Colchicine and Longitudinal Dynamics of Clonal Hematopoiesis

Abstract

BACKGROUND

Clonal hematopoiesis (CH) is an aging-related hematologic condition associated with increased risk for cardiovascular events. Larger CH clones associate more strongly with cardiovascular risk. Preclinical data indicate that inflammatory signaling drives expansion of CH clones and CH-associated cardiovascular disease. However, the effect of anti-inflammatory therapies on CH clonal dynamics in humans is unclear.

OBJECTIVES

The goal of this study was to test the association of randomization to colchicine vs placebo with CH growth in participants with chronic coronary artery disease. It also assessed the association of colchicine use with change in inflammatory biomarkers over time according to CH status.

METHODS

In this exploratory substudy of the LoDoCo2 (Low-Dose Colchicine 2) trial, high-coverage targeted sequencing was used to detect CH driver mutations and to quantify variant allele frequency at 4 timepoints: baseline, after a 30-day open-label colchicine run-in phase (0.5 mg daily), 1 year postrandomization to colchicine or placebo, and at end of study (median follow-up of 25.0 months). Clonal dynamics were assessed by using a generalized linear mixed model. High-sensitivity C-reactive protein and interleukin-6 were additionally measured at baseline, randomization, and 1 year postrandomization.

RESULTS

In total, 854 participants contributed 2,047 observations across 4 timepoints, including before and after the prerandomization colchicine run-in period. Randomization to placebo was associated with a 14.9% annual increase in CH clone size (β_time = 0.14; 95% CI: 0.08 to 0.21) vs a nonsignificant 6.3% increase with colchicine (β_time on colchicine: 0.06; 95% CI: −0.01 to 0.14), although this difference between treatment arms was not statistically significant (P_interaction = 0.13). Compared with placebo, colchicine was associated with attenuated clonal growth in TET2 CH (β_time on colchicine: 0.09 [95% CI: −0.04 to 0.22]; β_time placebo: 0.27 [95% CI: 0.16 to 0.37]; P_interaction = 0.04). Among individuals with non-DNMT3A CH, interleukin-6 levels increased to a lesser extent in those receiving colchicine vs placebo over 1 year (30.0% vs 98.1% increase, respectively; P_interaction = 0.01).

CONCLUSIONS

In this exploratory analysis, treatment with low-dose colchicine was associated with attenuated clonal expansion in TET2 CH. These findings suggest the potential for colchicine to curb the proliferative advantage of key CH driver mutations and to mitigate their associated risk of cardiovascular disease. Further validation in prospective studies is warranted.

Clonal hematopoiesis (CH) of indeterminate potential refers to the age-related expansion of hematopoietic stem cells with leukemogenic mutations detectable in the peripheral blood in the absence of hematologic malignancy.¹ CH of indeterminate potential (CHIP), traditionally defined as CH with a variant allele frequency (VAF) ≥2%, is common in the general population, affecting >10% of individuals aged >70 years.² In addition to increasing the risk for hematologic malignancies, CH is associated with up to a twofold increased risk of cardiovascular events and is linked to worse outcomes in heart failure and aortic valve stenosis.^3-7

In multiple prior studies, larger CH clones were associated with a greater risk of cardiovascular events, suggesting that clonal expansion promotes increased risk.^1,8-10 Clonal dynamics vary significantly among different CH driver mutations, the most common of which are DNMT3A, TET2, and ASXL1. DNMT3A non-R882 clones show the slowest growth, while clones with splicing factor mutations (eg, SF3B1, SRSF2, U2AF1) expand most rapidly.^11-13 DNMT3A is the most common CH driver gene but may confer less cardiovascular risk than non-DNMT3A genes such as TET2.⁵ The increased risk of cardiovascular events associated with CH has been attributed to an increased inflammatory state; this is particularly true in TET2 CH, for which data indicate that the NOD-, LRR-, and pyrin domain-containing protein 3 (NLRP3) inflammasome pathway may promote both clonal expansion and accelerated atherogenesis.^14-16 Systemic inflammation may confer relative proliferative advantage to hematopoietic stem and progenitor cells (HSPCs) harboring CH driver mutations, thereby driving an increase in clone size.¹⁷ Moreover, a recent post hoc analysis of the CANTOS (Canakinumab Antiinflammatory Thrombosis Outcome Study) trial indicated outsized cardiovascular benefit from canakinumab, an antiinterleukin 1-beta (IL-1β) antibody, among individuals with TET2 CH.¹⁸

Colchicine is a microtubule formation inhibitor and an anti-inflammatory drug that has been previously shown to reduce cardiovascular events in multiple clinical trials, at least partly by attenuating the NLRP3 inflammasome cascade.^19-21 Moreover, murine data suggested that colchicine attenuates TET2-driven accelerated atherosclerosis.²² However, the effect of colchicine on CH clonal dynamics in humans has not been studied to date. Colchicine reduces systemic inflammation; we therefore hypothesized that it attenuates clonal proliferation by dampening inflammatory signaling in HSPCs. The current analysis had 2 objectives: 1) to evaluate the association between colchicine and CH growth in a randomized trial of participants with chronic coronary artery disease (LoDoCo2 [Low-Dose Colchicine 2])¹⁹; and 2) to assess the longitudinal association between colchicine use and inflammatory biomarker levels of high-sensitivity C-reactive protein (hsCRP) and interleukin-6 (IL-6) according to CH status.

