Background Patients with colorectal cancer (CRC) with mismatch repair-deficient (dMMR) tumours without MLH1 methylation or germline MMR pathogenic variants (PV) were previously thought to have Lynch syndrome (LS). It is now appreciated that they can have double somatic (DS) MMR PVs. We explored the clinical characteristics between patients with DS tumours and LS in two population-based cohorts.
Methods We included patients with CRC from Ohio 2013–2016 and Iceland 2000–2009. All had microsatellite instability testing and/or immunohistochemistry (IHC) of MMR proteins, and MLH1 methylation testing when indicated. Germline next-generation sequencing was performed for all with dMMR tumours; tumour sequencing followed for patients with unexplained dMMR. Clinical characteristics of DS patients and patients with LS were compared.
Results Of the 232 and 51 patients with non-methylated dMMR tumours in the Ohio and Iceland cohorts, respectively, 57.8% (n=134) and 45.1% (n=23) had LS, 32.8% (n=76) and 31.4% (n=16) had DS PVs, 6% (n=14) and 9.8% (n=5) were unexplained and 4.3% (n=10) and 13.7% (n=7) had incorrect IHC. Age of diagnosis for DS patients was older than patients with LS (p=3.73×10−4) in the two cohorts. Patients with LS were more likely to meet Amsterdam II criteria (OR=15.81, p=8.47×10−6) and have multiple LS-associated tumours (OR=6.67, p=3.31×10−5). Absence of MLH1/PMS2 was predictive of DS PVs; isolated MSH6 and PMS2 absence was predictive of LS in both cohorts.
Conclusions Individuals with LS are 15× more likely to meet Amsterdam II criteria and >5× more likely to have multiple cancers as compared with those with DS tumours. Furthermore, isolated loss of MSH6 or PMS2 protein predicts LS.
- DNA repair system
- somatic mutation
- tumour testing
- Lynch-like syndrome
Statistics from Altmetric.com
Mismatch repair (MMR)-deficient colorectal cancer (CRC) accounts for 12%–15% of all localised CRC and 3%–4% of metastatic CRC.1 2 The primary cause of MMR-deficient CRC is acquired methylation of the MLH1 promoter.3 4 Identification of patients with MMR-deficient CRC via universal tumour screening is critical, as MMR deficiency provides insight into hereditary cancer risk for the patient and family members, as well as indicating which patients might benefit from immunotherapy.5–7 Individuals with MMR-deficient CRC can have Lynch syndrome (LS), caused by a germline pathogenic variant (PV) in any of the MMR genes: MLH1, MSH2 (EPCAM), MSH6 or PMS2. Individuals with LS have a significantly increased risk for developing cancers of the colon, endometrium, ovary, stomach and others.8 9 Identifying individuals with LS prevents future cancers, and provides information for family members by facilitating cascade testing and life-saving intensive surveillance and prophylactic surgeries among those who inherited LS.
Previous studies have shown that 2.5%–3.9% of CRC has unexplained MMR deficiency.4 5 10 Until recently, those patients were thought to have LS caused by an undetected germline MMR PV. In 2013, the term ‘Lynch-like syndrome’ (LLS) was coined to describe this indeterminate group of patients. The Spanish EPICOLON Consortium assessed 1705 unselected patients with CRC for family history of CRC and found that the incidence of CRC was lowest in sporadic CRC families, or those with proficient MMR (0.48), highest in LS families (6.04) and moderately elevated in LLS families (2.12).10 It was suggested that the LLS group was likely heterogeneous, including those with sporadic CRC and those with undetected LS, given the intermediate risk for CRC.11
It is now appreciated that many patients with LLS actually have double somatic (DS) MMR PVs in their tumour. While it has not been definitively proven that these tumours are biallelic, it has been presumed given the tumour phenotype in these patients (microsatellite instability-high (MSI-H) and abnormal immunohistochemistry (IHC)). In 2013, one study showed that 16.7% of patients with unexplained MMR-deficient CRC had DS MMR PVs, with an additional 27.8% of patients having one MMR PV identified, but loss of heterozygosity (LOH) was not assessed so many of the cases with one somatic MMR PV were likely DS with LOH as the second hit.12 This study also identified one patient (5.5%) with somatic mosaicism.12 In 2014, three additional studies were performed which determined that 52%–69% of unexplained MMR-deficient CRC was caused by DS MMR PV when both sequencing and LOH were evaluated.13–15 Other identified causes included previously undetected germline MMR PVs and incorrect tumour screening.13–15 To further characterise those with DS tumours, Mensenkamp et al reported that the age of patients with DS PVs was similar to that of patients with LS (p=0.055), and lower than that of patients with MLH1 methylation (p<0.0001).13 Other studies have shown that patients with DS MMR PVs can still have hereditary CRC caused by MUTYH-associated polyposis and other genes involved in DNA repair since they can lead to acquired PVs in the MMR genes.16–18 A recent study showed that the histology of DS CRC tumours is indistinguishable from those caused by LS.19
We sought to define the clinical characteristics differentiating patients with DS tumours from two population-based cohorts, and compare them with patients with LS.
