Abundance and balance of circulating leukocyte subsets and colorectal cancer survival Authors: Alicia R. Richards, Maria F. Gomez, Bianca I. Dowling, Catherine M. Bulka, Biljana Gigic, Jane C. Figueiredo, Christopher I. Li, David Shibata, Adetunji T. Toriola, Doratha A. Byrd, Cornelia M. Ulrich, Mingxiang Teng, Paul A. Stewart, Erin M. Siegel, Jacob K. Kresovich Published: 2026 ABSTRACT BackgroundCancer immunology research has traditionally focused on tumor-infiltrating leukocytes, but the role of the peripheral leukocytes remains understudied. Since tumor-infiltrating leukocytes are recruited from the blood, circulating immune profiles may provide prognostic insights obtainable before surgery, reflecting overall immune competence.MethodsWe analyzed treatment-naïve blood samples from 134 stage I-III colorectal cancer (CRC) patients using a nested case-control design (33 recurrences, 45 deaths; median follow-up: 7.4 years). Circulating leukocyte subsets were estimated using methylation cytometry applied to genome-wide methylation profiles. Cox regression models estimated associations between leukocyte metrics and disease-free and overall survival, adjusting for clinical factors.ResultsHigher counts of circulating neutrophils and T-regulatory cells were associated with worse disease-free and overall survival (neutrophil, disease-free HR: 1.32, 95% CI: 1.01, 1.72, P = 0.04; overall HR: 1.39, 95% CI: 1.00, 1.94, P = 0.05; T regulatory, disease-free HR: 1.32, 95% CI: 1.00, 1.74, P = 0.05; overall HR: 1.38, 95% CI: 1.00, 1.90, P = 0.05). Memory B cells were associated with worse disease-free survival (HR: 1.56, 95% CI: 1.31, 1.86, P < 0.0001), particularly in those diagnosed with rectal cancer. Higher basophil counts and proportions were associated with worse overall survival (count HR: 1.44, 95% CI: 1.05, 1.99, P = 0.02; proportion HR: 1.47, 95% CI: 1.06, 2.03, P = 0.02), with stronger associations in those diagnosed at earlier stages or with rectal tumors.ConclusionsPeripheral immune cell composition identifies CRC patients at higher risk for recurrence and death, providing insights into systemic immune contributions to CRC survival. Introduction Colorectal cancer (CRC) is a leading cause of cancer mortality in the United States [1]. Tumor-infiltrating lymphocytes, including B and T cell subsets, significantly influence CRC prognosis [2, 3]. For example, the ‘immunoscore,’ which quantifies the density of total (CD3+) T cell and cytotoxic (CD8+) T cell tumor infiltrates, has been validated as a clinical tool for colon cancer in multiple cohort studies [4]. However, the tumor microenvironment represents only one component of anticancer immunity. The peripheral immune system represents a complementary but understudied aspect of anticancer immunity. Since most tumor-infiltrating leukocytes are drawn from circulating populations produced by bone marrow and activated, expanded, and modulated in lymph nodes and tertiary lymphoid structures, blood-based immune profiling may provide insights into patient outcomes that reflect systemic immune competence. Critically, circulating immune profiles can be obtained before tumor resection, providing potentially important prognostic information before any therapeutic intervention is initiated. Immune responses to tumor development may drive alterations in peripheral leukocyte composition in ways that influence future outcomes. For example, elevated levels of granulocytes, specifically neutrophils, have been consistently associated with worse survival across cancer types [5]. Studies of CRC patients and healthy controls have reported several immune alterations of lymphocytes, including increased memory B cell percentages and lower CD8+ and CD4+ T cell counts [6–8], but the prognostic importance of these systemic changes remains unclear. While the peripheral immune system may reflect the body’s overall competence and response to tumor development, its relationship with CRC outcomes has not been examined in detail. Immune cell profiling approaches have traditionally relied on fresh blood samples and flow cytometry, which restricts their use in cohort studies of cancer survival. These studies typically bank frozen blood or DNA samples for future analysis once patient endpoints