Incident Cancer Risk and Signatures Among Older MUTYH Carriers: Analysis of Population-Based and Genomic Cohorts

The creation of 8-Oxoguanine (8-oxoG) after oxidative damage to guanine is one of the most common forms of DNA damage from reactive oxygen species. 8-oxoG binds with adenine and can result in G:C>T:A transversion mutations if not corrected (1). The MYH DNA glycosylase enzyme, encoded by MUTYH, is a critical component of base excision repair (BER) of 8-oxoG–related mutagenesis (2). Inheritance of two inactive alleles of MUTYH results in a colonic polyposis phenotype (MUTYH-associated polyposis, MAP) associated with a high lifetime risk of colorectal cancer, as well as hypermutant cancers occurring in other tissues such as the stomach and duodenum (3–6).

Given the 1% to 2% prevalence of MUTYH carriers among the European population, multiple investigations have been performed to estimate the cancer risk among monoallelic carriers (7, 8). These studies have identified a modest increase in colorectal cancer incident risk among older MUTYH carriers, or those with family histories of colorectal cancer (9, 10). However, previous prospective and population-based studies represent case–control analyses derived from cancer registries and/or kin-cohorts (11, 12). Often, such genetic studies overestimate risk estimates through ascertainment bias (13). Risk estimates presently remain lacking from population-based, prospective cohorts of unrelated individuals yet to develop cancer. As a result, present National Comprehensive Cancer Network (NCCN) guidelines only recommend more frequent screening for monoallelic carriers with personal or first-degree family histories of colorectal cancer (14). In addition, many practitioners default to current United States Preventative Task Force (USPTF) guidelines for older adults, which suggest individualized recommendations according to general health and prior screening results (15). Thus, robust risk estimates to inform screening strategies in older carriers are lacking, but needed given the increased colorectal cancer risk and diminished responses to DNA-damage in older persons (16, 17). This uncertainty also extends to extracolonic cancer risk among older carriers, with several studies demonstrating a nonstatistically significant trend for increased risk for several other tumor types (18, 19).

An unresolved aspect remains; the mechanism by which monoallelic MUTYH mutations predispose to cancer. Tumors of individuals with biallelic mutations arise from accumulation of genome-wide somatic G:C>T:A mutations in oncogenes and tumor suppressors. Among carriers, it remains unclear whether loss of heterozygosity (LOH) mediates this risk, and if these mechanisms are uniform among different tumor types. With the increasing availability of cancer genomic sequencing data, analyses of single base substitutions (SBS) have identified mutational signatures associated with MUTYH deficiency (1, 20). Colorectal cancers lacking MUTYH function demonstrate an abundance of COSMIC SBS signature 36, a measure of G:C>T:A mutagenesis found in colorectal and extracolonic cancers from those with biallelic MUTYH loss. COSMIC SBS signature 18, which is closely related to COSMIC Signature 36, has been noted in extracolonic malignancies possessing an enrichment of G:C>T:A transversions in the setting of NpCpA or NpCpT sequences due to oxidative damage (21). Somatic LOH has been reported in case series from tumor samples of MUTYH carriers affected by colorectal cancer, PNET, adrenocortical carcinoma, pancreatic ductal adenocarcinoma, and breast cancer, resulting in detection of COSMIC Signatures 18 and 36 (18, 20, 22).

To address some of these outstanding questions regarding monoallelic MUTYH risk and mechanisms of tumor evolution, we performed targeted high coverage germline DNA sequencing of a large population-based prospective cohort and evaluated the risk of monoallelic pathogenic mutations in MUTYH on cancer incidence. We analyzed Australian participants enrolled in the ASPirin in Reducing Events in the Elderly (ASPREE) clinical trial who represent healthy, community-dwelling individuals age ≥70 without dementia, disability, or preexisting disease expected to cause mortality within 5 years of enrollment (23). To elucidate possible mechanisms by which cancers from MUTYH monoallelic carriers evolved, we calculated SBS signatures from an initial pool of 13,033 cancers [The Cancer Genome Atlas (TCGA) and the International Cancer Genome Consortium (ICGC)] and compared with the relative contributions of SBS18 and SBS36 among monoallelic MUTYH carriers and those without such germline alterations.

