ABSTRACT
Tumor mutational burden (TMB) is a key biomarker for predicting the response to immune checkpoint inhibitors (ICIs). However, its predictive accuracy in real‐world clinical practice, particularly in Asian populations, remains inadequately evaluated. We addressed this issue by analyzing real‐world data from 63,952 patients registered in the Center for Cancer Genomics and Advanced Therapeutics (C‐CAT) database, which integrates genomic and clinical information from Japanese patients with various advanced solid tumors. We assessed the therapeutic efficacy of pembrolizumab in 1899 patients who underwent one of three comprehensive genomic profiling tests: FoundationOne CDx, the OncoGuide NCC Oncopanel System, or the GenMine TOP Cancer Genome Profiling System. Based on the reported TMB values, patients were classified as TMB‐high (≥ 10 mutations per megabase) or TMB‐low (< 10 mutations per megabase). The objective response rate (ORR) among 946 TMB‐high patients exceeded 30% and was significantly higher than that observed in 953 TMB‐low patients (16.8%, p < 0.001). Notably, patients with borderline TMB values (10 to less than 13 mutations per megabase) exhibited relatively modest responses (20.8%). The ORR improved when hotspot mutations were excluded from the TMB calculation, suggesting that this adjustment enhances the predictive accuracy of TMB. These findings support the clinical utility of TMB as a biomarker for predicting ICI response in routine oncology practice. In particular, excluding hotspot mutations from TMB calculations may improve response prediction in patients whose TMB values are near the threshold.
Keywords: comprehensive genomic profiling, immune checkpoint inhibitors, precision oncology, real‐world data analysis, tumor mutation burden
In this study, we investigated the predictive value of tumor mutational burden (TMB) for assessing the efficacy of pembrolizumab in a Japanese cohort. We analyzed real‐world data from 63,952 patients registered in the C‐CAT database who underwent comprehensive genomic profiling, and evaluated the therapeutic efficacy of pembrolizumab in 1899 of these patients. Our findings support the clinical utility of TMB as a predictive biomarker in routine oncology practice and underscore the importance of accounting for hotspot mutations, particularly in patients with TMB values near the clinical cutoff.
Abbreviations
- 95% CI
95% confidence interval
- C‐CAT
Center for Cancer Genomics and Advanced Therapeutics
- CGP
comprehensive genomic profiling
- CR
complete response
- F1CDx
FoundationOne CDx
- ICIs
immune checkpoint inhibitors
- MAF
minor allele frequency
- MSI
microsatellite instability
- NOP
OncoGuide NCC Oncopanel System
- ORR
objective response rate
- PD
progressive disease
- post‐CGP
post‐comprehensive genomic profiling treatment
- PR
partial response
- pre‐CGP
pre‐comprehensive genomic profiling treatment
- SD
stable disease
- SNP
single‐nucleotide polymorphisms
- TMB
tumor mutational burden
- TOP
GenMine TOP Cancer Genome Profiling System
- WES
whole exome sequencing
1. Introduction
Cancer immunotherapy with immune checkpoint inhibitors (ICIs) has become a major treatment modality for various cancers [1, 2, 3]. Tumor mutational burden (TMB), defined as the number of somatic mutations per megabase (mut/Mb) in genomic regions analyzed by next‐generation sequencing, is widely recognized as a key biomarker for predicting the response to ICIs [4, 5]. Whole‐exome sequencing (WES) was initially considered the gold standard for TMB assessment. However, its high cost and long turnaround time have limited its application in clinical practice. Instead, TMB is now commonly estimated using targeted panel sequencing that analyzes more than one hundred cancer‐related genes, an approach known as comprehensive genomic profiling (CGP) [6, 7].
In the international clinical trial KEYNOTE‐158 [8], pembrolizumab demonstrated a high objective response rate (ORR) of 32.6% (95% confidence interval [CI], 24.0%–42.7%) in patients with previously treated advanced or recurrent solid tumors that were classified as TMB‐high (≥ 10 mut/Mb), as determined by the FoundationOne CDx (F1CDx) CGP test (Foundation Medicine, Cambridge, MA, USA). Based on these results, a TMB threshold of ≥ 10 mut/Mb has been established as a criterion for identifying patients likely to benefit from ICI therapy using CGP tests. However, the predictive accuracy of TMB in real‐world clinical settings remains insufficiently characterized, as it may be influenced by multiple factors [9, 10], including cancer type and patient ethnicity [11, 12, 13]. Notably, most patients enrolled in KEYNOTE‐158 were of Caucasian descent, and therefore, the predictive performance of TMB in Asian populations remains largely unexplored. In addition, technical differences among CGP tests, such as variation in gene content and the size of the genomic regions analyzed, can lead to discrepancies in TMB estimation [14, 15]. The inclusion or exclusion of hotspot mutations in TMB calculations may also influence the classification of patients as TMB‐high, as we have recently reported [16].