A total of 3471 adults newly diagnosed with primary invasive CRC in Ohio between 1 January 2013 and 31 December 2016 were prospectively enrolled into the Ohio Colorectal Cancer Prevention Initiative (OCCPI; ClinicalTrials.gov identifier: NCT01850654). Written informed consent was obtained from all participants. Of the 3471 patients enrolled, 118 were deemed ineligible and seven withdrew. Primary reasons for ineligibility included insufficient tumour material, ineligible pathology type, diagnosed outside of the qualifying study period and not being diagnosed in Ohio. Of the 3346 active and eligible patients, testing was successful for 3310. Methods have previously been described,17 but briefly, all tumours were screened for MMR deficiency by MSI testing and/or IHC analysis. MSI testing was completed using the Promega MSI Analysis System (version 1.2), which includes five repeat markers (BAT-25, BAT-26, NR-21, NR-24, MONO-27). Tumours with ≥2/5 markers showing instability were classified as MSI-H. IHC of the MMR proteins was performed using the two-stain method as previously described.20 Staining for all four MMR proteins was done as routine clinical care for some patients, and attempted for all patients if MSI could not be performed or if the MSI and two-stain IHC results were discordant. Antibodies included MLH1 clone: Leica ES05 (mouse: NCL-L-MLH1), MSH2 clone: Calbiochem FE11 (mouse: NA27), MSH6 clone: Epitomics EP49 (rabbit: AC-0047) and PMS2 clone: BD Pharmingen A16-4 (mouse: 556415). Proteins with convincing stain in >1% of cells, or equivocal staining, were considered ‘present’. Methylation of the MLH1 promoter was assessed at four CpG sites using pyrosequencing21 when tumours were MSI-H and/or absent MLH1 and PMS2 proteins on IHC, with ≥15% methylation classified as positive. Patients with MMR deficiency without MLH1 methylation underwent germline next-generation sequencing (NGS) (ColoSeq or BROCA, University of Washington (UW)). Genomic regions were captured using biotinylated RNA oligonucleotides (SureSelect) and sequenced on an Illumina HiSeq2000 instrument.22 Large rearrangements were detected.23 Tumour sequencing with ColoSeq Tumor of the MMR genes followed for patients with unexplained MMR-deficient tumours. Data were created by the NGS Laboratory and Analytics group. Pathology reports were reviewed for all patients. The patients’ cancer history and first-degree relative cancer history were obtained at study enrollment, and three-generation pedigrees were obtained (when possible) after result disclosure. Each pedigree was assessed for clinical characteristics and if Amsterdam II and Revised Bethesda criterion were met. PREMM5 scores were obtained for each patient using current age and known family history as of February 2018. The following stipulations were also applied: for deceased patients, age at death was used as the current age in the calculation. For cases without an exact age of diagnosis (range provided), the middle number was used (eg, 40s–50s was calculated using 50). If age of diagnosis was not known, the average diagnosis age for CRC and endometrial cancer (EC) from the American Cancer Society was used (age 60 for EC, age 72 for women with CRC, age 68 for men with CRC). Some clinical characteristics of 51 cases have been previously reported.15 17
