Investigators have identified tumor mutational burden as a promising biomarker for predicting immunotherapy responses in patients with advanced stages of non–small cell lung cancer, but is it ready for clinical application?
Matthew Stein, MD
Investigators have identified tumor mutational burden (TMB) as a promising biomarker for predicting immunotherapy responses in patients with advanced stages of nonsmall cell lung cancer (NSCLC), but is it ready for clinical application? Several trials examining patient stratification based on TMB have found that high TMB was associated with positive responses to immune checkpoint inhibitors, in particular rates of progression-free survival (PFS) and objective response.
TMB scores have been examined in numerous clinical trials, including the randomized phase III CheckMate 026 study comparing nivolumab (Opdivo) with platinum-doublet chemotherapy as first-line therapy in treatment-naïve patients with stage IV or recurrent NSCLC. Patients with high TMB, defined as ≥243 missense mutations, had a PFS of 9.7 months with nivolumab versus 5.8 months in those who received chemotherapy (HR, 0.62; 95% CI, 0.38-1.00). High TMB was also predictive of a better objective response rate (ORR) with nivolumab compared with chemotherapy (47% vs 28%). However, no significant difference was observed in overall survival (OS) between the nivolumab and chemotherapy groups regardless of TMB, according to findings published in theNew England Journal of Medicine.1
Analyses from the phase II CheckMate 568 trial of nivolumab plus ipilimumab (Yervoy) in patients with treatment-naïve stage IV NSCLC were presented at the 2018 American Association of Cancer Research (AACR) Annual Meeting. The results identified a TMB of ≥10 mutations per megabase as an effective cutoff for selecting patients most likely to have a response, irrespective of tumor PD-L1 expression. ORRs were higher in patients with TMB of ≥10 mutations per megabase (n = 48; 43.8%) than in those with TMB <10 mutations per megabase (n = 50; 12.0%). In addition, the median PFS was longer in patients with high versus low TMB (7.1 vs 2.6 months).2
The effectiveness of using a TMB cutoff of ≥10 mutations per megabase was reinforced by data from the phase III CheckMate 227 trial comparing nivolumab plus ipilimumab and platinum-doublet chemotherapy, which were also presented at the 2018 AACR meeting3and published in theNew England Journal of Medicine.4The primary endpoint of OS in patients selected for PD-L1 expression was later revised to include an additional endpoint: PFS in TMB-selected patients.
“Based on the emerging data about TMB as an effective and independent biomarker for identifying those who [may receive] the greatest benefit with nivolumab plus ipilimumab, the study was amended prior to the initial analysis to include the second co-primary endpoint,” said Matthew D. Hellmann, MD, a medical oncologist at Memorial Sloan Kettering Cancer Center in New York, New York, during his presentation at the 2018 AACR meeting.
The endpoint of PFS superiority was met in patients with high TMB (≥10 mutations per megabase) regardless of PD-L1 expression. The preliminary findings showed a 1-year PFS rate of 42.6% for patients with high TMB assigned to the immunotherapy combination compared with 13.2% for those assigned to platinum-doublet chemotherapy. The median PFS was 7.2 months versus 5.5 months, respectively, representing a 42% reduction in risk of disease progression or death (HR, 0.58; 97.5% CI, 0.41-0.81;P<.001).4
Based on preliminary findings from the CheckMate 227 trial, the FDA accepted a supplemental biologics license application (sBLA) for nivolumab plus ipilimumab for the treatment of patients with NSCLC and TMB ≥10 mutations per megabase in June 2018.5
At that time, the OS data were not available, and some scientists expressed concerns that it was premature to seek FDA approval. “Tumor mutation burden has yet to prove its predictive or prognostic value for OS in order to become a reliable biomarker,” wrote Alfredo Addeo, MD, and colleagues in aJAMA Oncologyarticle published a year after the sBLA acceptance.6
Addeo mentioned several challenges in the biomarker field, including disparities between assays and cutoffs based on tumor type or immunotherapy. “Moreover, algorithms might influence TMB estimation, as they differ widely across gene panel platforms, and the mutation types considered for TMB assessment can vary from one assay to another.”