METHODS

STUDY POPULATION.

Longitudinal blood samples in the LoDoCo2 trial were collected at 3 participating Dutch sites from participants who voluntarily participated in this biobanking substudy.^21,23 For this exploratory LoDoCo2 substudy, we used all available blood samples from these participants, including end-of-study samples, for which the selection and study population have been described previously.²⁴ Participants provided separate written informed consent for DNA and biomarker analyses.

In brief, LoDoCo2 (ACRTN: 12614000093684) was a placebo-controlled, double-blind, international multicenter randomized trial.^19,25 The results have been published previously.¹⁹ A total of 5,522 participants between the ages of 35 and 82 years with chronic coronary artery disease were randomized to receive either colchicine 0.5 mg once daily or matching placebo. The primary composite endpoint included cardiovascular death, spontaneous myocardial infarction, ischemic stroke, or ischemia-driven revascularization. Compared with placebo, colchicine resulted in a 31% reduction in the primary endpoint.

Medical ethics approval for the LoDoCo2 trial and the LoDoCo2 substudies was obtained in both participating countries (Sir Charles Gairdner Group HREC [Perth, Australia] and MEC-U [Nieuwegein, the Netherlands]). All participants provided written informed consent.

HIGH-COVERAGE TARGETED SEQUENCING.

Wholeblood samples were collected at 4 timepoints: baseline, after a 30-day open-label run-in phase of colchicine 0.5 mg once daily, 1 year after randomization to receive colchicine or placebo, and at end of study (Figure 1). Targeted CH sequencing was performed on whole-blood samples from 854 LoDoCo2 participants by Vanderbilt University Medical Center on an NovaSeq 6000 platform (Illumina). A panel of 22 CH genes was prioritized and enriched in the DNA libraries, aiming for a depth of coverage approximately 2,000× in these common CH gene regions, robustly detecting variants at 1% VAF and lower (Supplemental Table 1).²⁶

FIGURE 1. Overview of the Study Design.

The figure depicts the 4 different timepoints at which samples were available. Baseline indicates the timepoint before the start of the 30-day open label run-in period. Participants were then randomized to receive colchicine or placebo (the randomization visit). The third visit was 1 year postrandomization, and the last visit was at the end of study. CH = clonal hematopoiesis; CHIP = clonal hematopoiesis of indeterminate potential; hsCRP = high-sensitivity C-reactive protein; IL-6 = interleukin-6; OD = once daily; VAF = variant allele frequency.

CH GENOTYPING.

Somatic mutations were identified from targeted CH sequencing data using GATK Mutect2 (Broad Institute) software in the Terra platform, as previously described.^16,27 A Panel of Normals (the 1000 Genomes PON) and the Genome Aggregation Database (gnomAD) were used to filter germline variants from the putative somatic mutation calls. Mutect2 calls were further filtered, and variants were retained for the following: 1) total depth of coverage ≥100; 2) number of reads supporting the alternate allele ≥10; 3) ≥2 reads in both forward and reverse direction supporting the alternate allele; and 4) not germline variants or recurrent sequencing artifacts. Sequencing depth was evaluated across all paired longitudinal samples. For variants not detected in all timepoints, 2 samples with insufficient depth (<100×) at positions where the variant was present in other visits were excluded to minimize potential false negatives.

BIOMARKERS.

hsCRP (#HK369; Hycult Biotech) and IL-6 (#D6050; R&D Systems) levels were measured in samples from baseline, after the 30-day colchicine run-in period, and 1 year postrandomization.

STATISTICAL ANALYSIS. CH dynamics.

A generalized linear mixed model was used to calculate the change in VAF over time, defined as allele depth divided by the depth of coverage, fitting a beta-binomial distribution with a logit link function. Because the vast majority of clones harbor a single mutation, models assume that the VAF for a given driver mutation represents the prevalence of a single clone, consistent with the approach used in previous longitudinal CH studies.^11,12,28 For variants not detected in paired samples from all timepoints, VAF was imputed as 0.1% for the timepoints where the variant was not observed.¹¹ We analyzed DNMT3A variants and non-DNMT3A variants separately because previous studies suggest differences in clonal dynamics and relevance to cardiovascular disease.¹² For the 30-day open-label run-in analyses, all clones of individuals with detectable CH at baseline or at the 30-day timepoint were included. Time (baseline and after 30-day run-in period) was included in the model as a fixed effect and subject and variant as random effects. Age and gene group (splicing factor, DNA damage repair, R882 DNMT3A, non-R882 DNMT3A, TET2, or other) were included as covariates. For the randomized treatment period, all clones of individuals with detectable CH at one of the 3 timepoints were included. Here, time was added as follow-up time in months for each individual, and a time-by-treatment interaction was added to the generalized linear mixed model, with subject and variant as random effects with the same covariates.