A total of 1182 patients with CRC diagnosed from 2000 to 2009 in Iceland were included. Written informed consent was obtained from all subjects participating in research studies at deCODE Genetics. Methods have been previously published,24 but briefly, IHC for all four MMR proteins was performed on tissue microarrays (two cores of 1 mm per tumour) using primary antibodies for MLH1 (Novocastra, Buffalo Grove, IL; NCL-L-MLH1; clone: ESO5; diluted 1:500), MSH2 (Calbiochem (Merck Biosciences), Basel-Land, Switzerland; NA-27; clone: FE11; diluted 1:3000), MSH6 (Epitomics, Burlingame, CA; AC-0047; clone: EP49; diluted 1:800) and PMS2 (BD Pharmingen, San Jose, CA; 556415; clone: A16-4; diluted 1:300). If the tumour was absent for MLH1/PMS2 immunostaining, MLH1 methylation testing was performed by pyrosequencing using the Pyromark Q96 CpG MLH1 kit (QIAGEN, Hilden, Germany) with ≥15% methylation classified as positive. All patients with abnormal IHC and no MLH1 methylation underwent germline testing for MMR variants found by genome sequencing (GS) of 8435 Icelanders. If no MMR PVs were identified, whole genome sequencing (WGS) was performed on blood samples with Illumina technology. In cases where blood DNA was not available, DNA from archived formalin-fixed paraffin-embedded normal tissue was subjected to Sanger sequencing of the MMR genes. Tumour sequencing using ColoSeq Tumor was performed in MMR-deficient cases that remained unexplained after negative germline testing and MLH1 methylation analysis. All Icelandic cases were previously reported.24 In cases of equivocal MMR staining with an identified germline PV, ColoSeq Tumor was done to determine the second pathogenic hit to the same MMR gene. The patient’s cancer history and first-degree relative cancer history were obtained from deCODE Genetics and the Icelandic Cancer Registry, and three-generation pedigrees were created. Each pedigree was assessed for clinical characteristics and if Amsterdam II and Revised Bethesda criterion were met. PREMM5 scores were obtained for each patient using current age at study enrolment or age at death if patient was deceased and known family history as of February 2018.
Classification of mutations for both cohorts
Our approach to MMR variant interpretation has been described previously.15 17 25 The Clinical Laboratory Improvement Amendments-approved laboratory (UW) adjudicated the pathogenicity of all germline mutations using criterion established by the American College of Medical Genetics and the International Agency for Research on Cancer guidelines.26 27 All variants were reviewed by at least two clinical lab directors prior to interpretation. For tumour sequencing, cases were considered DS if two pathogenic or likely pathogenic somatic mutations were identified or if one pathogenic or likely pathogenic somatic mutation was identified with associated LOH. For patients with MMR-deficient tumours and a germline MMR variant of uncertain significance (VUS), tumours were assessed for additional MMR mutations or LOH to attempt to clarify the pathogenicity of the variant. Variants were reclassified as likely pathogenic when tumour screening results supported pathogenicity and one additional pathogenic mutation was identified in the tumour using methods previously described.15
Statistics for both cohorts
Descriptive statistics were provided. Continuous data were tested for normality with Shapiro-Wilk test and for homogeneity of variances with Bartlett test. When the above assumptions are satisfied, multiple regression analysis was used to analyse continuous data. Non-parametric Wilcoxon rank-sum tests were used when the assumptions are violated. Pearson’s χ2 tests with continuity correction or Fisher’s exact tests were used to analyse dichotomous variables. Meta-analysis, to combine results from two cohorts, was performed by applying Mantel-Haenszel test and fixed-effect model with inverse variance method for dichotomous and continuous variables, respectively. All tests were two sided, and level of significance was set at 0.05. Of note, there are two pairs of patients with CRC in the Ohio cohort who are first-degree relatives (mother–daughter pair and sibling pair). Two of these patients (one from each pair) were excluded from statistical analysis as their close relation violated assumptions of the tests.
