Clonal hematopoiesis of indeterminate potential (CHIP) occurs when a hematopoietic stem cell acquires a somatic driver mutation in a leukemia-associated gene, with a variant allele fraction (VAF) exceeding 2% in peripheral blood.1 Higher VAF is associated with increased morbidity and mortality risk.2-5 Identifying factors influencing clonal expansion rate is crucial for risk stratification in CHIP patients. Recent cost-effective targeted assays have enabled serial sequencing and longitudinal profiling of CHIP dynamics.6 We present the largest longitudinal analysis of CHIP mutational dynamics to-date, examining 3,000 individuals from the Vanderbilt BioVU biobank. Our findings reveal that CHIP growth rates vary significantly by driver gene and are influenced by germline variants and medication exposures. Additionally, we demonstrate that monitoring blood counts may be more informative for risk stratification than frequent resequencing in CHIP patients.
We performed targeted, error-corrected sequencing on serial blood samples from 3,000 individuals in the Vanderbilt BioVU biobank, using custom-designed probes for 22 CHIP-associated genes (Online Supplementary Table S1; Online Supplementary Figure S1). Vanderbilt University Medical Center’s Institutional Review Board oversees BioVU and approved this project (IRB #201783). Unique molecular identifiers (UMI) were used for error correction, excluding mutations detected from a single UMI. The mean coverage depth was 1,725x after de-duplication. CHIP mutations were called for variants with ≥100x total read depth, ≥3 variant allele reads, and >2% VAF in at least one blood draw. We identified 893 CHIP mutations in 711 individuals (Figure 1A). Participants’ mean age at first draw was 70 years (range, 19-96), with a mean 5.7-year interval (range, 0.7-13) between samples. Mean VAF were 6.7% and 9.5% at first and second draws, respectively. Most individuals (79%) had a single CHIP mutation, 16% had two, and 4% had three or more (Figure 1B). DNMT3A and TET2 were the most frequently mutated genes. Of the 711 individuals, 74% had CHIP at both time points, while 26% had >2% VAF at only one draw, predominantly (78%) at the second draw.
We modeled the growth rate r with a compound interest formula -1. SRSF2/SF3B1 (splicing factor) driver mutations exhibited the fastest average growth rate (~25% annually), while DNMT3A driver mutations showed the slowest rate (~7% annually) (Figure 1C). Seventy-eight percent of mutations increased in VAF (annual growth rate >1%), consistent with prior studies.5,7 Decreases in VAF (annual growth rate <-1%) were observed in 30% of JAK2, 21% of ASXL1, 25% of DNMT3A, 19% of PPM1D, 19% of TET2, and 0% of SRSF2/SF3B1 mutations. While certain driver genes were more prone to expansion, CHIP expansion rates varied considerably among individuals with the same driver gene. This variability may necessitate larger sample sizes in prospective randomized controlled trials aiming to identify drugs that slow expansion.
We analyzed individuals with two CHIP mutations, which may co-occur as subclones or in distinct cells (Figure 1D). Mutation pairs were classified as subclones if their growth rates were within 0.6 standard deviations of each other; otherwise, they were considered distinct. Of 116 individuals with two mutations, 66 (57%) pairs were subclones. Subclones showed significantly lower growth rates than distinct clones (β=-0.15, P=0.02) (Figure 1E). This finding was robust for standard deviation thresholds from 0.4 to 0.8. Co-occurrence analysis of CHIP mutations by driver gene (Figure 1F) revealed that JAK2 and SF3B1/SRSF2 co-occurred significantly more often than expected (odds ratio [OR]=48.7, P=0.0006). Two mutations in TET2 (OR=6.6, P=0.001), DNMT3A (OR=3.8, P=0.006), and PPM1D (OR=8.7, P=0.02) also occurred more frequently than chance. Conversely, DNMT3A and TET2 co-occurred less often than expected based on frequency (OR=0.4, P=0.02) (Figure 1G).
