Pre-illness Clonal Hematopoiesis of Indeterminate Potential is an Independent Predictor of Morbidity and Mortality in Sepsis

Abstract

Rationale:

Sepsis is a life-threatening syndrome causing significant morbidity and mortality especially in the aging population. Clonal hematopoiesis of indeterminate potential (CHIP) is an age-related condition of clonal expansion of hematopoietic stem cells harboring somatic mutations associated with increased incidence of chronic illness and all-cause mortality.

Objective:

Evaluate the association of pre-illness CHIP with mortality and morbidity in patients admitted to the ICU with sepsis.

Methods:

We performed a retrospective study using a de-identified electronic health record linked with a DNA biorepository. We identified adult patients with sepsis who had DNA collected prior to ICU admission. We tested the association between CHIP status, determined from whole-genome sequencing, and ICU mortality, organ support-free days, and long-term survival adjusting for age, sex, race and Sequential Organ Failure Assessment (SOFA) score on ICU admission.

Measurements and Main Results:

Pre-illness CHIP was associated with increased sepsis mortality (OR = 1.54, 95% CI 1.13 to 2.07, P = 0.005) and fewer days alive and free of organ support (−1.7 days, 95% CI −3.2 to −0.2, P = 0.028) after adjusting for age, sex, race, and SOFA score. In sepsis survivors, CHIP was also associated with increased long-term mortality after discharge (HR 1.40, 95% CI 1.01 to 1.93, P = 0.041).

Conclusions:

Pre-illness CHIP was independently associated with increased mortality and morbidity in critically-ill adults with sepsis. These findings suggest that CHIP is a risk factor for sepsis severity. Elucidating the mechanism underlying this association could uncover new therapeutic interventions for sepsis.

Introduction:

Sepsis is a dysregulated host response to infection resulting in organ dysfunction(1). Despite timely implementation of appropriate resuscitative measures and antibiotics, sepsis in critically-ill patients continues to drive substantial morbidity and mortality(2, 3). Decades of research have advanced our understanding of the mechanisms underlying sepsis-induced organ injury and dysfunction, but improvements in mortality have been modest and are largely attributable to earlier recognition and increased adherence to best-practice interventions(4–6). Furthermore, overall sepsis mortality is increasing, a trend likely driven by the rising number of at-risk individuals including the elderly and those with multiple comorbidities(7, 8).

Advanced age is recognized as risk factor for severity of sepsis, in part due to this population’s frailty and burden of comorbid conditions(9). However, emerging evidence suggests that elderly patients also exhibit distinct biological responses to sepsis, including attenuated endothelial cell activation and dysregulated inflammatory signaling(10–12). Identifying the factors that underlie these altered physiological responses in older adults is therefore critical to the development of targeted therapies to reduce the burden of sepsis-related morbidity and mortality in this high-risk population.

Clonal hematopoiesis of indeterminant potential (CHIP) is an age-related phenomenon which occurs when somatic mutations in hematopoietic stem cells confer a clonal enrichment, resulting in detectable expansion of mutant blood cell populations at a variant allele fraction of 2% or greater, in the absence of overt hematologic malignancy or clonal disorder(13, 14). CHIP is associated with increased risk for numerous diseases including cardiovascular disease, autoimmune conditions, chronic kidney disease, malignancy and all-cause mortality(15, 16). The prevalence of CHIP increases exponentially with age with prevalence of around 5% in patients 60 years of age, but 30% in individuals 80 years or older(17–19). A growing body of evidence from animals and humans suggests that CHIP may contribute to disease risk via dysregulated expression of inflammatory cytokines, impaired neutrophil function, and enhanced monocyte-endothelial interactions(20–26), all mechanisms that could be important in the inflammatory response in sepsis.

In this study, we examined the relationship between CHIP and sepsis outcomes using a de-identified longitudinal electronic health record database linked to a DNA bio-repository. Given the association between CHIP and aberrant inflammatory responses, we hypothesized that patients with CHIP would experience heightened risk of adverse outcomes in sepsis. Additionally, we tested whether presence of CHIP correlated with altered blood-cell counts during sepsis, if its presence in sepsis survivors is associated with higher mortality, and whether sepsis accelerates the growth of CHIP mutations in sepsis survivors.