Details of the ASPREE cohort and clinical trial have been described previously (23, 24). Briefly, ASPREE was a double-blind, randomized, placebo-controlled trial aimed at determining the effect of low-dose aspirin on all-cause mortality, disability-free survival, and cardiovascular events in healthy older (≥70 years of age) individuals (23–26). A total of 19,114 individuals in the ASPREE study were followed prospectively after enrollment and monitored for a number of primary and secondary outcomes, including cancer incidence and cancer-related mortality. Demographic, lifestyle, clinical, and other health data from ASPREE participants were collected at baseline and at annual in-person study visits. At study enrollment and updated annually, participants provided self-reported personal medical history related to colon polyps (consisting of yes/no/unsure and number of polyps), cancer history (consisting of yes/no/unsure and cancer type), and cancer screening tests (by type).

Study participants were enrolled from March 2010 to December 2014, with follow-up data collected until June 2017 (mean follow-up 4.7 years). Of the total cohort 13,806 Australian study participants consented for genetic study, of those 13,131 germline DNA samples were available for sequencing (via stored biospecimens collected through the ASPREE Healthy Ageing Biobank). Differences between the sequenced cohort and the larger ASPREE cohort are presented in Supplementary Table S1. This study was approved by the human research ethics committee at the Alfred Hospital in Melbourne, Australia in accordance with Declaration of Helsinki. Informed written consent was obtained for all enrolled study participants.

Sample DNA was diluted to 5 ng/μL and libraries were prepared with the included Ion AmpliSeq Library Kit Plus reagents and the Ion Xpress Barcode Adapters 1–16 Kit (Thermo Fisher Scientific #4471250) to barcode each pair of samples, following standard manufacturer's instructions. Briefly, 10 ng of DNA was added to two master amplicon mixes, one for each primer pool of the DNA panel and the two primer pool reactions 54 were amplified in different wells of the same plate following the specified cycling conditions in the user guide. Following PCR amplification, the DNA amplification reactions from the two DNA primer pool reactions were 56 combined in wells and processed as follows. The primers were partially digested with the proprietary FuPa enzyme and each sample was barcoded with a unique IonExpress barcode. Finally, a 1.5× bead purification was performed with Agencourt AMPure XP Reagent (Beckman Coulter, #A63880) following the instructions in the protocol to clean up the sample and remove adapter dimers. At the end of the cleanup protocol, the samples were eluted and amplified with a master mix of library amplification mix and Equalizer primers provided in the Ion Library Equalizer Kit (Thermo Fisher Scientific, #4482298). The amplified libraries were then incubated and washed using Equalizer kit reagents following the protocol, resulting in all the libraries normalized to ∼100 pmol/L.

After priming the Ion Torrent chip per manufacturer's protocol, samples were sequenced on the Ion Torrent S5TM XL System with library read length set at 400 bp and 850 flows per chip, with all other instrument settings set to the manufacturer's default for the Ion 510/520/530 Kit. Analyses of sequencing raw data were performed with Ion Torrent Suite (version 5.6.0) using 81 the “coverageAnalysis,” “sampleID,” “variantCaller” (with somatic/low stringency settings for the “variantCaller”), and custom BED file plugins, with all other settings for the run report set to the manufacturer's default.

Torrent Suite 5.2 on Ion Torrent S5XL produced raw reads (average 200× coverage) and exported them as fastQ format. These raw reads on a per sample basis, were aligned to the human genome reference version GRCh37. Aligned bam files were passed onto the Torrent Variant Calling Suite for low-stringency variant calls. Variant calling was performed on each sample separately, which produced a single sample VCF file. Post variant calling, all per-sample VCF files were merged using bcftools merge option to produce a single multisample vcf file.

Variants were flagged for average quality scores <30 across the dataset, low reads (<50×), minor allele frequency, and/or homozygous genotype frequency >10× higher than gnomAD (Non-Finish Europeans), visible evidence of misalignment and/or errors around the variant site, evidence of allele-specific amplification bias, imbalanced stand coverage, or variant allele frequency (<25%), direct adjacency to homopolymers, and evidence of multi-allelic base calls. Variants with three or more flags were excluded as technical sequencing artifacts. Variants called as borderline were flagged for validation by Sanger sequencing. Study participants harboring rare, loss-of-function variants (nonsense, splice-site, or frameshift mutations presumed to cause nonsense mediated decay), or missense variants labeled as pathogenic in ClinVar were considered as MUTYH carriers.

BAM slices encompassing the MUTYH gene were individually downloaded, and analyzed with the Genome Analysis Toolkit Best Practices (27). TCGA samples were analyzed with the reference genome GRCh38, whereas ICGC were analyzed with GRCh37. Annotation was performed using SNPeff (28). Loss-of-function variants were defined as nonsense, essential splice-site, or frameshift mutations occurring 50 bp prior to the last exon–exon junction. Pathogenic missense variants were identified by those annotated in ClinVar (29). All qualifying variants were inspected through IGV to ensure no sequencing artifacts or homopolymer errors were evident (30).