We investigated individual determinants of CHIP expansion rate. When accounting for driver gene mutations, growth rate showed no significant association with age (P=0.07), body mass index (P=0.68), biological sex (P=0.89), or smoking status (P=0.70). Our targeting assay included probes for germline variants previously linked to CHIP prevalence and estimated growth rate (Figure 2A).1,8 Each additional G allele in rs1800057 (ATM) correlated with increased growth rate (β=0.46, 95% confidence interval [CI]: 0.23 to 0.67, P<0.001; Figure 2B), potentially explaining its association with CHIP prevalence.1 Consistent with NHLBI’s TOPMed cohort findings,9 each T allele at rs2887399 (in the TCL1A promoter) associated with increased DNMT3A expansion rate (P=0.04) and decreased TET2 expansion rate (P=0.02) (Figure 2C).
We investigated associations between medication exposure duration and CHIP growth rate. While no medications showed significant associations after multiple hypothesis correction, several demonstrated suggestive effects (Figure 2D). Exposure to three drugs suggested reduced annual growth rates: colchicine (β=-3.5%, 95% CI: -0.5 to -7, P=0.03), denosumab (β=-2.5%, 95% CI: -0.3 to -5, P=0.03), and methylprednisolone (β=-1.0%, 95% CI: -0.4 to -2, P=0.01). These associations have biological plausibility: methylprednisolone has anti-inflammatory effects, colchicine may abrogate cardiovascular risk in TET2 CHIP, and denosumab inhibits RANKL, which DNMT3A-mutated monocytes overexpress in bone.10 Conversely, hydroxychloroquine suggested increased annual growth rate (β=3.1%, 95% CI: 0.3-5, P=0.03). Moreover, we performed 3:1 case-control matching by age, sex, CHIP driver gene, baseline VAF, and drug indication for each drug and tested the association with growth rate. In this sensitivity analysis, we still observed an association between exposure duration between sequencing time points of colchicine and methylprednisolone and decreased growth rate. In external validation cohort Clonal Hematopoiesis and Inflammation in the Vasculature (CHIVE)11, four of five mutations exposed to colchicine and eight of ten exposed to methylprednisolone demonstrated reduced or unchanged VAF (Online Supplementary Figure S2). We found no associations between pre-existing diagnoses and CHIP expansion rate (Figure 2E). The lack of association between pre-existing atherosclerosis and CHIP growth rate supports a unidirectional association between clonal hematopoiesis and atherosclerosis.12
Figure 1.Cohort characteristics and co-occurring clonal hematopoiesis of indeterminate potential mutations. (A) Our cohort consists of 892 clonal hematopoiesis of indeterminate potential (CHIP) mutations in 711 individuals. We calculated growth rate using a compound interest formula for sequencing at 2 blood draws. (B) Larger bar plot demonstrates the number of CHIP with a mutation in a driver gene. Smaller bar plot shows the number of individuals with 1, 2, 3, and 4 CHIP mutations. (C) Box plot of growth rate, calculated with a compound interest formula, for each CHIP mutation by driver gene with number of individuals with mutations in the driver gene shown below the gene name. Red diamond represents the mean growth rate. Box represents the interquartile range of the growth rate. Middle line in the box represents the median of the growth rate. (D) Theoretical example of possible trajectories for individuals with two CHIP mutations. The mutations can either be in distinct cell populations or the same cell population. Variant allele fraction (VAF) trajectories for each of these scenarios should follow similar trends to this example. Each mutation is represented by color, and the line represents change in VAF between first and second blood draws. (E) Box plot showing the difference in average growth rate of the fastest growing mutations for the 116 individuals grouped into the distinct versus sub-clone categories. The red diamond represents the average of the growth rate. Box represents the interquartile range of the growth rate. Middle line in the box represents the median of the growth rate. The x-axis is the growth rate while the y-axis is the category. (F) Upset plot showing the different combinations of driver genes represented within the group of 116 individuals with multiple CHIP mutations. Above the upset plot is a bar plot showing the distribution between distinct versus sub-clone trajectories for each of the driver gene combinations, with driver gene on the x-axis and count on the y-axis. (G) Heat map showing the deviation from random for the co-occurrence of each driver gene pair for the 116 individuals with 2 CHIP mutations. Each gene pair was tested with a Fisher’s exact test to identify deviation from expected occurrence via random chance. The x- and y-axes represent each gene in a pair, the color of the box represents the direction of the deviation from random, and the darkness of the hue represents the magnitude of the deviation. For gene combinations with a P value >0.05, the odds ratio is displayed, and the box is marked with “*” and for gene combinations that passed multiple-hypothesis correction with a P value of <0.001, the odds ratio is displayed and the box is marked with “**”.