In this study, we investigated the predictive value of TMB for evaluating pembrolizumab efficacy in a Japanese cohort. We analyzed real‐world data from 63,952 patients registered in the Center for Cancer Genomics and Advanced Therapeutics (C‐CAT) database [17], which integrates genomic and clinical information from Japanese patients with a wide range of advanced solid tumors, who underwent CGP under the national health insurance system. We focused on 1899 patients submitted to one of the following CGP tests: F1CDx, the OncoGuide NCC Oncopanel System (NOP; Sysmex, Hyogo, Japan; approved in Japan in June 2019, approval number: 23000BZX00398000), or the GenMine TOP Cancer Genome Profiling System (TOP; GenMine Labs, Tokyo, Japan). Our findings support the clinical utility of TMB as a predictive biomarker in everyday oncology practice and underscore the importance of accounting for hotspot mutations, particularly in patients with TMB values near the threshold.
2. Material and Methods
2.1. Sample Assessment
Between June 1, 2019 and June 15, 2024, a total of 75,663 patients with solid tumors were registered in the C‐CAT database. Of these, 63,952 patients who had undergone one of the following CGP tests, F1CDx, NOP, or TOP, were included in the analysis. F1CDx is a tumor‐only sequencing test, while NOP and TOP are paired tumor‐normal matched tests that use blood‐derived DNA as the normal control. The characteristics of each test are summarized in Table S1. CGP test results were interpreted by molecular tumor boards in hospitals, known as the Expert Panel [18], which are composed of multidisciplinary specialists including clinical oncologists, pathologists, genome scientists, and bioinformaticians. Treatment decisions were made by the attending physicians based on the sequencing results and annotations provided by the Expert Panel. The C‐CAT database includes the treatment records both before and after the CGP test.
Clinical and genomic information, including specimen collection date, treatment history, cancer type, TMB, and somatic mutation profile, was extracted from the C‐CAT database. TMB was classified as TMB‐high (≥ 10 mut/Mb) or TMB‐low (< 10 mut/Mb). The efficacy of ICI treatment was evaluated based on data from patients who received pembrolizumab monotherapy as recorded in the C‐CAT database. Among patients who underwent tumor–normal matched sequencing with TMB ≥ 100, those in whom multiple called somatic mutations coincided with single‐nucleotide polymorphisms (SNP) having a minor allele frequency (MAF) ≥ 10% in the Tohoku Medical Megabank Organization (ToMMo) [19] database, and for which the probability of such coincidences occurring by chance was less than 0.1% were excluded. These patients were considered to represent spurious hypermutation, potentially caused by contamination with germline SNPs from other individuals. The probability of coincidental overlap was calculated based on the total number of somatic mutations estimated from TMB and the total number of ToMMo MAF ≥ 10% SNPs within the panel gene regions. CGP test‐specific analysis for TOP was not conducted owing to the limited number of TMB‐high patients treated with pembrolizumab (n = 5).
According to the Response Evaluation Criteria in Solid Tumors (RECIST) [20], responses were categorized as complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD). For ORR analysis, CR and PR were considered responses, and SD and PD were considered non‐responses. Patients with missing or non‐evaluable response data were excluded from the ORR analysis. For patients who received multiple treatments, the best observed response was used. Patients whose treatment started before the specimen collection date were excluded. The KEYNOTE‐158 ORR was calculated from 92 patients of the efficacy analysis TMB‐high population, after excluding non‐evaluable cases from the 102 patients [8].
All CGP tests were conducted under the Japanese National Health Insurance System and registered in the C‐CAT database with written informed consent from patients. The study was conducted in accordance with the Declaration of Helsinki and was approved by the Information Utilization Review Board of C‐CAT (CDU2023‐022N) and the Ethics Review Committee of Sysmex Corporation (2023‐005).
2.2. Weighted Average Objective Response Rate by Cancer Type
Cancer types were classified according to the first‐level categories in the OncoTree framework [21] as implemented in the C‐CAT database. To adjust for differences in cancer‐type distribution between test groups when evaluating the ORR, we calculated a weighted average ORR for the F1CDx group. This was done by applying the ORRs of each cancer type within the TMB‐high F1CDx subgroup to the corresponding cancer‐type proportions observed in the NOP group. For cancer types not present in the TMB‐high F1CDx subgroup, the overall ORR of the entire TMB‐high F1CDx group was used as a proxy.
2.3. Hotspot Mutation Filtering in TMB Calculation
Hotspot mutations were defined based on their detection frequency in the C‐CAT database and their occurrence count in the Catalogue of Somatic Mutations in Cancer (COSMIC), version 100 [22]. Somatic mutation frequencies, including point mutations, insertions, deletions, insertion‐deletions, and frameshifts, were aggregated from patients tested using tissue‐based CGP platforms (F1CDx, NOP, and TOP), based on data extracted from the C‐CAT database. For tumor‐only sequencing tests, somatic mutations were defined according to the European Society for Medical Oncology (ESMO) guidelines [23]. Specifically, point mutations with an allele frequency (AF) less than 0.3 and other mutations with an AF less than 0.2 were classified as somatic. Mutations were designated as hotspots if they had a detection frequency ≥ 0.06% in the C‐CAT database or a COSMIC count of ≥ 50. The number of hotspot mutations per patient was divided by the panel size of NOP (1.29 Mb; see Table S1) to derive a hotspot‐derived TMB, which was then subtracted from the original NOP TMB to calculate the hotspot‐filtered TMB.
