Unclear if there are systematic differences between those with and without data. Low RoB No concerns Low RoB No concerns
MSP 76–80 and 84–87 NR Low RoB No concerns Low RoB MGMT methylation by MSP was measured in 402 people (missing < 5% of the population). No information about reason for exclusion from this analysis is given. Low RoB No concerns Low RoB No concerns
PSQ 72–95 NR Low RoB Cut‐off based on median number of methylated CpG sites as resulted from the analysis. High RoB MGMT methylation by BiSEQ was measured in 312 people (> 25% is missing). No information about reason for exclusion from this analysis is given. Low RoB No concerns Low RoB No concerns
Lattanzio 2015 MSP 76–80 and 84–87 NR Low RoB No concerns Unclear Missing data for 6/46 enrolled people. No information to judge whether there were important differences between participants who completed the study and those who did not. N/A — Low RoB No concerns
MSP 76–80 and 84–87 NR Low RoB No concerns Low RoB Missing data for just 1 participant, and not due to missing data on MGMT status. N/A — Low RoB No concerns
PSQ 72–80 ≥ 9% Low RoB No concerns Low RoB Missing data for just 1 participant, and not due to missing data on MGMT status. N/A — Low RoB No concerns
PSQ 72–80 ≥ 9% Low RoB No concerns Low RoB Missing data for just 1 participant, and not due to missing data on MGMT status. N/A — Low RoB No concerns
Lechapt‐Zalcman 2012 IHC N/A < 15% Low RoB Single centre for MGMT testing, analysis of tumour specimens blinded to MSP data and clinical outcomes. Cut‐off was defined by the median value of reactivity – as done previously by other authors (referenced). Low RoB Low proportion of missing data. No information on differences between missing and non‐missing people. N/A — Low RoB No concerns
MSP 76–80 and 84–87 NR Low RoB Single centre for MGMT testing, analysis of tumour specimens blinded to participant identity, threshold not reported but mentions a "detailed protocol" so I think can assume this included a prespecified cut‐off. Low RoB Low proportion of missing data in univariate analysis. No information on differences between missing and non‐missing people. N/A — Low RoB No concerns
McDonald 2013 MSP 76–80 NR Low RoB No concerns Unclear Insufficient information Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 8% Low RoB Although cut‐off was determined post‐hoc, this was only to dichotomise the data and was not determined based on outcomes, which does not seem like it would increase risk of bias. Low RoB Only 2/78 missing data. Low RoB No concerns Low RoB No concerns
Melguizo 2012 IHC N/A < 25% Low RoB No concerns Low RoB Only 2.5% samples missing data. N/A — Low RoB No concerns
MSP 76–80 and 84–87 NR Low RoB Unclear as no information about threshold. Low RoB No concerns N/A — Low RoB No concerns
Nguyen 2015 FSQ‐MS‐PCR 76–80 and 84–87 > 15% High RoB Cut‐offs appeared to have been determined based on performance. Low RoB States that all clinical and molecular data were fully complete in all people. Low RoB — Low RoB No concerns
FSQ‐MS‐PCR 76–80 and 84–87 > 60% High RoB Cut‐offs appeared to have been determined based on performance. Low RoB States that all clinical and molecular data were fully complete in all people. N/A — Low RoB No concerns
Park 2011 MS‐MLPA NR > 0.1% High RoB Outcome‐based cut‐off chosen. Low RoB No missing data N/A — Low RoB No concerns
MS‐MLPA NR > 0.2 High RoB Outcome‐based cut‐off chosen. Low RoB No missing data N/A — Low RoB No concerns
MSP 76–80 and 84–86 NR Low RoB No concerns Low RoB No concerns N/A — Low RoB No concerns
Quillien 2014 (test) IHC N/A < 23% High RoB Threshold was not prespecified. Chosen based on performance. Low RoB 1/100 people excluded because of missing results on IHC due to a technical problem during the staining process (this participant was also excluded from all other analyses). Low RoB No concerns Low RoB No concerns
MSP 76–80 and 84–87 NR Low RoB Likely to be based on presence/absence of bands on a gel but threshold not reported. Low RoB 1/100 people excluded because of missing results on IHC due to a technical problem during the staining process (this participant was also excluded from all other analyses). Low RoB No concerns Low RoB No concerns
MethyLight‐MSP 75–86 > 0 High RoB Threshold was not prespecified. Chosen based on performance. Low RoB 1/100 people excluded because of missing results on IHC due to a technical problem during the staining process (this participant was also excluded from all other analyses). Low RoB No concerns Low RoB No concerns
PCR‐HRM 70–83 > 50% Low RoB No concerns Low RoB 1/100 people excluded because of missing results on IHC due to a technical problem during the staining process (this participant was also excluded from all other analyses). Low RoB No concerns Low RoB No concerns