Methods:

Study Population

We studied adult patients enrolled in BioVU, Vanderbilt’s DNA biobank linked to a de-identified longitudinal electronic health record (EHR). All samples were generated from excess blood collected during routine medical care(27). We identified patients with sepsis using either of two methods: 1) by applying a validated EHR algorithm (28) and 2) identifying the presence of an explicit ICD-9-CM and ICD-10-CM code in patients admitted to the ICU(29, 30) (codes are available in online supplement Supplemental Table E1). Whole genome sequencing for over 250,000 BioVU participants has previously been performed. Library preparation and sequencing methods are previously described(31). Presence of CHIP was classified as having a mutation in a CHIP driver gene with variant allele fraction (VAF) of 2% or more, consistent with established methods(32).

We selected patients who had DNA collected between two days and two years prior to ICU admission. This timeframe was implemented to ensure temporal proximity between DNA isolation and the ICU admission. A two-day lead-in period was used to ensure temporal precedence between CHIP status and ICU outcomes. Patients who had a concurrent or pre-existing diagnosis of myeloid neoplasm based on ICD-9-CM or ICD-10-CM codes were excluded. We also excluded patients who were transferred out of ICU within 24 hours of admission unless they died in the first 24 hours.

ICD-9-CM and ICD-10-CM codes during each hospital admission were used to generate Elixhauser comorbidity scores(33). Vital signs and clinical laboratory data including maximum creatinine, minimum platelets, and maximum bilirubin from the first 24 hours of ICU admission were used to calculate sequential organ failure assessment (SOFA) scores(34).

Outcomes

The primary outcome was in-hospital mortality, defined as death during the sepsis hospitalization or death occurring within two weeks of discharge to hospice to account for patients who were transitioned to an inpatient palliative care unit or to “hospice-in-place,” which is designated as a transfer to a separate hospice facility in our EHR.

We evaluated whether pre-illness CHIP was associated with worsened organ dysfunction characterized by fewer organ support-free days. Calculation of organ support-free days was adapted from published methods(35), defined as the number of days a patient remained alive without requiring organ support through day 28 of admission. Days requiring vasopressors, mechanical ventilation and dialysis were determined using medication records, ICD-10-PCS codes, and CPT procedure codes during each hospital stay (Supplemental Tables E3 and E4). Dialysis days were specifically calculated as the difference between the first and last date of dialysis events to reflect that some patients received intermittent dialysis and not continuous renal replacement therapy, and all dialysis analyses excluded patients with chronic renal failure requiring maintenance dialysis (ICD-9-CM and ICD-10-CM codes used listed in Supplemental Table E5).

Assessment of hematologic profiles

As CHIP results in the preferential expansion of myeloid lineages(36), we characterized its influence on the peripheral leukocyte profiles of patients with sepsis. We extracted peripheral leukocyte counts and differential data collected during routine clinical care from hospital admission through the first five days in the ICU. For baseline leukocyte assessments, we analyzed the laboratory values closest to the DNA collection date, ensuring it was recorded prior to admission and no more than 60 days away from the DNA collection.

Long-term survival

We assessed long-term survival among sepsis survivors using EHR-recorded death dates. Observation intervals for censored patients were defined as the time from ICU admission to the last documented EHR encounter.