Because larger clones may be more responsive to interventions and enable more precise quantification of change in VAF, sensitivity analyses were performed restricting the open-label analyses timeframe to clones with VAF ≥1% and ≥2% at baseline. For the postrandomization period, similar sensitivity analyses were performed restricted to clones with at least 1 timepoint with VAF ≥1% and ≥2%. In addition, because very large clones may reach a plateau of maximal growth, a sensitivity analysis was conducted excluding clones with VAF >30% at randomization.

All statistical analyses were performed by using R version 4.4.1 (R Foundation for Statistical Computing).

Biomarkers.

A linear mixed model was used with log₂+1-transformed hsCRP and IL-6 levels as outcome, time as a fixed effect, age as a covariate, and subject as random effects for samples drawn before and after the 30-day colchicine run-in period. For samples drawn before and after 1 year of randomized study treatment, a time-by-treatment interaction was added to the model. Estimated means (EMs) were calculated with R package emmeans for each timepoint by back transformation.

Due to the exploratory nature of our analyses, no adjustments were made for multiple testing, and findings should be viewed as hypothesis-generating. Two-sided P < 0.05 indicated statistical significance.

RESULTS

BASELINE CHARACTERISTICS.

The study design and included participants are summarized in Figure 1 and Supplemental Figure 1. The baseline characteristics of all 854 participants in this longitudinal analysis are presented in Supplemental Table 2. Overall, the mean ± SD age at baseline was 65.5 ± 8.8 years, and 149 (17.4%) were female. Of the 854 participants, 45 (5.3%) were not randomized (ie, they dropped out after the 30-day open-label run-in phase), 406 (47.5%) were allocated to receive colchicine, and 403 (47.2%) were allocated to receive placebo. There were no significant differences in baseline characteristics across these groups. The prevalence of all CH driver mutations (VAF >0) at any timepoint was 27.0% for DNMT3A CH, 10.4% for TET2, and 4.2% for ASXL1. Median VAF of all CH clones was 0.022 (Q1-Q3: 0.013-0.040), and the distribution of VAF per visit is shown in Supplemental Figure 2.

In total, 339 unique participants underwent sequencing before and after a 30-day open-label colchicine run-in period. Supplemental Table 3 shows the baseline characteristics of the participants with CH (VAF >0) vs no CH during the open-label run-in period. Those with CH had a higher mean age of 67.7 ± 8.8 years compared with 63.8 ± 8.7 years in those with no CH (P < 0.001). At baseline, presence of CH was associated with higher geometric mean IL-6 levels compared with those without CH (2.24 ng/L [95% CI: 1.82-2.72 ng/L] vs 1.74 ng/L [95% CI: 1.51-1.98 ng/L]; P = 0.03).

The most common CH gene at baseline was DNMT3A (51% of all clones) (Figure 2A), followed by TET2 (18% of all clones) and ASXL1 (8% of all clones). Of all participants with baseline samples, 25% had 1 CH variant, 9% had 2 variants, and <2% had ≥3 variants (Figure 2B). As expected, the prevalence of CH significantly increased with age across all VAF thresholds (Figure 2C). VAF distribution according to CH driver gene at each visit is depicted in Figure 2D.

FIGURE 2. Prevalence of CH in LoDoCo2 Targeted Sequencing.

(A) Frequency of CH driver mutations of all CH clones found in this analysis according to visit. (B) frequency of participants carrying ≥1 CH mutations of all participants with samples available at each timepoint. (C) CH prevalence by age by visit. (D) VAF distribution of all CH driver mutations detected; the black dashed line indicates the mean VAF of all variants detected across all visits. The number of unique participants sequenced at each visit is 335 at V1 (the visit before 30-day open-label colchicine), 318 at V2 (randomization visit [the visit after the run-in period]), 275 at V3 (1 year postrandomization), and 716 at end of study. Abbreviations as in Figure 1.

Baseline characteristics of participants randomized after the 30-day open-label run-in period with any CH (VAF >0) at any visit stratified according to treatment allocation are presented in Table 1. The table shows that randomization was largely preserved in the current analytic cohort.

TABLE 1.