2.4. Statistical Analysis
All statistical analyses were performed using R software (version 4.3.3; The R Foundation for Statistical Computing, Vienna, Austria). Comparisons of categorical variables were conducted using Fisher's exact test, as appropriate. A p‐value < 0.05 was considered statistically significant.
3. Results
3.1. Japanese CGP Cohort and Patient Characteristics
Among the 63,952 patients who underwent tissue‐based CGP and were registered in the C‐CAT database, 1899 received pembrolizumab monotherapy. Of these, 946 and 953 patients were classified as TMB‐high and TMB‐low, respectively (Figure 1A). Within the TMB‐high group, 339 patients received pembrolizumab before CGP testing (pre‐CGP), and 607 received it after CGP testing (post‐CGP). The clinical and genomic characteristics of these 946 patients are summarized in Table 1.
FIGURE 1.
Study design and analysis overview of TMB‐high patients. (A) Study flow diagram showing the inclusion criteria and classification of patients. Cancer‐type distribution among all tested TMB‐high patients, stratified by (B) pre‐CGP and (C) post‐CGP groups. Numbers in parentheses indicate the number of patient samples for each cancer type. (D) Frequency of TMB‐high patients among all tested patients, analyzed by cancer type. CGP, comprehensive genomic profiling; TMB, tumor mutational burden.
TABLE 1.
Clinical and demographic information of the TMB‐high group treated with pembrolizumab monotherapy.
| Category | Overall cohort n = 946 | Post‐CGP n = 607 |
|---|---|---|
| Age, median (range) | 65 (4–91) | 65 (4–90) |
| Gender, n (%) | ||
| Men | 485 (51) | 298 (49) |
| Women | 461 (49) | 309 (51) |
| Treatment line, n (%) | ||
| First line | 114 (12) | 57 (9.4) |
| Second line | 324 (34) | 157 (26) |
| Third line | 203 (22) | 133 (22) |
| Fourth line | 113 (12) | 81 (13) |
| Fifth line and beyond | 171 (18) | 159 (26) |
| No Data | 21 (2.2) | 20 (3.3) |
| MSI status, n (%) | ||
| MSI‐high | 274 (29) | 142 (23) |
| Non‐MSI‐high | 621 (66) | 441 (73) |
| No Data | 51 (5.4) | 24 (4.0) |
| CGP test, n (%) | ||
| FoundationOne CDx | 816 (86) | 526 (87) |
| OncoGuide NCC Oncopanel System | 125 (13) | 80 (13) |
| GenMine TOP Cancer Genome Profiling System | 5 (0.50) | 1 (0.20) |
| Cancer type, n (%) | ||
| Adrenal gland | 5 (0.50) | 3 (0.50) |
| Ampulla of vater | 1 (0.10) | 1 (0.20) |
| Biliary tract | 71 (7.5) | 62 (10) |
| Bladder/Urinary tract | 116 (12) | 26 (4.3) |
| Bowel | 168 (18) | 113 (19) |
| Breast | 52 (5.5) | 51 (8.4) |
| Cervix | 48 (5.1) | 42 (6.9) |
| CNS/Brain | 14 (1.5) | 13 (2.1) |
| Esophagus/Stomach | 25 (2.6) | 18 (3.0) |
| Head and neck | 31 (3.3) | 11 (1.8) |
| Kidney | 2 (0.20) | 0 (0) |
| Liver | 2 (0.20) | 2 (0.30) |
| Lung | 106 (11.2) | 44 (7.2) |
| Ovary/Fallopian tube | 39 (4.1) | 33 (5.4) |
| Pancreas | 35 (3.7) | 32 (5.3) |
| Penis | 5 (0.50) | 5 (0.80) |
| Peripheral nervous system | 1 (0.10) | 1 (0.20) |
| Peritoneum | 4 (0.40) | 4 (0.70) |
| Prostate | 37 (3.9) | 35 (5.8) |
| Skin | 28 (3.0) | 23 (3.8) |
| Soft tissue | 23 (2.4) | 20 (3.3) |
| Testis | 1 (0.10) | 1 (0.20) |
| Thymus | 3 (0.30) | 2 (0.30) |
| Thyroid | 1 (0.10) | 1 (0.20) |
| Uterus | 80 (8.5) | 28 (4.6) |
| Vulva/Vagina | 4 (0.40) | 3 (0.50) |
| Other | 44 (4.7) | 33 (5.4) |
Abbreviations: CGP, comprehensive genomic profiling; CNS, central nervous system; MSI, microsatellite instability; TMB, tumor mutational burden.