Longitudinal CHIP clonal dynamics

Prior studies suggest that IL-6 signaling drives the growth of CHIP variant alleles(37). Consequently, we hypothesized that the inflammation driven by sepsis may act as a catalyst for CHIP VAF expansion. To investigate this, we leveraged previously published longitudinal CHIP data from a subset of 3,000 individuals in BioVU with DNA samples collected at two points in time and sequenced with a deep, targeted sequencing assay(38, 39). As above, we selected individuals whose first blood draw occurred between two years and one week prior to an ICU admission for sepsis; patients also had to have a follow-up sample that was collected between two weeks and one-year post-admission. These intervals were chosen to more specifically capture CHIP expansion driven by sepsis-related inflammation rather than age-related expansion. Control subjects were patients in BioVU who had not been admitted to the ICU with sepsis and did not have ICD codes for leukemia. We performed case-control matching using the R package MatchIt. Control patients were matched (in a 2:1 ratio) first by exact CHIP gene and mutation type (e.g. missense, frameshift, nonsense, etc.) followed by nearest-neighbor matching for sex, age at first DNA collection, initial VAF, and the time interval between samples. CHIP growth rates were calculated based on a exponential growth formula as previously described(38).

Statistical Analysis

Statistical analyses were performed using R (version 4.5.1; R Core Team, 2025). Continuous variables, including SOFA scores, were compared using the Wilcoxon rank-sum test and reported as medians with interquartile ranges, while categorical variables were assessed using Chi-squared or Fisher’s exact test. Differences in Elixhauser comorbidities were evaluated through age-adjusted binary logistic regression. The association between pre-illness CHIP and mortality was analyzed using logistic regression, adjusting for sex, race, initial SOFA score, and age at time of admission. We also tested for interaction effects between CHIP status and initial SOFA score and age at time of admission on mortality. Organ support-free days (mean ± SD) were compared via age- and sex-adjusted linear regression. Long-term survival probabilities at one-, two-, and five-year timepoints were compared using a Z-score test and a hazard ratio was determined by Cox regression using the Survival R package, with age at time of admission and sex included as covariates.

Results:

Patient Characteristics

A total of 10,731 ICU patients with sepsis had available whole-genome sequencing. Among these, there were 3,230 patients who had their DNA collected between 2 years and 2 days prior to their ICU admission, did not have leukemia, and had ICU lengths of stay greater than 24 hours or died within the first day. Among them, 220 (6.8%) patients had CHIP mutations (Figure 1). As expected, CHIP prevalence increased exponentially with age, with 5% CHIP prevalence at age 60 years, and 30% prevalence at age 90 years (Supplemental Figure E1).

Figure 1.

Study flow-diagram.

There were no significant differences between sex or admitting ICU between patients with and without CHIP (Table 1). Expectedly, patients with CHIP were older than those without. There was a trend towards increased prevalence of CHIP in white compared to black patients which was likely explained, in part, by older age (median: 61 vs. 53 years; p<0.001). CHIP patients also had a small, but statistically significant lower body mass index (BMI). There was no significant difference between baseline indicators of respiratory function based on CHIP status as suggested by maximum respiratory rate, minimum SpO2, minimum paO2:FiO2 ratio, and need for mechanical ventilation. Similarly, there were no significant differences in baseline maximum bilirubin, minimum platelets, and maximum creatinine levels. Maximum pulse rate was similar between the groups. While there was a statistically significant difference in the median lowest recorded systolic blood pressure in patients with CHIP versus those without, there was no significant difference in need for vasopressor support. Overall, there were no differences in initial SOFA scores between patients with and without CHIP. After adjusting for age, there were also no major differences in Elixhauser comorbidities at the time of hospital admission (Table 2).

Table 1.