Baseline Characteristics of Participants With Any CH During the Randomized Treatment Period

	Colchicine (n = 155)	Placebo (n = 153)	SMD
Demographics
Age, y	67.81 ± 7.93	68.16 ± 8.23	0.043
Female	34 (21.9)	23 (15.0)	0.179
Clinical characteristics
Hypertension	79 (51.0)	79 (51.6)	0.013
Current smoker	16 (10.4)	18 (11.8)	0.044
Diabetes baseline	26 (16.8)	25 (16.3)	0.012
Creatinine clearance <60 mL/min/1.73 m2	9 (5.8)	15 (9.8)	0.149
Prior acute coronary syndrome	121 (78.1)	125 (81.7)	0.091
Prior coronary revascularization	147 (94.8)	142 (92.8)	0.084
History of atrial fibrillation	26 (16.8)	19 (12.4)	0.124
Coronary artery bypass grafting	14 (9.0)	26 (17.0)	0.238
Percutaneous coronary intervention	140 (90.3)	131 (85.6)	0.145
Medication
Single or dual antiplatelet therapy	134 (86.5)	138 (90.2)	0.117
Anticoagulant	31 (20.0)	21 (13.7)	0.168
Ezetimibe	31 (20.0)	29 (19.0)	0.026
Statin	146 (94.2)	140 (91.5)	0.105
High-dose statin	87 (56.1)	84 (54.9)	0.025
Renin angiotensin inhibitor	106 (68.4)	110 (71.9)	0.077
Beta-blocker	106 (68.4)	102 (66.7)	0.037
hsCRPa	1.97 (1.48-2.56)	2.24 (1.71-2.87)	0.123
IL-6a	1.83 (1.38-2.36)	2.30 (1.57-3.25)	0.181

Values are mean ± SD, n (%), or mean (95% CI). Baseline high-sensitivity C-reactive protein (hsCRP) and interleukin-6 (IL-6) values were available in 120 individuals. ^aGeometric mean of log₂+1 transformed data.

CH = clonal hematopoiesis; SMD = standardized mean difference.

CLONAL DYNAMICS DURING THE OPEN-LABEL RUN-IN PERIOD.

Clonal dynamics were examined in all available samples across the 30-day open-label run-in period in 339 participants. In total, 146 unique individuals had detectable CH (VAF >0) at 1 or 2 timepoints. Treatment with colchicine during the 30-day open-label run-in period was associated with a higher proportion of VAF reduction in non-DNMT3A CH clones (62.2% Δ VAF <0) compared with DNMT3A clones (42.1% Δ VAF <0; P < 0.01) (Figure 3). Among 420 CH clones in 146 individuals including all VAF, no statistically significant change was observed during the initial 30-day open-label colchicine phase for all CH clones (β_time = −0.026; 95% CI: −0.099 to −0.046; 2.6% decrease; P = 0.48) (Supplemental Figure 3A) or upon further stratification by CH driver gene. When restricting analysis to individuals with VAF ≥1% and ≥2% at baseline, non-DNMT3A exhibited a reduction in VAF (Supplemental Figures 3B and 3C).

FIGURE 3. Waterfall Plots of Absolute Change in VAF During the Open-Label Run-In Period in Paired Samples.

(A) Absolute change in VAF for DNMT3A clones in 107 paired CH clones where n = 45 (42.1%) had a negative change (ie, decrease) in VAF, n = 58 (54.2%) a positive change (ie, increase) in VAF, and n = 4 (3.7%) no change in VAF. (B) Absolute changes in VAF for non-DNMT3A clones in 90 paired CH clones where n = 56 (62.2%) had a negative change (ie, decrease) in VAF, n = 32 (35.6%) a positive change (ie, increase) in VAF, and n = 2 (2.2%) no change in VAF. The figures shows that VAF reduction was more prominent in non-DNMT3A CH clones compared with DNMT3A (calculated by using the test for equal proportions, P < 0.01). Abbreviations as in Figure 1.

CLONAL DYNAMICS AFTER RANDOMIZATION TO COLCHICINE OR PLACEBO.

Samples from 809 unique randomized participants (Supplemental Table 2) were available at any visit at or after the randomization visit (ie, the visit following the 30-day run-in period). Of these 809 participants, 308 had detectable CH (VAF >0 at any of the visits at or after randomization). We quantified change in clone size from randomization to end of study (median of 25.0 [Q1-Q3: 17.9-31.2] follow-up months) among 780 observations in 308 individuals. Absolute changes in VAF are summarized in Supplemental Table 4. After randomization, allocation to placebo was associated with an increase in VAF across visits for all CH variants (β_time in placebo group: 0.142 [95% CI: 0.078-0.206]; P < 0.0001) but not for the colchicine group (β_time in colchicine group: 0.064 [95% CI: −0.014 to 0.141]; P = 0.11). This difference between treatment arms, however, was not statistically significant (P_interaction = 0.13) (Figure 4).

FIGURE 4. Association of Randomization to Colchicine vs Placebo With Change in CH Clone Size.

Effect estimate for generalized beta-binomial mixed-effect models according to gene. A generalized mixed model was used to model the change over the follow-up period following randomization, fitting a beta-binomial distribution with a logit link function for VAF (defined as alternative reads/depth of coverage). The percent change was calculated using the estimated VAF at 1 year by the model relative to the randomization visit. All subjects were included with at least one measurement for the 3 time points. Models were adjusted for age and gene group (R882 DNMT3A, non-R882 DNMT3A, TET2, splicing factor, DNA damage repair, or other).

obs = observations; sub = subjects, var = variants; other abbreviations as in Figure 1.

