Lynch, Charles F.

AbstractTumor‐infiltrating lymphocytes in colorectal cancer (CRC) predict better survival. However, associations between T‐lymphocyte count in histologically normal tissues from patients with CRC and survival remain uncertain. We examined associations of CD3+ T‐cells in colorectal tumor and histologically normal tissues with CRC‐specific and all‐cause mortality in the prospective Iowa Women's Health Study. Tissue microarrays were constructed using paraffin‐embedded colorectal tissue samples from 464 women with tumor tissues and 314 women with histologically normal tissues (55–69 years at baseline) diagnosed with incident CRC from 1986 to 2002 and followed through 2014 (median follow‐up 20.5 years). Three tumor and two histologically normal tissue cores for each patient were immunostained using CD3+ antibody and quantified, and the counts were averaged across the cores in each tissue. Cox proportional hazards regression estimated hazard ratios (HR) and 95% confidence interval (CI) for CRC‐specific and all‐cause mortality. After adjustment for age at diagnosis, body mass index, smoking status, tumor grade, and stage, HRs (95% CI) for the highest versus lowest tertile of tumor CD3+ score were 0.59 (0.38–0.89) for CRC‐specific mortality and 0.82 (0.63–1.05) for all‐cause mortality; for histologically normal CD3+ score, the corresponding HRs (95% CI) were 0.47 (0.19–1.17) and 0.50 (0.27–0.90), respectively. The CD3+ score combining the tumor and histologically normal scores was inversely associated with CRC‐specific and all‐cause mortality. Although the association between tumor CD3+ score and all‐cause mortality was not significant, both higher CD3+ T‐lymphocyte counts in tumor and histologically normal scores tended to be associated with lower CRC‐specific and all‐cause mortality.

Abstract Background. The strongest established prognostic factors for follicular lymphoma are age and clinical factors. Biologic studies suggest a major role for the host immunologic environment in follicular lymphomagenesis, which is in part determined by host genetic background. Cytokines are key regulators of immune function and regulation, are highly polymorphic, and have been implicated in lymphoma etiology and prognosis. We evaluated the hypothesis that inherited variability in cytokine and related immune genes impact overall survival in follicular lymphoma. Methods. We genotyped 73 single nucleotide polymorphisms (SNPs) from 44 candidate cytokine and immune genes in 278 follicular lymphoma patients who participated in a population-based case-control study conducted from 1998–2000 in the Detroit, Iowa, Los Angeles and Seattle SEER (Surveillance, Epidemiology and End Results) cancer registries. Baseline clinical data and survival through early 2005 were obtained from cancer registry files. Cox proportional hazards analysis was used to estimate hazard ratios (HR) and corresponding 95% confidence intervals for the association between individual SNPs and overall survival, adjusting for age, demographic and clinical factors. We also used parallel modeling strategies to identify the best summary multi-SNP risk score to predict survival, and applied these results in a time-dependent receiver-operator characteristics (ROC) analysis. Results. The median age at diagnosis was 57 years (range, 25–74), and 59 (21%) of the patients died during follow-up, with a median follow-up of 59 months (range, 27 – 78 months) for surviving patients. In multivariate modeling, SNPs in IL8 (rs4073; HRTT=2.14, 1.26–3.63), IL2 (rs2069762; HRGT/TT=1.80, 1.06–3.05), IL12B (rs3212227; HRAA/CC=1.83, 1.06–3.06), and IL1RN (rs454078; HRAA=1.93, 1.11–3.34) were the strongest and most robust predictors of survival. A summary score of the number of deleterious genotypes from these four genes was strongly associated with survival (p=1.4 × 10−5) after accounting for demographic and clinical variables (HR=2.06 per deleterious genotype, 1.52–2.79). A combination of the four SNPs and the demographic and clinical risk scores was strongly associated with survival (p=1.8 × 10−11); the 5-year Kaplan Meier survival estimates were 96% (93%–100%), 72% (62%–83%) and 58% (48%–72%) for low, intermediate, and high risk groups respectively. In a time-dependent ROC analysis the three risk groups had an area under the curve (AUC) of 0.83 at 72 months of follow-up (95% CI 0.72–0.93). Conclusion: Host genetic variability in immune genes, particularly IL8, IL2, IL12B, and IL1RN, individually and particularly in combination, was associated with overall survival in follicular lymphoma after accounting for demographic and clinical factors. Host immunogenetics is a promising class of prognostic factors in follicular lymphoma that warrants further evaluation.