ICU sepsis cohort patient characteristics

	No CHIP N =3,0101	CHIP N = 2201	p-value2	q-value3
Gender			0.12	0.3
Female	1,341 (44.6%)	86 (39.1%)
Age at Admission	59.2 (47.8, 67.7)	70.2 (61.9, 77.2)	<0.001	<0.001
Admitting ICU			0.4	0.7
Cardiovascular	518 (17.2%)	32 (14.5%)
Medical	1,300 (43.2%)	107 (48.6%)
Other	472 (15.7%)	32 (14.5%)
Surgical	720 (23.9%)	49 (22.3%)
Race			0.065	0.3
Black	473 (15.7%)	22 (10.0%)
Other	110 (3.7%)	7 (3.2%)
White	2,427 (80.6%)	191 (86.8%)
Body Mass Index	27.9 (23.7, 33.0)	26.5 (23.3, 30.6)	0.006	0.041
Baseline Measures
Maximum Bilirubin, mg/dL	1.0 (0.6, 2.0)	1.1 (0.6, 1.8)	0.7	0.7
Maximum Creatinine, mg/dL	1.3 (0.9, 2.1)	1.3 (0.9, 2.3)	0.066	0.3
Minimum Platelets, 103/μL	165 (99, 243)	157 (98, 222)	0.4	0.7
Minimum Systolic Blood Pressure, mmHg	89 (79, 99)	85.0 (77, 96)	0.006	0.041
Maximum Respiratory Rate, per minute	27 (23, 31)	26 (22, 31)	0.3	0.7
Minimum SpO2,%	92 (90, 95)	92 (90, 94)	0.4	0.7
Minimum PaO2:FiO2 Ratio	202 (128, 312)	227 (119, 329)	0.7	0.7
Maximum Pulse Rate, per minute	116 (101, 132)	114 (100, 132)	0.5	0.7
Vasopressors at Baseline	1,107 (36.8%)	83 (37.7%)	0.8	0.8
Mechanical Ventilation at Baseline	1,498 (49.8%)	113 (51.4%)	0.6	0.7
SOFA Organ System Scores
Respiratory	2.0 (1.0, 3.0)	3.0 (1.0, 3.0)	0.6	0.7
Renal	1.0 (0.0, 2.0)	1.0 (0.0, 2.0)	0.11	0.3
Hepatic	0.0 (0.0, 2.0)	0.0 (0.0, 1.0)	0.5	0.7
Coagulation	0.0 (0.0, 2.0)	0.0 (0.0, 2.0)	0.4	0.7
Cardiovascular	1.0 (1.0, 3.0)	1.0 (1.0, 3.0)	0.2	0.4
Total SOFA Score	6.0 (4.0, 8.0)	6.0 (4.0, 8.0)	0.2	0.4

¹n (%); Median (Q1, Q3)

²Pearson’s Chi-squared test; Wilcoxon rank sum test

³False discovery rate correction for multiple testing

Table 2.

Patient cohort Elixhauser comorbidities.

	No CHIP N = 3,0101	CHIP N = 2201	Age-Adjusted P2
Comorbidities
Congestive Heart Failure	842 (28%)	72 (33%)	0.69
Arrhythmia	2,302 (77%)	182 (83%)	0.25
Valvular Disease	631 (21%)	48 (22%)	0.42
Pulm. Circulatory Disorders	505 (17%)	31 (14%)	0.39
Peripheral Vascular Disease	486 (16%)	59 (27%)	0.1
Uncomplicated Hypertension	1,571 (52%)	122 (55%)	0.13
Complicated Hypertension	458 (15%)	37 (17%)	0.66
Paralysis	174 (5.8%)	10 (4.5%)	0.83
Other Neurological Disorders	688 (23%)	56 (25%)	0.47
Chronic Pulm. Disease	947 (31%)	82 (37%)	0.71
Uncomplicated Diabetes	720 (24%)	73 (33%)	0.09
Complicated Diabetes	437 (15%)	25 (11%)	0.05
Hypothyroidism	472 (16%)	39 (18%)	0.5
Renal Failure	812 (27%)	65 (30%)	0.82
Liver Disease	753 (25%)	64 (29%)	0.06
Peptic Ulcer Disease	63 (2.1%)	9 (4.1%)	0.05
AIDS/HIV	52 (1.7%)	2 (0.9%)	0.82
Lymphoma	78 (2.6%)	4 (1.8%)	0.35
Metastatic Cancer	309 (10%)	15 (6.8%)	0.07
Solid Tumor	934 (31%)	85 (39%)	0.37
Rheumatoid Arthritis	197 (6.6%)	15 (6.8%)	0.95
Coagulopathy	1,077 (36%)	83 (38%)	0.61
Obesity	513 (17%)	24 (11%)	0.06
Weight Loss	1,034 (34%)	84 (38%)	0.48
Fluid/Electrolyte Disorder	2,417 (80%)	179 (81%)	0.88
Blood Loss Anemia	135 (4.5%)	13 (5.9%)	0.13
Deficiency Anemia	795 (26%)	60 (27%)	0.42
Alcohol Abuse	639 (21%)	44 (20%)	0.84
Drug Abuse	152 (5.1%)	7 (3.2%)	0.97
Psychoses	121 (4.0%)	10 (4.5%)	0.29
Depression	625 (21%)	44 (20%)	0.49
Elixhauser Comorbidity Score	20 (13, 28)	22 (15, 30)	0.26
Elixhauser Number of Comorbidities	7 (5, 9)	7 (6, 9)	0.43

