Effect of Clonal Hematopoiesis Mutations and Canakinumab Treatment on Incidence of Solid Tumors in the CANTOS Randomized Clinical Trial

Clonal hematopoiesis (CH) occurs when a hematopoietic stem cell (HSC) or group of HSCs contributes disproportionately to blood cell production; this may occur due to enhanced survival or proliferation mediated by the acquisition of somatic mutations that are associated with hematological malignancy (1, 2). Somatic mutations accumulate in HSCs over time, and as a result, CH is more common in older age (3, 4). At least 5% to 10% of people aged 65 to 70 years or older have mutations associated with hematological malignancy at a variant allele fraction (VAF) of ≥2% in individuals without a diagnosed hematologic disorder or unexplained cytopenia [thus meeting the definition of CH of indeterminate potential (CHIP)], compared with ∼1% of those younger than 50 years (3, 5, 6). Although this phenomenon is associated with (and increases the risk of) hematological cancers such as myelodysplastic syndromes, most people with CH mutations do not develop hematological malignancies or have hematological dysplasia or cytopenias (2). However, CHIP is associated with an increased risk of all-cause mortality, largely driven by cardiovascular events in the general population and by disease progression among patients with non-hematological malignancies (5, 7).

The most commonly mutated genes in CHIP are DNMT3A, TET2, and ASXL1, which have previously been implicated in hematological cancers (1, 2, 4, 5). The DNMT3A and TET2 genes encode enzymes involved in DNA methylation and epigenetic regulation, and mutations in these genes have been found to promote a pro-inflammatory state, leading to abnormal hematopoiesis and atherosclerotic cardiovascular disease (8–10). The presence of different CH mutations has been correlated with increased levels of inflammatory markers; for instance, SF3B1 and JAK2 mutations were associated with elevated circulating interleukin 18 (IL18), whereas mutations in the TET2 gene were associated with increased circulating levels of IL1β (4).

HSCs carrying CH mutations give rise to mature immune cells whose function may be altered; this suggests that the presence of CH mutations may influence the course of several diseases, particularly those with an inflammatory component such as atherosclerotic cardiovascular disease and certain malignancies (1). Whole-genome sequencing of blood-derived DNA from 97,691 individuals of diverse ancestries in the US National Heart, Lung, and Blood Institute TOPMed program revealed a correlation between the presence of CH mutations and a significantly increased risk of both hematological (myeloid neoplasms) and non-hematological (e.g., uterine leiomyoma and brain cancer) malignancies (4). Similar results were obtained from a phenome-wide association study on blood samples from 200,453 UK Biobank participants, which revealed an association between CH mutations and myeloid malignancies, as well as increased risks of hematological and non-hematological neoplasias including lymphoma, lung, and kidney cancer in patients with CH mutations (11). Longitudinal and Mendelian randomization analyses of 5,041 health traits from the UK Biobank revealed that CHIP is associated with solid cancers, including non-melanoma skin cancer and lung cancer (12). Finally, a study of whole-exome sequencing data from two different biobank sources found that CH mutations were associated with an increased risk of lung cancer, independent of known risk factors such as polygenic risk score, history of chronic obstructive pulmonary disease, or inflammatory status (13). However, the mechanisms underlying these effects are unknown and the role of CH mutations in non-hematological malignancies remains to be elucidated.