In the exploratory gene-specific analyses, DNMT3A CH did not exhibit clonal growth during follow-up, irrespective of treatment allocation (β_time in colchicine group: 0.044 [95% CI: −0.049 to 0.138; P = 0.35]; β_time in placebo group: 0.061 [95% CI: −0.025 to 0.146; P = 0.17]; P_interaction = 0.80). Significant clonal growth in non-DNMT3A variants was observed in the placebo group (β_time in placebo group: 0.216; 95% CI: 0.122-0.309; P < 0.0001) but not among individuals randomized to receive colchicine (β_time in colchicine group: 0.090; 95% CI: −0.032 to 0.212; P = 0.15); this difference did not reach statistical significance in interaction analysis (P_interaction = 0.11). In TET2 CH, use of colchicine vs placebo was associated with attenuated clonal growth (β_time colchicine: 0.090 [95% CI: −0.039 to 0.218]; β_time placebo: 0.265 [95% CI: −0.157 to 0.373]; P_interaction = 0.04). Sensitivity analyses restricted to clones with VAF ≥1% and ≥2% for at least one timepoint at or after the randomization visit and VAF ≤30% at the randomization visit displayed consistent findings (Supplemental Figures 4 to 6).

CHANGE IN INFLAMMATORY BIOMARKERS.

Biomarker levels of hsCRP and IL-6 were available in 145 individuals with CH and 181 without CH during the open-label run-in period. Of these, 11 participants had 1 missing observation each for hsCRP and 12 for IL-6. In those with CH, a significant reduction in hsCRP levels from baseline was observed during the open-label run-in period (P < 0.001) (Supplemental Table 5).

In further exploratory analyses of CH subtypes, a reduction in hsCRP was observed in individuals with non-DNMT3A CH with an estimated decrease of 35.2% (EM: 2.32 mg/L [95% CI: 1.80-2.93] vs 1.50 mg/L [95% CI: 1.10-1.98]; P < 0.001). A numerically smaller but significant change in hsCRP reduction was found for those without CH. During the run-in period, a borderline reduction in IL-6 levels was observed in individuals with non-DNMT3A CH (EM: 2.51 ng/L [95% CI: 1.89-3.26] vs 1.99 ng/L [95% CI: 1.44-2.65]; 20.8% decrease; P = 0.05), although this pattern was not seen in DNMT3A (EM: 1.87 ng/L [95% CI: 1.44-2.38] vs 1.93 ng/L [95% CI: 1.48-2.47]; 3.0% increase; P = 0.82).

Biomarker levels were measured at the randomization visit (the visit after the run-in period) and at 1 year postrandomization in 116 participants with CH (with 12 individuals missing 1 observation each for hsCRP and 6 for IL-6) and 161 without CH (with 5 missing 1 observation each for hsCRP). One year after randomization, no significant changes in hsCRP were observed compared with the randomization visit. In addition, there was no significant treatment-by-time interaction effect for hsCRP (Table 2). By contrast, IL-6 levels increased over time in individuals with and without CH overall, irrespective of treatment allocation. However, among those with non-DNMT3A CH, allocation to colchicine was associated with a blunted rise in IL-6 over time, with a nonsignificant increase in IL-6 in the colchicine arm (EM: 1.69 ng/L [95% CI: 0.93-2.77] at baseline vs 2.20 ng/L [95% CI: 1.28-3.48]; 30.0% increase; P = 0.33) vs an increase in IL-6 in the placebo arm (EM: 2.14 ng/L [95% CI: 1.31-3.25] vs 4.23 ng/L [95% CI: 2.86-6.09]; 98.1% increase; P < 0.0001; P_interaction = 0.01). This pattern of findings was consistent for both TET2 and ASXL1 CH clones.

TABLE 2.

Inflammatory Markers hsCRP and IL-6 at the Randomization Visit and After 1 Year of Randomized Treatment

