2.1 Evidence Acquisition
We adopted the Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines to conduct this study.
2.2 Searching strategy and data collected
Four distinct authors (AAB, VDN, EF, AT, and LG) searched English-written articles published on PubMed/Medline, Cochrane library, and Scopus until the 1st of May 2021. Keywords adopted for the research were: ‘’IDH’’ OR ‘’IDH1’’ OR ‘’IDH2’’ OR ‘’Isocitrate dehydrogenase’’ AND ‘’glioma’’. In addition, we searched also relevant abstracts of the main International oncological and neuro oncological meetings (American Society of Clinical Oncology, European Organisation of Research and Treatment of Cancer, European Association of Neuro-Oncology, European Society for Medical Oncology, Society of Neuro-Oncology).
In the case of multiple publications of the same cohort of patients, we included the most updated version with a longer follow-up. In case of the presence of both abstract and complete published version of the same cohort, we selected the complete publication.
We collected the following data from each study selected: 1) First author’s name, 2) Year of publication, 3) the overall number of patients, 4) the overall number of grade 2 and grade 3 tumors, 5) the overall number of IDH mutated grade 2 and grade 3 gliomas, 6) patients with IDH2 and IDH1 non-canonical mutations, 7) patients with 1p19q codeletion and IDH1 non-canonical mutation 8) the survival Hazard Ratio (HR) with 95% Confidence interval.
In addition, we collected data about the modality adopted for IDH assessment and age at diagnosis between subgroups with different IDH mutations. Finally, we recorded also the different grades of MGMT methylation according to the type of IDH mutation.
2.3 Outcomes of the meta-analysis
We were interested to investigate multiple issues (see below). In particular, all these outcomes were focused on patients with WHO grade 2 or grade 3 gliomas excluding all patients with Glioblastoma.
The main outcomes of the present analysis were:
1) The overall incidence of non-canonical mutations among patients with gliomas.
In particular, we were interested to assess the overall incidence of IDH non-canonical mutations and IDH1 non-canonical mutations among all Grade 2 and 3 gliomas and IDH mutated gliomas. To assess these issues, we selected studies reporting retrospective and prospective cohorts of patients with complete data of incidence. Thus, we did not include case-control studies for this outcome;
2) Incidence of IDH1 non-canonical mutations according to WHO tumor grade among patients with gliomas. For this outcome, we selected only studies reporting complete incidence data according to different tumor grades. For the risk-difference analysis (adopted to estimate the different incidence between patients with grade 2 and grade 3 tumors) we included also case-control studies.
3) Survival comparison between patients with IDH and IDH1 non-canonical mutations.
For this outcome, we selected only studies reporting complete data about survival. The preferred summarizing tool was Hazard Ratio (HR) with 95% Confidence Interval. When available we used the HR provided by the multivariate Cox regression model instead of that provided by the univariate log-rank test. Studies performing a survival comparison but not showing an HR were also reported in the text but not included in the analysis.
4) Difference between 1p19q codeletion incidence among patients with canonical IDH1 mutation or non-canonical IDH one.
5) For this outcome, we selected only studies reporting complete incidence data according to the different 1p19q incidence rates. Since we were interested in an incidence ratio between 1p19q patients and patients with IDH1 canonical/non-canonical mutations we included also case-control studies.
6) DNA methylation levels, localization of tumors on CNS, and age of diagnosis difference between patients with/without non-canonical mutations.
Studies reporting a comparison of methylation grade and age of diagnosis have been selected for this outcome.
2.3 Statistical Methods
All analyses have been performed through R statistical software. Packages adopted for the analysis were: tidyverse, dplyr, meta, and metaphor.
In survival analysis, we applied the inverse variance technique for HR assessment reporting both random and fixed effects models.
For incidence analysis, we used the Freeman-Tukey double arcsine transformation of proportions while inverse variance with the Der Simonian-Laird method adopted to estimate between-study variance has been employed.
Finally, the difference between proportions has been performed through a risk difference comparison. The inverse variance weighting has been used for pooling results.
2.4 Risk of Bias
We employed the Newcastle-Ottawa Scale (NOS) to assess the risk of bias of studies included in the meta-analysis. Four authors independently reviewed all studies (VDN, EF, AT, LG) rating each selected study. Studies with a score of 7 or more, 4–6, and lower than 4 were considered to have a low, moderate, and high risk of bias, respectively [10].