¹Median (Q1, Q3); n (%)

²Binary Logistic or Linear regression

Association of pre-illness CHIP with in-hospital mortality

Pre-illness CHIP was associated with increased in-hospital mortality (OR 1.54, 95% CI: 1.13 – 2.07, P = 0.005, Figure 2) after adjusting for age at the time of admission, sex, race, and initial SOFA score. There was no interaction between CHIP status and age at the time of admission (P> 0.9), suggesting the association of CHIP with mortality was consistent across all ages (Table 3). Furthermore, there was also no interaction between CHIP status and initial SOFA scores on sepsis mortality (P > 0.9, Table 3).

Figure 2.

Inpatient mortality in sepsis patients admitted to the intensive care unit. Predicted mortality by logistic regression for patients with and without clonal hematopoiesis of indeterminant potential (CHIP) after adjusting for sex, race, age at time of admission, and baseline sequential organ failure assessment (SOFA) scores. Shading represents 95% confidence intervals.

Table 3.

Logistic regression model results for odds of inpatient mortality in critically-ill sepsis patients.

	OR	95% CI	P-value1
CHIP Status
No CHIP	—	—
CHIP	1.54	1.13, 2.07	0.005
Gender
Female	—	—
Male	0.97	0.81, 1.16	0.7
Total SOFA Score	1.21	1.18, 1.24	<0.001
Age at Admission	1.01	1.01, 1.02	<0.001
Race
Black	—	—
Other	1.44	0.86, 2.36	0.2
White	1.08	0.84, 1.41	0.5
CHIP Status * Age at Admission	1.00	0.98, 1.03	>0.9
CHIP Status * Total SOFA Score	1.01	0.91, 1.13	0.9

OR = Odds Ratio, CI = Confidence Interval

¹Wald’s t-test

Previous studies have shown that in patients with CHIP, variant allele fraction (VAF) exerts a dose dependent effect on disease development. For example, a VAF of 10% or more was associated with the development of hematological malignancies(18) and coronary artery disease(40), but a VAF of less than 10% does not confer the same risk. While we observed a wide distribution of VAF in pre-illness CHIP in our cohort, VAF did not associate with sepsis mortality (Supplemental Figure E2A and B).

We also investigated the association between specific CHIP genes and hospital mortality. The distribution of specific mutated CHIP genes in our cohort was consistent with prior studies, with DNMT3A, TET2, and ASXL1 comprising the three most frequently observed mutated genes (Supplemental Figure E2C)(17–19). Due to low mutation frequencies for several CHIP genes in our cohort, we limited the mortality analysis to CHIP genes present in at least 10 patients. No significant differences in mortality were observed across the various CHIP genes (Supplemental Figure E2D).

Pre-illness CHIP is associated with fewer organ support-free days

After adjusting for age, sex, race, and initial SOFA score, patients with pre-illness CHIP had fewer organ support-free days compared to those without CHIP (1.7 fewer days, 95% CI: 3.2 – 0.3 fewer days, P = 0.018, Figure 3). This reduction included fewer days free from assisted ventilation (1.8 fewer days, 95% CI: 3.2 – 0.28 fewer days, P = 0.02), vasopressors (1.7 fewer days, 95% CI: 3.1 – 0.3 fewer days, P = 0.017), and renal replacement therapy (3 fewer days, 95% CI: 5.3 – 0.74 fewer days, P = 0.009, Figure 3).