Figure 2.Determinants of clonal hematopoiesis of indeterminate potential growth rate. (A) Forest plot of change in annual growth rate (%/yr) with each additional copy of the alternate allele (alt) for select germline single nucleotide polymorphisms (SNP) detectable with the targeted sequencing assay among individuals with clonal hematopoiesis of indeterminate potential (CHIP) mutations. Effect estimate represents the coefficient of a linear regression of growth rate and number of alternate alleles modeled additively as 0, 1, or 2, adjusting for age, sex, race, and variant allele fraction (VAF) at first sequencing. We computed the 95% confidence interval (95% CI) as the effect estimate +/- 1.96 standard error. The P value tests the null hypothesis of the effect estimate being 0. (B) Violin plot of %/yr for 0 versus >1 G allele for germline variant chr11:108272729:C:G in ATM. When modeled additively, each additional G allele is associated with greater %/yr in multiple linear regression by 0.46% per year (95% CI: 0.23-0.67, P<0.001). There were 619 CHIP mutations in individuals with 0 G alleles, 26 in individuals with 1 G allele, and 2 in individuals with 2 G alleles. (C) Violin plot of %/yr for 0, 1, and 2 T alleles for germline variant chr14:95714358:G:T in TCL1A for DNMT3A CHIP (left) and TET2 CHIP (right). When modeled additively, each additional T allele is associated with greater %/yr for TET2 CHIP mutations (P=0.03), but not for DNMT3A CHIP mutations (P=0.49). There were 132 TET2 mutations in individuals with 0 T alleles, 54 TET2 mutations in individuals with 1 T allele, and 6 TET2 mutations with 2 T alleles. There were 158 DNMT3A mutations in individuals with 0 T alleles, 95 DNMT3A mutations in individuals with 1 T allele, and 28 DNMT3A mutations in individuals with 2 T alleles. (D) Forest plot of change in annual growth rate (%/yr) with each additional prescription month of 4 selected drugs with suggestive associations: colchicine, denosumab, methylprednisolone, and hydroxychloroquine. Black effect estimate represents the coefficient of a linear regression of growth rate and number of months exposed to the drug adjusting for age, sex, race, and VAF at first sequencing with all data. Gray effect estimate represents the coefficient of a linear regression of growth rate and number of months exposed to the drug for 3:1 matched non-drug-exposed controls to drug-exposed cases (matched by age, sex, CHIP driver gene, VAF at first sequencing). We computed the 95% CI as the effect estimate +/- 1.96 standard error. The P value tests the null hypothesis of the effect estimate being 0. N represents the number of individuals prescribed the drug for at least 1 month. (E) Forest plot of change in %/yr among individuals with select pre-existing diagnoses based on phecodes. Individuals had to have their first diagnosis of the phecode before their second blood draw. Estimate represents the coefficient of a linear regression of growth rate and presence of diagnosis as a binary variable adjusting for age, sex, race, and VAF at first sequencing. We computed the 95% CI as the effect estimate +/- 1.96 standard error. The P value tests the null hypothesis of the effect estimate being 0. N represents the number of individuals with the diagnosis in each regression.
