ORLANDO — A low-cost gene expression assay using FFPE biopsies can classify multiple lymphoma subtypes as well as non-lymphoma diagnoses with high accuracy in a Guatemalan sample, according to research presented at the ASH Annual Meeting and Exposition.
“There’s a major gap between higher-income countries and low-/middle-income countries in the outcomes of patients with different kinds of cancer. Not only are there shortcomings in the availability of treatments, but there are major issues with accurate diagnosis,” David Weinstock, MD, from Dana-Farber Cancer Institute, told HemOnc Today. “In many places, there simply aren’t trained pathologists who are capable of looking at biopsies, even if there were adequate resources to be able to do histochemistry, fluorescence in situ hybridization (FISH) and other things. There’s clearly a need for technologies or other advances that can overcome the shortage of expert pathology and lower cost.”
Weinstock and colleagues reviewed 900 different biopsies performed in the major public cancer hospital in Guatemala because the clinician had a suspicion of lymphoma. Then, all biopsies were sent to Stanford University for diagnostics, like histology, FISH assessments and other approaches, to determine the WHO diagnosis for each biopsy.
The researchers searched the literature for trials where researchers had previously conducted some form of transcriptional profiling and had identified differences between lymphoma subsets. Based on gene expression, they created a very parsimonious gene signature of 37 genes plus a few controls, according to Weinstock. Once they had the WHO diagnosis, they divided the biopsies into a discovery cohort and validation cohort and performed a chemical ligation-based probe amplification assay on all the samples to quantify the gene expression
“We created an artificial intelligence-based model using the discovery cohort for measuring the gene expression,” Weinstock said. “There are over 80 different lymphoma diagnoses in the WHO, but many of them have very small differences and don’t make a difference based on treatment; instead, we created ‘bins’ that were very treatment-driven and then allowed the model to figure out how to ‘bin’ the patient based on their gene expression.”
These bins included aggressive B-cell lymphoma, diffuse large B-cell lymphoma (DLBCL), follicular lymphoma, Hodgkin lymphoma, mantle cell lymphoma, marginal zone lymphoma, natural killer T-cell lymphoma, T-cell lymphoma or non-lymphoma.
When the investigators applied the model to a blinded validation cohort, it performed above 90% accuracy for a range of different types of lymphoma, including DLBCL, mantle cell lymphoma, Hodgkin lymphoma and NK/T cell lymphoma, according to Weinstock. Additionally, the model was able to ‘bin’ non-lymphoma cases – which included lymphoid hyperplasia, normal skin and others – with over 90% accuracy, which could prevent patients from receiving inappropriate chemotherapy.
“Now we see there’s a very useful diagnostic that could be applied broadly across lower- and middle-income countries and we’re pursuing a number of different approaches to try to understand how we can develop this into a prospectively-applied test across Latin America, India and other places where we think it could be potentially transformative,” he told HemOnc Today. ‘There’s no reason that a similar assay wouldn’t also work for breast cancer, lung cancer, or others – it just has to be developed.”
Also, for high-income countries, such as the U.S., Weinstock said this assay has potential as a second opinion.
“We know that there are major issues with diagnosis of lymphomas and other types of cancer when you compare multiple pathologists’ assessments of the same tissue,” he said. “Theoretically, a low-cost assay like this can be performed as a kind of second opinion to be certain that there was concordance with the original diagnosis and if there’s discordance, that might just draw more attention and the need for additional pathological review.” – by Savannah Demko
Briercheck E, et al. Abstract 409. Presented at: ASH Annual Meeting and Exposition; December 7-10, 2019; Orlando, Florida.
Disclosures: Weinstock reports research funding from Celgene and Verastem Oncology.