NEW YORK – Researchers at Memorial Sloan Kettering Cancer Center have determined that for many cancers the effect that tumor mutational burden (TMB) has on patients' overall survival prognosis depends on whether patients receive immune checkpoint inhibitors as part of their treatment regimen.
In a paper published in Nature Genetics earlier this week, the researchers showed that in the absence of immune checkpoint inhibitors, high TMB status is actually a marker conferring poor prognosis for patients, whereas after treatment with checkpoint inhibitors, the prognosis of TMB-high patients improves.
A number of studies have shown that patients with high TMB tend to respond particularly well to checkpoint inhibitors compared to those with low TMB. Last year, the US Food and Drug Administration recognized the value of TMB as a predictive biomarker and granted accelerated approval to pembrolizumab (Merck's Keytruda) for refractory solid tumor patients, regardless of histology, and high TMB, defined as 10 or more mutations per megabase.
Following that approval, however, some oncologists remained unconvinced that immunotherapy is the driver behind the improved outcomes seen in TMB-high patients, as opposed to the possibility that these patients have an improved prognosis in general. According to MSK's Luc Morris, one of the authors of the Nature Genetics paper, this question — whether high TMB confers a good prognosis on its own or whether immunotherapy is an essential part of the equation — had not been addressed in the study that was the basis of pembrolizumab's tissue-agnostic approval.
"This question has been around for a while," Morris said in an interview. "I was a bit surprised that the FDA approval came through before it had been answered."
Part of the reason for this, Morris explained, is that checkpoint inhibitors have become standard treatment for many cancer types, making it difficult to enroll a cohort of high-TMB patients who would not receive immunotherapy in a comparator arm.
"So many patients with cancer now are eligible one way or another to receive checkpoint inhibitors, that if you did a study of just non-checkpoint inhibitor-treated patients, it would be very unusual," he said.
Morris and his colleagues were able to work around this challenge because they had access to a group of 10,233 patients with 17 different tumor types who had had their tumors profiled by the MSK-IMPACT next-generation sequencing assay from 2015 to 2018. Grouping together multiple cancer types helped ensure that standard-of-care checkpoint inhibition in certain cancer types would not bias the results.
Then, using the genomic and clinical information from these patients, the researchers built a statistical model that could home in on the overall survival outcomes for patients in a time-dependent manner. Essentially, Morris explained, the statistical model allowed the researchers to assess overall survival without immune checkpoint inhibitors, even in patients who ultimately would go on to receive immunotherapy as a later line of treatment.
"[The statistical model] allowed us to look at what happened to them if they never got [immunotherapy] or before they got it, and that allowed for a more representative group of patients to answer this question," Morris said. "The model that we developed uses a time-varying variable, which basically allows you to designate the immunotherapy status of a patient."
Instead of looking at whether patients simply received or didn't receive immunotherapy, researchers considered whether they received checkpoint inhibitors at any point in their care timeline. This allowed researchers to "harvest [patients'] survival data until the moment they receive immunotherapy," at which point, these patients became part of the immunotherapy-treated group, he explained.
Initially, in the non-immune checkpoint inhibitor group, the researchers excluded patients whose tumors had high or unknown microsatellite instability (MSI), which also results in a high tumor mutation burden and has known prognostic implications, so that this wouldn't confound the results.
After implementing these measures, the researchers ran the data through the statistical model. Importantly, for the purpose of defining "high TMB," they chose to select patients whose mutation burden fell in the top 20th percentile for their specific cancer type. This was significant, Morris explained, because TMB varies from cancer to cancer, and the jury is still out as to whether assigning a universal cutoff, such as the 10 mutations/megabase cutoff the FDA approved as part of pembrolizumab's tissue-agnostic indication, can accurately identify patients who are likely to benefit from checkpoint inhibitors.
Among the 8,356 patients with microsatellite stable tumors who had not (or not yet) received immune checkpoint inhibitors, high TMB was associated with worse overall survival with a 26 percent greater risk of death compared to those with low TMB. Among all patients who did receive immune checkpoint inhibitors, high TMB was associated with improved overall survival with a 26 percent decreased risk of death compared to those with low TMB.
When the researchers ran additional analyses accounting for factors that can generally affect prognosis, such as age, sex, cancer type, and tumor stage, but excluded patients who previously received potentially mutagenic therapies like chemo, radiation, or specific alkylating agents, they repeatedly arrived at the same conclusion: high TMB is associated with worse prognosis in the absence of checkpoint inhibitor therapy, but improved prognosis with it.
"The analyses in our paper support the biological hypothesis that TMB reflects neoantigen load in tumors and likelihood of response to immunotherapy in most but probably not all cancer types," Morris said. "We confirmed the biological mechanism that TMB is related to neoantigens and is related to immunotherapy response in most types of cancer."
Tumor type exceptions, remaining questions
When the researchers performed their analyses in all patients, including those with MSI high tumors, they found that colorectal cancer, endometrial cancer, and bladder cancer were outliers in that high TMB was associated with better prognosis regardless of whether or not patients received immune checkpoint inhibitors.
"Considering these data, we conclude that high TMB may have prognostically favorable associations with survival in colorectal, endometrial, and bladder cancer," Morris and co-authors wrote in their Nature Genetics paper. "However, in all other cancer types analyzed, high TMB was associated with unchanged or poorer prognosis among patients not receiving [ICI] therapy."
Colorectal cancer is one tumor type where a high mutation burden is likely to be due to MSI. The FDA last year approved pembrolizumab as a first-line option for metastatic colorectal cancer patients with high MSI or mismatch repair deficiency. The agency has also approved a tissue-agnostic indication for pembrolizumab based on these same biomarkers.
As Morris explained, the effects of MSI status on prognosis likely account for the outliers, though the fact that there are exceptions to the findings at all underscores the need for additional, histology-specific research. "This is not a black and white situation, and it's not a universal truth," he said in reference to the prognostic trends with high TMB documented in the paper. "It's not true in every cancer. It depends on the cancer type and on MSI status, and the very definition of high TMB is different from cancer type to cancer type."
In terms of this varying definition of high TMB, Morris emphasized that additional research is needed. "The optimal cutoff value for differentiating responders and non-responders [to checkpoint inhibition] is likely to be different between different cancer types," he said.
Earlier research published by Morris and his colleagues similarly suggested that the ideal TMB cutoff for identifying best responders to immunotherapy may be tumor-specific. Because of this and other research, the FDA's TMB-based tissue agnostic approval for pembrolizumab was controversial among oncologists who felt the 10 mutations/megabase cutoff was arbitrary and would lead to overuse of immunotherapy in patients unlikely to benefit.
"The next step, which is a clinical question, is whether measuring this number helps you decide how to treat a patient," Morris said, adding that this question will likely require clinical trials involving patients with specific tumor types above and below certain TMB thresholds and treating them to see whether TMB is a useful predictive marker.
"We've answered some questions," he said. "But this does not make TMB like a light switch. The definition of high and low still needs to be clarified and the optimal cutoffs for different cancer types still have to be defined."