Randomized	No. of Observations	No. of Participants	Treatment	Estimate for Time (SE)	Estimated Mean (95% CI) V2	Estimated Mean (95% CI) V3	% Change	P Value	P Value (Interaction)
hsCRP
Any CH	220	116	Colchicine	−0.215 (0.130)	1.52 (1.12-2.00)	1.17 (0.82-1.60)	−23.0	0.36	0.99
			Placebo	−0.213 (0.127)	1.76 (1.34-2.25)	1.38 (1.00-1.83)	−21.5	0.34
DNMT3A	146	76	Colchicine	−0.241 (0.170)	1.56 (1.10-2.13)	1.17 (0.76-1.66)	−25.3	0.49	0.89
			Placebo	−0.207 (0.179)	1.74 (1.22-2.39)	1.38 (0.91-1.95)	−21.0	0.66
Non-DNMT3A	109	58	Colchicine	−0.035 (0.130)	1.23 (0.74-1.84)	1.18 (0.70-1.78)	−4.2	0.99	0.78
			Placebo	−0.089 (0.126)	1.52 (1.02-2.15)	1.38 (0.89-2.00)	−9.5	0.90
TET2	38	20	Colchicine	−0.255 (0.286)	1.36 (0.58-2.52)	0.98 (0.31-2.00)	−28.1	0.81	0.62
			Placebo	−0.473 (0.320)	2.62 (1.32-4.64)	1.60 (0.64-3.14)	−38.7	0.47
ASXL1	27	14	Colchicine	0.240 (0.285)	1.16 (−0.12 to 4.32)	1.55 (0.04-5.28)	33.8	0.83	0.41
			Placebo	−0.035 (0.156)	1.72 (0.78-3.17)	1.66 (0.73-3.08)	−3.9	1.00
Other	64	34	Colchicine	−0.053 (0.111)	1.22 (0.63-2.03)	1.14 (0.57-1.93)	−6.5	0.96	0.30
			Placebo	0.111 (0.111)	1.02 (0.53-1.66)	1.18 (0.64-1.90)	15.9	0.74
No CH	317	161	Colchicine	−0.131 (0.095)	1.25 (0.98-1.55)	1.05 (0.81-1.33)	−15.6	0.52	0.09
			Placebo	0.105 (0.098)	1.44 (1.14-1.78)	1.63 (1.30-2.00)	12.8	0.71
IL-6
Any CH	226	116	Colchicine	0.397 (0.135)	1.77 (1.21- 2.47)	2.65 (1.90-3.58)	49.6	0.02	0.58
			Placebo	0.500 (0.129)	2.14 (1.53-2.90)	3.44 (2.58-4.52)	60.8	<0.01
DNMT3A	148	76	Colchicine	0.451 (0.181)	1.67 (1.08-2.44)	2.65 (1.83-3.71)	58.7	0.07	0.99
			Placebo	0.447 (0.195)	1.81 (1.14-2.68)	2.83 (1.93-4.01)	56.4	0.11
Non-DNMT3A	113	58	Colchicine	0.246 (0.144)	1.69 (0.93-2.77)	2.20 (1.28-3.48)	30.0	0.33	0.01
			Placebo	0.739 (0.132)	2.14 (1.31-3.25)	4.23 (2.86-6.09)	98.1	<0.0001
TET2	39	20	Colchicine	0.195 (0.212)	1.52 (0.49-3.25)	1.88 (0.70-3.89)	24.1	0.79	0.15
			Placebo	0.661 (0.224)	2.44 (0.93-5.14)	4.44 (2.05-8.71)	81.9	0.04
ASXL1	28	14	Colchicine	0.169 (0.509)	2.06 (0.51-5.20)	2.44 (0.70-5.97)	18.5	0.99	0.22
			Placebo	0.913 (0.266)	1.84 (1.02-2.89)	4.34 (2.81-6.49)	136.4	0.02
Other	66	34	Colchicine	0.220 (0.164)	1.69 (0.67-3.33)	2.13 (0.95-4.04)	26.2	0.55	0.14
			Placebo	0.561 (0.154)	2.02 (0.98-3.62)	3.46 (1.91-5.82)	71.0	<0.01
No CH	322	161	Colchicine	0.322 (0.082)	1.70 (1.39-2.06)	2.38 (1.99-2.83)	39.7	<0.001	0.23
			Placebo	0.465 (0.085)	1.70 (1.37-2.08)	2.73 (2.28-3.25)	60.4	<0.0001

The analysis shows the estimated mean in inflammatory biomarkers hsCRP and IL-6 at the randomization visit (the visit after the run-in period; V2) and 1 year after the randomization visit (V3), stratified according to CH status. P values were adjusted for age and calculated using a linear mixed model on log₂(X+1) transformed biomarker levels; they indicate the change between the 2 timepoints. Changes were calculated by estimating the back-transformed mean from the log₂(X+1) scale at each visit with 95% CI. Any CH was defined as variant allele frequency >0 at either visit; no CH was defined as variant allele frequency = 0 at both visits. Participants (n = 16) with no CH at 1 visit and missing data at the other visit were excluded. Bold indicates statistical significance for interaction testing.

Abbreviations as in Table 1.

DISCUSSION

In this exploratory substudy of a randomized clinical trial of low-dose colchicine in individuals with chronic coronary artery disease, colchicine was associated with attenuated clonal growth of CH clones, with evidence of heterogeneity among subtypes of CH (Central Illustration). Specifically, randomization to receive colchicine was associated with attenuated clonal expansion in TET2 mutations. Although colchicine use was associated with numerically lower clonal expansion in non-DNMT3A CH clones compared with placebo, differences between treatment arms were not significant for CH subtypes other than TET2. Furthermore, among those with non-DNMT3A CH, randomization to colchicine vs placebo was associated with less increase over time in IL-6, part of the NLRP3 inflammasome axis that has been implicated in CH-associated atherogenesis.^10,14,29-32 These findings represent preliminary evidence supporting the hypothesis that use of colchicine or targeted anti-inflammatory therapies may blunt the proliferative advantage of key CH driver mutations and mitigate their associated risk of cardiovascular disease.

