Clonal Hematopoiesis of Indeterminate Potential Predicts Adverse Outcomes in Patients With Atherosclerotic Cardiovascular Disease

Abstract

BACKGROUND

Clonal hematopoiesis of indeterminate potential (CHIP)–the age-related clonal expansion of blood stem cells with leukemia-associated mutations–is a novel cardiovascular risk factor. Whether CHIP remains prognostic in individuals with established atherosclerotic cardiovascular disease (ASCVD) is less clear.

OBJECTIVES

This study tested whether CHIP predicts adverse outcomes in individuals with established ASCVD.

METHODS

Individuals aged 40 to 70 years from the UK Biobank with established ASCVD and available whole-exome sequences were analyzed. The primary outcome was a composite of ASCVD events and all-cause mortality. Associations of any CHIP (variant allele fraction ≥2%), large CHIP clones (variant allele fraction ≥10%), and the most commonly mutated driver genes (DNMT3A, TET2, ASXL1, JAK2, PPM1D/TP53 [DNA damage repair genes], and SF3B1/SRSF2/U2AF1 [spliceosome genes]) with incident outcomes were compared using unadjusted and multivariable-adjusted Cox regression.

RESULTS

Of 13,129 individuals (median age: 63 years) included, 665 (5.1%) had CHIP. Over a median follow-up of 10.8 years, any CHIP and large CHIP at baseline were associated with adjusted HRs of 1.23 (95% CI: 1.10-1.38; P < 0.001) and 1.34 (95% CI: 1.17-1.53; P < 0.001), respectively, for the primary outcome. TET2 and spliceosome CHIP, especially large clones, were most strongly associated with adverse outcomes (large TET2 CHIP: HR: 1.89; 95% CI: 1.40-2.55; P <0.001; large spliceosome CHIP: HR: 3.02; 95% CI: 1.95-4.70; P < 0.001).

CONCLUSIONS

CHIP is independently associated with adverse outcomes in individuals with established ASCVD, with especially high risks observed in TET2 and SF3B1/SRSF2/U2AF1 CHIP.

Cardiovascular disease remains the leading cause of death globally.¹ Although chronologic age is the dominant risk factor for cardiovascular conditions such as atherosclerotic cardiovascular disease (ASCVD), mechanisms by which aging leads to disease remain incompletely understood.^2,3 Large-scale next-generation sequencing studies of blood DNA have revealed that acquired leukemogenic mutations occur in apparently healthy subjects.⁴ This phenomenon, termed clonal hematopoiesis of indeterminate potential (CHIP), increases exponentially with age, affecting >10% of individuals aged 70 years and older.⁵ The epigenetic regulators DNMT3A, TET2, and ASXL1 represent the 3 most commonly mutated CHIP-associated genes.⁵ CHIP is associated with accelerated atherosclerosis and first ASCVD events in humans and aggravates experimental atherosclerosis in mice.^6–9 Although this association is largely independent of traditional cardiovascular risk factors, heightened inflammation partially mediates the link between some forms of CHIP and ASCVD.^6,9–12

Recent studies have suggested that CHIP may adversely influence prognosis in the setting of heart failure, aortic stenosis, ST-segment elevation myocardial infarction, and cardiogenic shock.^13–19 However, it is unknown whether CHIP is prognostic in the broader population of individuals with established ASCVD. Furthermore, it remains unclear whether all CHIP driver mutations confer a similar degree of risk because studies to date have generally lacked adequate power to examine differences across gene-specific CHIP subtypes. As any precision medicine clinical trials in individuals with CHIP would likely begin in those with established ASCVD, understanding the associations of CHIP and CHIP subtypes with adverse outcomes in this population is critical.

To address these gaps, we leveraged the UK Biobank to test the associations of CHIP with ASCVD events, all-cause mortality, and cause-specific mortality among individuals with established ASCVD. In addition, we assessed which gene-specific CHIP subtypes associated most strongly with adverse outcomes. Finally, we tested whether high-sensitivity C-reactive protein (hsCRP) and neutrophil-to-leukocyte ratio (NLR), biomarkers of inflammation, differed across CHIP subtypes or explained CHIP-associated event risks.

METHODS

STUDY COHORT.

The UK Biobank is a population-based cohort study of approximately 500,000 adults in the United Kingdom.²⁰ Individuals aged 40 to 70 years were enrolled between 2006 and 2010 and are followed prospectively via linkage to national health records. After providing informed consent, participants provided detailed information on demographics, lifestyle, health history, and medication use; provided blood sequencing samples for DNA and other biomarkers; and underwent measurement of vital signs and anthropomorphic features. The present study included patients with established ASCVD (ie, coronary artery disease [CAD], ischemic stroke, and/or peripheral artery disease) before UK Biobank enrollment, defined using qualifying International Classification of Diseases (ICD) or procedure codes (Supplemental Table 1), and available whole-exome sequencing (WES) from blood-derived DNA. We excluded individuals with known hematologic malignancy (n = 213) at baseline, missing or unknown history of cancer (n = 71), or key missing covariates (Figure 1). Follow-up occurred through March 2020. The UK Biobank received approval from the North West Multi-center Research Ethics Committee, and the Mass General Brigham Institutional Review Board approved these secondary data analyses (UK Biobank application 7089).

