A scoping review and meta-analysis on the prevalence of pan-tumour biomarkers (dMMR, MSI, high TMB) in different solid tumours

Immune checkpoint inhibitors have been approved in the USA for tumours exhibiting mismatch repair deficiency (dMMR), microsatellite instability (MSI), or high tumour mutational burden (TMB), with regulatory and reimbursement applications in multiple other countries underway. As the estimated budget impacts of future reimbursements depend on the size of the potential target population, we performed a scoping review and meta-analysis of the prevalence of these pan-tumour biomarkers in different cancers. We systematically searched Medline/Embase and included studies reporting the prevalence of dMMR/MSI/high TMB in solid tumours published 01/01/2018–31/01/2021. Meta-analyses were performed separately for the pan-cancer prevalence of each biomarker, and by cancer type and stage where possible. The searches identified 3890 papers, with 433 prevalence estimates for 32 different cancer types from 201 studies included in meta-analyses. The pooled overall prevalence of dMMR, MSI and high TMB (≥ 10 mutations/Mb) in pan-cancer studies was 2.9%, 2.7% and 14.0%, respectively. The prevalence profiles of dMMR/MSI and high TMB differed across cancer types. For example, endometrial, colorectal, small bowel and gastric cancers showed high prevalence of both dMMR and MSI (range: 8.7–26.8% and 8.5–21.9%, respectively) and high TMB (range: 8.5–43.0%), while cervical, esophageal, bladder/urothelial, lung and skin cancers showed low prevalence of dMMR and MSI (< 5%), but high prevalence of high TMB (range: 23.7–52.6%). For other cancer types, prevalence of all three biomarkers was generally low (< 5%). This structured review of dMMR/MSI/high TMB prevalence across cancers and for specific cancer types and stages provide timely evidence to inform budget impact forecasts in health technology assessments for drug approvals based on these pan-tumour biomarkers.


Managing overlap of data sources for meta-analyses
a) Some patient cohorts have been included in biomarker prevalence estimates in multiple original research studies or systematic reviews. To avoid data duplication in the meta-analyses in this review, we identified studies with overlapping data sources and only included at most one estimate based on a specific patient cohort. b) For each biomarker and each major data source that was included multiple original research studies on the prevalence of dMMR/MSI/high TMB in specific cancer(s) and pan-cancer analyses (e.g., Foundation Medicine database, Memorial Sloan Kettering Cancer Centre patients, analysis of the Cancer Genome Atlas), we used the following approach.
• Pan-cancer overall prevalence: the study with the largest sample size was included in the meta-analysis.
• Cancer-specific overall prevalence: the study with the largest sample size for the specific cancer was included in the meta-analysis for "overall" cancer-specific estimates.
• Cancer-specific prevalence for early-stage or advanced-stage cancers: the study with the largest sample size for the specific cancer type and stage was included in the metaanalysis. c) Systematic reviews and/or meta-analyses • For some cancers (e.g., colorectal cancer, endometrial cancer), we identified multiple systematic reviews and/or meta-analyses that reported the prevalence of dMMR/MSI/high TMB to address a specific research question. For each cancer type, we included the study with the largest sample size. If the sample size was similar across multiple systematic reviews and/or meta-analyses, we included the study for which the research question was most aligned with our scoping review.
• To avoid further data duplication, if a meta-analysis in this review included estimates from a previously published meta-analysis, the underlying original studies included in the previously published meta-analysis were excluded from the corresponding meta-analysis in this review.
See Section 2 below for the list of studies with overlapping data sources that reported the prevalence of dMMR/MSI/high TMB and the rationale for their inclusion or exclusion in meta-analyses. • TMB • MSI

Studies with overlapping data sources and the rationale for their inclusion or exclusion in meta-analyses
• The prevalence of high TMB was not included in the data synthesis due to a data-driven high TMB cut-off (75 th percentile) • The combined prevalence of MSI from both the TCGA and two local hospitals were included in the data synthesis o The TCGA cohort is larger than that included in Qu et al. 2 and Ren et al. 3

Gastric cancer
Overall • Prevalence of TMB was provided in graphical format only and excluded from the data synthesis • Prevalence of MSI was excluded from the data synthesis due to the smaller sample size than Fan et al. 1 Li (2020) 4 Gastric cancer Overall • MSI • Excluded from the data synthesis o The TCGA cohort is likely to be a subset of the cohort reported in Fan et al. 1 o The GEO cohort did not satisfy the minimum cancer-specific sample size cut-off (400+) Dai (2020) 5 Gastric cancer IB-III • TMB (≥20 mut/Mb) • Prevalence of high TMB (≥10 mut/Mb) was included in the data synthesis after combining estimates from both primary and metastatic site given metastatic sites include lymph node metastasis only and not limited to distant metastasis o Primary site (3.9%), metastatic site (6.2%) Necchi (2020) 16 Testicular cancer Germ cell tumours only Advanced (relapsed after CT) • Included in the data synthesis after combining estimates from both seminomas and non-seminomas since this study focused on advanced testicular cancer only • Patients who have experienced a relapse after at least one cisplatin-based CT regimen were included in the study, however tumour samples could have been obtained at any time during the treatment course and from any site of disease. Patel (2020) 17 Brain tumour • Paediatric tumours only (age ≤ 21 years) • TMB (≥20 mut/Mb) • Excluded from the data synthesis o Likely to be a subset of the cohort reported by Yoshino et al. 8 (408 non-gliomas and 800 gliomas in paediatric patients) o Yoshino et al. 8 was the only one study reported the prevalence of dMMR/MSI/high TMB and data synthesis in the prevalence of the pan-tumour biomarkers in paediatric solid tumours was not performed Chow (2020) 18  • MSI • TMB (≥20 mut/Mb) • Prevalence of MSI was excluded from the data synthesis since this study was included in a systematic review by Luchini et al. 21 • Prevalence of TMB (≥20 mut/Mb) was included in the data synthesis given Yoshino et al. 8 did not report the prevalence of TMB (≥20 mut/Mb) in pancreatic cancer Huang (2020) 22 Breast cancer • TMB (≥9 mut/Mb) • Exclude in the data synthesis due to the uncommon high TMB cut-off

3) Studies reporting the prevalence of dMMR/MSI/high TMB based on data from the Memorial Sloan Kettering Cancer Center
Author (  • dMMR/MSI • Included in the data synthesis for the combined prevalence of dMMR and MSI given high concordance between dMMR and MSI in colorectal cancer o Pooled analysis of patients from six adjuvant CT trials (MOSAIC, C07, C08, PETACC8, N0147, AVANT) treated with fluorouracil, leucovorin, and oxaliplatin and included in the ACCENT database o Larger sample size than Taieb et al. 65 Taieb (2019)