PSQ 74 > 4% High RoB Threshold was not prespecified. Chosen based on performance. Low RoB 1/100 people excluded because of missing results on IHC due to a technical problem during the staining process (this participant was also excluded from all other analyses). Low RoB No concerns Low RoB No concerns
PSQ 74 > 8% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 8% High RoB Threshold was not prespecified. Chosen based on performance. Low RoB 1/100 people excluded because of missing results on IHC due to a technical problem during the staining process (this participant was also excluded from all other analyses). Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 9% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 74‐89 > 11% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 75 > 11% High RoB Threshold was not prespecified. Chosen based on performance. Low RoB 1/100 people excluded because of missing results on IHC due to a technical problem during the staining process (this participant was also excluded from all other analyses). Low RoB No concerns Low RoB No concerns
PSQ 75–79 > 8% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 76 > 4% High RoB Threshold was not prespecified. Chosen based on performance. Low RoB 1/100 people excluded because of missing results on IHC due to a technical problem during the staining process (this participant was also excluded from all other analyses). Low RoB No concerns Low RoB No concerns
PSQ 76 > 5% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 76–79 > 8% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 76–80 > 9% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 77 > 6% High RoB Threshold was not prespecified. Chosen based on performance. Low RoB 1/100 people excluded because of missing results on IHC due to a technical problem during the staining process (this participant was also excluded from all other analyses). Low RoB No concerns Low RoB No concerns
PSQ 77 > 7% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 77–81 > 8% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 78 > 4% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 78 > 5% High RoB Threshold was not prespecified. Chosen based on performance. Low RoB 1/100 people excluded because of missing results on IHC due to a technical problem during the staining process (this participant was also excluded from all other analyses). Low RoB No concerns Low RoB No concerns
PSQ 78–82 > 9% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 79 > 7% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 79–83 > 8% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 80 > 4% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 81 > 8% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 82 > 16% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 83 > 10% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 84 > 9% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 84–88 > 17% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 84–89 > 22% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 85 > 5% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 85–89 > 13% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 86 > 11% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 87 > 25% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 88 > 4% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 89 > 12% High RoB Cut‐off of methylation based on outcome. Low RoB No concerns Low RoB No concerns Low RoB No concerns
Quillien 2014 (validation) PSQ 74–78 > 9% Low RoB Cut‐off of methylation based on outcome of testing cohort. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 10% Low RoB Cut‐off of methylation based on outcome of testing cohort. Low RoB No concerns Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 28% Low RoB Cut‐off of methylation based on outcome of testing cohort. Low RoB No concerns Low RoB No concerns Low RoB No concerns
Quillien 2016 PSQ 74–78 > 6% High RoB Data driven – based on the results of ROC analysis. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 8% Low RoB Threshold found in a previous publication. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 12% Unclear Unclear why this was chosen/no justification. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 13% Unclear Unclear why this was chosen/no justification. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 16% High RoB Data driven – based on the results of ROC analysis. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 6% High RoB Data driven – based on the results of ROC analysis. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 8% Low RoB Threshold found in a previous publication. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 12% or 13% Unclear Unclear why this was chosen/no justification. