A blueprint for pursuing therapeutic interventions and early phase clinical trials in clonal haematopoiesis

Definitions

Clonal haematopoiesis (CH) is a somatic mutational event arising from an error in genomic replication within a small number of clonal haematopoietic stem cells.¹ While CH is present in nearly all adults at small variant allele frequencies (VAF),² CH of indeterminate potential (CHIP) refers to a growth of clones to a clinically meaningful size, usually considered ≥2% VAF.^{2, 3} CHIP occurs recurrently in a relatively small number of genes which are also commonly mutated in myeloid malignancies.^{1, 2, 4-6} CHIP is more commonly detected with increasing age (>10% over age 70), male sex and a history of tobacco smoking.^{1, 4, 5, 7} CHIP is associated with an increased risk of developing a haematological malignancy (~0.5% per year) and an increased all-cause mortality, mostly from adverse cardiovascular events.^{5, 8} Beyond cardiovascular disease, higher rates of CH have been implicated in other medical conditions, including chronic inflammatory diseases.⁹ The most commonly mutated genes in CH are DNMT3A, TET2 and ASXL1 (DTA), all of which alter epigenetic patterns and chromatin conformation, influencing transcriptional profiles.^{1, 4, 5, 7}

While the risk of transformation of CHIP to an overt malignancy remains small in the general population, several studies have explored predictors of a higher risk. In one study, nearly 75% of acute myeloid leukaemia (AML) patients who had sequencing performed on pre-AML blood samples had somatic mutations detectable before AML diagnosis in putative driver genes.¹⁰ Having a greater number of mutations at a younger age, higher VAF and CHIP-specific mutations including U2AF1, SRSF2, SF3B1 or TP53 were associated with an increased risk of progression to AML.¹⁰ Similarly, the presence of CH mutations in a Women's Health Initiative study found that higher VAF (>10%), mutations in IDH1, IDH2, TP53, DNMT3A, TET2 or spliceosome factor genes increased the probability of developing AML.¹¹

Clonal cytopenia(s) of undetermined significance (CCUS) is a term used to describe CH in conjunction with clinically relevant blood cytopenias.^{3, 12} While CCUS is heterogeneous, it is associated with decreased overall survival (OS) and course similar to that of patients with lower risk myelodysplastic syndromes (MDS, expanded as myelodysplastic neoplasms in the 2022 World Health Organization classification).^{3, 13} CCUS progression risk appears higher with non-DNMT3A/TET2 mutations such as spliceosome mutations, high VAF and the presence of multiple mutations.^{14, 15} While CCUS usually connotes an increased risk over CH alone, some individuals with CCUS (e.g. DTA single mutants with low VAF and mild cytopenias) have a low risk of transforming into an overt myeloid malignancy (Figure 1). Determining which individuals with CH are at risk for disease progression is critical. Early intervention in this subset is most desirable if disease progression a be prevented, as progression to MDS and AML is often incurable with currently available treatment.¹⁶

Risk stratification: Not all CH mutations have the same leukaemogenic potential

People are discovered to have CHIP when genetic sequencing is performed on blood in individuals without known blood disorders, either as part of testing for non-myeloid cancer (e.g. cell-free DNA), during testing for germline predispositions to breast cancer and other neoplasms, or during cardiovascular risk evaluation including polygenic risk score assessment.^17-19 CHIP can be discovered by clinicians who are unfamiliar with counselling, which has led to the development of many CHIP-focused clinics at large referral centres. Early reports described the logistics, goals and challenges of such a clinic as well as the benefits of prospective evaluation and counselling from specialists.^{20, 21} There is currently no clear recommended screening or monitoring criteria.