Figure 3.Phenotypic consequences of growth rate. (A) On the left, Kaplan-Meier curve of time to low myeloid counts - defined as thrombocytopenia (platelet count <169.06x109 cells/L), anemia (red blood cell count <3.96x1012 cells/L) or neutropenia (neutrophil count <1.47x109 cells/L) for individuals with a CHIP growth rate >16% annually (red) and <16% annually (gray). A Cox proportional hazard model of time to low myeloid counts and rank-inverse normalized growth rate, when adjusting for age, variant allele fraction (VAF) at the first blood draw, and sex was significant (hazard ratio [HR]=1.20, 95% confidence interval [CI]: 1.05-1.36, P<0.001). On the right, Kaplan-Meier curve of time to high myeloid counts - defined as thrombocytosis (platelet count >397.1x109 cells/L), polycythemia (red blood cell count >5.5x1012 cells/L), or elevated neutrophil count (neutrophil count >7.06x109 cells/L) for individuals with a clonal hematopoiesis of indeterminate potential (CHIP) growth rate >16% annually (red) and <16% annually (gray). A Cox proportional hazard model of time to high myeloid counts and rank-inverse normalized growth rate, when adjusting for age, VAF at the first blood draw, and sex was significant (HR=1.14, 95% CI: 1.02-1.27, P=0.003). (B) Forest plot showing the association between growth rate and time to event for the listed phenotypes. Out of the 711 individuals in the study, individuals who had a phenotype for the first time after the second blood draw are included, with each phenotype sample size listed in the figure. (C) Heatmap showing Clonal Hematopoiesis Risk Score (CHRS) category at the first blood draw (TP1) versus risk at second blood draw (TP2) for N=134 individuals with data available to compute a CHRS (blood counts, mean corpuscular volume, and red cell distribution width). As per Weeks et al.43, low risk was defined as CHRS ≤9.5, intermediate risk as 10< CHRS <12, and high risk as CHRS ≥12.5. Darker hues represent higher numbers of individuals, and exact counts are displayed for each category. (D) Heatmap showing the change in each of the CHRS criteria for the N=30 individuals that shifted into a different risk category between first and second blood draw. The x-axis represents each individual and the y-axis represents the CHRS components, which are faceted based on what clinical test would need to be performed to ascertain that information (i.e., none, blood panel, and sequencing). The color of the box represents the direction of the change (red is positive and blue is negative) and the darkness of the hue represents the magnitude of the change.
We investigated the consequences of CHIP expansion rate on blood counts and diagnoses after the second blood draw. Analyzing associations between growth rate and time to blood cell count abnormalities (as defined in Niroula et al.13), we found that rank-inverse normalized growth rate, adjusted for age, baseline VAF, and sex, was associated with shorter time to high myeloid cell parameters (hazard ratio [HR]=1.14, 95% CI: 1.02-1.27) and low myeloid cell parameters (HR=1.20, 95% CI: 1.05-1.36). No associations were found with lymphocytosis (P=0.48) or lymphopenia (P=0.53). Annual growth rate >16% increased risk of both high and low myeloid cell parameters (P=0.003 and P<0.001, respectively) (Figure 3A). A 1% increase in annual growth rate suggested a 1.6-fold increased risk of myeloproliferative disorders (95% CI: 1.1-2.5, P=0.02), though this association was not significant after multiple-hypothesis correction (Figure 3B). These findings indicate that faster growth rate heightens disease risk in individuals with CHIP.
We next examined changes in clinical risk scores over time, focusing on the Clonal Hematopoiesis Risk Score (CHRS) for 138 individuals with complete blood count data within 6 months of sequencing.14 Of 115 individuals initially classified as low risk (CHRS ≤9.5), 24 (21%) were reclassified as intermediate risk (10< CHRS <12) at the second blood draw (Figure 3C). Among 21 initially intermediate-risk individuals, four (19%) were reclassified as low risk, and one (5%) as high risk (CHRS ≥12.5). Only two individuals were classified as high risk overall. Risk category changes were primarily driven by alterations in blood counts (Figure 3D). Every individual who shifted CHRS category had new alterations in RDW, MCV, or blood counts. Only five (17%) exceeded the 20% VAF threshold, but all five also had blood count changes. These findings suggest that repeated blood sequencing may be unnecessary without accompanying laboratory abnormalities. Our study has notable limitations. Despite being the largest cohort of longitudinal CHIP clonal dynamics to-date, we lack power to detect clinical consequences of CHIP growth rate due to the short follow-up time and low-risk population. The exclusion of individuals with hematologic malignancies further limits our ability to detect associations between CHIP growth rate and these phenotypes. Additionally, associations between medication exposure and growth rate are susceptible to confounding by indication, although we observed no associations between pre-existing diagnoses and growth rate for the medications of interest and colchicine and methylprednisolone had a significant negative association in matched controls.
In conclusion, we present the largest longitudinal study of CHIP dynamics to date. Our findings demonstrate that germline genetics, medications, driver genes, and mutation numbers significantly influence CHIP growth rates, with considerable individual variation. As targeted CHIP panels become more accessible, clinicians may leverage these identified risk factors and CHIP trajectories to better assess clinical risk in individuals with CHIP.