FIGURE 1. Study Flow Chart.

This study included UK Biobank (UKB) participants with a diagnosis of atherosclerotic cardiovascular disease (ASCVD) (ie, coronary artery disease, stroke, and/or peripheral artery disease) before enrollment and available whole-exome sequencing from whole blood-derived DNA. Participants with prevalent hematologic malignancy, unknown/missing history of cancer, and missing key covariates were excluded.

BMI = body mass index; CHIP = clonal hematopoiesis of indeterminate potential; HDL = high-density lipoprotein.

EXPOSURES: CHIP AND GENE-SPECIFIC CHIP SUBTYPES.

WES was derived from whole blood using the Illumina NovaSeq 6000 platform at the Regeneron Genetics Center (Tarrytown, New York, USA).²¹ CHIP-related phenotypes were ascertained from blood-derived WES as recently described.²² Briefly, somatic variants were identified using the Mutect2 tool from the Genome Analysis Toolkit.^23,24 CHIP mutations were called in a previously curated list of 58 myeloid driver genes (Supplemental Table 2), and variants were removed if: 1) total read depth was <20²²; 2) minimum read depth for the alternate allele was <5; or 3) variant support in both forward and reverse sequencing reads were lacking. Sequencing artifacts and germline variants were removed as described previously.²²

The coprimary study exposures were the presence of any CHIP, defined as a variant allele fraction (VAF) ≥2%, and the presence of a large CHIP clone, defined as VAF ≥10%, as prior work suggests that large clones may associate more strongly with adverse outcomes.^6,11,25 In addition, we separately tested the most frequently mutated genes in the data set, both overall and restricted to large clones (ie, VAF ≥10%): DNMT3A, TET2, ASXL1, JAK2, DNA damage repair genes (PPM1D/TP53), and spliceosome genes (SF3B1/SRSF2/U2AF1). Consistent with previous CHIP-focused analyses, for the 0.3% of individuals with multiple detectable CHIP clones, the VAF of the composite CHIP exposure was assigned based on the VAF of the largest CHIP clone in primary analysis⁷; secondary analyses separately examined individuals with multiple clones (VAF ≥2%) and those with multiple large clones (VAF ≥10%).

OUTCOMES.

The primary outcome was a composite of ASCVD events during follow-up and all-cause mortality. ASCVD events were ascertained from qualifying codes (Supplemental Table 3) and included: 1) CAD events requiring invasive coronary angiography or coronary revascularization; 2) acute ischemic cerebrovascular events; and 3) peripheral artery disease diagnoses and events, including peripheral artery revascularization and lower extremity amputation. We examined composite ASCVD and all-cause mortality separately, as well as cause-specific mortality (cardiovascular death and cancer death), incident hematologic malignancy, and incident acute myeloid leukemia as secondary outcomes. Cause-specific mortality was defined using ICD codes linked to death register records that were listed as the primary cause of death. All participants who died during follow-up were categorized by organ system based on the listed cause of death using phecode groupings.²⁶ ICD codes used for defining cause-specific mortality are listed in Supplemental Table 4, whereas those for hematologic malignancy and acute myeloid leukemia are listed in Supplemental Table 5.

STATISTICAL ANALYSIS.

The Shapiro-Wilk test was used to verify data normality. Continuous variables were compared using the Student’s t-test or Wilcoxon rank-sum test and categorical variables using the Pearson chi-square test or Fisher exact test as appropriate. Cox proportional hazards models tested the association of CHIP with primary and secondary outcomes, with follow-up starting at UK Biobank enrollment. Multivariable-adjusted models were adjusted for age, age², sex, race, Townsend deprivation index, current or former smoking, diabetes, body mass index, systolic blood pressure, antihypertensive medication use, total and high-density lipoprotein cholesterol, and cholesterol-lowering medication use (Supplemental Methods, Supplemental Table 6). Follow-up from study enrollment was truncated at 12 years due to departures from proportional hazards beyond this timepoint. The proportional hazards assumption was verified using Schoenfeld residuals.

We conducted sensitivity analyses: 1) excluding individuals with each underlying ASCVD subtype; 2) excluding individuals with cancer at baseline; and 3) further adjusting for time from first ASCVD diagnosis. Subgroup analyses tested associations between CHIP and the primary outcome stratified by: 1) age <60 vs ≥60 years; and 2) female vs male sex. In addition, we tested differences in hsCRP and NLR by driver mutation and performed Cox regression further adjusted for these variables, given previous evidence that inflammation mediates the association of CHIP with outcomes.