Abstract Background: Several studies suggest that smoking and obesity may increase the risk of developing NHL, but the impact of these factors on survival is relatively unexplored. One recent population-based Italian study found that smoking may negatively impact overall survival. Methods: We evaluated the association of body mass index (BMI) and cigarette smoking on overall survival in 1,286 patients who participated in a population-based case-control study conducted through the Surveillance, Epidemiology, and End Results (SEER) cancer registries in Detroit, Iowa, Seattle and Los Angeles. BMI (on 1189 participants) and a smoking history (on 471 participants) were obtained through in-person interviews. Histology, stage, presence of B symptoms, first course of therapy, date of last follow-up, and vital status were derived from linkage to registry databases at each study site in the spring of 2005. Hazard Ratios (HRs) and 95% confidence intervals were estimated using Cox proportional hazards models, while adjusting for demographic and clinical factors. Results: The median age at diagnosis was 58 years (range: 16 – 74), and the most common histologies were DLBCL (33%), follicular (25%), SLL/CLL (12%), marginal zone (9%), mantle cell (4%) and Peripheral T-cell (3%). The median BMI was 26.6 (range: 16.2 – 47.3), and 26% of patients were classified as obese (i.e. BMI >30). Of patients with a smoking history, 47% had never smoked, 19% were active smokers at diagnosis, and 34% were former smokers. No significant differences were observed between those with and without tobacco use data in terms of primary lymphoma characteristics. Overall survival (OS) was defined as time from diagnosis to death due to any cause. Median follow-up on all living patients was 58 months (range: 4 – 78), and 333 patients had died. When BMI was assessed as a continuous measure, it was borderline significantly associated with OS (p=0.050). An indicator of whether or not BMI > 30 was also significantly associated with OS (HR=1.30; 95% CI: 1.01 – 1.68; p=0.042). Ever vs. never smokers also exhibited a survival disadvantage (HR=1.66; 95% CI: 1.12 – 2.44; p=0.011). When broken down further, former smokers who had smoked for >10 years (HR=1.61; 95% CI: 1.02 – 2.52; p=0.039) and current smokers (HR=1.79; 95% CI: 1.07 – 3.02; p=0.028) were at a survival disadvantage over those who had never smoked. When both smoking and BMI status were included in the model, only the indicator of being a current smoker remained statistically significant (p=0.031), with a trend toward decreased survival for former smokers who had smoked >10 years (p=0.061) and those with BMI >30 (p=0.058). Conclusions: NHL patients who smoke or are obese had a poorer overall survival. These observations warrant further investigation, including whether interventions for smoking cessation or weight loss can impact NHL survival.

Abstract Background: Mantle cell lymphoma (MCL) is an incurable non-Hodgkin lymphoma. MCL is characterized by the chromosomal translocation t(11;14) leading to overexpression of cyclin D1 (CCND1), a regulator of the cell cycle. NF-κB plays a central role in key cellular processes, including cell cycle regulation and apoptosis, and it has been shown to be constitutively activated in MCL cell lines and in patient biopsy samples and to play a key role in the growth and survival of MCL cells. In an exploratory study, we evaluated the hypothesis that germline variation in candidate genes from the cell cycle and NF-κB pathways predict overall survival in MCL. Methods: We genotyped 235 SNPs from 29 cell cycle genes and 447 SNPs from 55 NF-κB genes in 39 MCL patients aged 38–64 years who participated in a population-based study conducted from 1998–2000 through the Surveillance, Epidemiology, and End Results (SEER) cancer registries in Detroit, Seattle, Iowa, and Los Angeles county. Tagging single nucleotide polymorphisms (SNPs) were selected from HapMap. Stage, presence of B-symptoms, first course of therapy, date of last follow-up, and vital status were derived from linkage to registry databases at each study site. To assess the statistical significance of a gene, we first used principal components analysis to capture the genetic variation of the SNPs within a gene. Cox proportional hazards analysis was then used to assess the association between the top principal components from a gene with overall survival. Results: Through early 2005, there were 22 deaths in 39 patients (56%). The median follow-up of the 17 surviving patients was 56 months (range 36–70). Eleven of the 55 genes (TNFSF14 p=0.0007, NF-κBIA p=0.003, IL1R2 p=0.01, TNFRSF25 p=0.02, TNFSF13B p=0.02, FAS p=0.03, TLR2 p=0.04, TNFSF10D p=0.04, TNFSF10 p=0.06, MAP3K2 p=0.06) in the NF-κB pathway were associated with overall survival. In the cell cycle pathway, only 3 of the 29 genes (CASP5 p=0.06, CASP9 p=0.09, BCL2 p=0.09) were associated with overall survival, and a previously reported SNP in CCND1 (rs603965) showed no association (p=0.5). Permutation p-values for observing at least this number of associated tests in each pathway due to chance (from multiple testing) was p=0.8 for the cell cycle pathway and p=0.03 for NF-κB pathway. The small sample size precluded multi-gene modeling. Conclusion: In a population series of MCL patients, we found suggestive evidence that host genetic variation in candidate NF-κB pathway genes affects overall survival. These results complement tumor expression data implicating this pathway in the pathogenesis of MCL. In addition, such findings further support the value of developing therapeutic targets from this pathway.