The pre‐CGP group included patients with bladder or urinary tract, head and neck, PD‐L1 positive lung, skin, MSI‐high colorectal, and MSI‐high endometrial cancer (Figure 1B). These cancer types are approved for pembrolizumab monotherapy by the United States Food and Drug Administration (FDA), either based on specific biomarkers or as tumor‐type–specific approvals regardless of biomarker status. This suggests that patients in the pre‐CGP group were likely treated based on alternative predictive biomarkers. By contrast, the post‐CGP group included patients with a broader range of cancer types, indicating that TMB‐high status may have served as the primary biomarker guiding treatment decisions in this population (Figure 1C). Among the TMB‐low patients, colorectal, bladder, lung, head and neck, and skin cancers constituted the majority in the pre‐and post‐CGP groups (Figure S1).
The frequency of TMB‐high patients varied substantially across cancer types (Figure 1D). Lung cancer exhibited the highest frequency of TMB‐high cases, consistent with previous reports that smoking‐associated lung cancers tend to show higher TMB values [24, 25].
3.2. Response to Pembrolizumab According to TMB‐High/Low Status
The ORR to pembrolizumab monotherapy was 30.5% (95% CI, 27.7%–33.6%) in the overall C‐CAT cohort of TMB‐high patients (Figure 2). The ORR was 34.2% (95% CI, 29.4%–39.4%) in the pre‐CGP group and 28.5% (95% CI, 25.1%–32.2%) in the post‐CGP group. By contrast, TMB‐low patients had ORRs of 16.7% (95% CI, 14.3%–19.4%) in the pre‐CGP group and 17.4% (95% CI, 11.0%–26.4%) in the post‐CGP group. The differences in ORR between TMB‐high and TMB‐low patients were significant in the overall cohort (p < 0.001) as well as the pre‐CGP (p < 0.001) and post‐CGP (p = 0.0316) subgroups. In the KEYNOTE‐158 trial, the ORR among evaluable TMB‐high patients was 32.6% (95% CI, 24.0%–42.7%), while the ORR for TMB‐low patients was 6.9% (95% CI, 5.2%–9.2%) [8]. The ORR for the C‐CAT TMB‐high group was similar to that in KEYNOTE‐158, although the post‐CGP group showed a slightly but non‐significantly lower ORR.
FIGURE 2.
ORR by TMB status. ORR among all tested TMB‐high patients and TMB‐low patients. Dotted line indicates the ORR reported in KEYNOTE‐158 trial. ORR, objective response rate; TMB, tumor mutational burden. *p < 0.05.
3.3. Response to Pembrolizumab by TMB Range
We further stratified TMB‐high patients into three TMB categories (TMB ≥ 10 mut/Mb and < 13 mut/Mb, TMB 13 mut/Mb and < 40 mut/Mb, and TMB ≥ 40 mut/Mb) and evaluated the corresponding ORRs. In both analyses of the overall cohort and post‐CGP groups, patients with TMB ≥ 13 mut/Mb exhibited ORRs comparable to those in KEYNOTE‐158 (dotted line in Figure 3A, p > 0.05). By contrast, patients with TMB ≥ 10 mut/Mb and < 13 mut/Mb showed a significantly lower ORR (20.8%, Figure 3A) compared to KEYNOTE‐158 (p = 0.019 for overall cohort and p = 0.0025 for post‐CGP group). This trend was consistent across CGP test types (Figure 3B,C), with particularly low ORRs (12.5%, Figure 3C) observed in post‐CGP patients tested with NOP. The cancer‐type distribution differed notably between F1CDx and NOP, especially TMB ≥ 10 mut/Mb and < 13 mut/Mb category of post‐CGP patients, where NOP‐tested patients were enriched for bowel (mostly colorectal) and pancreatic cancers (Figure S2B).
FIGURE 3.
ORRs across TMB categories and CGP tests. (A) ORR by TMB category across all CGP tests, (B) in the F1CDx test, (C) in the NOP test, and (D) ORR among all tested TMB‐high patients, analyzed by cancer type. Dotted line indicates the ORR reported in the KEYNOTE‐158 trial. ORR by TMB category in (E) F1CDx test after adjustment for cancer‐type distribution bias and (F) NOP test after filtering hotspot mutations. CGP, comprehensive genomic profiling; F1CDx, FoundationOne CDx; NOP, OncoGuide NCC Oncopanel System; ORRs, objective response rates; TMB, tumor mutational burden. *p < 0.05.
The ORR to pembrolizumab monotherapy also varied across cancer types (Figure 3D). High ORRs were observed in skin and uterine cancers, in line with previous studies [8, 26]. Notably, TMB‐high patients with colorectal cancer demonstrated a low ORR (20.2%, Figure 3D: Bowel). To correct for cancer‐type bias, we calculated weighted ORRs for F1CDx tested patients. After this adjustment, the ORR for TMB ≥ 10 mut/Mb and < 13 mut/Mb category in post‐CGP patients was reduced to 14.8% in the F1CDx group (Figure 3E), which narrowed the gap with the NOP group (12.5%, Figure 3C).
3.4. Correction of TMB Overestimation in NOP Testing
The NOP test includes hotspot mutations in TMB calculation, whereas F1CDx excludes them (Table S1). As a result, TMB values may be overestimated in NOP‐tested patients, as reflected by the higher frequency of TMB‐high cases across cancer types (Figure S3). To correct for this, we recalculated TMB values in NOP‐tested patients by filtering out hotspot mutations based on data from the C‐CAT and COSMIC databases.