Given examination of 6 gene-specific CHIP subtypes, 2-sided P < 0.0083 (ie, 0.05/6) indicated statistical significance. For secondary analyses, P values and 95% CIs were not adjusted for multiple testing, and inferences drawn from these statistics may not be reproducible for secondary analyses. Analyses were performed in R version 4.1.3.

RESULTS

STUDY COHORT.

The final analytic cohort included 13,129 individuals with established ASCVD (Figure 1), of whom 665 (5.1%) had any and 454 (3.5%) large CHIP (ie, VAF ≥10%). DNMT3A CHIP was detected in 337 participants (2.6% of overall cohort), TET2 CHIP in 92 (0.7%), ASXL_l CHIP in 141 (1.1%), JAK2 CHIP in 9 (0.1%), PPM1D/TP53 CHIP in 41 (0.3%), and SF3B1/SRSF2/U2AF1 CHIP in 30 (0.2%). DNMT3A, TET2, and ASXL_l CHIP collectively constituted 83.9% (n = 558 of 665) of CHIP cases (Supplemental Figure 1). Twenty-six individuals (3.9% of CHIP carriers) had mutations in >1 driver gene (Supplemental Figure 2).

The median age of the study cohort was 63 years (IQR: 59-66 years), and 3,210 (24.4%) participants were women (Table 1). Individuals with CHIP were older than those without (65 years [IQR: 62-67 years] vs 63 years [IQR: 59-66 years]; P < 0.001). Median time from first ASCVD diagnosis to UK Biobank enrollment was approximately 4.6 years and was slightly longer in those with CHIP (5.1 years [IQR: 2.2-8.3 years] vs 4.6 years [IQR: 2.1-7.8 years]; P = 0.03). Participants with vs without CHIP were more likely to be current or former smokers. Consistent with inclusion of a secondary prevention population, rates of antihypertensive and cholesterol-lowering medication prescription were high and similar in both groups. Underlying ASCVD diagnoses were CAD in 11,507 (87.6%), ischemic stroke in 1,152 (8.8%), and peripheral artery disease in 1,172 (8.9%) of all participants at baseline. Peripheral artery disease was slightly more common in those with CHIP (12.0% vs 8.8%, P = 0.005). Baseline characteristics were similar between participants carrying CHIP with VAF <10% vs ≥10% (Supplemental Table 7).

Table 1.

Baseline Characteristics of UK Biobank Participants With Prevalent Atherosclerotic Cardiovascular Disease at Study Enrollment

	CHIP (n = 665)	No CHIP (n = 12,464)	P Value
Age at enrollment, y	65 (62-67)	63 (59-66)	<0.001
Time from first diagnosis to enrollment, y	5.1 (2.2-8.3)	4.6 (2.1-7.8)	0.03
Female	158 (23.8)	3,052 (24.5)	0.71
Race/ethnicity	–	–	0.21
Asian	15 (2.3)	464 (3.7)	–
Black	3 (0.5)	119 (1.0)	–
White	641 (96.4)	11,743 (94.2)	–
Mixed/other	6 (0.9)	138 (1.1)	–
Smoking status	–	–	<0.001
Current	92 (13.8)	1,610 (12.9)	–
Former	382 (57.4)	6,394 (51.2)	–
Never	191 (28.7)	4,480 (35.9)	–
Body mass index, kg/m2	28.3 (25.8-31.1)	28.6 (26.0-31.9)	0.09
Systolic blood pressure, mm Hg	140 (128-154)	139 (127-153)	0.06
Diastolic blood pressure, mm Hg	78 (71-86)	79 (72-86)	0.62
Total cholesterol, mg/dL	167.0 (145.2-188.4)	169.5 (147.2-194.7)	0.02
High-density lipoprotein cholesterol, mg/dL	45.2 (37.3-55.0)	44.9 (38.1-53.5)	0.95
Triglycerides, mg/dL	144.7 (103.2-206.3)	145.5 (102.1-207.3)	0.95
Low-density lipoprotein cholesterol, mg/dL	98.8 (83.0-114.3)	100.9 (85.0-119.8)	0.008
Apolipoprotein B, mg/dL	80.1 (69.5-92.2)	82.5 (70.7-96.1)	0.006
High-sensitivity C-reactive protein, mg/L	1.71 (0.91-3.66)	1.53 (0.73-3.21)	0.002
Neutrophil-to-lymphocyte ratio	2.44 (1.87-3.21)	2.33 (1.80-3.05)	0.002
Antihypertensive medication use	133 (20.0)	2,458 (19.7)	0.90
Cholesterol-lowering medication use	580 (87.2)	10,802 (86.7)	0.73
Townsend deprivation index	−1.5 (−3.4 to 1.7)	−1.6 (−3.3 to 1.6)	0.77
History of cancer	67 (10.1)	1,065 (8.5)	0.19
Diabetes mellitus	135 (20.3)	2,498 (20.0)	0.91
Prevalent coronary artery disease	564 (84.8)	10,943 (87.8)	0.03
Prevalent ischemic stroke	59 (8.9)	1,093 (8.8)	0.98
Prevalent peripheral artery disease	80 (12.0)	1,092 (8.8)	0.005

Values are median (IQR) or n (%). Continuous characteristics were compared using the Student’s t-test or Wilcoxon rank-sum test as appropriate. Categorical characteristics were compared using the Pearson chi square test or Fisher exact test as appropriate. No corrections for multiple testing were applied.