The fully human, neutralizing anti-IL1β antibody canakinumab has been approved for the treatment of diseases of inflammatory origin such as cryopyrin-associated periodic syndromes and juvenile idiopathic arthritis (14). The Canakinumab Anti-inflammatory Thrombosis Outcome Study (CANTOS; ClinicalTrials.gov NCT01327846) evaluated canakinumab versus placebo in 10,061 randomized patients with a history of myocardial infarction and persistent inflammation (baseline levels of high-sensitivity C-reactive protein (hsCRP) ≥2 mg/L; ref. 15). Patients were administered canakinumab at different doses or placebo and followed up for a median of 3.7 years. Treatment with canakinumab resulted in a dose-dependent reduction in cardiovascular events; moreover, canakinumab treatment across all doses was also associated with significantly lower all-cancer mortality versus placebo (P = 0.0007; refs. 16, 17). Lung cancer mortality was also significantly reduced in the canakinumab treatment group (all doses pooled) versus placebo (P = 0.0002). Although there was no significant decrease in the incidence of cancer at other primary sites, the incidence of lung cancer decreased dose-dependently for patients receiving canakinumab [hazard ratios vs. placebo: 0.74 (95% CI, 0.47–1.17; P = 0.20) for the 50 mg dose, 0.61 (95% CI, 0.39–0.97; P = 0.034) for the 150 mg dose, and 0.33 (95% CI, 0.18–0.59; P < 0.0001) for the 300 mg dose; ref. 16]. In addition, a CANTOS exploratory genomic substudy revealed that canakinumab treatment resulted in larger decreases in major adverse cardiovascular events in patients with TET2 mutations than in those without CH mutations, suggesting that targeting the IL1β pathway may provide an enhanced clinical benefit to patients with certain CH mutations (18).

The clinical benefit of canakinumab in terms of reduced incidence of non-hematological malignancies in the context of CH mutations remains to be elucidated. Here, we aimed to retrospectively determine whether CH mutational status and canakinumab treatment were associated with the incidence of non-hematological malignancy rates among patients enrolled in CANTOS.

CANTOS (ClinicalTrials.gov NCT01327846) was a randomized, double-blind, placebo-controlled phase III study conducted between April 2011 and June 2017. The study design has been described in full in a previous publication (15). Briefly, eligible patients were adults with a history of myocardial infarction and circulating hsCRP ≥2 mg/L. Patients were excluded from the trial if they had any prior malignancy other than basal cell skin carcinoma, a history of recurrent or chronic infectious disease and/or immunosuppression, or ongoing use of anti-inflammatory systemic treatments. A total of 10,061 patients were enrolled across 39 countries and randomized 1.5:1:1:1 to receive placebo or canakinumab at doses 50, 150, or 300 mg subcutaneously every 3 months. The primary endpoint was the first occurrence of nonfatal myocardial infarction, any nonfatal stroke, or cardiovascular death; the results of the primary analysis have been published (15). Local institutional review board approval was required and obtained. Written informed consent was provided by all patients and the study was conducted in accordance with Declaration of Helsinki and Consolidated Standards of Reporting Trials.

Targeted genomic sequencing was carried out as previously described in the subset of patients who provided their consent for DNA collection and analysis (18). Patients who were not included in this genomic substudy either declined to provide consent or had various reasons for not participating. Briefly, 3,946 DNA samples isolated from whole blood taken from patients at the baseline visit were sequenced using the SureSelect custom gene sequencing panel (Agilent) on a HiSeq 2500 sequencer (Illumina), which targeted a prespecified list of variants in 74 genes; these variants have repeatedly been described in hematological malignancies and have been employed in the identification of CH (4, 5, 19). Of all sequenced samples, 3,923 passed quality control analysis and were included in further studies. Incidence of malignancy was assessed as per safety data collected in CANTOS: all reported cancer events are presented; categories and classification levels were based on the Medical Dictionary for Regulatory Activities, version 20.0. Patients may have had multiple cancer events reported.

Demographic and other baseline characteristics were summarized using mean (standard deviation) and median (interquartile range). Patients across all doses of canakinumab (50, 150, or 300 mg) were pooled for all analyses. For each malignancy, incidence was summarized by canakinumab treatment and CHIP status. A Kaplan–Meier plot was used to estimate the cumulative incidence of malignancy events and a log-rank test was used to calculate the P-value for comparison of CHIP status or canakinumab treatment. A multivariable analysis using a logistic regression model was also carried out to check and adjust any prognostic/confounding factors. The response variable used was at least one reported malignancy (yes vs. no), whereas baseline covariates tested in the model included CH mutation status, hsCRP levels at baseline, age, gender, smoker status (current smoker and former smoker vs. never smoker), and treatment (pooled canakinumab vs. placebo). All analyses were conducted using SAS version 9.4 TS1M6 (Statistical Analysis System, RRID: SCR_008567) and R version 4.1.0 (R Base, RRID:SCR_001905).

The CANTOS study was approved by institutional review boards at all sites, and all patients provided written informed consent.