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 12% Low RoB From Quillien 2016: Standard Operating Procedure and 10 quality control samples for the determination of MGMT promoter methylation were sent to the different centres as way of standardisation of the process throughout the multiple centres. This approach reduced risk of bias due to different setting for prognostic factor measurement. Cut‐off of methylation based on outcome. Unclear From the original cohort, 10 people who had successful initial PSQ could not have the Thera PSQ. Therefore, the number of people included in this analysis was 102, but should have been 112. Authors commented in paper: "These data are almost identical to those obtained for the overall population (n=112 people, 49%, 44% and AUCROC values of 0.69 and 0.70), indicating the absence of bias in the selection of the 102 people for the present cohort." Low RoB No concerns Low RoB No concerns
PSQ 74–78 > 16% High RoB Data driven – based on the results of ROC analysis. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
PSQ 76–79 > 8% Low RoB Analysis of intralaboratory reproducibility of Thera showed highly reproducibility of results from the different centres. Unclear From the original cohort, 10 people who had successful initial PSQ could not have the Thera PSQ. Therefore, the number of people included in this analysis was 102, but should have been 112. Authors commented in paper: "These data are almost identical to those obtained for the overall population (n=112 people, 49%, 44% and AUCROC values of 0.69 and 0.70), indicating the absence of bias in the selection of the 102 people for the present cohort." Low RoB No concerns Low RoB No concerns
PSQ 76–79 > 12% Low RoB Analysis of intralaboratory reproducibility of Thera showed highly reproducibility of results from the different centres. Unclear From the original cohort, 10 people who had successful initial PSQ could not have the Thera PSQ. Therefore, the number of people included in this analysis was 102, but should have been 112. Authors commented in paper: "These data are almost identical to those obtained for the overall population (n=112 people, 49%, 44% and AUCROC values of 0.69 and 0.70), indicating the absence of bias in the selection of the 102 people for the present cohort." Low RoB No concerns Low RoB No concerns
SQ‐MSP 76–80 and 84–87 > 12% Unclear Unclear why this was chosen/no justification. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
SQ‐MSP 76–80 and 84–87 > 13% Unclear Unclear why this was chosen/no justification. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
SQ‐MSP 76–80 and 84–87 > 23% High RoB Data driven – based on the results of ROC analysis. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
SQ‐MSP 76–80 and 84–87 > 13% High RoB Data driven – based on the results of ROC analysis. Although also the cut‐off that corresponds to best concordance which we rated as unclear elsewhere. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
SQ‐MSP 76–80 and 84–87 > 23% High RoB Data driven – based on the results of ROC analysis. Unclear Numerous missing data but unclear if those with missing data different to those without. Low RoB No concerns Low RoB No concerns
Thon 2017 MSP 76–80 and 84–87 NR Low RoB MSP is usually evaluated by visibility of a band indicating methylation, therefore, we judged as prespecified threshold in the absence of a clear description in the text. Low RoB 1 participant lost to follow‐up after 6 months. Low RoB No concerns Low RoB No concerns
Sequencing 75–99 (unclear) > 50% Low RoB No concerns Low RoB 1 participant lost to follow‐up after 6 months. Low RoB No concerns Low RoB No concerns
Yamashita 2018 MSP 76–80 and 84–87 NR Low RoB No cut‐off/presence or absence of band. Low RoB 1 participant lost to follow‐up after 6 months. High RoB Model included other MGMT status using alternative method. Low RoB No concerns
PCR‐HRM 72–89 > 5% High RoB Data driven – based on the results of ROC analysis. This was not the optimal cut‐off. Low RoB No missing data for MGMT status or OS. Multivariate analyses reportedly for all 75 participants, although 1 participant with missing IDH1 status. N/A — High RoB 2 other primer sets were used for PCR‐HRM. Only have ROC curve data for these (discarded as not as predictive as primer set 1).
PCR‐HRM 72–89 > 8% High RoB Data driven – based on the results of ROC analysis. This was not the optimal cut‐off. Low RoB No missing data for MGMT status or OS. Multivariate analyses reportedly for all 75 participants, although 1 participant with missing IDH1 status. N/A — High RoB 2 other primer sets were used for PCR‐HRM. Only have ROC curve data for these (discarded as not as predictive as primer set 1).