Abstract Background. Intracellular one-carbon transfer reactions are essential for nucleotide synthesis and methylation of biologic compounds including DNA. Previous studies have linked genetic variants in one-carbon metabolism genes with risk of developing NHL, but little is known regarding the impact of these variants on disease outcome. We evaluated the hypothesis that inherited genetic variation in genes involved in one-carbon metabolism is associated with overall survival in DLBCL. Methods. We genotyped 30 single nucleotide polymorphisms (SNPs) from 18 candidate one-carbon metabolism genes in 215 DLBCL patients who participated in a population-based case-control study conducted from 1998–2000 using the SEER (Surveillance, Epidemiology and End Results) cancer registries in the Detroit, Iowa, Los Angeles and Seattle. Stage, B-symptoms, first course of therapy, date of last follow-up and vital status through early 2005 were obtained from cancer registry files. Cox proportional hazards analysis was used to estimate hazard ratios (HR) and corresponding 95% confidence intervals for the association between individual SNPs and overall survival, adjusting for age, demographic and clinical factors. We also used parallel modeling strategies to identify the best summary multi-SNP risk score to predict survival. Results. The median age at diagnosis was 57 years (range, 20–74), and 50 (23%) of the patients died during follow-up, with a median follow-up of 57 months (range, 31–78 months) for surviving patients. After adjusting for demographic and clinical variables, SNPs in SHMT1 (rs1979276; HRCT/TT=2.47, 1.31–4.67), BHMT (rs585800; HRAT/TT=2.02, 1.16–3.54), and TCN1 (rs526934; HRTT=1.86, 1.04–3.33) were the strongest and most robust predictors of survival. A summary score of the number of deleterious genotypes (0–3) from these three genes was strongly associated with survival (p=7.9 x 10−5) after accounting for demographic and clinical variables (HR=2.58 per deleterious genotype, 95% CI 1.75–3.80). A risk score combining the three SNPs with clinical and demographic variables (score of 0 to 5) was even more strongly associated with survival (p=1.4 x 10−13); the Kaplan Meier survival curves are shown in the Figure. In a time-dependent ROC analysis, the combined risk score had a concordance index of 0.75 at 5 years of follow-up (95% CI 0.69–0.81). Conclusion: Host genetic variation in the one-carbon metabolism genes SHMT1, BHMT, and TCN1, individually and particularly in combination, was associated with overall DLBCL survival after accounting for clinical and demographic factors, supporting a role for this pathway in disease progression. Future work should evaluate interactions of genes from this pathway with dietary nutrients and therapeutic agents in DLBCL prognosis. Figure Figure

Abstract Small lymphocytic lymphoma (SLL) is an incurable indolent lymphoma, and there are relatively few prognostic biomarkers for this important NHL subtype. We evaluated the hypothesis that inherited variability in cytokine and immune-related genes was associated with overall survival in SLL. We genotyped 73 SNPs in 44 candidate genes in 140 SLL patients aged 20–70 years who participated in a population-based case-control study conducted from 1998–2000 through the Surveillance, Epidemiology, and End Results (SEER) cancer registries in Detroit, Seattle, Iowa, and Los Angeles. DNA was extracted from a venous blood sample or mouthwash buccal cell sample. All genotyping was conducted at the National Cancer Institute Core Genotyping Facility using the Taqman platform. Histology, stage, presence of B-symptoms, first course of therapy, date of last follow-up, and vital status were derived from linkage to registry databases at each study site in the spring of 2005. Cox proportional hazards analysis was used to estimate the hazard ratio (HR) and 95% confidence interval for the association between individual SNPs and overall survival, adjusted for age, demographic and clinical factors. Multiple simultaneous modeling strategies were used to identify the best summary multi-SNP risk score to predict survival. At last follow-up, there were 45 deaths in 140 patients (32%). The median follow-up of the 95 surviving patients was 58 months (range 24–75 months). In multivariate modeling, SNPs in IL13 (rs1800925; HRCC=2.36, 1.24–4.49), IL7R (rs1494555; HRGG =2.20, 0.84–5.76), and TNF-alpha (rs1800630; HRAC/AA=1.73, 0.95–3.17) were the strongest and most robust predictors of survival. Two or more deleterious genotypes from these three SNPs increased the risk of death (HR=2.2, 1.2–4.1) compared with one or fewer deleterious genotypes (p=0.009). Three groups (low, intermediate, and high risk) were defined by combining the SNP score and a clinical/demographic score. The 5-year Kaplan-Meier survival estimates for these groups were 85%, 65%, and 11%, respectively. Compared to patients with a low risk score, patients with intermediate (HR=2.5, 1.2–5.1) or high (HR=11, 4.3–27.7) risk scores had poorer overall survival (p=3.9 x 10−8). Our preliminary results suggest that SNPs in IL13, IL7R, and TNF-alpha alone and in combination predict overall survival in SLL, lending support to the hypothesis that host genetic background is a promising class of prognostic biomarkers in SLL.