After filtering, the TMB‐high frequency and TMB distribution in the NOP group became more aligned with those of the all‐test cohort (Figure S3). In the recalculated TMB ≥ 10 mut/Mb and < 13 mut/Mb category, the ORR increased to 25.8% (95% CI, 13.7%–43.3%) in the overall cohort and 18.2% (95% CI, 7.3%–38.5%) in the post‐CGP group (Figure 3F), levels comparable to those of F1CDx‐tested patients. This increase was largely owing to the reclassification of 15 non‐responders, many of whom had colorectal or pancreatic cancer. In colorectal cancer, filtered hotspot mutations included TP53, KRAS, and APC, while in pancreatic cancer, TP53 and KRAS mutations were the most frequently excluded (Table S2).
4. Discussion
In this study, we investigated the predictive value of TMB for evaluating pembrolizumab efficacy in a Japanese cohort by using real‐world data. Importantly, our analysis focused on a Japanese population, addressing a significant gap in existing data, as most previous studies have primarily involved patients of European descent. Given that hotspot mutation profiles differ between Japanese and Caucasian populations, as previously reported [27, 28, 29], the distribution of TMB and the composition of patient subgroups likely to respond may also vary across populations. Therefore, the Japanese population‐specific insights may contribute to refining the use of TMB as a biomarker in clinical practice, thereby facilitating more precise identification of patient subgroups most likely to benefit from ICI therapy.
The majority of patients in the post‐CGP group likely received ICI treatment based on their TMB status. Therefore, the clinical relevance of TMB as a predictive biomarker is particularly important in this group, and the observed outcomes directly reflect its real‐world applicability in Japan. Notably, the ORR to pembrolizumab monotherapy among Japanese patients with TMB‐high tumors was comparable to that reported in the KEYNOTE‐158 trial [8]. This consistency suggests that the clinical trial results are reproducible in real‐world settings, reinforcing the validity of using TMB to guide patient selection during tumor board discussions. Accordingly, our findings provide a valuable reference for interpreting TMB in future clinical decision‐making.
The commonly used TMB cutoff of 10 mut/Mb was originally established to broaden access to pembrolizumab therapy, given the relatively low prevalence of TMB‐high tumors. In KEYNOTE‐158, however, response varied with TMB level: patients with TMB ≥ 13 mut/Mb had an ORR of 37%, whereas those with TMB ≥ 10 mut/Mb and < 13 mut/Mb had a much lower ORR of 13% [30]. In its Approval Summary based on KEYNOTE‐158, the FDA acknowledged the lower response rate in the TMB ≥ 10 mut/Mb and < 13 mut/Mb subgroup and noted that the use of the 10 mut/Mb threshold remains controversial. A similar pattern was observed in our cohort, where patients with TMB values between TMB ≥ 10 mut/Mb and < 13 mut/Mb exhibited lower response rates than those with higher TMB. This parallel further supports the reproducibility of the KEYNOTE‐158 findings in real‐world practice.
While overall consistency was observed, the ORR in the post‐CGP group was slightly lower than that in the trial, although not significantly so. Stratified analysis by CGP test revealed that patients tested with the NOP panel, especially those with TMB values near the cutoff, tended to have lower ORRs. Although we adjusted for differences in cancer‐type distribution, these discrepancies could not be fully explained by tumor type alone. Rather, they likely reflect differences in how each CGP test handles hotspot mutations in TMB calculation.
According to the Friends of Cancer Research TMB harmonization study [15], more than half of the 16 CGP tests evaluated did not apply hotspot mutation filtering, indicating that tests without filtering are frequently used in clinical practice. In our study, we retrospectively applied hotspot filtering, using C‐CAT and COSMIC annotations, to tests that originally lacked this feature and directly assessed its impact on TMB‐based prediction of pembrolizumab efficacy. The ORR was higher in the TMB‐high group, redefined after filtering, than in the unfiltered group. The average TMB reduction after filtering was −1.5 mut/Mb, and all patients who were reclassified from TMB‐high to TMB‐low had baseline TMB values near the cutoff (TMB ≥ 10 mut/Mb and < 13 mut/Mb). Notably, most of these reclassified patients were non‐responders and had colorectal or pancreatic cancers (Table S2). When tumor types were ranked by the number of recurrently detected mutations (mutation frequency ≥ 1%) in the C‐CAT database, colorectal (27 mutations) and pancreatic (23 mutations) cancers ranked among the highest (Figure S4). These findings suggest that hotspot mutation filtering may enhance the predictive accuracy of TMB, particularly for tumor types with a high burden of recurrent mutations and for patients with borderline TMB values.
This study has several limitations. First, the sample size, especially in subgroup analyses, was relatively small, limiting statistical power and possibly contributing to some non‐significant findings. Second, treatment histories were not fully captured; factors such as prior ICI exposure or rechallenge were not accounted for. These variables may have affected treatment outcomes and contributed to the underestimation of ORR in the post‐CGP group. Future studies with more detailed treatment data will be essential to accurately assess real‐world ICI effectiveness.