Figure 3.

Clonal hematopoiesis of indeterminant potential (CHIP) is associated with fewer organ support-free days. Organ support-free days were calculated as days free of vasopressors, assisted ventilation, or dialysis. Dialysis free days were only calculated for patients with acute renal failure and not already undergoing dialysis for chronic kidney disease. The data are presented as mean days and standard deviation (SD). Forest plot displays the mean difference in days (CHIP versus no CHIP). Beta values are adjusted for age and sex based on linear regression with 95% confidence intervals (CI). P values are based on Wald’s t-test.

Pre-illness CHIP is associated with altered peripheral hematologic profiles

The cell count data were adjusted for age at time of admission, time from ICU admission, sex, race, and initial SOFA score. Total leukocyte counts were significantly higher at all time points in patients with CHIP when compared to those without CHIP (mean increase of 2.8 × 10³ cells/uL, 95% CI: 1.5 – 4.0 × 10³ cells/uL, P < 0.001, Figure 4A). This result appeared to be largely related to increases in absolute neutrophil counts (mean increase of 2.8 × 10³ cells/uL in patients with CHIP, 95% CI: 1.7 – 3.9 × 10³ cells/uL, P < 0.001), with an associated trend towards increased neutrophil percentages (mean increase of 2% in patients with CHIP, 95% CI: 0% – 4%, P = 0.082, Figure 4B, C). While there was no measured difference in absolute monocyte counts (Figure 4D), patients with CHIP had nominally lower monocyte percentages (mean decrease of 1%, 95% CI −2% – 0%, P = 0.019, Figure 4E). Patients with CHIP also had lower absolute lymphocyte counts (mean decrease of 0.16 × 10³ cells/uL, 95% CI:−0.31 – −0.01 × 10³ cells/uL, P = 0.038), without a difference in lymphocyte percentages (Figure 4F, G). There were no differences in eosinophil counts or percentages (Figure 4H, I).

Figure 4.

Association of clonal hematopoiesis of indeterminant potential (CHIP) with cell-count differentials. Regression lines from peripheral leukocyte count and differentials collected during routine care between 2 days prior to ICU admission and 5 days after admission are plotted for A) total white blood cell count, B, C) neutrophil count and percentage, D, E) monocyte count and percentage, F, G) lymphocyte count and percentage, and H, I) eosinophil count and percentage. Shading represents 95% confidence interval. P values represent results of linear regression Wald t-test after adjusting for time of blood collection relative to time of ICU admission, age at time of admission, race, initial SOFA score, and sex.

To assess if these differences in hematologic profiles during sepsis preceded admission to the ICU, we also analyzed peripheral leukocyte count and differential data in 2,105 patients (149 [7.1%] with CHIP) who had outpatient labs performed within 60 days of DNA collection for CHIP sequencing. Overall, there were no differences in any cell counts between patients with and without CHIP at the time that their DNA was isolated (Supplemental Figure E3).

Sepsis survivors with CHIP have reduced long-term survival

Among 2,577 patients (156 [6%] with CHIP) who survived their initial critical illness event, CHIP was associated with decreased one-year (55% vs. 75%, P < 0.001), two-year (46% vs. 69%, P = 0.049), and five-year survival (39% vs. 60%, P = 0.037, Table 4). This association remained significant after adjusting for age at admission, sex, race and baseline SOFA score (HR 1.40, 95% CI: 1.01 – 1.93, P = 0.041; Figure 5).

Table 4.

Survival probability in sepsis survivors with and without clonal hematopoiesis of indeterminant potential (CHIP).

Survival at	Survival Probability (95% CI)		P value 1
	CHIP	No CHIP
1 year	0.55 (0.45, 0.70)	0.75 (0.72, 0.77)	<0.001
2 years	0.46 (0.35, 0.60)	0.69 (0.66, 0.72)	0.049
5 years	0.39 (0.28, 0.54)	0.60 (0.57, 0.63)	0.037

¹Z-test

Figure 5.