have been accumulated. DNA methylation (DNAm)-based deconvolution, also known as ‘methylation cytometry,’ is an attractive method for indirectly estimating circulating immune cell homeostasis that overcomes the limitations of traditional cell-sorting techniques [9–11]. Methylation cytometry can be applied to archived biospecimens and leverages cell-type-specific patterns of DNAm to provide proportion estimates of numerous circulating leukocyte subsets, which can be summed into higher-order subsets. These proportion estimates can be further multiplied by total leukocyte counts to calculate the absolute abundances of specific circulating subsets. Here, we profiled genome-wide DNAm profiles from archived, treatment-naïve blood samples from a retrospectively sampled nested case-control study of newly diagnosed CRC patients. By sampling a nested case-control study from the parent cohort, we were able to enrich for survival endpoints, thereby maximizing statistical power to test associations with CRC outcomes. Our approach demonstrates new opportunities for combining innovative immune profiling techniques with established molecular epidemiologic designs to understand how peripheral immune cell composition contributes to CRC outcomes. Materials and methods Study population The ColoCare Study is a prospective study of adults aged 18 years and older who have been newly diagnosed with primary CRC (Clinicaltrials.gov Identifier: NCT02328677). To be eligible for enrollment, patients had to be diagnosed with invasive CRC (stage I–IV) and have no prior history of invasive cancer. For this analysis, only stage I–III patients who received care at Moffitt Cancer Center and had an available treatment-naïve (collected prior to surgery, chemotherapy, or radiation) blood sample stored in the institutional biobank were included. Patient demographic and clinical data at diagnosis were self-reported or abstracted from medical records and included information on age and year at diagnosis (years), body mass index (BMI, kg/m2), sex (male, female), race (White, non-White), stage at diagnosis (I–III), tumor microsatellite status (stable, unstable, unknown), tumor site (colon, rectal). Total leukocyte count (K cells/µL) was measured at the same time as the pre-treatment blood draw via automated counter and was abstracted from medical records. Recurrence and vital status were ascertained through September 2024 using electronic medical records and Moffitt’s Cancer Registry, which links to the National Death Index. In February 2024, a nested case-control set of patients was retrospectively sampled from the eligible Moffitt ColoCare cohort, with frequency matching of recurrence cases to controls at approximately 1:2, based on age, year of diagnosis, sex, race, BMI group, tumor stage, and tumor site. Two patients were missing total leukocyte count and were excluded from downstream analyses, resulting in an analytic sample of 134 patients (Supplementary Fig. 1). All participants provided written informed consent, and the study protocol was approved by the Advarra Institutional Review Board (MCC 16028). Whole blood DNA methylation assessment and quality control Genomic DNA from whole blood (n = 16; 12%) or buffy coat (n = 118; 88%) was extracted using the Qiagen DNA extraction kit at Moffitt’s Tissue Core. The quality and quantity of extracted DNA were evaluated using TapeStation, and one microgram of DNA was bisulfite-converted using the EZ DNA methylation Kit. Samples were tested for complete bisulfite conversion, and 500 ng of bisulfite-converted DNA was profiled at Moffitt’s Molecular Genomics Core using Illumina’s Infinium MethylationEPIC v2 BeadChips. Preprocessing of the raw DNAm data was performed using the ENmix pipeline, as it is reported to outperform alternative preprocessing methods [12, 13]. The ENmix preprocessing steps include background correction, dye-bias correction, inter-array normalization, and probe-type bias correction [14–16]. No samples failed quality control after methylation profiling, as all samples had bisulfite intensities > 5,000, fewer than 5% of probes with low-quality methylation values, and beta value distributions within expected limits. Immune cell composition estimation and metric calculation