Of the 232 patients with an MMR-deficient tumour without methylation, 57.8% (n=134) had LS, 32.8% (n=76) had DS PVs, 6% (n=14) were unexplained and 4.3% (n=10) had incorrect IHC (table 1). Of patients with a true MMR-deficient tumour without methylation or a germline MMR PV in the gene corresponding with their absent protein in cases with abnormal IHC, 82% (73/89) have DS PVs, which is the highest percentage reported to date. The majority (60%) of patients with absence of MLH1/PMS2 on IHC had DS PVs, and the other staining patterns (absence of MSH2/MSH6, isolated absence of MSH6, isolated absence of PMS2, normal IHC with an MSI-H tumour) were more predictive of LS (table 1). Two patients are counted twice in table 1: one has both LS (germline PMS2 PV) and DS MSH6 PVs (case 409149), and one has both LS (germline MLH1 PV) and unexplained absence of MSH6 on IHC (case 417591). These two patients are only counted once in table 2, under the LS column. The clinical characteristics of patients with LS compared with DS and unexplained patients are presented in table 2; germline and somatic mutations in the LS and DS cases are found in the online supplementary tables 1 and 2. The average age of diagnosis for DS patients was older than patients with LS (58.8 vs 52.4, p=1.68×10−3), but still younger than what would be expected in the general population.28 Compared with patients with DS PVs, patients with LS were more likely to have synchronous or metachronous LS-associated tumours (33.9% vs 4.5%, p=1.64×10−6), have a first-degree relative with CRC or EC (57.3% vs 18.2%, p=1.43×10−7), meet Amsterdam criteria II (25.8% vs 1.5%, p=4.40×10−6) and revised Bethesda criteria (87.1% vs 48.5%, p=2.47×10−8) and have a higher median PREMM5 score (9.4% vs 2.45%, p=6.97×10−12) (tables 2 and 3). Overall, 19 patients had an MMR-deficient CRC plus a germline PV in a non-MMR gene (eight LS, nine DS, two unexplained; online supplementary table 3). These patients were included in table 1 but not in table 2, as it is possible that their germline PV may have contributed to their clinical characteristics and family history (particularly for the DS and unexplained patients). Online supplementary table 4 details the cases in the unexplained group. Case 506793 had a germline VUS in the MMR gene that was consistent with their missing proteins on IHC, plus one somatic PV in their tumour suggesting that this VUS could be pathogenic. In the DS group, case 369991 had a germline VUS in the MMR gene that was consistent with their missing proteins on IHC, plus two clearly pathogenic somatic PVs in their tumour suggesting that this VUS could be benign. MMR IHC results corresponded to the affected MMR gene in all cases except for the 11 with intact IHC and five with mutations in the unexpected member of the heterodimer pair (eg, two cases with germline MSH6 mutations had absence of MSH2/MSH6 in their tumour, see online supplementary tables 1 and 2).
Of all MMR-deficient CRCs without methylation, 45.1% (n=23) had LS, 31.4% (n=16) had DS PVs, 9.8% (n=5) were unexplained and 13.7% (n=7) had incorrect staining (see table 1). Of patients with an MMR-deficient tumour without methylation or a germline MMR PV, 57.1% (16/28) have DS PVs. Absence of MLH1 or MSH2 on IHC was more predictive of DS than LS. Average age of diagnosis was not significantly different for DS compared with LS (69 vs 62, p=0.12; see table 2). Patients with LS were significantly more likely to have higher median PREMM5 scores (6.0% vs 2.2%, p=7.49×10−2) and were borderline statistically more likely to fulfil Amsterdam II (34.8% vs 6.7%, p=6.11×10−2) and revised Bethesda criteria (82.6% vs 60%, p=1.50×10−1) (table 3). There was no significant difference between metachronous (4.3% vs 0%, p=1) and synchronous tumours (8.7% vs 0%, p=5.09×10−1) in the two groups but a trend towards more first-degree relatives with CRC or EC was seen in the LS group (60.9% vs 40%, p=3.20×10−1). All patients with MMR-deficient CRC with DS PVs underwent germline GS and one patient was found to have a germline mutation in a non-MMR gene (CHEK2), as detailed in online supplementary table 3. This patient is included in table 1 but not in table 2. MMR IHC results corresponded to the affected MMR gene.