CHIP = clonal hematopoiesis of indeterminate potential.

ABSOLUTE RISKS OF INCIDENT ATHEROSCLEROTIC CARDIOVASCULAR EVENTS AND DEATH.

Over a median follow-up duration of 10.8 years (IQR: 10.0-11.6 years), incident ASCVD events or death (ie, the primary outcome) occurred in 4,580 (36.7%) without CHIP, 309 (46.5%) with any CHIP, and 223 (49.1%) with large CHIP. Incident ASCVD events during follow-up occurred in 3,249 (26.1%) without CHIP, 215 (32.3%) with any CHIP, 159 (35.0%) with large CHIP. All-cause death occurred in 2,114 (17.0%) individuals without CHIP, 164 (24.7%) with any CHIP, and 120 (26.4%) with large CHIP. Most deaths were cardiovascular (n = 831 of 2,278 [36.5%]) or cancer-related (n = 751 of 2,278 [33.0%]) (Supplemental Figure 3). Increased risk of adverse outcomes was consistent across the primary outcome, its individual components (Figure 2, Supplemental Table 8), cardiovascular mortality (Figure 3), and incident hematologic malignancy and acute myeloid leukemia (Supplemental Figure 4). Cancer mortality was comparable in those with large and small CHIP clones, but both were increased vs those without CHIP. Respiratory illness, the third-most common cause of death, was also increased in individuals with large CHIP clones (Supplemental Table 9).

FIGURE 2. Cumulative Incidence of ASCVD Events or Death by CHIP Status.

Cumulative incidence plots were constructed using the Kaplan-Meier method and represent (A) the primary outcome of ASCVD events or all-cause mortality, (B) ASCVD events, and (C) all-cause mortality during a median follow-up of 10.8 years (IQR: 10.0 to 11.6 years). The shaded area indicates the 95% CI. Multivariable-adjusted models were adjusted for age at the start of follow-up, age², sex, race (White vs non-White), Townsend deprivation index, current or former smoking, systolic blood pressure, antihypertensive medication use, total cholesterol, HDL cholesterol, cholesterol-lowering medication use, diabetes status, and BMI. Individuals without CHIP constituted the reference group in all analyses. No corrections for multiple testing were applied. *P<0.05. **P<0.01. ***P<0.001. Ref = reference; VAF = variant allele fraction; other abbreviations as in Figure 1.

FIGURE 3. Cumulative Incidence of Cardiovascular and Cancer Mortality by CHIP Status.

Cumulative incidence plots were constructed using the Kaplan-Meier method and represent (A) cardiovascular mortality and (B) cancer mortality. See Figure 2 legend for details on the cumulative incidence plots and multivariable-adjusted models used. No corrections for multiple testing were applied. *P<0.05. **P<0.01. ***P<0.001. Abbreviations as in Figures 1 and 2.

MULTIVARIABLE-ADJUSTED ASSOCIATION OF CHIP AND GENE-SPECIFIC CHIP SUBTYPES WITH OUTCOMES.

Any vs no CHIP was associated with a multivariable-adjusted HR of 1.23 (95% CI: 1.10-1.38; P < 0.001) for the primary composite outcome of ASCVD events or all-cause death (Figure 4), with consistent associations observed for ASCVD events (HR: 1.24; 95% CI: 1.08-1.43; P = 0.002) and all-cause death (HR: 1.28; 95% CI: 1.09-1.51; P = 0.002) when examined separately (Table 2). Hazards associated with large CHIP were greater for composite ASCVD events or death (HR: 1.34; 95% CI: 1.17-1.53; P < 0.001), ASCVD (HR: 1.38; 95% CI: 1.18-1.62; P < 0.001), and all-cause death (HR: 1.39; 95% CI: 1.16-1.67; P < 0.001). The presence of multiple CHIP and multiple large CHIP driver genes was associated with HRs of 2.49 (95% CI: 1.62 to 1.83; P < 0.001) and 3.03 (95% CI: 1.93 to 4.78, P < 0.001), respectively.

FIGURE 4. Associations of CHIP Subtypes With Incident ASCVD or Death.