Footnotes
- Received August 22, 2024
- Accepted November 6, 2024
Correspondence
Disclosures
AGB has received honoraria for advisory board membership from, and holds equity in TenSixteen Bio. MRS has received honoraria for advisory board membership or consultancy from Bristol Myers Squibb, CTI, Forma, Geron, GlaxoSmithKline/Sierra Oncology, Karyopharm, Ryvu Therapeutics, and Taiho Pharmaceutical; has received research funding from ALX Oncology, Astex Pharmaceuticals, Incyte Corporation, Takeda, and TG Therapeutics; holds equity in Empath Biosciences, Karyopharm, and Ryvu Therapeutics; and has been reimbursed for travel expenses by Astex. AK has received honoraria for advisory board membership or consultancy from Geron, Sobi, Incyte Corporation, Morposys, Rigel, and Servier Pharmaceuticals. All other authors have no conflicts of interest to disclose.
Funding
References
- Bick AG, Weinstock JS, Nandakumar SK. Inherited causes of clonal haematopoiesis in 97,691 whole genomes. Nature. 2020; 586(7831):763-768. https://doi.org/10.1038/s41586-020-2819-2|Google Scholar
- Jaiswal S, Natarajan P, Silver AJ. Clonal hematopoiesis and risk of atherosclerotic cardiovascular disease. N Engl J Med. 2017; 377(2):111-121. https://doi.org/10.1056/NEJMoa1701719|Google Scholar
- Jaiswal S, Fontanillas P, Flannick J. Age-related clonal hematopoiesis associated with adverse outcomes. N Engl J Med. 2014; 371(26):2488-2498. https://doi.org/10.1056/NEJMoa1408617|Google Scholar
- Vlasschaert C, Mack T, Heimlich JB. A practical approach to curate clonal hematopoiesis of indeterminate potential in human genetic data sets. Blood. 2023; 141(18):2214-2223. https://doi.org/10.1182/blood.2022018825|Google Scholar
- Fabre MA, de Almeida JG, Fiorillo E. The longitudinal dynamics and natural history of clonal haematopoiesis. Nature. 2022; 606(7913):335-342. https://doi.org/10.1038/s41586-022-04785-z|Google Scholar
- Mack T, Vlasschaert C, Von Beck K. Cost-effective and scalable clonal hematopoiesis assay provides insight into clonal dynamics. J Mol Diagn. 2024; 26(7):563-573. https://doi.org/10.1016/j.jmoldx.2024.03.007|Google Scholar
- Uddin MM, Zhou Y, Bick AG. Longitudinal profiling of clonal hematopoiesis provides insight into clonal dynamics. Immun Ageing. 2022; 19(1):23. https://doi.org/10.1186/s12979-022-00278-9|Google Scholar
- Kessler MD, Damask A, O’Keeffe S. Common and rare variant associations with clonal haematopoiesis phenotypes. Nature. 2022; 612(7939):301-309. https://doi.org/10.1038/s41586-022-05448-9|Google Scholar
- Weinstock JS, Gopakumar J, Burugula BB. Aberrant activation of TCL1A promotes stem cell expansion in clonal haematopoiesis. Nature. 2023; 616(7958):755-763. https://doi.org/10.1038/s41586-023-05806-1|Google Scholar
- Wang H, Divaris K, Bohu P. Clonal hematopoiesis driven by mutated DNMT3A promotes inflammatory bone loss. Cell. 2024; 187(14):3690-3711.e19. https://doi.org/10.1016/j.cell.2024.05.003|Google Scholar
- Shannon ML, Heimlich JB, Olson S. Clonal hematopoiesis and inflammation in the VasculaturE (CHIVE): a prospective, longitudinal cohort and biorepository. Blood Adv. 2024; 8(13):3453-3463. https://doi.org/10.1182/bloodadvances.2023011510|Google Scholar
- Díez-Díez M, Ramos-Neble BL, De La Barrera J. Unidirectional association of clonal hematopoiesis with atherosclerosis development. Nat Med. 2024; 30(10):2857-2866. https://doi.org/10.1038/s41591-024-03213-1|Google Scholar
- Niroula A, Sekar A, Murakami MA. Distinction of lymphoid and myeloid clonal hematopoiesis. Nat Med. 2021; 27(11):1921-1927. https://doi.org/10.1038/s41591-021-01521-4|Google Scholar
- Weeks LD, Niroula A, Neuberg D. Prediction of risk for myeloid malignancy in clonal hematopoiesis. NEJM Evid. 2023; 2(5):10. https://doi.org/10.1056/EVIDoa2200310|Google Scholar
Data Supplements
Figures & Tables