It is becoming increasingly recognised that DS MMR PVs represent a group of cancers distinct from LS with an MMR-deficient phenotype that is caused by somatic PVs. In this paper, we describe the clinical characteristics and family history of those with DS PVs and compare with those with LS from two large population-based studies that used similar algorithms for testing. We aimed to explore whether certain characteristics might predict LS over DS or vice versa (once MLH1 methylation has been ruled out as appropriate in MLH1/PMS2-deficient tumours).
Because the germline mutational landscape is so different between the two populations, it is important to compare the DS cohorts with the LS cohorts from that population. Therefore, our results are presented separately for each population, as well as combined, in tables 2 and 3. In the Ohio cohort, the most predictive characteristics of LS include meeting Amsterdam II and/or Revised Bethesda criterion, multiple primary tumours, PREMM5 score, a first-degree relative with CRC or EC, or having absence of MSH2/MSH6 on IHC or isolated absence of MSH6 or PMS2 on IHC. In the Icelandic cohort, the most predictive characteristics of LS include PREMM5 score, fulfilling Amsterdam II criteria and having isolated absence of MSH6 or PMS2 on IHC. See table 3 for a meta-analysis of the two cohorts. DS PVs were not as common as LS in either cohort. DS PVs were found in patients of all ages, ranging from 27 to 96 in the Ohio cohort and 41 to 88 in the Icelandic cohort with a median age higher than that of patients with LS.
In the Ohio cohort, younger age at diagnosis and having a metachronous or synchronous tumour was more predictive of LS, while the same was not true for the Icelandic cohort. Prior studies in Ohio have revealed that PV in MSH2 is the most common,5 and MSH2 PV carriers have a higher lifetime risk of cancers as well as a younger age at diagnosis as compared with MSH6 and PMS2 PV carriers. In Iceland, MSH6 and PMS2 PV are the most common with 96% of all LS-related CRC being related to these genes.24 Therefore, it is not surprising to see less metachronous and synchronous tumours as well as a higher age at diagnosis in patients with LS from Iceland given the reduced penetrance of those genes. It was rare to see synchronous or metachronous CRC in both the Ohio and Icelandic DS patients.
In both cohorts, DS PVs occurred more frequently in the MLH1 and MSH2 genes as compared with MSH6 and PMS2. However, due to differences in LS mutational landscapes, staining pattern predictability differs between the two populations. In Ohio, an absence of MLH1/PMS2 is more likely related to DS while MSH2/MSH6 absence is more likely related to LS. In Iceland, however, MLH1/PMS2 and MSH2/MSH6 absence is more likely to predict DS than LS. In both populations, sole absence of MSH6 or PMS2 is more likely to be due to LS.
Defining clinical characteristics between patients with LS and DS patients will be helpful in the clinical setting, especially for counselling patients who are identified with abnormal IHC during routine universal tumour screening (see table 1). For example, knowing that patients with absence of MLH1/PMS2 without MLH1 methylation and little family history are unlikely to have LS, clinicians could consider ordering paired tumour-normal NGS since the majority of these patients will have DS PVs. Likewise, those with positive family history and absence of MSH2/6, MSH6 or PMS2 could proceed with germline testing first given the higher likelihood of LS compared with DS mutations.
The tumour location bears a somewhat different resemblance in the two populations where 76.8% of tumours in the DS groups are right sided vs 59.1% in the LS groups. Dr. Lynch described years ago that LS-related tumours, while having a predilection for the right colon, could also arise on the left side and a rectal tumour location should not preclude the diagnosis of LS.29 Mas-Moya et al compared 45 patients with LS with 16 patients with LLS (presumably most of these were DS) and had similar findings with patients with LLS being more likely to have a right-sided location, less likely to have isolated MSH6 loss and less likely to have synchronous and metachronous tumours.30
PREMM5 scores were calculated for all patients with non-methylated MMR-deficient tumours included in table 2. PREMM5 scores were clearly higher in patients with LS in both cohorts (online supplementary tables 1 and 2). Of the Ohio patients, 87.9% of patients with LS had at least a 2.5% predicted probability of carrying an MMR PV, which is the risk deemed appropriate for referral for genetic evaluation by the model creators,31 while 50% of DS and 42.9% of unexplained patients had a score of ≥2.5. Of the Icelandic patients, 78.3% of patients with LS had a PREMM5 score of ≥2.5% while 46.7% of DS and 25% of unexplained patients had a score of ≥2.5.