Abstract DLBCL is a curable subtype of non-Hodgkin lymphoma, although a significant number of patients do not achieve a remission or they relapse with conventional chemotherapy. While clinical variables (e.g., IPI), tumor (somatic) genetic alterations, and gene expression profiling have all been shown to predict outcome, there remains a need for additional prognostic biomarkers. One understudied class of biomarkers is host genetic background. We evaluated the hypothesis that germline variability in 73 SNPs from 44 candidate immune genes was associated with overall survival in DLBCL. We addressed this hypothesis in 365 DLBCL patients aged 20–70 years who participated in a population-based case-control study conducted from 1998–2000 (prior to the use of R-CHOP) through the Surveillance, Epidemiology, and End Results (SEER) cancer registries in Detroit, Seattle, Iowa, and Los Angeles. Germline DNA was extracted from a venous blood sample or mouthwash buccal cell sample, which was collected a median of 4.8 months after diagnosis in this population-based study. All genotyping was conducted at the National Cancer Institute Core Genotyping Facility using the Taqman platform, and was successful in over 95% of the DNA samples for the SNPs evaluated. Histology, stage, presence of B-symptoms, first course of therapy, date of last follow-up, and vital status were derived from linkage to registry databases at each study site in the spring of 2005. Cox proportional hazards analysis was used to evaluate the association between individual SNPs, adjusted for age, demographic and clinical factors. Parallel modeling strategies were used to identify the best summary multi-SNP risk score to predict survival. At a median follow-up of 56 months (range, 27-78 months) for surviving patients, there were 96 deaths in 365 patients (26%). In multivariate modeling, SNPs in IL1A (rs1800587; HRCT/TT=1.90, 1.26–2.87), IL6 (rs1800795; HRGG=1.48, 0.99–2.23), IL-10 (rs1800896; HRAG/GG=1.48, 0.91–2.38), and IFNGR2 (rs2070385; HRAG/GG=1.35, 0.86–2.11) were the strongest and most robust predictors of overall survival. A summary score of the number of deleterious genotypes from these four genes in combination with clinical and demographic variables was strongly associated with survival (p=9.3 x 10−12); Kaplan-Meier 5-year survival estimates for low, intermediate, and high risk patients were 89%, 68%, and 47% respectively. In conclusion, host genetic background as measured by germline polymorphisms in immune genes including IL1A, IL6, IL10, and IFNGR2 were associated with overall survival in DLBCL after accounting for clinical and demographic factors. These promising results require confirmation and need further evaluation in patients treated with R-CHOP in conjunction with tumor and other prognostic biomarkers.