CENTRAL ILLUSTRATION. Colchicine and Longitudinal Dynamics of Clonal Hematopoiesis.

This central illustration summarizes the study design and main findings of the article. CAD = coronary artery disease; CH = clonal hematopoiesis; LoDoCo2 = Low-Dose Colchicine 2; VAF = variant allele frequency.

Notably, our findings are aligned with those reported in a recent preprint investigating the effect of colchicine on CH in COLCOT (Colchicine Cardiovascular Outcomes Trial).³³ Using ultra-deep sequencing, the authors identified 1,954 CH variants in 848 COLCOT participants who underwent sequencing at 2 timepoints. Over a median 19.5 months of follow-up, those with TET2 CH randomized to the colchicine group had less increase in VAF vs those with TET2 CH randomized to the placebo group (estimated 10.3% reduction in VAF with colchicine vs 9.1% increase in VAF with placebo; P_interaction = 0.001). Although LoDoCo2 included a stable coronary disease population while COLCOT enrolled a postacute myocardial infarction population, both studies suggest that TET2 clonal dynamics may be modifiable by low-dose colchicine. In addition, the COLCOT investigators also reported a similar pattern of altered clonal trajectories associated with colchicine use for CH driven by mutations in SF3B1, a spliceosome gene, and TP53, a DNA damage repair gene.³³

Larger CH clones are more strongly linked with risk of adverse clinical outcomes, including both cardiovascular disease and cancer.^1,34 Our findings support the premise of attenuated long-term expansion in TET2 CH with use of colchicine, which is known to dampen NLRP3 inflammasome–related signaling. This is consistent with prior preclinical evidence in a macaque model with genome-edited CH showing that attenuated IL-6 signaling with tocilizumab suppressed TET2-mutant clonal expansion.³⁰ In addition, preclinical evidence in another common non-DNMT3A CH driver gene, ASXL1, revealed that CH mutant clones and their progeny exhibit increased inflammatory signaling and that clonal fitness was associated with resistance to the deleterious effects of inflammation in mutated cells.³⁵ These findings suggest that the proliferative advantages of CH clones may be driven by increased inflammation originating from mutated clones. It is important to note, however, that IL-6 is less specific to the NLRP3 inflammasome vs IL-1β and IL-18 and that colchicine has other effects via microtubule inhibition.^36,37 Further mechanistic studies are needed to clarify mechanisms linking colchicine to modulation of clonal growth.

In the current analysis, 30 days of colchicine treatment were associated with an apparent reduction in clone size in those with a VAF ≥1%. However, this most likely does not reflect a true change in VAF, as it takes weeks for HSPCs to develop into mature circulating blood cells. This finding may instead be attributable to alteration of the peripheral blood cell differential and reduction of the neutrophil-to-lymphocyte ratio by colchicine, as previously reported.^38,39 Non-DNMT3A mutations are most commonly found in myeloid lineage cells but rarely seen in lymphoid cells, whereas DNMT3A mutations are typically present in both myeloid and lymphoid cells; thus, a relative reduction of circulating myeloid cells with colchicine use may appear as a reduction in VAF while this actually reflects neutrophil suppression.^38,40

Gene-specific CH subtypes display heterogeneous clonal dynamics. Similar to previous findings, we found that DNMT3A clones postrandomization were associated with the slowest growth, with an annual estimated growth of 4.4% for colchicine and 6.8% for placebo.¹² Growth rates in the placebo group for individuals with non-DNMT3A variants in the current analysis were numerically higher than those reported in previous studies. ASXL1 and TET2 have been reported to have an annual growth rate of approximately 10%, while our study observed annual growth of 19% and 30%, respectively. Considering the numerical decline in VAF during the 30-day open-label period, one possible explanation for this increased clonal growth rate could be a “rebound” effect after stopping colchicine. A similar rebound effect in clonal growth was observed previously in some macaques with TET2-deficient bone marrow after stopping tocilizumab.³⁰ Another possibility is that the current cohort included individuals with established coronary artery disease, which could represent a specific population with accelerated clonal growth related to accelerated hematopoietic stem cell proliferation in the setting of atherosclerosis.¹⁷