See Figure 2 legend for details on the multivariable-adjusted models used. No corrections for multiple testing were applied. Abbreviations as in Figures 1 and 2.

Table 2.

Multivariable-Adjusted Associations of CHIP With ASCVD and Death During Follow-Up

	ASCVDa,b			All-Cause Mortality
	Incidence Rate (Per 1,000 Person-Years)	Unadjusted HR (95% CI)	Multivariable-Adjusted HR (95% CI)	Incidence Rate (Per 1,000 Person-Years)	Unadjusted HR (95% CI)	Multivariable-Adjusted HR (95% CI)
CHIP status at enrollment
No CHIP (n = 12,464)	29.0	Ref.	Ref.	16.5	Ref.	Ref.
Any CHIP (n = 665)	38.5	1.33 (1.16-1.53)c	1.24 (1.08-1.43)d	24.7	1.51 (1.29-1.78)c	1.28 (1.09-1.51)d
Large CHIP (ie, VAF ≥10%) (n = 454)	42.6	1.47 (1.26-1.73)c	1.38 (1.18-1.62)c	26.7	1.65 (1.37-1.98)c	1.39 (1.16-1.67)c
Any CHIP (VAF ≥2%)
DNMT3A CHIP (n = 337)	33.5	1.15 (0.94-1.41)	1.10 (0.90-1.35)	18.7	1.13 (0.88-1.45)	0.99 (0.77-1.27)
TET2 CHIP (n = 92)	51.6	1.78 (1.28-2.46)c	1.75 (1.26-2.42)c	31.9	1.97 (1.36-2.87)c	1.80 (1.24-2.61)d
ASXL1 CHIP (n = 141)	43.8	1.52 (1.15-2.01)d	1.34 (1.01-1.78)e	28.6	1.77 (1.29-2.44)c	1.34 (0.97-1.84)
JAK2 CHIP (n = 9)	41.9	1.43 (0.46-4.43)	1.61 (0.52-5.01)	49.8	3.16 (1.18-8.42)e	3.44 (1.29-9.20)e
PPM1D/TP53 CHIP (n = 41)	37.8	1.29 (0.73-2.27)	1.10 (0.62-1.93)	47.3	3.00 (1.89-4.77)c	2.47 (1.55-3.93)c
SF3B1/SRSF2/U2AF1 CHIP (n = 30)	79.5	2.76 (1.66-4.58)c	2.41 (1.45-4.00)c	61.5	4.07 (2.49-6.66)c	3.01 (1.84-4.94)c
Large CHIP (ie, VAF ≥10%)
Large DNMT3A CHIP (n = 199)	38.2	1.32 (1.03-1.68)e	1.24 (0.97-1.59)	18.5	1.12 (0.82-1.55)	0.96 (0.70-1.33)
Large TET2 CHIP (n = 75)	53.4	1.84 (1.29-2.62)c	1.89 (1.33-2.69)c	35.1	2.16 (1.46-3.21)c	2.07 (1.40-3.08)c
Large ASXL1 CHIP (n = 104)	46.5	1.62 (1.18-2.22)d	1.44 (1.05-1.97)e	29.8	1.86 (1.29-2.66)c	1.42 (0.99-2.04)
Large JAK2 CHIP (n = 9)	41.9	1.43 (0.46-4.43)	1.61 (0.52-5.01)	49.8	3.16 (1.18-8.42)e	3.44 (1.29-9.20)e
Large PPM1D/TP53 CHIP (n = 27)	46.9	1.59 (0.83-3.07)	1.36 (0.71-2.62)	55.2	3.56 (2.07-6.15)c	2.78 (1.61-4.81)c
Large SF3B1/SRSF2/U2AF1 CHIP (n = 25)	81.5	2.82 (1.60-4.98)c	2.52 (1.43-4.45)d	64.3	4.27 (2.52-7.22)c	3.22 (1.90-5.47)d

^aThe composite outcome of ASCVD and all-cause mortality is the primary outcome of the present study.

^bASCVD includes coronary artery disease, ischemic stroke, and peripheral artery disease. See Figure 2 Legend for details on the multivariable-adjusted models used. No corrections for multiple testing were applied.

^cP < 0.001.

^dP < 0.01.

^eP < 0.05.

ASCVD = atherosclerotic cardiovascular disease; CHIP = clonal hematopoiesis of indeterminate potential; Ref = reference; VAF = variant allele fraction.