In conclusion, our findings confirm the clinical utility of TMB as a predictive biomarker in a Japanese real‐world setting and underscore the importance of accounting for hotspot mutations in TMB assessment, particularly in patients with TMB values near the cutoff.
Author Contributions
Tomoyo Yasuda: investigation, data curation, formal analysis, writing – original draft, writing – review and editing. Mio Yumura: investigation, methodology, writing – review and editing. Azusa Hamasaki: data curation. Tongyi Fei: data curation. Takashi Kubo: writing – review and editing. Hitoshi Ichikawa: writing – review and editing. Takashi Kohno: writing – review and editing. Kuniko Sunami: writing – review and editing.
Funding
This work was supported by joint research funding from Sysmex Corporation [C2019‐071].
Ethics Statement
Approval of the research protocol by Review Board: This study was approved by the Information Utilization Review Board of C‐CAT (CDU2023‐022N) and the Ethics Review Committee of Sysmex Corporation (2023‐005).
Consent
All participants provided written informed consent.
Conflicts of Interest
Tomoyo Yasuda, Mio Yumura, Azusa Hamasaki, and Tongyi Fei are employees of Sysmex Corporation. Takashi Kohno and Kuniko Sunami received research funds from Sysmex Corporation. Takashi Kohno and Kuniko Sunami are editorial board members of Cancer Science.
Supporting information
Figure S1: Cancer‐type distributions among TMB‐low patients. Cancer‐type distributions among all tested TMB‐low patients in (A) pre‐CGP group and (B) post‐CGP group. Numbers in parentheses indicate the number of patient samples for each cancer type. CGP, comprehensive genomic profiling; TMB, tumor mutational burden.
Figure S2: Cancer‐type distributions among patients near the TMB cutoff. Cancer‐type distribution in post‐CGP patients with TMB ≥ 10 mut/Mb and < 13 mut/Mb in (A) F1CDx and (B) NOP. Numbers in parentheses indicate the number of patient samples for each cancer type. CGP, comprehensive genomic profiling; F1CDx, FoundationOne CDx; NOP, OncoGuide NCC Oncopanel System; TMB, tumor mutational burden.
Figure S3: Changes in TMB‐high frequency following hotspot mutation filtering. Frequency of TMB‐high patients across cancer types, comparing All tests, original NOP, and NOP recalculated after hotspot mutation filtering. NOP, OncoGuide NCC Oncopanel System; TMB, tumor mutational burden.
Figure S4: Ranking of cancer types by the number of frequently detected mutations in CGP tests. Mutations with a detection frequency of ≥ 1% are included. Red bars indicate notable hotspot mutations with a detection frequency of ≥ 5%. CGP, comprehensive genomic profiling.
Table S1: Characteristics of F1CDx, NOP, and TOP. F1CDx, FoundationOne CDx; NOP, OncoGuide NCC Oncopanel System; TOP, GenMine TOP Cancer Genome Profiling System.
Table S2: Impact of hotspot mutation filtering on TMB value and classification in the post‐CGP NOP testing group. CGP, comprehensive genomic profiling; NOP, OncoGuide NCC Oncopanel System; TMB, tumor mutational burden.
Acknowledgments
Editorial support, in the form of medical writing, assembly of tables, creation of high‐resolution images based on authors' detailed directions, collating of author comments, copyediting, fact‐checking, and referencing, was provided by Editage, Cactus Communications. The results published here are in part based upon data generated by C‐CAT (https://www.ncc.go.jp/en/c_cat/about). The authors would like to thank Mamoru Kato, Eisaku Furukawa, and Momoko Nagai for sharing their time during regular clinical sequence meetings, which offered helpful opportunities for discussion.
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.