Proportions of twelve circulating leukocyte subsets were deconvolved using the EPIC IDOL-Ext library for MethylationEPIC v2 BeadChips [11]. This library estimates the proportions of leukocyte subsets from both the myeloid (neutrophils, basophils, eosinophils, and monocytes) and lymphoid lineages (naïve and memory B cells, CD4+ and CD8+ T cells, T regulatory [Treg] cells, and natural killer cells). Leukocyte proportions were summed into higher-order subsets based on their hematopoietic lineage: granulocytes (neutrophils, basophils, eosinophils); total B cells (naïve and memory); total CD8+ T cells (naïve and memory); and total CD4+ T cells (naïve, memory, and Treg cells), resulting in sixteen related leukocyte metrics for investigation. Although these metrics are DNAm-inferred, for simplicity, we refer to the leukocyte subsets in this study using their standard immunological nomenclature. Validation studies from both the original deconvolution paper and external datasets show high correlations between DNAm-inferred and directly measured leukocyte subsets [11, 17]. Three types of immune metrics were calculated from the deconvolved DNAm data. All cell proportions were multiplied by 100 to calculate percentages of circulating subsets, representing ‘peripheral immune system balance.’ All subset percentages were multiplied by total leukocyte count to calculate counts of circulating subsets, representing ‘peripheral immune cell abundance’ (K cells/µL). Finally, ratio metrics representing ‘systemic inflammation’ (neutrophil-to-lymphocyte ratio), ‘immune system aging’ (naïve-to-total B cell, CD4+ T cell, and CD8+ T cell ratios), ‘cytotoxic-to-helper T cell balance’ (CD8+ T cells-to-CD4+ T cell ratio), and ‘immune suppression’ (Treg-to-total CD4+ T cell ratio) were derived. Statistical analysis Patient characteristics were described using means and standard deviations (SD) for continuous variables or counts and proportions for categorical variables, overall and stratified by recurrence status. A Spearman correlation matrix of leukocyte subset counts and age was created to examine the statistical interdependencies among these variables. Associations of circulating leukocyte subsets with CRC survival outcomes, scaled per 1-SD increase, were estimated using Cox regression models weighted by inverse probability of selection to account for the sampling scheme, and reported as hazard ratios (HRs) with 95% confidence intervals (CIs) and 2-sided P-values. Because the study design preserves the time and event information for each patient, time since diagnosis was treated as the timescale in all analyses. The timescale began at the CRC diagnosis date for analyses of overall survival and the treatment completion date for analyses of disease-free survival. Events were defined as deaths in analyses of overall survival, and disease recurrence, new primary CRC, or death in analyses of disease-free survival. Patients were censored at the end of the study period (September 2024) or loss-to-follow-up, whichever occurred first. Sensitivity analyses were conducted for events occurring within three years of the pre-treatment blood draw to examine associations with short-term survival. All models were adjusted for age at diagnosis and DNA source (buffy coat, whole blood) (reduced model). Additional adjustments were performed for known clinical predictors of CRC prognosis, namely: sex, race, BMI group, stage at diagnosis, microsatellite stability (stable, unstable, missing), and tumor site (full model). Effect modification of leukocyte subset associations was examined by stratifying models on age at diagnosis (< 65 years vs. ≥ 65 years), BMI ( < 30 kg/m2 vs. ≥ 30 kg/m2), sex (males vs. females), stage at diagnosis (I and II vs. III), and tumor site (colon vs. rectum). Statistical interaction was calculated using an interaction term of the leukocyte subset and the modifier of interest. Primary survival analyses used a statistical significance threshold of P ≤ 0.05, and Benjamini-Hochberg false discovery rates are reported in the supplemental material. Given the limited sample size in each stratum, the statistical interaction analyses were considered exploratory, and a significance threshold of P ≤ 0.10 was used. All analyses were