Among the 6 gene-specific CHIP subtypes examined, the strongest and most consistently significant associations were observed for large TET2 CHIP (primary outcome: HR: 1.89; 95% CI: 1.40-2.55; P < 0.001) and large SF3B1/SRSF2/U2AF1 CHIP (primary outcome: HR: 3.02; 95% CI: 1.95-4.70; P < 0.001). Although PPM1D/TP53 CHIP was strongly associated with all-cause mortality (HR: 2.78; 95% CI: 1.61-4.81; P < 0.001), associations with the primary outcome (Figure 4) or ASCVD (Table 2) did not reach significance. Although overall DNMT3A CHIP did not associate significantly with primary or secondary outcomes, in-frame deletions/insertions in DNMT3A vs no CHIP were associated with a trend toward increased risk of the primary outcome (HR: 2.31; 95% CI: 0.87-6.16; P = 0.09) and with significantly increased risk of incident ASCVD (HR: 3.25; 95% CI: 1.21-6.68; P = 0.02) in exploratory analyses (Supplemental Figure 5, Supplemental Table 10). There was a large, significant association between CHIP and incident hematologic malignancy, with the highest risks observed for JAK2 and SF3B1/SRSF2/U2AF1 CHIP (Supplemental Table 11).

We conducted several sensitivity analyses to probe the robustness of our findings. Associations with outcomes were consistent after excluding each subtype of prevalent ASCVD (Supplemental Table 12); after excluding individuals with prevalent cancer, notably including associations of outcomes with PPM1D/TP53 CHIP (Supplemental Table 13); and after further adjustment for time from first ASCVD diagnosis to UK Biobank enrollment (Supplemental Table 14). Subgroup analyses suggested a larger magnitude of association with outcomes in adults <60 vs ≥60 years of age for TET2 (P_interaction = 0.06) and SF3B1/SRSF2/U2AF1 CHIP (P_interaction = 0.04) (Supplemental Table 15). In sex-stratified analyses, PPM1D/TP53 CHIP associated more strongly with outcomes in women vs men (P_interaction < 0.001); this interaction was also significant for large PPM1D/TP53 CHIP clones and persisted after excluding individuals with a cancer diagnosis before enrollment (P_interaction < 0.001 for both).

C-REACTIVE PROTEIN LEVELS AND THE NLR BY CHIP DRIVER MUTATION.

As prior evidence implicates the NLRP3 and AIM2 inflammasomes and downstream inflammatory cytokines in CHIP-associated acceleration of atherosclerosis, we examined whether elevated hsCRP and/or NLR accompanied CHIP in individuals with ASCVD.^9–11,27,28 Those with vs without CHIP had only modestly higher hsCRP levels (median: 1.71 mg/L [IQR: 0.91-3.66 mg/L] vs 1.53 mg/L [IQR: 0.73-3.21 mg/L]; P = 0.002). Levels of hsCRP were ≥2 mg/L in 5,042 (40.4%) individuals with no CHIP, 95 (45.0%) with small CHIP, and 195 (43.0%) with large CHIP (P = 0.22). NLR was modestly elevated in individuals with CHIP (median: 2.44 [IQR: 1.87-3.21] vs 2.33 [IQR: 1.80-3.01]; P = 0.002). After log-transformation to achieve a normal distribution and adjustment for age and sex using linear regression, the association with CHIP attenuated for NLR (P = 0.09) but not hsCRP (P = 0.003).

In examination of gene-specific CHIP subtypes, hsCRP levels were nominally higher in PPM1D/TP53 CHIP (median: 2.15 mg/L [IQR: 1.15-4.25 mg/L]; P = 0.02 vs individuals without CHIP) (Figure 5); this finding persisted after exclusion of individuals with prevalent cancer (median: 2.13 mg/L [IQR: 0.97-4.57 mg/L]; P = 0.03), log-transformation (P = 0.02) (Supplemental Table 16), or adjustment for age and sex (P = 0.02). Additionally, the distribution of gene-specific CHIP subtypes did not substantially differ between individuals with hsCRP levels ≥2 mg/L vs <2 mg/L (Supplemental Figure 6). Furthermore, NLR values were significantly elevated in those with JAK2 (median: 4.5 [IQR: 3.0-6.2]; P < 0.001 vs those without CHIP), including after exclusion of individuals with prevalent cancer (median: 4.7 [IQR: 3.6-6.3]; P < 0.001), log-transformation (P = 0.004) or adjustment for age and sex (P < 0.001). Although hsCRP and NLR associated significantly with outcomes (Supplemental Tables 17 and 18), associations of CHIP and CHIP subtypes with outcomes did not change materially after further adjustment for hsCRP or NLR (Supplemental Tables 19 and 20).

FIGURE 5. Distribution of hsCRP and NLR by CHIP Driver Gene.

Violin plots show the distribution of high-sensitivity C-reactive protein (hsCRP) levels and neutrophil-to-leukocyte ratio (NLR) by CHIP driver mutation. The plots represent all (A) hsCRP and (B) NLR values between 0-11.10 mg/L and 0-8, corresponding to 95% and 99% of all measurements, respectively. The upper and lower bounds of the white boxes represent the IQR, whereas the horizontal line in the box represents the median. P values represent comparisons of gene-specific CHIP subtypes vs no CHIP using the Wilcoxon rank-sum test. hsCRP and NLR values were available for n = 13,088 and n = 12,812, respectively. No corrections for multiple testing were applied.