Novartis is committed to sharing with qualified external researchers, access to patient-level data and supporting clinical documents from eligible studies. These requests are reviewed and approved by an independent review panel based on scientific merit. All data provided are anonymized to respect the privacy of patients who have participated in the trial in line with applicable laws and regulations.

As previously reported, there was a significant decrease in lung cancer incidence among patients in CANTOS treated with canakinumab versus those treated with placebo (16); this difference was not significant for other cancer types. Supplementary Table S1 summarizes the incidence of all non-hematological malignancies reported in patients in CANTOS by treatment type [canakinumab (all doses pooled, 50, 150, and 300 mg doses] or placebo]. A dose-dependent reduction in lung cancer incidence was observed for canakinumab-treated patients.

Approximately 8% of patients in CANTOS had CH mutations; these patients were equally distributed between the pooled canakinumab doses and placebo arms (Supplementary Table S2). Mutations in TET2 and DNMT3A were the most common, accounting for approximately 30% and 25% of the total CH mutations, respectively. As noted in previous studies (4, 5, 12), the majority of patients bear a solitary mutation, and only 8.2% (28 out of 338 patients) exhibit multiple mutations associated with CH.

To assess the impact of specific CH mutations on the incidence of malignancy, patients were stratified by type of CH mutation into the following groups: no CH mutations, TET2 mutations, DNMT3A mutations, and other CH mutations (defined as CH mutations other than mutations in TET2 or DNMT3A; ref. 18). These groups were subsequently stratified by treatment received; patients treated with canakinumab at different doses were pooled for the analysis. Baseline characteristics of patients stratified by type of CH mutations and treatment are presented in Supplementary Table S3. As previously reported (20), patients with TET2, DNMT3A, or other CH mutations were older than patients without CH mutations. Sex and the presence of comorbidities such as diabetes were balanced across patient groups, whereas the presence of CH mutations was associated with higher circulating levels of IL6 and tumor necrosis factor-α (TNF-α) at baseline, suggesting CH-mediated activation of systemic inflammation. The proportion of current smokers varied across patient groups, with the highest proportion observed among patients with DNMT3A mutations treated with canakinumab. Baseline characteristics between all patients in CANTOS and this genomic substudy are comparable and well-balanced, as shown in Supplementary Table S4.

In terms of specific CH mutations, patients with TET2 mutations treated with canakinumab had the lowest incidence of malignancy across cancer types compared with any other patient group (Table 1). The incidence of at least one malignancy was lower among patients with TET2 mutations treated with canakinumab versus those treated with placebo (Fig. 1A). The highest incidence rates of lung malignancy were observed in placebo-treated patients with DNMT3A and TET2 mutations; treatment with canakinumab was associated with a reduction in the risk of lung malignancy in these patients, since incidence rates were lower in these groups (Fig. 1B). Similarly, the highest incidence rate of non-melanoma skin cancer was observed in patients with TET2 mutations treated with placebo; this rate was significantly higher than for patients with any other mutations or without CH mutations treated with placebo. Similar to lung cancer, treatment with canakinumab was significantly associated with decreased incidence of non-melanoma skin cancer in patients with TET2 mutations (Fig. 1C).

A multivariate analysis was also carried out to identify potential prognostic or confounding factors (Fig. 2). For patients with at least one reported malignancy, the significant prognostic factors at baseline identified with the model were age, sex, hsCRP levels at baseline, and smoker status. With every increase of 1 year in age, the odds of having at least one malignancy increased by 1.05-fold, whereas every increase of 1 unit in hsCRP levels increased these odds by 1.02-fold. Patients with current smoker status had increased odds (2.16-fold) of having at least one malignancy compared with patients who had never smoked; former smokers also showed an increase in the risk of malignancy compared with never smokers, but this was not statistically significant. The study did not identify an association between the presence of CH mutations and the incidence of malignancy. This lack of association is likely attributed to the small sample size, despite the consistent trend observed in both previous reports (4, 11–13) and the current study.