performed using Stata version 18.5 (College Station, TX). Results Sample population Over a median follow-up of 7.4 years (range, 0.3–13.8 years), 45 (34%) patients died, and 33 (25%) experienced disease recurrence. Of the 45 patients who died, 23 (51%) had a disease recurrence. The mean age at diagnosis was 63 years (SD = 13) (Table 1). By design, the sample was nearly evenly split by sex (46% female, 54% male) and tumor site (54% rectum, 46% colon), while most participants identified as White (93%) and had BMIs greater than 25 (78%). Compared to those without a recurrence, individuals who experienced a recurrence were more likely to be non-White (15% vs. 4%, P = 0.03) or have microsatellite-stable tumors (70% vs. 46%, P = 0.02).Table 1Patient characteristics by recurrence status (n = 134).CharacteristicOverallRecurrenceNo (N = 101)Yes (N = 33)Age at diagnosis, mean years (SD)62.8 (13)62.4 (13)63.7 (12)BMI, mean kg/m2 (SD)28.8 (6)28.5 (6)29.8 (7)Sex, n (%) Female62 (46)48 (48)14 (42) Male72 (54)53 (52)19 (56)Race, n (%) Non-White9 (7)4 (4)5 (15) White125 (93)97 (96)28 (86)BMI group, n (%) <2530 (22)23 (23)7 (21) 25–3061 (46)47 (47)14 (42) 30+43 (32)31 (31)12 (36)Stage, n (%) I25 (19)19 (19)6 (18) II42 (31)34 (34)8 (24) III67 (50)48 (48)19 (58)MSI status, n (%) Stable69 (52)46 (46)23 (70) Unstable11 (8)11 (11)0 (0) Missing54 (40)44 (44)10 (30)Tumor Site, n (%) Rectum72 (54)52 (52)20 (61) Colon62 (46)49 (49)13 (39) Table 1Patient characteristics by recurrence status (n = 134).CharacteristicOverallRecurrenceNo (N = 101)Yes (N = 33)Age at diagnosis, mean years (SD)62.8 (13)62.4 (13)63.7 (12)BMI, mean kg/m2 (SD)28.8 (6)28.5 (6)29.8 (7)Sex, n (%) Female62 (46)48 (48)14 (42) Male72 (54)53 (52)19 (56)Race, n (%) Non-White9 (7)4 (4)5 (15) White125 (93)97 (96)28 (86)BMI group, n (%) <2530 (22)23 (23)7 (21) 25–3061 (46)47 (47)14 (42) 30+43 (32)31 (31)12 (36)Stage, n (%) I25 (19)19 (19)6 (18) II42 (31)34 (34)8 (24) III67 (50)48 (48)19 (58)MSI status, n (%) Stable69 (52)46 (46)23 (70) Unstable11 (8)11 (11)0 (0) Missing54 (40)44 (44)10 (30)Tumor Site, n (%) Rectum72 (54)52 (52)20 (61) Colon62 (46)49 (49)13 (39) Patient characteristics by recurrence status (n = 134). Overall, granulocytes were the most abundant circulating leukocyte subset (mean: 5.43 K/µL), followed by CD4+ T cells (mean: 0.90 K/µL), monocytes (mean: 0.75 K/µL), CD8+ T cells (mean: 0.34 K/µL), natural killer cells (mean: 0.27 K/µL), and B cells (mean: 0.26 K/µL). Basophils, eosinophils, Treg cells, naïve CD8+ T cells, and naïve CD4+ T cells showed right-skewed distributions and represented the least abundant circulating subsets (Supplementary Fig. 2). Interdependencies of age and circulating leukocyte counts Age was inversely correlated with counts of circulating naïve CD8+ T cell (r = -0.33) and naïve CD4+ T cell subsets (r = -0.12) and positively correlated with counts of circulating basophils (r = 0.15), monocytes (r = 0.13), memory CD8+ T cell (r = 0.14), and memory B cell subsets (r = 0.18) (Fig. 1). Among the leukocyte subsets, counts of granulocytes were positively correlated with neutrophils (r = 0.99); similarly, counts of total CD8+ and CD4+ T cells were positively correlated with their corresponding memory subsets (r = 0.99 for CD8+ and r = 0.93 for CD4+ T cells). The circulating count of total B cells was positively correlated with naïve subsets (r = 0.89).Fig. 1Correlation matrix of leukocyte counts and age.Spearman correlation coefficient matrix displaying the interdependences between age and the counts of circulating leukocyte subsets. The color scale represents the strength of the correlation, with yellow representing the strongest positive correlation and purple representing the strongest inverse correlation. Correlation matrix of leukocyte counts and age. Spearman correlation coefficient matrix displaying the interdependences between age and the counts of circulating leukocyte subsets. The color scale represents the strength of the correlation, with yellow representing the strongest positive correlation and purple representing the strongest inverse correlation. Circulating leukocyte abundance, balance and disease-free survival In Cox regression models adjusted for age at diagnosis and DNA source, counts of circulating granulocytes