DISCUSSION

In this large study of adults with established ASCVD, CHIP associated significantly with recurrent ASCVD events and all-cause mortality (Central Illustration). The highest risk was observed in participants with driver mutations in TET2 or SF3B1/SRSF2/U2AF1. These findings have implications for identifying individuals at risk for having CHIP and for the development of novel ASCVD prevention strategies, including those targeting the secondary prevention population, which has the highest risk of future ASCVD events.

CENTRAL ILLUSTRATION. Clonal Hematopoiesis of Indeterminate Potential Predicts Adverse Outcomes in Patients With Established Atherosclerotic Cardiovascular Disease.

Among participants with established atherosclerotic cardiovascular disease (ASCVD), clonal hematopoiesis of indeterminate potential (CHIP) was associated with higher incidence of ASCVD events and all-cause mortality vs those without CHIP, with the highest incidence observed for those carrying large clones. The highest risk was observed in participants with driver mutations in TET2 or spliceosome genes (ie, SF3B1, SRSF2, and U2AF1). VAF = variant allele fraction.

CHIP IS A RISK FACTOR FOR ADVERSE OUTCOMES IN PATIENTS WITH ESTABLISHED ASCVD.

This study shows that CHIP associates with adverse outcomes in a large cohort of individuals with prevalent ASCVD, extending the findings of previous research linking CHIP to worse prognosis in patients with ST-segment elevation myocardial infarction and cardiogenic shock to the broader population with ASCVD.^17–19 Our findings align with multiple studies establishing a link between CHIP and cardiovascular disease.^6,7,9,11,29 Importantly, this evidence includes mouse experiments supporting causal effects of TET2, JAK2, and TP53 CHIP on acceleration of atherosclerosis.^9,28,30 Nevertheless, recent analyses from the UK Biobank suggested this association in humans may be weaker than suggested by early reports, possibly due to differences in cohort characteristics.^31–33 In recent analyses to explain the heterogeneity in CHIP-related associations across studies, Vlasschaert et al³⁴ found that the association between CHIP and cardiovascular disease is affected by the parameters used to filter CHIP calls. Specifically, when calls are filtered stringently to reduce sequencing artifacts and falsepositive results, the association with cardiovascular disease is stregthened.^22,34 Furthermore, recent experimental data suggest that atherosclerosis and CHIP may show bidirectional causality, with more advanced atherosclerosis leading to increased bone marrow turnover and increased chance of developing detectable CHIP–a vicious cycle in which both factors aggravate each other.³⁵ Whether CHIP is unidirectionally or bidirectionally linked to atherosclerosis, it is independently associated with adverse outcomes, including in the secondary ASCVD prevention population.

We observed the highest risks in individuals carrying large clones with mutations in TET2 or in spliceosome genes (SF3B1/SRSF2/U2AF1). In addition, CHIP driven by mutations in DNA damage repair genes (PPM1D/TP53) was associated with all-cause mortality; this association was unexpectedly stronger in women vs men, and future efforts to replicate this sex difference and define underlying mechanisms are needed. By contrast, DNMT3A CHIP, the most common type of CHIP across the general population, was not significantly associated with the primary outcome. Although this finding aligns with observational evidence that DNMT3A CHIP is linked less strongly to CAD vs other CHIP types such as TET2, previous research in smaller cohorts has also shown that DNMT3A mutations portend a worse prognosis in the setting of ST-segment elevation myocardial infarction and chronic ischemic heart failure.^{13,15,17,36,37} As recent experimental data indicate that mutations in DNMT3A hasten progression of fibrosis, relevant mechanisms by which CHIP accelerates disease may differ across cardiovascular disease subtypes (eg, atherothrombosis vs heart failure).³⁸

Our findings underscore that CHIP consists of a heterogeneous set of gene-specific phenotypes rather than representing one dichotomous risk factor.³¹ A subgroup analysis of CANTOS (Canakinumab Antiinflammatory Thrombosis Outcomes Study) suggested that this heterogeneity may have important implications for clinical translation.¹⁰ The investigators observed that participants with the TET2 CHIP derived greater reduction in major adverse cardiovascular events from canakinumab compared with participants with other CHIP subtypes or no CHIP. Our findings reinforce the notion that evaluating CHIP phenotypes at the gene or mutation level may enable precision medicine for patients with ASCVD, similar to the genetic risk stratification commonly used in hematologic malignancies. Individuals with TET2 or other higher-risk CHIP subtypes (eg, SF3B1/SRSF2/U2AF1) may constitute an ideal population for prospective trials of CHIP-guided novel therapeutics.

CONVENTIONAL BLOOD MARKERS OF INFLAMMATION DO NOT CAPTURE EXCESS RISK IN INDIVIDUALS WITH CHIP.