Somatic mutational signatures were calculated from whole-exome and whole-genome sequencing data processed by the MuTect somatic variant caller (31). Mutect2 processed VCF files of TCGA samples were downloaded from the National Cancer Institute Genomic Data Commons, and International Cancer Genome Consortium (ICGC) samples were downloaded from the Ontario Institute for Cancer Research Collaboratory Repository and analyzed through a nonnegative matrix factorization algorithm to decompose mutational spectra. The DeconstructSigs package in R version 3.4 was used to compare components in the context of the reference signatures identified from the Wellcome Trust Sanger Institute (WTSI) Mutational Signature Framework version 3 (32, 33). For TCGA and ICGC samples, the germline variants in MUTYH were called from BAM file slices of the gene downloaded from their respective repository. The tumor types analyzed, and their respective repositories are presented in Supplementary Table S2. OncoKB was used to identify catalogued driver somatic mutations that have been previously described as oncogenic (34).

Comparison of continuous and categorical variables between MUTYH gene variant carriers and noncarriers were tested with the Student t test and Chi-squared test, respectively. OR and 95% confidence intervals (CI) were calculated for the association between pathogenic MUTYH mutations and history of all cancers, colonic polyps, and colorectal cancer. The crude and the adjusted OR were estimated using univariate and multivariable logistic regression analysis. Hierarchical logistic regression methods were used to explore the associations between history of prior cancers and MUTYH carrier status. Model 1 included pathogenic MUTYH variant status as a risk factor. Model 2 included pathogenic MUTYH variant status, as well as age, sex, smoking status, and alcohol use.

Incidence of any cancer (excluding nonmelanoma skin cancers) and colorectal cancer were calculated by dividing the number of individuals who had developed the incident case by years at risk during the study follow-up. Factors associated with incident case were analyzed with Cox regression model, which included only those factors that were significant in the univariate analysis. We used Schoenfeld residual tests to evaluate the proportional hazard assumption and checked interactions between variables. This model generates regression coefficients for the independent variables, the exponents of which reflect hazard ratios. Model 1 consisted of a univariate analysis of pathogenic MUTYH variant status stratified by sex. Model 2 also incorporated age, smoking status, and alcohol use; Model 3 additionally incorporated baseline BMI and randomization to aspirin use; and Model 4 further incorporated history of cancer (for all cancer and extracolonic cancers), history of colorectal cancer (for incident colorectal cancer), and history of colonic polyp (for incident colorectal cancer). Aspirin use was incorporated into the latter models given that the results of ASPREE demonstrated an increase in cancer-related mortality associated with randomization to aspirin (23). All epidemiologic data were analyzed with Stata software, version 15.1 and two-sided P values less than 0.05 were considered statistically significant.

Outliers for SBS18 and SBS36 were prospectively designated as cancers with either greater than 50% of mutational fractions attributable to SBS18/SBS36, or with contributions of these signatures within the top quartile of tumor type and absolute contribution of at least 10%.

TCGA dataset is available through dbgap accession phs000178.v11.p8. International Cancer Genome Consortium datasets can be accessed through https://daco.icgc.org/. The human sequence data generated in this study are not publicly available due to patient privacy requirements but are available upon reasonable request from the corresponding authors. Other data generated in this study are available within the article and its supplementary data files.

Two-hundred and sixteen individuals among the 13,131 sequenced study participants (1.64%) harbored monoallelic pathogenic variants in MUTYH as detailed in Supplementary Table S3. No individuals with biallelic pathogenic mutations were discovered. No monoallelic carriers harbored concurrent mutations in Lynch Syndrome (35). Table 1 gives the characteristics of sequenced participants stratified by monoallelic MUTYH carrier status, which were balanced among carriers and noncarriers. Similar percentages were observed for carriers and noncarriers with prior history of colonic polyps (16.7% vs. 20.2%), prior history of colon cancer (1.8% vs. 2.6%), and prior history of any cancer (19.4% vs. 19.8%). Given the varied colorectal and extracolonic cancer risk estimates up to the age of 70 previously reported for MUTYH carriers, we performed a multivariable logistic regression analysis adjusting for age, sex, smoking status, alcohol use, and baseline BMI as presented in Supplementary Table S4. Despite this adjustment, no significant differences were observed between carriers and noncarriers for prior histories of colonic polyps, past history of colorectal cancer, or any cancer type.