In an effort to define risk factors associated with CHIP that predispose towards the development of myeloid neoplasia (MN), Weeks et al. analysed sequenced exomes of healthy U.K. Biobank (UKBB) participants and developed the CH risk score (CHRS).²² Important determinants of risk were single DNMT3A mutations, high-risk mutations, two or more mutations, a VAF of ≥20%, age ≥65 years, having CCUS versus CHIP and abnormal red blood cell indices including mean corpuscular volume and red cell distribution width.²² CHRS correlated not just with progression to MN but also with OS and risk of ischaemic cardiovascular disease.²² Similarly, the MN prediction system utilized annotated genetic data from the UKBB to outline time-dependent risk of transformation to specific haematological malignancies including MDS, myeloproliferative neoplasms and AML.²³ The clonal cytopenia risk score was developed by performing multivariate analysis on multicentre CCUS patient data and uses three variables (thrombocytopenia, presence of splicing factor mutations and ≥2 mutations) to calculate a score that predicts CCUS progression risk to an overt MN.²⁴ While large population databases were used to describe incidence of CHIP and clinical associations, recognition of CHIP in real-world clinical practice more often occurs in patients who have reason for genetic testing, thus accounting for an increased rate of CCUS and non-MN. Nonetheless, CHRS is still useful in the prediction of progression of CHIP to MN,²⁵ and other prognostic systems will be validated in this context as well to help identify higher risk individuals who might benefit from participation in CH intervention trials.

Attempts to develop guidelines for the management of individuals with CH, such as determining who might be at higher risk for malignant progression, are underway.²¹ Accumulating real-world clinical outcomes data for people with CH is refining risk evaluation, and therapeutic intervention may be indicated in select subpopulations at higher risk for an adverse event. There is currently a lack of consensus on therapeutic intervention for CHIP, including clarity on which subpopulations to treat, the investigational interventions, the goals and objectives in early intervention and validated end-points to measure that potential benefit.

Detecting CHIP

While common in the general population in large cohort studies, most people with CH in clinical practice today are managed within academic research centres by oncologists, geneticists, haematologists and cardiologists within established CHIP clinics. The assays that detect CH, generally error-corrected sequencing or c-DNA testing, will differ based on institution, with varying depth, quality and reproducibility of results. Assays may have varying numbers of genes detected, potentially missing clinically relevant or understudied mutations. Confirmatory germline testing may not be routinely performed at all centres. Whole genome sequencing and whole exome sequencing used in research assays are not available in clinical practice. The variation of testing modalities brings to forefront the need to consider standardized clinical testing approaches to do large-scale prospective studies feasible and generalizable.²⁶ As technology improves, historical data will need to be re-analysed, and partnering with multiple institutions, federal authorities and/or industry may be needed for far reaching harmonization of detection within samples. In addition, large-scale testing and novel approaches to conduct gene testing more efficiently may lower costs, making testing accessible to a broader population in both clinical practice as well as in practical collaborative research opportunities.²⁷

Patient selection

Defining the intended treatment population is a particularly important in clinical trials in which participants are healthy individuals. Criteria for selecting individuals with CH to be considered for clinical intervention should be predefined to ensure safety and interpretability between studies. Selection criteria should first and foremost be risk-informed, selecting populations with high-risk genomic features, clinically relevant cytopenias or high cardiovascular risk profiles. Examples of a selected CH population may be by a specific high-risk feature, such as a specific mutation (such as TP53); by number of mutations (more than 2); or by a high VAF (such as >10%).^{10, 11} Other features that may be of interest may be those with CH and predisposing cardiovascular risk factors, or CH and those undergoing intensive chemotherapy.¹⁴ Future studies may also choose specific populations, such as an older population, or those with a specific comorbidity such as autoimmune disease that is associated with higher rates of CH.^28-38 Efforts to homogenize eligibility criteria across different planned clinical studies will help validate the future implementation of therapeutic interventions. In other haematological malignancy precursor conditions such as smouldering myeloma, the lack of uniform entry criteria had been recognized as a barrier to implementation of potentially effective interventions.^39-41

Risk of progression to an adverse outcome may also affect population choice. Using a standardized approach for assessing risk, especially in a community/population setting, may be helpful in making clinical decisions for eligibility. An attractive stepwise approach may be choosing the highest risk patients for initial studies, such as individuals with CHIP who already have cardiovascular disease as tertiary prevention, prior to treating lower risk CHIP patients as secondary prevention, as the tolerance for adverse events in such a setting is lower. However, criteria should also be dependent on the goal/intent of the intervention. For example, enrolling extremely high-risk individuals for prevention studies may work against the intervention if the intent is prevention and the disease process is too advanced to halt. Risk stratification for therapy will change as more prospective cohorts are analysed.