PCR‐HRM 72–89 > 10% High RoB Data driven – based on the results of ROC analysis. Low RoB No missing data for MGMT status or OS. Multivariate analyses reportedly for all 75 participants, although 1 participant with missing IDH1 status. High RoB Model includes other MGMT status using alternative method High RoB 2 sets were used for PCR‐HRM. Only have ROC curve data for these (discarded as not as predictive as primer set 1).
PCR‐HRM 72–89 > 12% High RoB Data driven – based on the results of ROC analysis. This was not the optimal cut‐off. Low RoB No missing data for MGMT status or OS. Multivariate analyses reportedly for all 75 participants, although 1 participant with missing IDH1 status. N/A — High RoB 2 other primer sets were used for PCR‐HRM. Only have ROC curve data for these (discarded as not as predictive as primer set 1).
PCR‐HRM 72–89 > 15% High RoB Data driven – based on the results of ROC analysis. This was not the optimal cut‐off. Low RoB No missing data for MGMT status or OS. Multivariate analyses reportedly for all 75 participants, although 1 participant with missing IDH1 status. N/A — High RoB 2 other primer sets were used for PCR‐HRM. Only have ROC curve data for these (discarded as not as predictive as primer set 1).
Yang 2012 IHC N/A < 10% Low RoB To a degree we can set the cut‐off in this study. Low RoB No missing data Low RoB No concerns Unclear HRM analyses also performed but extractable data not presented.
MSP 76–80 and 84–87 NR Low RoB To a degree we can set the cut‐off in this study. Low RoB No missing data Low RoB No concerns Unclear HRM analyses also performed but extractable data not presented.
Yoshioka 2018 SQ‐MSP 76–80 and 84–87 > 0 Low RoB No concerns Unclear No information regarding 4 missing samples. Low RoB No concerns Low RoB No concerns
SQ‐MSP 76–80 and 84–87 > 2 Low RoB No concerns Unclear No information regarding 4 missing samples. Low RoB No concerns Low RoB No concerns
SQ‐MSP 76–80 and 84–87 > 4 Low RoB No concerns Unclear No information regarding 4 missing samples. Low RoB No concerns Low RoB No concerns
SQ‐MSP 76–80 and 84–87 > 6 Low RoB No concerns Unclear No information regarding 4 missing samples. Low RoB No concerns Low RoB No concerns
SQ‐MSP 76–80 and 84–87 > 8 Low RoB No concerns Unclear No information regarding 4 missing samples. Low RoB No concerns Low RoB No concerns
AUROC: area under receiver operating characteristic; CpG: 5'‐cytosine‐phosphate‐guanine‐3'; DIF: double immunofluorescence; FSQ‐MS‐PCR: fluorescent semi‐quantitative methylation‐specific polymerase chain reaction; GBM: glioblastoma; HR: hazard ratio; IDH: isocitrate dehydrogenase; IHC: immunohistochemistry; MGMT: O6‐methylguanine–DNA methyltransferase; mRNA: messenger ribonucleic acid; MS‐MLPA: methylation‐specific multiplex ligation‐dependent probe amplification; MS‐RE‐qPCR: methylation‐specific restriction enzyme quantitative polymerase chain reaction; MSP: methylation‐specific polymerase chain reaction; N/A: not applicable; NR: not reported; OS: overall survival; PCR: polymerase chain reaction; PCR‐HRM: polymerase chain reaction with high‐resolution melting; PCR‐mRNA: polymerase chain reaction‐messenger ribonucleic acid; PF: prognostic factor; PSQ: pyrosequencing; QF‐IHC: quantitative fluorescence immunohistochemistry; qMSP: quantitative methylation‐specific polymerase chain reaction; qMSP‐PSQ: quantitative methylation‐specific polymerase chain reaction with pyrosequencing; RCT: randomised controlled trial; RoB: risk of bias; ROC: receiver operating characteristic; sel. rep.: selective reporting; SQ‐MSP: semi‐quantitative methylation‐specific polymerase chain reaction; TMZ: temozolomide.