In October 2018, Bristol-Myers Squibb (BMS), the company that manufactures both agents, released an updated analysis of OS rates in response to nivolumab/ipilimumab versus chemotherapy from the CheckMate 227 trial. The data showed little difference between the high- and low-TMB groups, with hazard ratios for OS of 0.77 (95% CI, 0.56-1.06) and 0.78 (95% CI, 0.61-1.00) in patients with TMB ≥10 and <10 mutations per megabase, respectively.7
The median OS in patients with high TMB was 23.03 months for the nivolumab/ ipilimumab arm and 16.72 months for the chemotherapy arm. In patients with low TMB, the median OS was 16.20 months and 12.42 months, respectively.7
The FDA determined that the new data represented a major amendment to the sBLA and extended the review period by 3 months, moving the Prescription Drug User Fee Act goal date to May 20, 2019.7
However, BMS voluntarily withdrew its sBLA in January 2019, stating that it needed “further evidence on the relationship between TMB and PD-L1...to fully evaluate the impact of Opdivo plus Yervoy on OS in first-line [treatment of patients with] NSCLC,” according to a press release.8The complete findings from part 1 of CheckMate 227 were presented at the 2019 ESMO conference. A longer duration of overall survival was observed with nivolumab plus ipilimumab compared with chemotherapy, regardless of PD-L1 expression. In the primary analysis, patients with PD-L1>1% treated with nivolumab plus ipilimumab had a median overall survival of 17.1 months (95% CI, 15.0-20.1) compared with 14.9 months (95% CI, 12.7- 16.7) in the chemotherapy group (HR 0.79, 97.72% CI, 0.65-0.96;P= .007).9
In the accompanyingNew England Journal of Medicinearticle, Hellmann an colleagues found “a similar degree of overall survival benefit” for nivolumab and ipilimumab regardless of tumor mutational burden (stratified at 10 mutations per megabase).10In subset analyses, the combination of PD-L1 and TMB did not reveal a subgroup with an increased magnitude of benefit for nivolumab and ipilimumab versus chemotherapy, with the authors noting a decrease in sample size for this evaluation. In summary, the authors stated that “further understanding of the role of tumor mutational burden, if any, as a biomarker is warranted before the integration of this factor into clinical practice.
A recent timeline of events pertaining to CheckMate 227 can be seen in theFIGURE).5,7-9The efficacy of other combinations, such as the PD-L1 inhibitor durvalumab (Imfinzi) and the antiCTLA-4 agent tremelimumab, has been examined in patients with high TMB. In the open-label phase III MYSTIC trial of durvalumab with or without tremelimumab versus platinum-doublet chemotherapy, blood TMB (bTMB) was analyzed as a predictive biomarker of OS in patients with metastatic NSCLC. Levels of bTMB were evaluated using the GuardantOMNI platform from Guardant Health, and PD-L1 tumor cell (TC) expression was examined with the Ventana PD-L1 (SP263) immunohistochemistry assay.11
Preliminary results showed that compared with chemotherapy, treatment with durvalumab alone produced a clinically meaningful improvement in OS among patients with PD-L1 TC ≥25% across all bTMB levels. Additionally, the durvalumab/tremelimumab combination improved OS compared with chemotherapy in patients with bTMB ≥16 and ≥20 mutations per mega-base across different PD-L1 TC expression levels.11
Naiyer A. Rizvi, MD, and colleagues concluded in a published abstract that the preliminary data “support PD-L1 TC expression as an appropriate predictive biomarker for OS with durvalumab versus chemotherapy, while suggesting bTMB as a predictive biomarker for OS with durvalumab plus tremelimumab in metastatic NSCLC. These biomarkers appear to be independent and both may be important for NSCLC treatment decisions.” The results may be limited by small sample sizes, the authors stated, adding that further research is needed.11An analysis led by Matthew Stein, MD, a medical oncology fellow at West Cancer Center and The University of Tennessee Health Science Center in Memphis, aimed to explore whether differences exist in TMB relative to tissue location.12
“Before potential implementation in a clinical setting, TMB needs to be better characterized. We showed TMB is heterogeneous in NSCLC, with novel biopsy-site and histologic distinctions,” Stein toldTargeted Therapies in Oncologyin an interview.