During the study follow-up period, 1,328 incident cancers developed among 10% of sequenced study participants, of which 177 (13.3%) were colorectal cancers. The distribution of extracolonic cancers by sex and genotype are presented in Table 2. Among male MUTYH carriers, we observed a significantly increased risk for incidence of any cancer type (HR, 1.66; 95% CI, 1.04–2.66; P = 0.035; Table 3) and colorectal cancer incidence (HR, 3.50; 95% CI, 1.43–8.79; P = 0.006) in univariate analysis. Adjustment for age, smoking status, alcohol use, baseline BMI, cancer history, colorectal cancer history, polyp history, and randomization to aspirin did not alter the increased risk seen with any cancer type (multivariable HR, 1.66; 95% CI, 1.03–2.68; P = 0.038) or colorectal cancer (multivariable HR, 3.55; 95% CI, 1.42–8.78; P = 0.007). To determine if the total incident cancer risk was primarily driven by incident colorectal cancers, we calculated the incident risk of extracolonic cancers and observed no significant increase among monoallelic carriers (multivariable HR, 1.40; 95% CI, 0.81–2.44; P = 0.229). In contrast, female MUTYH carriers did not demonstrate any increased risk in overall cancer incidence (multivariable HR, 1.45; 95% CI, 0.84–2.51; P = 0.187), colorectal cancer incidence (multivariable HR, 0.81; 95% CI, 0.11–5.82; P = 0.113), or extracolonic cancer incidence (multivariable HR, 1.55; 95% CI, 0.88–2.75; P = 0.130).

Given the sex-dependent difference observed in risk of incident overall cancer and colorectal cancer, we next hypothesized that MUTYH LOH resulting in the widespread G:C>T:A transversions, as seen in colorectal cancers associated with MAP, was unlikely the predominant mechanism of tumorigenesis seen in cancers from MUTYH carriers. To test this hypothesis, we analyzed SBS signatures in MUTYH carriers from TCGA and ICGC cancer samples (Fig. 1). Among 10,483 individuals from 33 different TCGA cancer projects, and 2,550 participants from 26 ICGC cancer projects (Supplementary Table S2), MUTYH monoallelic carriers were found among 185 TCGA participants (1.76%) and 51 ICGC participants (2.0%). In total, 41 different TCGA and ICGC cancer projects with MUTYH carriers lacked any measurable contribution from SBS18 or SBS36, and did not depend upon sex. Of the 72 tumors from MUTYH carriers with appreciable SBS18 or SBS36 components, only 21 cancers were found to either have greater than half of the mutations attributable to SBS18/36 or have mutation fractions of SBS18/36 of at least 10% that were in the top quartile in their respective tumor type. The age distributions of TCGA and ICGC study participants in the context of SBS18/36 contributions are presented in Supplementary Fig. S1.

No colorectal or gastric cancers from MUTYH carriers displayed widespread G:C>T:A alterations similar to those seen in MAP or possessed key early oncogenic events driven by such mutations. Furthermore, no differences in SBS18 or SBS36 were noted between males and females in cancer types, including colorectal or gastric cancers (Supplementary Figs. S2 and S3). Among the cancer types with detectable SBS18 or SBS36, considerable intertumor heterogeneity existed among the mutation fractions attributable to these signatures (Fig. 2A and B). Although some breast, colon, prostate, pancreatic adenocarcinoma, and ovarian cancers from MUTYH carriers demonstrated outlier contributions of SBS18 or SBS36 in their respective tumor type, none of these cancer types demonstrated genome-wide G:C>T:A transversions to the magnitude seen in colorectal cancers arising from individuals who inherit biallelic mutations. However, a few tumor types displayed consistent and widespread G:C>T:A transversions to such levels. Three out of four MUTYH carriers with pancreatic neuroendocrine tumors, and 2/2 MUTYH carriers with adrenocortical carcinomas demonstrated high fractions. Cases with high fractions consisted of males and females in each tumor type. In addition, all recognizable oncogenic events for these tumors, as identified by OncoKB, occurred via G:C>T:A mutations (Supplementary Table S5). Among gastrointestinal epithelial tumors, only one esophageal adenocarcinoma demonstrated extensive G:C>T:A alterations, with large contributions from both SBS18 and SBS36. Examination of the oncogenic mutations allowed us to infer that LOH likely occurred after transformation into adenocarcinoma (Supplementary Fig. S4). The oncogenic events associated with the development of Barrett esophagus with high-grade dysplasia and early adenocarcinoma (KRAS and TP53) were not induced by G:C>T:A alterations, whereas a pathogenic mutation in SMAD4 that typically arises after the development of invasive adenocarcinoma appeared to be induced by such a transversion (36).

