Assessing the clinical value of targeted massively parallel sequencing in a longitudinal, prospective population-based study of cancer patients

Abstract

Introduction: Recent discoveries in cancer research have revealed a plethora of clinically actionable mutations that provide therapeutic, prognostic and predictive benefit to patients. The feasibility of screening mutations as part of the routine clinical care of patients remains relatively unexplored as the demonstration of massively parallel sequencing (MPS) of tumours in the general population is required to assess its value towards the health-care system.

Methods: Cancer 2015 study is a large-scale, prospective, multisite cohort of newly diagnosed cancer patients from Victoria, Australia with 1094 patients recruited. MPS was performed using the Illumina TruSeq Amplicon Cancer Panel.

Results: Overall, 854 patients were successfully sequenced for 48 common cancer genes. Accurate determination of clinically relevant mutations was possible including in less characterised cancer types; however, technical limitations including formalin-induced sequencing artefacts were uncovered. Applying strict filtering criteria, clinically relevant mutations were identified in 63% of patients, with 26% of patients displaying a mutation with therapeutic implications. A subset of patients was validated for canonical mutations using the Agena Bioscience MassARRAY system with 100% concordance. Whereas the prevalence of mutations was consistent with other institutionally based series for some tumour streams (breast carcinoma and colorectal adenocarcinoma), others were different (lung adenocarcinoma and head and neck squamous cell carcinoma), which has significant implications for health economic modelling of particular targeted agents. Actionable mutations in tumours not usually thought to harbour such genetic changes were also identified.

Conclusions: Reliable delivery of a diagnostic assay able to screen for a range of actionable mutations in this cohort was achieved, opening unexpected avenues for investigation and treatment of cancer patients.

Figure 1

VCR Reference for ascertainment bias in the Cancer 2015 cohort. (A) The Cancer 2015 Cohort by cancer type, compared with the VCR 2011 census of solid-cancers only (with removal of paediatric and haematological cancers). Note: melanoma incidences represent advanced stages only. (B) The Reference Cohort obtained from the VCR 2011 census of solid-cancers data segmented into various cancer staging groups compared with the Cancer 2015 Cohort as a percentage of each random sample number.

Figure 2

Mutational landscape of actionable mutations and pathways in the Cancer 2015 cohort. (A) Landscape of actionable mutations from the Cancer 2015 cohort. Tracks are (from outside in): Gene name, Exon label with type of cancer gene (green: tumour-suppressor gene, orange: oncogene), exon size shown as a blue tile, amplicon covered by the TSACP platform (grey tiles) and variants occurring >10 times in the filtered data. Variants are colour-coded based on the type of actionable mutation: (I) sensitive or resistant to an, approved drug/treatment (IA) or experimental drug/treatment (IB). (II) Provides prognostic or diagnostic information based on significant functional or clinically characterisation, (III) Unknown significance due to lack of biological/functional evidence or (IV) benign. Recurrent mutations are also highlighted. (B) Tumour classification by the actionable pathway. Variants from patients were stratified based on known associated pathways, detailed in Supplementary Table 1. The overall percentage of variants in any particular pathway is shown in the x axis. In some cases, a gene was associated with multiple pathways, for example, NRAS for PI3K-Akt and Ras-Raf pathways. In some cases, multiple genes were mutated in the same pathway. Multiple variants in the same gene from the same patient were only counted once. Only tumour streams with more than five patients mutated in a pathway are shown, with other cases combined to the other subset.

Figure 3

The prevalence of mutations in common cancer genes from the Cancer 2015 cohorts compared with other institutional-based series. Reported prevalence of mutations in colorectal adenocarinoma, breast-invasive carcinoma, lung adenocarcinoma and head and neck squamous cell carcinoma as reported by public mutational catalogues (COSMIC and TCGA) compared with the Cancer 2015 cohort.