Stein and his co-investigators compared the rate of high TMB, defined as ≥10 mutations per megabase, among 3424 unmatched NSCLC samples, including 2351 lung adenocarcinomas (LUADs) and 1073 lung squamous cell carcinomas (LUSCs). A key finding was that both LUAD and LUSC metastases were more likely to have high TMB compared with their respective primary tumors (LUAD, 38% vs 25%;P<.001; LUSC, 41% vs 35%;P= .038). LUSC primary tumors were more likely than LUAD primary tumors to have high TMB (35% vs 25%, respectively;P<.001).
In both LUAD and LUSC, poorly differentiated primary tumors were more likely to have high TMB than other primary tumors. Similarly, high TMB was more common in poorly differentiated metastases than in other metastases.
In addition to the observation that the rate of high TMB increases with metastasis, the authors found considerable TMB variability with respect to anatomic site in both LUAD and LUSC. Specifically, of 6 metastasis site subgroups, they found brain (61%) and adrenal lesions (51%) most likely, and bone (19%) and liver lesions (31%) least likely to have a high TMB in LUAD.
The study also evaluated whether any anatomic differences exist when combining TMB and PD-L1 as biomarkers. Consistent with previous research, high TMB did not correlate with the percentage of PD-L1 expression for either LUAD (Spearmanr= 0.018;P= .407) or LUSC (Spearmanr= 0.008;P= .785). However, primary LUSC and metastatic LUAD cases were more likely to be both PD-L1 positive and TMB high. Adrenal and other extrathoracic metastases were also identified as subgroups in LUAD that were more likely to be TMB high and have PD-L1 expression of >1%.
“It has previously been described that patients with tissue samples that are PD-L1 positive and TMB high may derive increased benefit from PD-1/PD-L1 inhibitors. While exploratory, our data revealed site-specific differences in this biomarker combination which should be further studied,” said Stein.
He underscored that the finding of TMB being highest in brain metastases is hypothesis-generating. “We postulate whether a site-specific differential response to nivolumab and ipilimumab exists for metastatic sites like brain and adrenal lesions with high TMB. This inquiry could be particularly important for patients with brain metastases, which historically portends a poor prognosis.”
Stein summarized: “I believe this site-dependent heterogeneity should be taken into consideration if TMB is to be utilized as a biomarker for dual checkpoint inhibition [in NSCLC]. Practically, this may manifest as obtaining a biopsy from a specific metastatic site or using blood-based testing to capture high TMB.”
Limitations of the study included a reduced number of LUSC metastases available for analysis and that TMB was not correlated with whole exome sequencing from a subset of samples.Stein cited several drawbacks of using TMB as a biomarker, including the lack of a standardized TMB cutoff.
“TMB is a continuous variable for which a wide range of cut points have been used to date. While clinical benefit has been reported for various TMB-high thresholds and significant collaboration has occurred between platforms to harmonize TMB, the optimal value for use in NSCLC remains elusive.”
Another challenge is that “not all mutations are necessarily meaningful. The premise behind a high TMB as a predictor for immunotherapy response is that a higher number of mutations leads to more neoantigens which translates to an anti-tumor T-cell response. However, this is not straightforward, as certain genetic aberrations like indels [insertions and deletions] may carry more predictive weight. An area of ongoing research is investigating the role of total indel burden and how it might supplement total mutational burden,” Stein said.
For example, a recent study in theAnnals of Oncologyexamined accuracy limitations inherent in gene panel sequencing estimates of TMB. The authors identified indel burden as an independent marker of different tumor types and genetic subgroups and suggested that the accuracy of TMB estimates could be improved by adding indel analysis to missense mutation burden, the type of genetic alteration traditionally considered in panel-based TMB calculations. In addition, by utilizing a derived mathematical law to capture the imprecision of panel-based TMB, the authors introduced a novel 3-tier TMB classification scheme. In contrast with the current 2-tier system with high and low TMB values separated by a single cutoff point, they argued identifying a so-called “gray zone” for TMB could aid future panel designs, trial design, and clinical decision making.13