Identification of patients as potential clinical trial candidates is essential to ensure robust accrual. Patients with CCUS are often referred to haematologists; however, individuals with asymptomatic CHIP may be identified by medical oncologists, geneticists, primary care physicians, cardiologists or others that are not necessarily seen in a setting where trial enrolment is feasible. While many centres have developed CHIP clinics, some referred patients have competing risks including solid tumours more urgently requiring therapy and these patients may not be appropriate candidates for an interventional secondary or tertiary prevention study. A high incidence of CHIP (~25%) in patients evaluated in the cardiac catheterization lab suggests that this population might be appropriate for more systematic screening, and potential enrolment into interventional protocols.^{19, 36, 42} Older adults, a group inherently enriched for CH, can be insufficiently included in clinical trials despite explicit Federal Drug Administration (FDA) guidance regarding their inclusion.⁴³ It also important to consider how the exclusion criteria may affect enrolment (e.g. excluding those with solid tumours from participating in a CHIP study).

CH may have different epidemiology across populations, and representation of diverse groups is important in interventional studies. The largest data sources defining the spectrum of mutations and epidemiology of CHIP are from European populations (e.g. UKBB) or from US-based cohorts with a majority of persons of European descent. In these cohorts and others, CHIP has been observed to have differential prevalence in different populations.^{5, 8, 26, 44} In particular, Hispanic individuals have been identified as having lower rates of CHIP than European individuals.⁴⁵ The basis for these differential prevalence is incompletely understood and likely has both genetic and environmental contributors. Several studies have identified germline genetic variants that are enriched in different populations as conferring differential risk of CHIP.⁴⁶ For example, the JAK2 46/1 risk haplotype that is associated with an increased risk of JAK2-mutated CHIP (and MPN) is more common in Northern European ancestry individuals than other populations.⁴⁶ A germline variant in a distal enhancer of TET2 that increases the risk of CHIP is more common in African Americans.⁴⁴ Although each of these individual germline variants confers a very small increased or decreased risk of CHIP, it is possible that, in aggregate, they contribute to differential population risk and warrant further study.

It is critical to keep in mind that most CH is incidental, may never need treatment and, in some situations, may provide clonal fitness and protection from certain diseases. Progressive haematopoietic oligoclonality is inescapable with ageing and mutations in such epigenetic modifiers as DNMT3A and TET2 may be necessary to maintain adequate haematopoietic output in the setting of ageing-associated stem cell attrition.⁴⁷ CH has been associated with the decreased rates of Alzheimer's dementia, particularly at higher CH VAF burden, and may be related to a more efficient microglial replacement.^{47, 48} CHIP does not seem to have a meaningful impact in key areas intuitively thought to be associated with increased risk. For example, there was no difference in the rates of CH among patients with sickle cell anaemia compared to age-matched controls, including in those treated with hydroxyurea.⁴⁹ In the context of allogeneic haematopoietic cell transplant, donor CH was associated with improved outcomes in the recipient in one study⁵⁰ and, in another, had no net effect on outcomes⁵¹; while recipient CH had no effect on long-term transplant outcomes,⁵² and inferior progression-free survival in another.⁵³ It is possible that heterogeneity of studies or populations lead to these discrepant findings, but currently the impact of CH on allogeneic stem cell transplant is unclear.⁵⁴ There is an increase in TET2-mutant CHIP in solid-organ transplant recipients not due to an increased risk of receiving a transplant in the first place.⁵⁵ While this may be partly explained by a clear correlation between certain types of immunosuppression and emergence of TET2-mutant CHIP, there is also evidence from a preclinical mouse model of cardiac transplant that TET2-null haematopoietic cells promote allograft tolerance, although no such association has been described in the limited number of clinical cohort studies completed to date.⁵⁶ These studies suggest there exists subsets of CH that are better left alone, particularly when not associated with any cytopenias (Table 1).^57-60

Selecting the right therapeutic end-points

In pursuing interventions for a predisease state, it is important to highlight the expected outcomes, the timeframe and clinical significance the intervention would have on subject outcomes, which can be challenging. While OS is a critical end-point, this is likely to be difficult to measure in CH trials given the low risk of a death event during the first years of follow-up, and competing risks in an elderly population that would be enriched for persons with CH. Relevant alternative clinical end-points measurable within a shorter timeframe include progression to MN, improvement of clinically significant cytopenias and major adverse cardiac events (MACE) (Table 2). Measuring time to progression to MN would not describe the severity of MN as OS would; for example, low-risk MDS and TP53-mutated MN have strikingly different clinical courses.