The cumulative incidence of non-hematological malignancy over time was assessed by Kaplan–Meier analysis using time to first malignancy event for different patient groups. When all patients were analyzed, there were no significant differences in the cumulative incidence of at least one reported malignancy following treatment with canakinumab or placebo, regardless of the presence of CH mutations and canakinumab doses (Fig. 3A; Supplementary Fig. S1A ); this is in agreement with previously published results (16). However, the cumulative incidence of at least one reported malignancy was significantly lower for patients with TET2 mutations treated with canakinumab versus those treated with placebo (P = 0.019; Fig. 3B). No significant differences were observed between treatment arms for patients with other CH mutations or for patients without mutations. Furthermore, no dose-dependent response was noted in other mutations or in the absence of mutations. However, a lower incidence of malignancy was observed in patients with TET2 mutations who received a higher dose of canakinumab compared to the placebo (P = 0.03; canakinumab 300 mg vs. placebo, Supplementary Fig. S1B).

While incident lung cancer was significantly less frequent in both the 150 and 300 mg groups among all CANTOS patients (16), it is noteworthy that a reduction in incident lung cancer was specifically observed in the 300 mg group among patients without CH mutations (Supplementary Fig. S2). This observation suggests that individuals without CH mutations may not be as responsive to canakinumab, and their clinical benefits might not be as pronounced as those with CH mutations. The evaluation of dose dependency in patients with CH mutations was not conducted due to the small sample size.

Previous studies assessing the impact of CH mutations in patients with cancer found that CH mutations were associated with a high mortality rate. The observed increase in mortality in patients with non-hematological cancers and CH mutations could be attributed to primary malignancy progression (7). This exploratory analysis of the prospective CANTOS trial assessed cancer incidence as a surrogate for tumor development in the presence or absence of CH mutations and measured the impact of IL1β inhibition on cancer incidence. Our results show that patients with CH mutations may have a higher risk of non-hematological malignancy across different cancer types compared with those who do not harbor these mutations. This is in agreement with previously reported results (4, 11–13). However, the multivariate analysis did not reveal a significant association of CH mutation with malignancy, despite identifying age, hs-CRP levels, and smoking status as factors linked to an increased risk of having at least one malignancy. Considering the association between age and CH, we conducted a multivariate analysis excluding age. Although the association of CH mutation with malignancy became more robust (OR, 1.25; 95% CI, 0.86–1.81), it still did not reach statistical significance. Given the consistent trend observed in previous reports, the lack of a significant association is likely attributable to the limitations of the sample size.

Inhibition of IL1β appeared to have an impact on the incidence of malignancies in the context of CH mutations. Patients with TET2 mutations in the canakinumab treatment arm had the lowest overall cancer incidence of all patient groups. The observed preferential response to canakinumab in patients with TET2 mutations aligns with a recent analysis involving patients with concurrent anemia and CH mutations treated with canakinumab. This analysis demonstrated a more pronounced improvement in anemia among patients with CH mutations treated with canakinumab (20). These findings suggest a cooperative interaction between CH mutations and inflammatory pathways. Patients with CH mutations may exhibit increased sensitivity to IL1β inhibition, particularly when the inflammatory pathway is implicated in the specific disease of interest, such as anemia of inflammation (20, 21) or lung cancer (22).

Our results show an association between the presence of CH mutations and increased incidence of non-hematological malignancies but do not provide proof of causation or evidence of a specific underlying mechanism due to the exploratory nature of the analysis. However, these findings are in agreement with previous genomic studies (4, 11, 13). CH mutations have been associated with defects in DNA repair and the accumulation of further mutations (4); this could be a possible mechanism through which CH mutations increase the risk of malignancy. Moreover, accumulation of damaged DNA can also induce tumor-promoting inflammation (23). The presence of CH mutations has also been correlated with a pro-inflammatory state, with increased levels of circulating IL6 and IL1β; these cytokines have in turn been associated with different steps in the development of malignancy, such as mutagenesis, stimulation of cancer cell proliferation, cell survival, angiogenesis, invasiveness, modulation of the tumor microenvironment, and evasion of the antitumor immune response (23–26). Indeed, molecular characterization of patients in CANTOS who developed lung cancer during the study showed that patients with increased inflammation markers associated with the IL1β pathway at baseline had a shorter time to diagnosis of lung cancer, demonstrating the importance of IL1β inhibition for these patients (17). It has also previously been shown that inhibition of cytokine signaling can decrease tumorigenesis and tumor growth (23). Thus it seems possible that canakinumab treatment could result in a reduced incidence of malignancies. In particular, mutations in the TET2 gene have been associated with increased circulating levels of IL1β in humans (4) and increased levels of IL1 and IL6 cytokines in mice (10). It is, therefore, likely that patients with these mutations might benefit the most from IL1β inhibition. In addition, a recent report showed that air pollutants mediate macrophage influx and release of IL1β into the lung, which eventually promotes tumorigenesis within EGFR-mutant lung epithelial cells (22); this suggests that pro-inflammatory signals mediated by IL1β can enhance lung cancer development. It is conceivable that CH mutations can cooperate with environmental factors and other genetic mutations to promote lung cancer via systemic activation of inflammatory signals such as IL1β, which may be targeted by canakinumab to suppress lung cancer development.