CH has emerged as a novel risk factor for cardiovascular disease, at least partly due to increased production of inflammatory mediators by macrophages.⁶ Specifically, CH is associated with elevated levels of IL-1β and IL-6, which are components of the NLRP3 inflammasome pathway, whereas associations with hsCRP have been less consistent.^8,16 This was shown in a cohort study of 4,131 participants, which showed elevated levels of IL-6, but not hsCRP, in individuals with non-DNMT3A CH subtypes.⁴¹ Consistent with these findings, our analysis similarly showed increased IL-6 but not hsCRP levels at baseline in participants with non-DNMT3A CH. The role of IL-6 in CH-associated cardiovascular risk is further supported by studies showing reduced cardiovascular risk in individuals with CH in the context of genetically decreased IL-6 signaling.¹⁰ Moreover, preclinical evidence showed that colchicine attenuated accelerated atherosclerosis and reduced IL-1β levels in a mouse model of TET2 mutant CH to a greater extent than in wild-type control mice. In line with these findings, among individuals with non-DNMT3A CH, IL-6 levels increased to a lesser extent in those receiving colchicine vs placebo over 1 year of randomization treatment in the current study. However, IL-6 findings were not fully aligned with observed clonal dynamics, suggesting that IL-6 is likely not the sole inflammatory mechanism contributing to clonal growth.

CLINICAL IMPLICATIONS.

The addition of low-dose colchicine to optimal medical therapy has been shown to reduce residual cardiovascular risk in chronic coronary artery disease.¹⁹ These findings represent the first in-human evidence suggesting that colchicine may modulate CH clonal dynamics, particularly for TET2 CH, and may, in turn, mitigate the risk linked to larger CH clones. CH is associated with residual risk, especially TET2 CH among more common CH subtypes.⁸ These findings align with a post hoc analysis of the CANTOS trial, which showed that individuals with TET2 CH derived the greatest benefit from IL-1β blockade, while those with non-TET2 CH or no CH seemed to experience a lesser magnitude of benefit.¹⁸ Taken together, these findings suggest that there may be potential for anti-inflammatory therapies, such as colchicine, as a precision medicine strategy in CH-guided therapy to reduce residual cardiovascular risk.

Notably, targeted IL-6 inhibition is currently under investigation using the IL-6 inhibitor ziltivekimab in several cardiovascular outcome trials (eg, ZEUS [A Research Study to Look at How Ziltivekimab Works Compared to Placebo in People With Cardiovascular Disease, Chronic Kidney Disease and Inflammation]; NCT05021835) and may warrant dedicated study in the CH population. Lower cost and wider availability represent potential practical advantages of colchicine over monoclonal antibodies. However, it is important to acknowledge that the observed effect sizes in the current analysis were modest and CIs were wide. It remains unknown whether small changes in VAF also result in reduction of CH-associated cardiovascular risk. Larger prospective studies are needed with prespecified CH-related outcomes and clinical endpoints to strengthen the premise for CH-guided colchicine precision therapy. Of note, a prospective clinical trial testing the effect of IL-1β inhibition on vascular imaging endpoints and clonal dynamics is planned in TET2 CH (TECTONIC [Effects of IL-1 Beta Inhibition on Vascular Inflammation in TET2 Clonal Hematopoiesis]; NCT06691217).

STUDY LIMITATIONS.

First, the findings during the 30-day open-label run-in period should be interpreted with caution as this period lacks a placebo arm for reference and, as mentioned earlier, a true decrease in VAF is unlikely in this short timeframe.³⁸ Leukocyte differential was not consistently measured in the LoDoCo2 trial. Second, because all randomized participants were exposed to colchicine prior to randomization, we cannot exclude the possibility that a “legacy effect” of the 30-day colchicine run-in period may have attenuated or otherwise affected observed differences between the colchicine and placebo groups after randomization. Third, due to the small sample size, our analysis had limited power to detect an interaction even when effect estimates diverged considerably, or to associate CH with cardiovascular events. Fourth, the current analysis lacked measurements of other cytokines more specific to the NLRP3 inflammasome, such as IL-1β or IL-18. Fifth, during follow-up in the LoDoCo2, the initiation of new therapies was not consistently ascertained and was therefore not considered in the current analysis. Finally, as noted earlier, this analysis is post hoc and exploratory using samples from the subset of LoDoCo2 sites that conducted biobanking; findings should be seen as hypothesis-generating and require further validation in prospective studies with prespecified CH-related outcomes and clinical endpoints.

CONCLUSIONS.

In this exploratory LoDoCo2 substudy, treatment with low-dose colchicine was associated with attenuated clonal expansion in TET2 CH. These findings suggest the potential for colchicine to curb the proliferative advantage of key CH driver mutations and mitigate their associated risk of cardiovascular disease. These findings require further validation in prospective studies.

ABBREVIATIONS AND ACRONYMS

ASXL1

additional sex combs-like 1

CH

clonal hematopoiesis

DNMT3A

DNA methyltransferase 3A

EM

estimated mean

hsCRP

high-sensitivity C-reactive protein

HSPC

hematopoietic stem and progenitor cell

IL

interleukin

NLRP3

NOD-, LRR-, and pyrin domain-containing protein 3

TET2

ten-eleven translocation methylcytosine dioxygenase 2

VAF

variant allele frequency

APPENDIX

For supplemental figures and tables as well as the statistical analysis plan, protocol, and additional material, please see the online version of this article.