Clonal hematopoiesis of indeterminant potential (CHIP) in sepsis survivors is associated with increased long-term mortality. The survival curve represents 156 patients with CHIP and 2421 without. The hazard ratio (HR) was adjusted for age at time of admission, sex, race, and initial sequential organ failure assessment (SOFA) score using Cox regression analysis.

Sepsis accelerates the growth of CHIP allele fractions

We identified 11 BioVU patients who had CHIP sequencing performed between 2 years and 1 week before their ICU sepsis admission and subsequently had a second DNA sample sequenced between 2 weeks and 1 year after their admission. These patients were compared to BioVU participants with longitudinal CHIP sequencing who were never hospitalized between CHIP measurements (Figure 6A). After matching for sex, age at first sample, race, initial VAF, gene mutation, mutation type, and time interval between the two samples (Supplemental Table E6), we observed an accelerated rate of growth in CHIP VAF among sepsis survivors (+9% per year, 95% CI: 1% - 17%, P = 0.024, Figure 6B). To account for the possibility that changes in myeloid and granulocyte counts after sepsis drove the observed CHIP VAF growth, we also analyzed cell count and differentials obtained from the times of first and second DNA collection and found no significant differences between control and sepsis patients (Supplemental Table E7). Thus, the VAF expansion was not fully explained by changes in cell count and differentials.

Figure 6.

Clonal hematopoiesis of indeterminant potential (CHIP) variant allele fraction (VAF) expansion accelerates in sepsis patients who survive an ICU admission. A) Diagram of patient selection. Patients were selected who had two DNA samples collected; one that was collected between two years and two days prior to their admission and a second sample collected between two weeks and one year after their admission. Controls were selected in a case-control fashion, first enforcing for exact CHIP gene and mutation type, then matching for sex, race, time interval between the two samples, patient age at time of first sample draw, and allele fraction of the first sample. B) Boxplots of the growth rates for control and sepsis patients. Growth rates are calculated based on a compound interest function between two samples. P value based on linear regression Wald’s t-test.

Discussion:

To our knowledge, this study is the first to establish a significant association between CHIP and outcomes in critically-ill sepsis patients. In adult patients with sepsis, CHIP was independently associated with prolonged organ support and increased in-hospital mortality across all age groups and baseline SOFA scores. Furthermore, the impact of CHIP extended beyond the period of acute illness, as sepsis survivors with CHIP had worse long-term survival and an acceleration in CHIP clonal expansion. Collectively, these data underscore the importance of CHIP as a marker of a particularly vulnerable patient population in both the ICU and post-ICU care settings.

Our results demonstrate that there is an altered circulating white blood cell profile in sepsis patients who have CHIP. Specifically, we observed elevated absolute neutrophil counts and reduced absolute lymphocyte counts. These findings are consistent with existing literature establishing that neutrophilia and lymphopenia during sepsis are associated with increased organ dysfunction and death(41, 42). Neutrophils act as primary effectors in septic acute inflammation, where an exacerbated response drives tissue destruction, subsequent organ injury, and coagulopathy(43). Furthermore, increased granulopoiesis during sepsis in humans is associated with immunosuppression, more severe disease, and early mortality(44). While our findings do not provide direct evidence for dysregulated function of white blood cells in patients with CHIP, they indicate patients with CHIP have altered hematologic profiles during sepsis.

Beyond its known association with all-cause mortality(15), CHIP is linked to a wide spectrum of non-malignant chronic conditions, including cardiovascular disease(40), chronic obstructive pulmonary disease(45), and chronic kidney disease. Emerging evidence also links CHIP to acute illnesses such as acute kidney injury(46) and increased supplemental oxygen requirements in SARS-CoV-2 pneumonia(47). Our findings, which link CHIP to worse outcomes and prolonged organ dysfunction during sepsis, both align with and expand upon these previous findings. Furthermore, while retrospective data indicates that sepsis survivors generally face increased one-year mortality(48), our results identify CHIP as a distinct driver that exacerbates this long-term risk.