Mechanistic insights from animal and human studies suggest that the relationship between CHIP and atherosclerosis is partially mediated through inflammatory pathways.^6,9–11,27 For instance, Tet2-deficient mice exhibit increased levels of circulating interleukin (IL)-1β and IL-6, both of which are downstream mediators of the NLRP3 inflammasome.^9,12,27 Human TET2 CHIP carriers exhibit increased concentrations of circulating IL-1β compared to those without CHIP, but data are less consistent for IL-6.^10,39 Although much mechanistic research has focused on TET2 CHIP, the mechanisms by which spliceosome-associated mutations may lead to increased rates of ASCVD events remain largely uncharacterized. Preclinical studies have shown that mutations in SF3B1, SRSF2, and U2AF1 are associated with increased inflammation.^40,41 For instance, Pollyea et al⁴¹ have shown that myeloid cell lines with mutations in SF3B1, SRSF2, or U2AF1 were enriched for IL-6 mRNA, which they replicated in blood samples from patients with myelodysplastic syndrome driven by mutations in these genes. Future research is needed to define pathways linking these diverse mutations to increased atherothrombotic risk.

This study affirms that conventional blood markers of inflammation do not fully capture CHIP-associated cardiovascular risk and may provide only limited utility in identifying individuals with CHIP. Few clinical markers beyond age and prior exposure to cancer-directed therapies robustly predict the presence of CHIP in a given individual.⁴² Indiscriminate population screening for CHIP is neither practical nor currently recommended, although screening populations at a heightened risk for developing CHIP may be more cost-effective.⁴³ Further research is needed to determine clinical and germline genetic factors that signal sufficiently high risk to guide CHIP diagnosis, which will become increasingly relevant and actionable as sequencing costs fall and evidence-based CHIP-focused precision therapies emerge.

STUDY LIMITATIONS.

First, the proportion of participants with CHIP (5.1%) was relatively low in comparison with previous studies of CHIP in populations at heightened cardiovascular risk where the prevalence ranged from ~12% to ~18%.^13,15,17 These studies often ascertained CHIP using targeted deep sequencing methods, which are known to be more sensitive than WES. For example, in their study of clonal hematopoiesis in chronic ischemic heart failure, Assmus et al¹⁵ used sensitive methods to show that clonal hematopoiesis with VAF >0.5% was prevalent in ~56% of study participants. Here, we used biobank-scale WES data and used a stringent protocol to designate CHIP, which–although likely misclassifying a proportion of individuals carrying CHIP clones with VAF <10%–minimizes the number of false-positive CHIP calls due to sequencing artifacts and germline variants and yields a high-confidence CHIP callset.^22,34 This protocol yields a CHIP prevalence of 3.4% in the full UK Biobank, which suggests that CHIP is roughly 1.5-fold more common in those with ASCVD.²² However, the present study may underestimate the true frequency of low-VAF CHIP (ie, VAF <10%), and possibly its impact on disease progression, cardiovascular risk, and outcome in individuals with established ASCVD. Second, ~95% of individuals included in this study were White, limiting the generalizability of our findings to other racial/ethnic groups. Third, the cross-sectional nature of our CHIP analysis precluded assessment of the longitudinal dynamics of CHIP. Fourth, few participants had CHIP driven by mutations other than those included in this analysis, precluding adequately powered analyses of these rarer gene-specific subtypes. Only 9 participants in the study cohort carried JAK2 mutations, which have been strongly associated with risk of incident cardiovascular disease in previous analyses.^6,31 Roughly one-third of those with established ASCVD and JAK2 CHIP mutations were excluded for prevalent diagnoses of hematologic malignancy. This study design may lead to underestimation of event risk in those with JAK2 CHIP.

CONCLUSIONS

The current study identifies CHIP as an independent risk factor for recurrent ASCVD events and all-cause mortality in a large cohort of individuals with established ASCVD. Greatest risk was observed in participants with mutations in TET2 or SF3B1/SRSF2/U2AF1. Further research is needed to elucidate novel mechanisms by which specific CHIP subtypes increase cardiovascular risk and define effective risk mitigation strategies for ASCVD patients with CHIP.

PERSPECTIVES.

COMPETENCY IN MEDICAL KNOWLEDGE:

In patients with established ASCVD, CHIP is associated with an increased risk of ischemic events and all-cause mortality.

TRANSLATIONAL OUTLOOK:

Future studies should seek to define the mechanisms by which subtypes of CHIP are associated with differential cardiovascular risk to facilitate development of risk reduction strategies for patients with ASCVD.

ABBREVIATIONS AND ACRONYMS

ASCVD

atherosclerotic cardiovascular disease

CAD

coronary artery disease

CHIP

clonal hematopoiesis of indeterminate potential

hsCRP

high-sensitivity C-reactive protein

VAF

variant allele fraction

WES

whole exome sequencing