References
- 1. Emens L. A., Romero P. J., Anderson A. C., et al., “Challenges and Opportunities in Cancer Immunotherapy: A Society for Immunotherapy of Cancer (SITC) Strategic Vision,” Journal for Immunotherapy of Cancer 12, no. 6 (2024): e009063, 10.1136/jitc-2024-009063. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Zemek R. M., Anagnostou V., Pires da Silva I., Long G. V., and Lesterhuis W. J., “Exploiting Temporal Aspects of Cancer Immunotherapy,” Nature Reviews. Cancer 24, no. 7 (2024): 480–497, 10.1038/s41568-024-00699-2. [DOI] [PubMed] [Google Scholar]
- 3. Waldman A. D., Fritz J. M., and Lenardo M. J., “A Guide to Cancer Immunotherapy: From T Cell Basic Science to Clinical Practice,” Nature Reviews Immunology 20, no. 11 (2020): 651–668, 10.1038/s41577-020-0306-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Budczies J., Kazdal D., Menzel M., et al., “Tumour Mutational Burden: Clinical Utility, Challenges and Emerging Improvements,” Nature Reviews. Clinical Oncology 21, no. 10 (2024): 725–742, 10.1038/s41571-024-00932-9. [DOI] [PubMed] [Google Scholar]
- 5. Dall'Olio F. G., Marabelle A., Caramella C., et al., “Tumour Burden and Efficacy of Immune‐Checkpoint Inhibitors,” Nature Reviews Clinical Oncology 19, no. 2 (2022): 75–90, 10.1038/s41571-021-00564-3. [DOI] [PubMed] [Google Scholar]
- 6. Sha D., Jin Z., Budczies J., Kluck K., Stenzinger A., and Sinicrope F. A., “Tumor Mutational Burden as a Predictive Biomarker in Solid Tumors,” Cancer Discovery 10, no. 12 (2020): 1808–1825, 10.1158/2159-8290.CD-20-0522. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Özdoğan M., Papadopoulou E., Tsoulos N., et al., “Comprehensive Tumor Molecular Profile Analysis in Clinical Practice,” BMC Medical Genomics 14, no. 1 (2021): 105, 10.1186/s12920-021-00952-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Marabelle A., Fakih M., Lopez J., et al., “Association of Tumour Mutational Burden With Outcomes in Patients With Advanced Solid Tumours Treated With Pembrolizumab: Prospective Biomarker Analysis of the Multicohort, Open‐Label, Phase 2 KEYNOTE‐158 Study,” Lancet Oncology 21, no. 10 (2020): 1353–1365, 10.1016/S1470-2045(20)30445-9. [DOI] [PubMed] [Google Scholar]
- 9. Killock D., “TMB – A Histology‐Agnostic Predictor of the Efficacy of ICIs?,” Nature Reviews Clinical Oncology 17, no. 12 (2020): 718, 10.1038/s41571-020-00438-0. [DOI] [PubMed] [Google Scholar]
- 10. Addeo A., Friedlaender A., Banna G. L., and Weiss G. J., “TMB or Not TMB as a Biomarker: That Is the Question,” Critical Reviews in Oncology/Hematology 163 (2021): 103374, 10.1016/j.critrevonc.2021.103374. [DOI] [PubMed] [Google Scholar]
- 11. Huang T., Chen X., Zhang H., et al., “Prognostic Role of Tumor Mutational Burden in Cancer Patients Treated With Immune Checkpoint Inhibitors: A Systematic Review and Meta‐Analysis,” Frontiers in Oncology 11 (2021): 706652, 10.3389/fonc.2021.706652. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Horie S., Saito Y., Kogure Y., et al., “Pan‐Cancer Comparative and Integrative Analyses of Driver Alterations Using Japanese and International Genomic Databases,” Cancer Discovery 14, no. 5 (2024): 786–803, 10.1158/2159-8290.CD-23-0902. [DOI] [PubMed] [Google Scholar]
- 13. Wu L., Yao H., Chen H., et al., “Landscape of Somatic Alterations in Large‐Scale Solid Tumors From an Asian Population,” Nature Communications 13, no. 1 (2022): 4264, 10.1038/s41467-022-31780-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Merino D. M., McShane L. M., Fabrizio D., et al., “Establishing Guidelines to Harmonize Tumor Mutational Burden (TMB): In Silico Assessment of Variation in TMB Quantification Across Diagnostic Platforms: Phase I of the Friends of Cancer Research TMB Harmonization Project,” Journal for Immunotherapy of Cancer 8, no. 1 (2020): e000147, 10.1136/jitc-2019-000147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Vega D. M., Yee L. M., McShane L. M., et al., “Aligning Tumor Mutational Burden (TMB) Quantification Across Diagnostic Platforms: Phase II of the Friends of Cancer Research TMB Harmonization Project,” Annals of Oncology 32, no. 12 (2021): 1626–1636, 10.1016/j.annonc.2021.09.016. [DOI] [PubMed] [Google Scholar]
- 16. Nishino T., Yumura M., Sunami K., et al., “Leveraging Non‐Coding Regions to Guarantee the Accuracy of Small‐Sized Panel‐Based Tumor Mutational Burden Estimates,” Cancer Science 115, no. 12 (2024): 4045–4053, 10.1111/cas.16342. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Kohno T., Kato M., Kohsaka S., et al., “C‐CAT: The National Datacenter for Cancer Genomic Medicine in Japan,” Cancer Discovery 12, no. 11 (2022): 2509–2515, 10.1158/2159-8290.CD-22-0417. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Sunami K., Naito Y., Aimono E., et al., “The Initial Assessment of Expert Panel Performance in Core Hospitals for Cancer Genomic Medicine in Japan,” International Journal of Clinical Oncology 26, no. 3 (2021): 443–449, 10.1007/s10147-020-01844-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Tadaka S., Hishinuma E., Komaki S., et al., “jMorp Updates in 2020: Large Enhancement of Multi‐Omics Data Resources on the General Japanese Population,” Nucleic Acids Research 49, no. D1 (2021): D536–D544, 10.1093/nar/gkaa1034. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Eisenhauer E. A., Therasse P., Bogaerts J., et al., “New Response Evaluation Criteria in Solid Tumours: Revised RECIST Guideline (Version 1.1),” European Journal of Cancer 45, no. 2 (2009): 228–247, 10.1016/j.ejca.2008.10.026. [DOI] [PubMed] [Google Scholar]