Abstract Background: Different molecules within coffee may be pro-carcinogenic, while others may have anti-oxidant and anti-neoplastic effects. Hence, it can be hypothesized that coffee may have differing risk associations with specific colorectal cancer (CRC) subtypes. The aim of this study was to evaluate associations between coffee consumption and incident CRC overall, by anatomic subsite, and by molecularly defined tumor subtypes, in the population-based Iowa Women's Health Study (IWHS). Methods: IWHS participants, ages 55-69 years, provided dietary intake estimates using self-reported food frequency questionnaires at baseline (1986). From 35,216 women included the final analysis, 1298 developed CRC (673 with proximal CRC, 597 with distal CRC) through December 31, 2002 (end of the IWHS molecular epidemiology project study period). Archived, paraffin-embedded tissue specimens for 732 representative CRC cases were collected and analyzed to determine if coffee consumption (none or <1 cup/month, <1 cup/week, 1 cup/day, 2-3 cups/day and ≥4 cups/day) was associated with microsatellite instability (MSI-L/MSS or MSI-H), CpG island methylator phenotype (CIMP-negative or CIMP-positive), BRAF mutation status (BRAF-wildtype or BRAF-mutated), KRAS mutation status (KRAS-wildtype and KRAS-mutated), p53 protein expression level (p53 absent, low or high) and estrogen receptor β protein expression level (ERβ absent, low or high). Multivariable Cox regression models were fit to estimate relative risks (RR) and 95% confidence intervals (CI) with adjustment for age, smoking, alcohol use, diabetes, body mass index and waist-hip ratio, physical activity, estrogen use, dietary intake of fat, calcium, vitamin E, sucrose, red meat and methionine. P-values were not adjusted for multiple testing. Results: There was no significant difference in the risk of overall (RR, 1.05; 95% CI, 0.85-1.29; Ptrend=0.50), proximal (RR, 0.91; 95% CI, 0.68-1.21; Ptrend=0.57) or distal CRC (RR, 1.18; 95% CI, 0.86-1.64; Ptrend=0.19) in women who consumed ≥4 cups of coffee/day (caffeinated or decaffeinated), as compared to non-consumers. By CRC molecular subtypes for these same two exposure groups, women with ≥4 cups of coffee/day were more likely to develop ERβ-low CRC (RR, 2.98; 95% CI, 1.04-8.51; Ptrend=0.06), though no significant associations were observed for caffeinated or decaffeinated coffee specifically. In contrast, women who consumed ≥4 cups of decaffeinated coffee/day were less likely to develop MSI-L/MSS (RR, 0.57; 95% CI, 0.36-0.92; Ptrend=0.04) and BRAF-WT (RR, 0.69; 95% CI, 0.44-1.06; Ptrend=0.06) CRCs than non-consumers. Women who consumed ≥4 cups of caffeinated coffee/day were also less likely to develop p53-low CRCs (RR, 0.53; 95% CI, 0.24-1.15; Ptrend=0.02). Null associations were observed between decaffeinated coffee consumption and risk of CIMP-, KRAS- and p53- defined CRC subtypes, as well as between caffeinated coffee consumption and risk of MSI-, BRAF-, CIMP- or KRAS-defined CRC subtypes. Conclusion: In this prospective cohort study of older women, though coffee consumption was not associated with risk of overall or site-specific CRC, unique associations were observed with molecularly defined CRC subtypes, supporting a need for further investigation of the potential pro- and anti-carcinogenic effects of coffee constituents on biologic targets involved in colorectal carcinogenesis. Citation Format: Siddharth Singh, Robert Vierkant, Lori Tillmans, Alice Wang, Daniel Weisenberger, Peter Laird, Maki Inoue-Choi, Andrea Mariani, Charles Lynch, Kristin Anderson, Amy French, Robert Haile, Lisa Harnack, John Potter, Susan Slager, Thomas Smyrk, Stephen Thibodeau, James Cerhan, Paul Limburg. Coffee consumption and colorectal cancer risk by molecularly defined subtypes in a prospective cohort study of older women. [abstract]. In: Proceedings of the Twelfth Annual AACR International Conference on Frontiers in Cancer Prevention Research; 2013 Oct 27-30; National Harbor, MD. Philadelphia (PA): AACR; Can Prev Res 2013;6(11 Suppl): Abstract nr A44.