Figure 1.Cohort characteristics and co-occurring clonal hematopoiesis of indeterminate potential mutations. (A) Our cohort consists of 892 clonal hematopoiesis of indeterminate potential (CHIP) mutations in 711 individuals. We calculated growth rate using a compound interest formula for sequencing at 2 blood draws. (B) Larger bar plot demonstrates the number of CHIP with a mutation in a driver gene. Smaller bar plot shows the number of individuals with 1, 2, 3, and 4 CHIP mutations. (C) Box plot of growth rate, calculated with a compound interest formula, for each CHIP mutation by driver gene with number of individuals with mutations in the driver gene shown below the gene name. Red diamond represents the mean growth rate. Box represents the interquartile range of the growth rate. Middle line in the box represents the median of the growth rate. (D) Theoretical example of possible trajectories for individuals with two CHIP mutations. The mutations can either be in distinct cell populations or the same cell population. Variant allele fraction (VAF) trajectories for each of these scenarios should follow similar trends to this example. Each mutation is represented by color, and the line represents change in VAF between first and second blood draws. (E) Box plot showing the difference in average growth rate of the fastest growing mutations for the 116 individuals grouped into the distinct versus sub-clone categories. The red diamond represents the average of the growth rate. Box represents the interquartile range of the growth rate. Middle line in the box represents the median of the growth rate. The x-axis is the growth rate while the y-axis is the category. (F) Upset plot showing the different combinations of driver genes represented within the group of 116 individuals with multiple CHIP mutations. Above the upset plot is a bar plot showing the distribution between distinct versus sub-clone trajectories for each of the driver gene combinations, with driver gene on the x-axis and count on the y-axis. (G) Heat map showing the deviation from random for the co-occurrence of each driver gene pair for the 116 individuals with 2 CHIP mutations. Each gene pair was tested with a Fisher’s exact test to identify deviation from expected occurrence via random chance. The x- and y-axes represent each gene in a pair, the color of the box represents the direction of the deviation from random, and the darkness of the hue represents the magnitude of the deviation. For gene combinations with a P value >0.05, the odds ratio is displayed, and the box is marked with “*” and for gene combinations that passed multiple-hypothesis correction with a P value of <0.001, the odds ratio is displayed and the box is marked with “**”.
Figure 2.Determinants of clonal hematopoiesis of indeterminate potential growth rate. (A) Forest plot of change in annual growth rate (%/yr) with each additional copy of the alternate allele (alt) for select germline single nucleotide polymorphisms (SNP) detectable with the targeted sequencing assay among individuals with clonal hematopoiesis of indeterminate potential (CHIP) mutations. Effect estimate represents the coefficient of a linear regression of growth rate and number of alternate alleles modeled additively as 0, 1, or 2, adjusting for age, sex, race, and variant allele fraction (VAF) at first sequencing. We computed the 95% confidence interval (95% CI) as the effect estimate +/- 1.96 standard error. The P value tests the null hypothesis of the effect estimate being 0. (B) Violin plot of %/yr for 0 versus >1 G allele for germline variant chr11:108272729:C:G in ATM. When modeled additively, each additional G allele is associated with greater %/yr in multiple linear regression by 0.46% per year (95% CI: 0.23-0.67, P<0.001). There were 619 CHIP mutations in individuals with 0 G alleles, 26 in individuals with 1 G allele, and 2 in individuals with 2 G alleles. (C) Violin plot of %/yr for 0, 1, and 2 T alleles for germline variant chr14:95714358:G:T in TCL1A for DNMT3A CHIP (left) and TET2 CHIP (right). When modeled additively, each additional T allele is associated with greater %/yr for TET2 CHIP mutations (P=0.03), but not for DNMT3A CHIP mutations (P=0.49). There were 132 TET2 mutations in individuals with 0 T alleles, 54 TET2 mutations in individuals with 1 T allele, and 6 TET2 mutations with 2 T alleles. There were 158 DNMT3A mutations in individuals with 0 T alleles, 95 DNMT3A mutations in individuals with 1 T allele, and 28 DNMT3A mutations in individuals