This study has several limitations. The first is its exploratory nature: the incidence of malignancy was not a primary endpoint of the CANTOS trial. CH mutations were assessed retrospectively, and even though the dataset was large, genomic samples were available from only 39% of patients overall, those who consented to biomarker genomic substudies. However, overall clinical characteristics are well balanced between all patients in CANTOS and patients included in this genomic substudy (Supplementary Table S3). Furthermore, it is worth noting that CANTOS did not specifically assess the presence of primary malignancies in enrolled patients before the initiation of study treatment, such as through dedicated imaging methods. Therefore, it is possible that some patients may have had undiagnosed malignancies at the time of trial enrollment. Additionally, the incidence of cancer was not formally specified as a study endpoint in CANTOS, which could introduce potential confounding factors. However, the randomized design of the study significantly reduces the likelihood of confounding. As previously reported (16), the careful monitoring of study participants indicated minimal differences between groups in terms of visit attendance, dropouts, and loss to follow-ups. This underscores the study’s robustness and lends support to the validity of the results. The small sample size of observations in some treatment groups and CH mutation groups is also a limitation; some correlations may not have reached significance due to the low patient numbers in these groups, as shown in the multivariate analysis. As is well known, malignancies may take a long time to develop, and may not have been detected in the relatively limited (3.7-year median) follow-up time of CANTOS. Finally, although we believe our findings regarding incidence of malignancy relate to the presence of CH mutations and IL1β inhibition, we acknowledge that they may be the result of unknown or unconsidered factors.

In summary and in agreement with previous findings (1, 4, 11, 13), our study showed that CH mutations were associated with a higher incidence of non-hematological malignancies. Moreover, canakinumab treatment was associated with a lower reported incidence of non-hematological malignancies among patients with CH mutations compared to those given a placebo. These results are hypothesis-generating and will need further study to understand underlying molecular mechanisms.

J. Woo reports other support from Novartis during the conduct of the study. H. Xu reports as a full-time employee of Novartis. D.P. Yates reports other support from Novartis outside the submitted work and Novartis employee as well as restricted share holder. D.P. Steensma reports other support from Novartis during the conduct of the study; other support from Ajax Therapeutics outside the submitted work. No disclosures were reported by the other authors.

J. Woo: Conceptualization, data curation, formal analysis, supervision, writing—original draft, writing—review and editing. T. Zhai: Formal analysis, writing—review and editing. F. Yang: Formal analysis, writing—review and editing. H. Xu: Data curation, writing—review and editing. M.L. Healey: Formal analysis, investigation, writing—review and editing. D.P. Yates: Formal analysis, investigation, writing—review and editing. M.T. Beste: Formal analysis, investigation, writing—review and editing. D.P. Steensma: Formal analysis, investigation, writing—review and editing.

The authors would like to thank Paul M. Ridker, MD, MPH, at the Brigham and Women’s Hospital, Harvard Medical School, US, for his scientific and intellectual contribution to the CANTOS trial, and Vanesa Martinez Lopez, PhD, of Novartis Ireland Ltd. and Helen Findlow, PhD, of Novartis UK, for providing medical writing/editorial support in accordance with Good Publication Practice (GPP2022) guidelines (https://www.ismpp.org/gpp-2022). This study was funded by Novartis Pharmaceuticals.