In this study, we demonstrate through a large population-based, prospective cohort of community-dwelling adults that older, male MUTYH carriers are at an increased risk for the development of cancer driven primarily by an increased risk of colorectal cancer. To further investigate the mechanisms of this sex-dependent risk, we examined the mutational signatures of cancers from MUTYH carriers deposited in the TCGA and ICGC repositories. We discovered that the widespread G:C>T:A transversions seen in colorectal cancers arising from biallelic inheritance of pathogenic MUTYH mutations are largely absent in the cancers of carriers, particularly of those of gastrointestinal epithelial origin. These novel epidemiologic and genomic findings provide further insight into the risks, as well as the heterogeneous mechanisms of carcinogenesis, associated with inherited monoallelic disruption of MUTYH.

Previously reported colorectal cancer risk estimates in carriers have been on the basis of case–control analyses derived from individuals and family members enrolled in colon cancer registries with matched controls or kin-cohort studies. Although risk estimates derived from these studies have prompted the NCCN to issue enhanced colorectal cancer screening recommendations for only those carriers with family histories of colorectal cancer, there is little guidance for monoallelic carriers at age 75 or over. In this study, we present the first reported risk estimates from a population-based, prospective cohort selected prior to the development of cancers, and our data suggest that healthy, older male MUTYH carriers represent a high-risk group that may benefit from continued colorectal cancer screening beyond age 75.

Although we report a strong sex-dependent cancer risk among older MUTYH carriers, a trend of much smaller magnitude has been observed in prior case–control studies (10). Moreover, the elevated risk for gastrointestinal cancers among males has been noted in other cancer-predisposing syndromes such as Lynch Syndrome, and observed in esophageal adenocarcinoma that typically afflict older adults (37–39). In our study, we incorporated several extrinsic exposures in our analyses that are known to differ by sex, including history of polyps, BMI, smoking, and alcohol use. Adjusting for these covariates, we nevertheless observed a sex-dependent risk.

Given these results, we next hypothesized that the mechanisms for carcinogenesis among MUTYH carriers were likely pleiotropic, and not uniform to base-excision repair. To begin to answer this question, we identified and analyzed cancers from all MUTYH carriers present in two large cancer genome repositories (TCGA and ICGC), thereby performing the largest analysis known to date. Cancers from those who inherit biallelic mutations display G:C>T:A transversions early in development. In the cases of colorectal cancers from patients with MAP, APC and KRAS are uniformly mutated by these transversions (3, 40). LOH has been the presumed mechanism in monoallelic carriers, given the frequent loss of chromosome 1p observed in colorectal cancers; however, these prior studies were contaminated by analyzing large numbers of nonpathogenic, common variants in MUTYH prior to large-scale genetic epidemiology studies (41, 42).

In our analysis, we observed that in carriers, no cancers arising from a gastrointestinal epithelial origin demonstrated widespread G:C>T:A transversions that were initiated early in tumorigenesis due to LOH. However, in select cancer types, such as adrenocortical carcinomas, pancreatic neuroendocrine tumors, and thymomas, widespread G:C>T:A transversions occurred early in carcinogenesis, and contributed to a similar fraction of the overall mutagenesis seen in MAP-associated colorectal cancers. These biallelic alterations were seen in men and women, and at variable ages. The large differences in G:C>T:A transversions among tumor types suggests that cell intrinsic factors, along with functions other than base-excision repair are likely relevant in the ability of MUTYH to suppress tumor formation. MUTYH has been linked to DNA-damage signaling, PARP1 activation, and SIRT6 activity; many of these processes have notable sex-dependent effects (43, 44). Additional studies are required to examine the mechanisms by which aging, sex, and MUTYH monoallelic disruption interact to promote carcinogenesis.