While previous studies established a variant allele fraction (VAF) of greater than 10% as the threshold for increased risk of cardiovascular disease and hematologic cancer(18, 40), VAF was not associated with hospital mortality in this cohort. As CHIP VAF was quantified prior to sepsis onset, the lack of VAF dependence could be due to unobserved changes in VAF between DNA sampling and ICU admission. Another possibility is that an acute clonal expansion, driven by the systemic inflammation and stress hematopoiesis of sepsis, is more relevant than the baseline VAF(49–52) However, confirming this hypothesis would require serial DNA collection and CHIP VAF measurements throughout the course of sepsis. Previous studies have indicated that the long-term growth rates of CHIP variant alleles are slower in individuals carrying IL6R mutations. Specifically, the expansion rate of TET2 CHIP variants shows an inverse relationship with serum levels of soluble IL-6 receptor (sIL-6R) and a direct correlation with serum levels of IL-6(37). Consistent with this close association with IL-6 signaling, our findings indicate that sepsis may accelerate the expansion of CHIP clones, potentially heightening the risk of known CHIP-associated diagnoses and overall all-cause mortality(15, 17, 18, 40). However, our small sample size for this analysis of CHIP expansion limits this conclusion and additional longitudinal analyses will be required to validate this association.

This study has several strengths. A primary strength is the utilization of a large and well-established institutional EHR biobank with over 20 years of longitudinal data, providing us the statistical power necessary to study a relatively uncommon genetic exposure(53). An additional strength is that we were able to precisely correlate pre-illness CHIP status with sepsis outcomes. Unlike most critical-care genetic studies that rely on DNA collected at the time of or during ICU admission, the longitudinal data and biospecimens in BioVU allowed assessment of pre-illness risk. Another strength is the use of validated EHR algorithms that have higher sensitivity than claims data alone to identify patients with sepsis (28).

There are also some limitations. First, data collection from a single quaternary referral center inherently enriches the patient population for sicker individuals, which limits the generalizability of our results to broader populations. Second, since we relied on institutional EHR data, information on prior hospitalizations at other centers not linked to our EHR is unavailable. A third limitation is that while the DNA biospecimens used were not collected during the acute sepsis encounter, they could have been obtained during other periods of illness. While BioVU generally collects DNA from samples obtained during outpatient care encounters, it is not fully discernable if patients were at their true baseline health at the time of collection. Another limitation is that some patients could have developed CHIP in the time between their pre-illness DNA biospecimen and their ICU admission, which could result in misclassification bias if some of the non-CHIP patients did in fact have CHIP by the time of their ICU admission. Finally, assessment of CHIP in this cohort was based on approximately 35X sequencing depth rather than the “gold standard” 1000X coverage of deep sequencing (longitudinal CHIP sequencing data), which could lead to an underestimation of CHIP and effect sizes(54).

In conclusion, pre-illness CHIP in critically ill adults with sepsis was associated with higher mortality, fewer organ support-free days, and worse long-term survival. Furthermore, CHIP was associated with altered hematologic profiles during sepsis, suggesting a maladaptive inflammatory response, and sepsis was associated with increased expansion of mutant CHIP clones in longitudinal follow up. As genomic data become more integrated into clinical care, detecting CHIP could improve risk stratification and long-term management for critically ill patients.

Impact of this work:

This study identifies pre-existing clonal hematopoiesis of indeterminate potential (CHIP) as a significant, independent risk factor for increased mortality and morbidity in ICU patients with sepsis, showing that affected individuals face higher risks regardless of age or initial severity of illness. Two possible mechanisms underlying the poorer sepsis outcomes are fewer organ support-free days and elevated long-term post-discharge mortality.

Data availability

This study used data from the Alliance for Genomic Discovery (AGD), a partnership between Nashville Biosciences, Illumina, and Vanderbilt University Medical Center (VUMC). Data access is subject to restrictions and are made available to commercial researchers via NashBio and academic researchers via BioVU.