were statistically significantly associated with worse disease-free survival (HR: 1.34, 95% CI: 1.06, 1.69, P = 0.01) (Fig. 2, top panel). This association appeared to be largely driven by neutrophil subsets (HR: 1.35, 95% CI: 1.08, 1.69, P = 0.01) (Supplementary Table 1). In fully adjusted models, the associations of granulocytes and neutrophils with worse disease-free survival were consistent. Furthermore, in fully-adjusted models, higher counts of circulating Treg cells were associated with worse disease-free survival (HR: 1.32, 95% CI: 1.00, 1.74, i = 0.05) (Supplementary Table 1). Higher counts of B cells were also associated with worse disease-free survival (HR: 1.30, 95% CI: 1.01, 1.68, P = 0.04) (Fig. 2, top panel), with the strongest association observed for the memory subsets (memory B cell, HR: 1.56, 95% CI: 1.31, 1.86, P < 0.0001; naïve B cell, HR: 1.01, 95% CI: 0.78, 1.32, P = 0.93) (Supplementary Table 1). The memory B cell association remained statistically significant in the analysis restricted to events occurring within the first three years of the pre-treatment blood draw (HR: 1.63, 95% CI: 1.18, 2.28, P = 0.003). We did not observe violations in proportional hazard assumptions in our models.Fig. 2Abundance and balance metrics of the major circulating leukocyte subsets and disease-free CRC survival.Hazard ratios and 95% CIs for the relationships between the major leukocyte subsets and disease-free CRC survival, scaled per 1-SD increase in the leukocyte subset. Reduced models are adjusted for age at diagnosis and DNA source only; full models additionally adjust for sex, race, BMI group, stage at diagnosis, microsatellite stability, and tumor site. Abundance and balance metrics of the major circulating leukocyte subsets and disease-free CRC survival. Hazard ratios and 95% CIs for the relationships between the major leukocyte subsets and disease-free CRC survival, scaled per 1-SD increase in the leukocyte subset. Reduced models are adjusted for age at diagnosis and DNA source only; full models additionally adjust for sex, race, BMI group, stage at diagnosis, microsatellite stability, and tumor site. Proportions of circulating leukocytes generally displayed fewer associations with disease-free survival than the leukocyte count associations. In reduced models, proportions of circulating monocytes were associated with better disease-free survival (HR: 0.77, 95% CI: 0.61, 0.98, P = 0.03) (Fig. 2, bottom panel). However, this association was no longer statistically significant after adjustment for prognostic clinical factors. Similar to the associations of circulating counts, a higher circulating proportion of memory B cells was associated with worse disease-free survival (HR: 1.31, 95% CI: 1.00, 1.72, P = 0.05) (Supplementary Table 2). Stratified analyses were performed to explore the potential effect modification of the associations between peripheral immune cell metrics and disease-free survival. Counts of circulating memory B cells appeared to be more strongly related to disease-free survival among those diagnosed with rectal cancer (rectal tumors, HR: 3.21, 95% CI: 1.68, 6.10, P = 0.0004; colon tumors, HR: 1.71, 95% CI: 1.29, 2.27, P = 0.0002; P-interaction= 0.03) (Supplementary Table 3). Circulating leukocyte abundance, balance and overall survival In Cox models adjusted for age at diagnosis and DNA source, counts of circulating granulocytes were statistically significantly associated with worse overall survival (HR: 1.47, 95% CI: 1.15, 1.87, P = 0.002) (Fig. 3 top panel). When more refined granulocyte subsets were examined, higher counts of both neutrophils and basophils were associated with worse overall survival (neutrophils, HR: 1.47, 95% CI: 1.16, 1.87, P = 0.002; basophils, HR: 1.33, 95% CI: 1.05, 1.68, P = 0.02) (Supplementary Table 4). The granulocyte and neutrophil associations were slightly attenuated in fully-adjusted models, whereas the basophil association became slightly stronger. The fully-adjusted models also showed that higher circulating counts of Treg cells were associated with worse overall survival (HR: 1.38, 95% CI: 1.00, 1.90, P = 0.05) (Supplementary Table 4). We did not observe violations in proportional hazard assumptions in our models.Fig. 3Abundance and