There is a precedent of measuring MACE in the CH setting. In the CANTOS study of secondary prevention of cardiovascular events and death in patients who previously had a myocardial infarction and had elevated levels of C-reactive protein (CRP), the use of canakinumab was associated with a mild reduction of cardiovascular outcomes and death in all patients and greater than a 60% reduction in the subgroup of patients with TET2-mutated CHIP.⁶¹ Canakinumab therapy was associated with a lower incidence of non-myeloid malignancies; the reduction was greatest in individuals with TET2 mutations.^{62, 63} CHIP was associated with signatures of inflammation and infection defence, which were exacerbated by concomitant anaemia and ameliorated by canakinumab therapy.⁶³ For patients with CCUS, improvement in cytopenias, changes in patient-reported outcomes and reduction in infections and transfusion needs are likely to be important.

Surrogate end-points in CHIP such as reduction in VAF, improvement in inflammatory markers (e.g. CRP, erythrocyte sedimentation rate, interleukin-6 [IL-6], IL-8) or other relevant biomarkers (e.g. those relevant to specific genotypes, such as increases in GDF15 in a TP53 mutant population or 2-hydroxyglutarate levels in those with IDH1/IDH2 mutations) may be easier to observe in a shorter time but are not yet clinically validated, so the clinical relevance of these end-points is unknown. The balance between a clinically meaningful outcome with a feasible measurable outcome will be a fundamental challenge in designing therapeutic trials in this setting.

Outcomes in long-term interventional studies may be difficult to attribute to a single intervention, especially if the duration of intervention is limited. In large studies of diabetes prevention in high-risk individuals, both lifestyle interventions and metformin were efficacious interventions for long-term diabetes prevention; however, neither group had a reduction in cardiovascular events after more than 20 years of follow-up.⁶⁴ Therefore, changes in clinical practice based on surrogate end-points or interim analyses should be used with caution until clear evidence is available. Examples include altering decisions about allogeneic Hematopoietic stem cell transplant based on donor or recipient CH status or modifying standard chemotherapy or radiotherapy regimens for non-myeloid cancer patients due to incidentally detected CH.

Outcomes in non-interventional studies in CHIP

The majority of CH in the general population will be considered lower risk and may not be the prime focus of therapeutic interventional clinical trials. Low-risk CH may include those with a normal CBC, as well as a single mutation in DNMT3A or TET2 with low VAF, which are less likely to transform to an advanced MN. This group of people would be excellent candidates for clinical studies in previvorship, natural history studies, understanding risks or extrinsic factors that may predispose to clonal evolution and assessing quality of life and OS (Table 1). As these are important outcomes that cannot be captured in retrospective studies, a focus on innovative measures such as telehealth platforms and surveys may be utilized, as well as working with data managers to collect data uniformly.

Study design considerations

As a condition that is prevalent in the general population, with nuances in diverse and varied populations that remain understudied, efforts to design CHIP clinical trials that will lead to high enrolment and reduce barriers are very important. Efforts to include participants within various clinical practice settings beyond academic centres should be considered, including community practices, inner city and rural populations. Telehealth visits have been increasingly incorporated into clinical practice since the COVID-19 era, and novel clinical trial designs have included telehealth clinical trial design including which may increase access, including the IDH1 targeting CCUS study (NCT05030441). International collaborations as part of larger studies and cooperative groups may also increase enrolment.