Our study has several strengths. The incident risk estimates associated with MUTYH carrier status are derived from a large, healthy, older, population-based, prospective cohort with detailed covariate data. In addition, the genomic analyses of cancers tissues from MUTYH gene carriers are the largest yet performed given the size of TCGA and ICGC cohorts. The genomic analyses identified which oncogenic driver changes were specifically induced by the G:C>T:A transversions on a genome-wide scale; prior studies have used LOH of MUTYH to infer pathogenicity, which fails to distinguish its impact as an initiating, driver event or later-stage passenger alteration. Several limitations of this study should also be noted. Medical history prior to enrollment was self-reported. Second, ASPREE participants represent a healthy subset of the older population, and thus any lack of association with history of prior cancers before enrollment may reflect selection bias of healthy volunteers. Also, we lack granular data on colon cancer screening prior to enrollment. As a result, it may be possible that the lack of risk seen in elderly, female MUTYH may be attributable to higher overall higher rates of participation in colorectal cancer screening in women. Finally, although the number of the incident number of colorectal cancers among MUTYH carriers may appear small, the deep phenotypic and longitudinal characterization of participants and cohort selection prior to the development of cancers ensure that risk estimates are less likely affected by ascertainment and selection biases.

The decision to pursue and cease colorectal cancer screening in older individuals is a complex, decision-making process that balances patient preferences, medical history, functional status, life expectancy, and the risk of colorectal cancer that increases with age (45, 46). In this study, we identified a promising genetic biomarker that is strongly associated with increased colorectal cancer risk among otherwise healthy, older men. Given its prevalence in the European population, monoallelic, pathogenic MUTYH mutations are among the most frequent genetic alterations identified during genetic testing for cancer predisposition. Additional studies will be required to establish if enhanced or continued colorectal cancer screening for older, male MUTYH carriers results in decreased cancer incidence and mortality.

J.M. Downie reports grants from American College of Gastroenterology during the conduct of the study. A.T. Chan reports grants and personal fees from Pfizer Inc., Bayer Pharma AG, and Boehringer Ingelheim outside the submitted work. S.G. Orchard reports grants from National Institute on Aging, NCI, National Health and Medical Research Council of Australia, and Victorian Cancer Agency, Australia, during the conduct of the study. R.P. Sebra reports other support from Sema4 outside the submitted work; and Robert Sebra is VP of Technology Development at Sema4. A.M. Murray reports grants from NIH during the conduct of the study. F. Macrae reports Manager, InSiGHT DNA variant databases and Member, InSiGHT ClinGen MUTYH Variant Curation Expert Panel. E. Schadt reports other support from Sema4 outside the submitted work. R.L. Woods reports grants from NIH and NHMRC during the conduct of the study. M. Gala reports grants from American College of Gastroenterology and National Institute of Diabetes and Digestive and Kidney Diseases during the conduct of the study; other support from New Amsterdam Genomics, Inc outside the submitted work. No disclosures were reported by the other authors.

J.M. Downie: Formal analysis, investigation, writing–original draft, writing–review and editing. M. Riaz: Formal analysis, investigation, writing–original draft. J. Xie: Data curation, formal analysis, supervision, investigation, methodology. M. Lee: Resources, data curation, formal analysis, investigation, project administration. A.T. Chan: Resources, writing–review and editing. P. Gibbs: Resources, writing–review and editing. S.G. Orchard: Writing–original draft, writing–review and editing. S.E. Mahady: Writing–review and editing. R.P. Sebra: Formal analysis, investigation, writing–review and editing. A.M. Murray: Funding acquisition, writing–review and editing. F. Macrae: Writing–review and editing. E. Schadt: Formal analysis, funding acquisition, writing–review and editing. R.L. Woods: Funding acquisition, project administration, writing–review and editing. J.J. McNeil: Resources, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing. P. Lacaze: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing. M. Gala: Conceptualization, resources, data curation, supervision, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing.

We thank the trial staff in Australia and the United States, the participants who volunteered for this trial, and the general practitioners and staff of the medical clinics who cared for the participants. This study was funded by the NIH (U01AG029824 to J.J. McNeil and A.M. Murray, K23DK103119 to M. Gala, and U19AG062682 to R.L. Woods, A.M. Murray, and J.J. McNeil), American College of Gastroenterology (Resident Clinical Research Award to J.M. Downie), the National Health and Medical Research Council of Australia (grant nos. 334047 to J.J. McNeil and 1127060 to R.L. Woods and J.J. McNeil), Monash University (Australia) to R.L. Woods, J.J. McNeil, and P. Lacaze and the Victorian Cancer Agency (Australia) to R.L. Woods, J.J. McNeil, and P. Lacaze. The ASPREE Healthy Ageing Biobank was supported by a Flagship cluster grant (including the Commonwealth Scientific and Industrial Research Organization, Monash University, Menzies Research Institute, Australian National University, and University of Melbourne) to J.J. McNeil and P. Lacaze.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.