Abstract Background: The increased number of T-lymphocytes residing in the tumors of colorectal cancer (CRC) patients is consistently shown to predict better survival of CRC patients independent of stage. In addition to tumor itself, T-lymphocytes also infiltrate normal tissues adjacent to tumors; however, little is known about how the inter-individual T-cell variability in tumor-adjacent tissues affects CRC prognosis. Our goal was to characterize total T-cells in colorectal tumor and tumor-adjacent tissues and study their associations with all-cause and CRC-specific survival in the Iowa Women’s Health Study. Methods: We constructed tissue microarrays and quantified CD3 antibody staining (Clone F.7.238; Dako), an established marker for total T-cells, in paraffin-embedded tissue samples from 463 women diagnosed with incident CRC from 1986-2002 (mean age at diagnosis was 74 years). Up to 3 tumor and 3 tumor-adjacent cores were immunostained for each person. An experienced pathologist quantified CD3+ T cells in each area using 4 categories: non-detected, mild (1-10 cells per 0.28 mm2); moderate (11-29 cells per 0.28 mm2); and strong infiltration (≥30 cells per 0.28 mm2). The obtained scores were averaged over all quantified cores for each person. Cox regression was used to estimate the hazard ratio (HR) and 95% CI for all-cause and CRC mortality in relation to (1) tumor score; (2) tumor-adjacent score; and (3) the ratio of tumor to tumor-adjacent score, hereafter called Score ratio. All scores were categorized in quartiles. The final model adjusted for age at diagnosis, SEER stage, tumor grade, body mass and smoking history. Results: During follow-up for a maximum of 25 years, 67% of participants diagnosed with CRC died (~30% died from CRC). There was a weak but significant correlation between CD3 tumor and tumor-adjacent scores (Spearman coefficient: r=0.15, p=0.02). CD3 tumor score was inversely associated with stage and was higher in those with proximal colon cancer, and for MSI-high, CIMP-high and BRAF-mutated tumors. The HRs (95%CI) for the highest versus lowest category of CD3 tumor score were 0.61 (0.44-0.84) (p-trend=0.003) for all-cause mortality and 0.27 (0.15-0.48) (p-trend <0.0001) for CRC-specific mortality. The tumor-adjacent score was not associated with any characteristics or death. The associations between the Score ratio and all-cause and CRC survival mirrored those for the tumor score but were weaker. The HRs (95%CI) for the highest versus lowest quartile of the Score ratio were 0.70 (0.45-1.09) (p-trend=0.09) for all-cause and 0.45 (0.23-0.87) (p-trend=0.003) for CRC-specific mortality. Conclusions. Based on data from prospectively identified CRC cases in the IWHS cohort, colorectal tumor T-cell infiltration is associated with improved survival. Further investigation is needed to determine the T-cell subtypes that are most predictive of CRC survival outcomes. Citation Format: Anna E. Prizment, Robert A. Vierkant, Thomas C. Smyrk, Lori S. Tillmans, Heather H. Nelson, Charles F. Lynch, Stephen N. Thibodeau, Timothy R. Church, James R. Cerhan, Kristin E. Anderson, Paul J. Limburg. Associations between total T-lymphocytes and colorectal cancer survival in a prospective cohort study of older women [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 2247. doi:10.1158/1538-7445.AM2017-2247

Abstract Background: Host immune response may predict the course of colorectal cancer. We examined the survival of 468 colorectal cancer patients associated with two tumor-infiltrating immune biomarkers, the number of cytotoxic T lymphocytes (CTLs), and the activated CTLs, as reflected by the number of cells expressing granzyme B (GZMB) in the prospective Iowa Women's Health Study. Methods: Using paraffin-embedded tissue samples, we constructed and immunostained tumor microarrays with CD8 (for CTL) and GZMB antibodies. We scored CTL and GZMB densities in tumor epithelial and stromal tissues and also created a composite score for each biomarker (sum of the scores across tissue compartments). Cox regression estimated the HR and 95% confidence intervals (CI) for all-cause and colorectal cancer–specific death associated with each composite score. Results: CTL and GZMB composite scores were positively correlated (r = 0.65) and each biomarker was inversely correlated with stage at diagnosis. Both composite scores were higher in proximal colon tumors and tumors characterized by MSI-high, CIMP-high, or BRAF mutation status. HRs (95% CI) were 0.53 (0.38–0.75; Ptrend = 0.0004) and 0.66 (0.51–0.86; Ptrend = 0.002) for all-cause death, respectively, and 0.30 (0.18–0.51; Ptrend < 0.0001) and 0.41 (0.27–0.63; Ptrend < 0.0001) for colorectal cancer–related death, respectively. Including CTL and GZMB scores simultaneously in the model significantly improved the predictive performance of the models for all-cause and colorectal cancer–related death. Conclusions: Higher tumor infiltration with CTL and GZMB cells is associated with improved all-cause and cancer-specific survival of colorectal cancer patients. Impact: Both the number of CTLs and GZMB appear to be useful prognostic factors in colorectal cancer, irrespective of stage. Cancer Epidemiol Biomarkers Prev; 26(4); 622–31. ©2016 AACR.