balance metrics of the major circulating leukocyte subsets and overall CRC survival.Hazard ratios and 95% CIs for the relationships between the major leukocyte subsets and overall CRC survival, scaled per 1-SD increase in the leukocyte subset. Reduced models are adjusted for age at diagnosis and DNA source only; full models additionally adjust for sex, race, BMI group, stage at diagnosis, microsatellite stability, and tumor site. Abundance and balance metrics of the major circulating leukocyte subsets and overall CRC survival. Hazard ratios and 95% CIs for the relationships between the major leukocyte subsets and overall CRC survival, scaled per 1-SD increase in the leukocyte subset. Reduced models are adjusted for age at diagnosis and DNA source only; full models additionally adjust for sex, race, BMI group, stage at diagnosis, microsatellite stability, and tumor site. Similar to the analysis of leukocyte proportions and disease-free survival, proportions of circulating leukocyte subsets generally showed fewer associations with overall survival than counts. A higher proportion of circulating basophils was associated with worse overall survival in both reduced- and fully-adjusted models (reduced HR: 1.40, 95% CI: 1.13, 1.75, P = 0.002; full HR: 1.47, 95% CI: 1.06, 2.03, P = 0.02) (Supplementary Table 5). Stratified analyses were performed to explore the potential effect modification of the associations between peripheral immune cell metrics and overall survival. Higher counts of circulating basophils were associated with worse overall survival among those diagnosed with rectal tumors (rectal, HR: 1.75, 95% CI: 1.26, 2.44, P = 0.0009; colon, HR: 0.90, 95% CI: 0.50, 1.62, P = 0.71; P-interaction= 0.07) and among those diagnosed with earlier stage disease (stage I and II, HR: 1.85, 95% CI: 1.31, 2.62, P = 0.0005; stage III, HR: 0.74, 95% CI: 0.45, 1.23, P = 0.24; P-interaction= 0.03) (Supplementary Table 6). Discussion Using methylation cytometry to systematically assess peripheral immune cell composition from archived biospecimens, we identified novel circulating leukocyte subsets associated with CRC survival. Specifically, after adjusting for multiple comparisons, we observed that higher pre-treatment counts of circulating memory B cells were associated with worse disease-free survival, whereas higher counts of circulating neutrophils and higher proportions of circulating basophils were associated with worse overall survival. Notably, circulating leukocyte count metrics showed stronger survival associations than proportion metrics, possibly because proportion metrics are constrained to sum to 100% and thus cannot vary independently of one another. We also observed largely similar associations in both reduced and fully adjusted models, suggesting a limited role for confounding by the model covariates. Certain associations showed anatomical specificity, with basophil and memory B cell associations appearing stronger in rectal tumors compared to colon tumors. Overall, these findings confirm methylation cytometry as an innovative technique for comprehensively profiling peripheral immune cell subsets, offering new opportunities to investigate systemic immune contributions to cancer progression and outcomes in epidemiologic studies. Methylation cytometry is increasingly being used in studies of cancer risk and survival. In the Sister Study cohort, our group applied it to archived biospecimens and identified shifts in leukocyte subsets occurring before a breast cancer diagnosis [18] and across breast cancer development and treatment [19]. More recently, it has been used to explore biomarkers of prognosis in head and neck cancer patients treated with immune checkpoint inhibitors [20, 21]. In the context of CRC, methylation cytometry was used to find that higher lymphocyte proportions were associated with better survival, whereas higher granulocyte proportions were associated with worse survival [22]. However, this earlier study applied an older methylation cytometry algorithm that estimated only six leukocyte subsets and lacked total leukocyte count information, preventing analyses of circulating leukocyte counts [22]. Our study addressed these limitations by applying the top-performing algorithm (EPIC