Which method of determining MGMT promoter methylation best predicts survival in people with glioblastoma treated with temozolomide?

What was the aim of this review?

Glioblastoma is a very aggressive type of brain cancer. People with glioblastoma are usually treated with surgical removal of the tumour followed by radiotherapy, chemotherapy, or both. The standard chemotherapy is a medicine called temozolomide. Some glioblastoma tumours have a particular modification in their DNA (which contains the genetic code of organisms), and knowing whether a person has this modification is useful to predict how long the person may live after their diagnosis with cancer and how they may respond to temozolomide. The modification is known as 'methylation of the MGMT promoter region' and it can also affect MGMT protein expression (the way MGMT is made and modified). There are many ways to work out whether a tumour has this modification. In this review, we attempted to work out which method is best.

What we found

We identified 32 studies comparing different ways to measure whether the MGMT promoter region is methylated. The main three methods were called 'methylation-specific polymerase chain reaction (PCR),' 'pyrosequencing' (both of which look directly at the MGMT promoter region) and 'immunohistochemistry' (which looks at MGMT protein expression). We found that methylation-specific PCR and pyrosequencing are better at predicting overall survival than immunohistochemistry. Methylation-specific PCR and pyrosequencing can be carried out by targeting different parts of the tumour DNA. Pyrosequencing can be performed using different cut-off thresholds to determine whether a tumour is methylated or unmethylated. We did not identify very clear signals in terms of the best parts of the DNA to target or which are the best cut-off thresholds.

How reliable are results of the studies in this review?

We rated our confidence in the evidence as 'moderate' for our conclusions about methylation-specific PCR, but as 'low' for pyrosequencing. Although there were many studies, they all looked at different variants of the methods, so it is difficult to work out exactly which variant is best.

What are the implications of this review?

Our review indicates both methylation-specific PCR and pyrosequencing provide better predictions of survival than immunohistochemistry. There is some evidence that pyrosequencing may be better than methylation-specific PCR at predicting overall survival, depending on the DNA targets and cut-off thresholds used. We documented the most frequent DNA targets used in methylation-specific PCR and pyrosequencing. We described cut-off thresholds used in pyrosequencing, although it is unclear which of these is best.

Authors' conclusions: 

PSQ and MSP appear more prognostic for overall survival than IHC. Strong evidence is not available to draw conclusions with confidence about the best CpG sites or thresholds for quantitative methods. MSP has been studied mainly for CpG sites 76 to 80 and 84 to 87 and PSQ at CpG sites ranging from 72 to 95. A threshold of 9% for CpG sites 74 to 78 performed better than higher thresholds of 28% or 29% in two of three good-quality studies making such comparisons.

Read the full abstract...
Background: 

Glioblastoma is an aggressive form of brain cancer. Approximately five in 100 people with glioblastoma survive for five years past diagnosis. Glioblastomas that have a particular modification to their DNA (called methylation) in a particular region (the O6-methylguanine–DNA methyltransferase (MGMT) promoter) respond better to treatment with chemotherapy using a drug called temozolomide.

Objectives: 

To determine which method for assessing MGMT methylation status best predicts overall survival in people diagnosed with glioblastoma who are treated with temozolomide.

Search strategy: 

We searched MEDLINE, Embase, BIOSIS, Web of Science Conference Proceedings Citation Index to December 2018, and examined reference lists. For economic evaluation studies, we additionally searched NHS Economic Evaluation Database (EED) up to December 2014.

Selection criteria: 

Eligible studies were longitudinal (cohort) studies of adults with diagnosed glioblastoma treated with temozolomide with/without radiotherapy/surgery. Studies had to have related MGMT status in tumour tissue (assessed by one or more method) with overall survival and presented results as hazard ratios or with sufficient information (e.g. Kaplan-Meier curves) for us to estimate hazard ratios. We focused mainly on studies comparing two or more methods, and listed brief details of articles that examined a single method of measuring MGMT promoter methylation. We also sought economic evaluations conducted alongside trials, modelling studies and cost analysis.

Data collection and analysis: 

Two review authors independently undertook all steps of the identification and data extraction process for multiple-method studies. We assessed risk of bias and applicability using our own modified and extended version of the QUality In Prognosis Studies (QUIPS) tool. We compared different techniques, exact promoter regions (5'-cytosine-phosphate-guanine-3' (CpG) sites) and thresholds for interpretation within studies by examining hazard ratios. We performed meta-analyses for comparisons of the three most commonly examined methods (immunohistochemistry (IHC), methylation-specific polymerase chain reaction (MSP) and pyrosequencing (PSQ)), with ratios of hazard ratios (RHR), using an imputed value of the correlation between results based on the same individuals.

Main results: 

We included 32 independent cohorts involving 3474 people that compared two or more methods. We found evidence that MSP (CpG sites 76 to 80 and 84 to 87) is more prognostic than IHC for MGMT protein at varying thresholds (RHR 1.31, 95% confidence interval (CI) 1.01 to 1.71). We also found evidence that PSQ is more prognostic than IHC for MGMT protein at various thresholds (RHR 1.36, 95% CI 1.01 to 1.84). The data suggest that PSQ (mainly at CpG sites 74 to 78, using various thresholds) is slightly more prognostic than MSP at sites 76 to 80 and 84 to 87 (RHR 1.14, 95% CI 0.87 to 1.48). Many variants of PSQ have been compared, although we did not see any strong and consistent messages from the results. Targeting multiple CpG sites is likely to be more prognostic than targeting just one. In addition, we identified and summarised 190 articles describing a single method for measuring MGMT promoter methylation status.