Targeting inflammation

Inflammatory pathways are an appealing target in CH, due to preclinical models and other evidence pointing to a reciprocal relationship between clonal expansion in CH and upregulation of a variety of inflammatory signalling cascades, including NLRP3 inflammasome and its downstream cytokines (e.g. interleukin-1β [IL1 β], interleukin 18). For example, the increase in atherosclerosis seen in the TET2 knockout murine models was mediated by IL1 β.^{8, 65} High doses of vitamin C modified TET2 function in murine models, leading to a prospective study of vitamin C in CCUS (NCT03418038), which showed no responses at 20 weeks (Table 3).^{66, 67} The CANTOS study described previously has motivated a prospective trial in CCUS with canakinumab (NCT05641831).⁶¹ NLRP3 inflammasome mediators are now being tested in the clinic for MDS and could be explored for CH/CCUS if safety patterns permit (NCT06097663). In addition, luspatercept treatment in the front-line COMMANDS MDS study showed downregulation of IL-1, IL-1R, IFNα, IL-6 pro-inflammatory pathways suggesting a modulation of inflammatory bone marrow microenvironment along with a reduction in NT-proBNP and may be attractive target in patients with CCUS.⁶⁸ As the inflammatory mediators and clinical consequences of CH are genotype-specific, such as the link between DNMT3A-mutated CH and osteoporosis, studies should aim to collect data encompassing a variety of CH driver mutation and end-organ consequences.³²

Addressing select mutations

Another therapeutic strategy is earlier intervention with small-molecule inhibitors or other agents already approved for the treatment of MDS or AML, such as IDH1 and IDH2 inhibitors, low-dose/short course DNA methyltransferase inhibitors (hypomethylating agents), lenalidomide, luspatercept, imetelstat or haematopoietic growth factors in at-risk individuals. This approach does have appeal, as specific CH mutations such as IDH1 and IDH2 mutations, for example, are associated with an increased risk of disease progression, and the safety profile of IDH inhibitors is well established; however, IDH mutations are uncommon in CH. Since these treatments are associated with treatment-related toxicity (e.g. cytopenias with hypomethylating agents, differentiation syndrome with IDH inhibitors), justification of treatment needs to be supported by longer term safety data and/or a clear definition of risk. If targeted interventions show clear promise in long-term prevention of morbidity and mortality in particularly adverse populations, novel approaches such as autologous gene editing or other techniques to prevent disease can be considered in the future, based on success in diseases such as the haemoglobinopathies.⁶⁹ The duration of therapy with targeted agents is also an important consideration particularly if disease occurrence can take many years, potentially exposing individuals to drug toxicities versus waiting for disease to occur and then treating with targeted agents. Delay of progression to MN becomes an important criterion in such cases.

Nutritional targets

The role of dietary interventions and naturally occurring compounds in reducing clonal expansion in CH is attractive given the low tolerance for toxicity and the strong precedent of effective non-pharmacologic intervention in a variety of similar precursor states. Diet has been shown to influence stem cell fitness in several preclinical studies, and observational cohort data indicate that individuals with less healthy eating habits have a greater prevalence of CHIP.⁷⁰ An ongoing randomized placebo controlled study of a plant-based dietary intervention in patients with precursor myeloma states is ongoing, and a similar dietary intervention in the CH population, to assess changes in clonal trajectory, is attractive.⁷⁰

Repurposing drugs

Novel therapies that may affect oxidative phosphorylation or metabolomics or epigenetics may shed new light in prevention of progression in a predisease state. Repurposing drugs approved drugs such as HMG Co-A reductase inhibitors, particularly when targeting associated cardiovascular and other chronic disease of ageing may also play a role and warrant further investigation. There is preclinical evidence to support studying mTOR inhibition by rapamycin in DNMT3A- and ASXL1-mutated CH.^{71, 72} Statins and metformin are already in clinical investigation in CCUS (Table 4).

Targeting the fitness landscape

There may be a role for considering interventions specifically for populations that may be at higher risk for MN progression, such as people with pre-existing or therapy-related CH within DNA damage response genes (e.g. TP53 or PPM1D), who will undergo additional chemotherapy or radiation for a non-haematological indication. It is well established that select genotoxic and immunomodulatory drugs can promote the expansion of pre-existing TP53- and PPM1D-mutated haematopoietic clones; thus, it is crucial to exercise caution when considering the study of modulating treatment decisions for the primary malignancy.^73-75 Thoughtful interventions that will likely not compromise the success of eradicating the existing disease (e.g. another cancer) will need to be considered in these novel trial designs.