IDOL-Ext) to estimate over twice as many leukocyte subsets and incorporating information on total leukocyte counts. This approach revealed that distinct subsets of granulocytes (neutrophils and basophils) and lymphocytes (memory B and Treg cells) were associated with CRC prognosis, providing a higher-resolution view into the associations between peripheral immune cell composition and cancer survival. The biological mechanisms underlying the observed associations between circulating leukocyte subsets and CRC outcomes likely reflect systemic processes that can influence cancer progression, including immunosuppression and systemic inflammation. Neutrophils have traditionally served as a defense against microbial infections, but there is a growing appreciation that tumor growth can perturb neutrophil heterogeneity via cancer-dependent priming [23]. Specifically, tumors are reported to contribute to emergency granulopoiesis [24], leading to increased neutrophil production with reduced function and immunosuppressive properties [25, 26]. This process may be driven by metabolic reprogramming toward more oxidative states [27]. Moreover, higher levels of circulating neutrophils, commonly measured relative to lymphocyte levels to represent systemic inflammation, have been consistently associated with poor cancer outcomes, including for CRC [28]. Memory B cells are responsible for long-term humoral immunity but may influence CRC prognosis when they acquire immunosuppressive qualities. Specifically, memory B cells can differentiate into regulatory B cells [29], which produce interleukin (IL)-10, an anti-inflammatory cytokine that suppresses immune responses [30]. Circulating levels of IL-10 are elevated in CRC patients [31] and are associated with worse disease-free survival [32, 33], suggesting that higher counts of circulating memory B cell subsets may reflect systemic immunosuppression rather than protective immunity. This study is not without limitations. First, the modest sample size limited our statistical power in both full sample and stratified analyses. For example, although some evidence of effect modification by tumor site and stage at diagnosis was observed, the findings should be considered exploratory. Second, the sample population lacked racial and ethnic diversity, limiting the generalizability of the leukocyte-CRC survival associations across different racial and ethnic groups. However, this homogeneity protects against confounding by race and ancestry, potentially increasing the validity of findings among the enrolled groups. Third, we lacked information on patients’ infection status, which may be associated with both circulating immune cell characteristics and CRC survival. Lastly, like most leukocyte subtyping techniques, methylation cytometry is less accurate for rarer subsets. For instance, in a validation study comparing methylation cytometry to traditional sorting techniques, the reliability of basophil proportion and count estimates was lowest [17]. As such, the basophil associations with CRC survival warrant further investigation and validation in external populations. Despite these limitations, this study is strengthened by its retrospective sampling design, enriched survival endpoints, extended patient follow-up, and the application of methylation cytometry to archived biospecimens, allowing the comprehensive profiling of peripheral immune cell composition. In summary, we find that the abundance and balance of circulating leukocyte subsets are associated with CRC survival. While extensive research has focused on the immune composition within the tumor microenvironment, comparatively fewer studies have examined peripheral immune cell composition. Since most tumor-infiltrating leukocytes arrive via circulation, characteristics of the peripheral immune system appear to provide a complementary view into patient prognosis, reflecting both systemic immune competence and accessibility before surgery. Future studies examining changes in peripheral immune cell composition across cancer treatment may further refine CRC survival prediction and provide additional insights into anticancer responses from the peripheral immune system. Supplementary information Supplemental Material Supplemental Material