with 2 T alleles. (D) Forest plot of change in annual growth rate (%/yr) with each additional prescription month of 4 selected drugs with suggestive associations: colchicine, denosumab, methylprednisolone, and hydroxychloroquine. Black effect estimate represents the coefficient of a linear regression of growth rate and number of months exposed to the drug adjusting for age, sex, race, and VAF at first sequencing with all data. Gray effect estimate represents the coefficient of a linear regression of growth rate and number of months exposed to the drug for 3:1 matched non-drug-exposed controls to drug-exposed cases (matched by age, sex, CHIP driver gene, VAF at first sequencing). We computed the 95% CI as the effect estimate +/- 1.96 standard error. The P value tests the null hypothesis of the effect estimate being 0. N represents the number of individuals prescribed the drug for at least 1 month. (E) Forest plot of change in %/yr among individuals with select pre-existing diagnoses based on phecodes. Individuals had to have their first diagnosis of the phecode before their second blood draw. Estimate represents the coefficient of a linear regression of growth rate and presence of diagnosis as a binary variable adjusting for age, sex, race, and VAF at first sequencing. We computed the 95% CI as the effect estimate +/- 1.96 standard error. The P value tests the null hypothesis of the effect estimate being 0. N represents the number of individuals with the diagnosis in each regression.
Figure 3.Phenotypic consequences of growth rate. (A) On the left, Kaplan-Meier curve of time to low myeloid counts - defined as thrombocytopenia (platelet count <169.06x109 cells/L), anemia (red blood cell count <3.96x1012 cells/L) or neutropenia (neutrophil count <1.47x109 cells/L) for individuals with a CHIP growth rate >16% annually (red) and <16% annually (gray). A Cox proportional hazard model of time to low myeloid counts and rank-inverse normalized growth rate, when adjusting for age, variant allele fraction (VAF) at the first blood draw, and sex was significant (hazard ratio [HR]=1.20, 95% confidence interval [CI]: 1.05-1.36, P<0.001). On the right, Kaplan-Meier curve of time to high myeloid counts - defined as thrombocytosis (platelet count >397.1x109 cells/L), polycythemia (red blood cell count >5.5x1012 cells/L), or elevated neutrophil count (neutrophil count >7.06x109 cells/L) for individuals with a clonal hematopoiesis of indeterminate potential (CHIP) growth rate >16% annually (red) and <16% annually (gray). A Cox proportional hazard model of time to high myeloid counts and rank-inverse normalized growth rate, when adjusting for age, VAF at the first blood draw, and sex was significant (HR=1.14, 95% CI: 1.02-1.27, P=0.003). (B) Forest plot showing the association between growth rate and time to event for the listed phenotypes. Out of the 711 individuals in the study, individuals who had a phenotype for the first time after the second blood draw are included, with each phenotype sample size listed in the figure. (C) Heatmap showing Clonal Hematopoiesis Risk Score (CHRS) category at the first blood draw (TP1) versus risk at second blood draw (TP2) for N=134 individuals with data available to compute a CHRS (blood counts, mean corpuscular volume, and red cell distribution width). As per Weeks et al.43, low risk was defined as CHRS ≤9.5, intermediate risk as 10< CHRS <12, and high risk as CHRS ≥12.5. Darker hues represent higher numbers of individuals, and exact counts are displayed for each category. (D) Heatmap showing the change in each of the CHRS criteria for the N=30 individuals that shifted into a different risk category between first and second blood draw. The x-axis represents each individual and the y-axis represents the CHRS components, which are faceted based on what clinical test would need to be performed to ascertain that information (i.e., none, blood panel, and sequencing). The color of the box represents the direction of the change (red is positive and blue is negative) and the darkness of the hue represents the magnitude of the change.
Article Information
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.