- 21. Kundra R., Zhang H., Sheridan R., et al., “OncoTree: A Cancer Classification System for Precision Oncology,” JCO Clinical Cancer Informatics 5 (2021): 221–230, 10.1200/CCI.20.00108. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Tate J. G., Bamford S., Jubb H. C., et al., “COSMIC: The Catalogue of Somatic Mutations in Cancer,” Nucleic Acids Research 47, no. D1 (2019): D941–D947, 10.1093/nar/gky1015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Kuzbari Z., Bandlamudi C., Loveday C., et al., “Germline‐Focused Analysis of Tumour‐Detected Variants in 49,264 Cancer Patients: ESMO Precision Medicine Working Grouprecommendations,” Annals of Oncology 34, no. 3 (2023): 215–227, 10.1016/j.annonc.2022.12.003. [DOI] [PubMed] [Google Scholar]
- 24. Zaarour R. F., Sharda M., Azakir B., et al., “Genomic Analysis of Waterpipe Smoke‐Induced Lung Tumor Autophagy and Plasticity,” International Journal of Molecular Sciences 23, no. 12 (2022): 6848, 10.3390/ijms23126848. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Nagahashi M., Sato S., Yuza K., et al., “Common Driver Mutations and Smoking History Affect Tumor Mutation Burden in Lung Adenocarcinoma,” Journal of Surgical Research 230 (2018): 181–185, 10.1016/j.jss.2018.07.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Robert C., Schachter J., Long G. V., et al., “Pembrolizumab Versus Ipilimumab in Advanced Melanoma,” New England Journal of Medicine 372, no. 26 (2015): 2521–2532, 10.1056/NEJMoa1503093. [DOI] [PubMed] [Google Scholar]
- 27. Nagahashi M., Wakai T., Shimada Y., et al., “Genomic Landscape of Colorectal Cancer in Japan: Clinical Implications of Comprehensive Genomic Sequencing for Precision Medicine,” Genome Medicine 8, no. 1 (2016): 136, 10.1186/s13073-016-0387-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Egashira A., Morita M., Kakeji Y., et al., “ p53 Gene Mutations in Esophageal Squamous Cell Carcinoma and Their Relevance to Etiology and Pathogenesis: Results in Japan and Comparisons With Other Countries,” Cancer Science 98, no. 8 (2007): 1152–1156, 10.1111/j.1349-7006.2007.00524.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Ko R., Kenmotsu H., Serizawa M., et al., “Frequency of EGFR T790M Mutation and Multimutational Profiles of Rebiopsy Samples From Non‐Small Cell Lung Cancer Developing Acquired Resistance to EGFR Tyrosine Kinase Inhibitors in Japanese Patients,” BMC Cancer 16, no. 1 (2016): 864, 10.1186/s12885-016-2902-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Marcus L., Fashoyin‐Aje L. A., Donoghue M., et al., “FDA Approval Summary: Pembrolizumab for the Treatment of Tumor Mutational Burden–High Solid Tumors,” Clinical Cancer Research 27, no. 17 (2021): 4685–4689, 10.1158/1078-0432.CCR-21-0327. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure S1: Cancer‐type distributions among TMB‐low patients. Cancer‐type distributions among all tested TMB‐low patients in (A) pre‐CGP group and (B) post‐CGP group. Numbers in parentheses indicate the number of patient samples for each cancer type. CGP, comprehensive genomic profiling; TMB, tumor mutational burden.
Figure S2: Cancer‐type distributions among patients near the TMB cutoff. Cancer‐type distribution in post‐CGP patients with TMB ≥ 10 mut/Mb and < 13 mut/Mb in (A) F1CDx and (B) NOP. Numbers in parentheses indicate the number of patient samples for each cancer type. CGP, comprehensive genomic profiling; F1CDx, FoundationOne CDx; NOP, OncoGuide NCC Oncopanel System; TMB, tumor mutational burden.
Figure S3: Changes in TMB‐high frequency following hotspot mutation filtering. Frequency of TMB‐high patients across cancer types, comparing All tests, original NOP, and NOP recalculated after hotspot mutation filtering. NOP, OncoGuide NCC Oncopanel System; TMB, tumor mutational burden.
Figure S4: Ranking of cancer types by the number of frequently detected mutations in CGP tests. Mutations with a detection frequency of ≥ 1% are included. Red bars indicate notable hotspot mutations with a detection frequency of ≥ 5%. CGP, comprehensive genomic profiling.
Table S1: Characteristics of F1CDx, NOP, and TOP. F1CDx, FoundationOne CDx; NOP, OncoGuide NCC Oncopanel System; TOP, GenMine TOP Cancer Genome Profiling System.
Table S2: Impact of hotspot mutation filtering on TMB value and classification in the post‐CGP NOP testing group. CGP, comprehensive genomic profiling; NOP, OncoGuide NCC Oncopanel System; TMB, tumor mutational burden.
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.