Although DS MMR-deficient tumours are now well described as a separate entity from LS, it is still unclear what causes these tumours and a lingering question remains as to whether these patients could have an unidentified inherited PV in an MMR or another cancer susceptibility gene. This is crucial to determine, as the cancer screening programmes for those with DS MMR-deficient tumours, and their family members, could resemble those with sporadic CRC. Of note, all patients in the Ohio cohort underwent germline genetic testing with a panel of cancer susceptibility genes (ColoSeq, 12–22 genes or BROCA, 66 genes), with nine DS patients having mutations in other genes (two biallelic MUTYH, one RPS20, one GALNT12, one RAD51D and MUTYH heterozygote, two CHEK2, one NTHL1 heterozygote, one POT1). Of those nine patients, only two were diagnosed before age 50 and six did not have a family history concerning for hereditary predisposition. Similarly, all patients in the Icelandic cohort underwent GS and one DS patient was found to have a germline PV in another cancer susceptibility gene (CHEK2). This patient was under age 50 at diagnosis and had a PREMM5 score of 14.5. Over 10% (10/92) of DS MMR-deficient tumours in the two cohorts had a germline PV in a non-MMR cancer susceptibility gene. Aside from the two MUTYH-associated polyposis cases,16 it is not known whether these germline mutations contribute to the development of DS MMR-deficient tumours or are simply secondary findings. However, due to the high likelihood of a non-MMR cancer susceptibility PV being found in patients with CRC with DS PVs, a large panel of hereditary cancer susceptibility genes should always be considered to ensure patients with other syndromes are detected before establishing a sole diagnosis of DS PVs.
In our investigation of clinical characteristics, there are no findings that suggest that DS tumours are inherited. On the contrary, these individuals had less family history than those with LS, and very low rates of synchronous and metachronous CRC, and most do not fulfil Amsterdam II criteria. Of course, for DS patients who were diagnosed at a very young age or those who have a strong family history of cancer, an unknown hereditary syndrome cannot be ruled out and heightened surveillance may be warranted in those cases. The National Comprehensive Cancer Network (version 1.2018) recommends that patients with MMR-deficient tumours without a germline mutation be managed based on family history.32
In some cases, it is possible that DS tumours could be related to prior radiation or chemotherapy. A recent publication found DS MMR PVs in 70% of non-MLH1 methylated MMR-deficient CRC in Hodgkin’s lymphoma survivors from the Netherlands, which was higher than in a Dutch CRC population-based cohort.33 Interestingly, 18.2% of the Ohio DS patients and 50% of the Iceland DS patients had a history of other malignancies. At least three of eight patients in the Iceland cohort did receive intra-abdominal radiation (for prostate cancer, anal cancer and Hodgkin’s disease), so it possible that these were indeed treatment related.
The remaining group of unexplained patients is likely a heterogeneous mix of individuals with missed germline MMR PVs (LS) and missed somatic PVs (DS) (see online supplementary table 4). PREMM5 scores among the unexplained cases were higher than in DS, although we had few cases so no statistical comparisons were undertaken. We also included one patient in the unexplained group who had a germline VUS in the MMR gene that matched their abnormal IHC. This patient had just one somatic MMR PV in addition to the germline MMR VUS, so it is possible that this patient actually does have LS. In addition, the Iceland cohort reported the first case of LS due to a chromosome translocation involving the MLH1 gene, which cannot be detected by NGS.24 There are certainly other structural rearrangements like this and the Boland inversion (MSH2 exons 1–734) that are not being detected by our current testing methodologies, which could potentially explain some of these unanswered cases.
The strengths of this study include the large size of this cohort, which is the largest DS patient cohort presented to date, as well as the fact that these cases were unselectively obtained from large population-based studies that screened all CRC cases for MMR deficiency. Limitations include the fact that prior cancer and family history was obtained by patient report in the Ohio study, and the Icelandic study had few LS and DS cases, limiting the power of any statistical comparison between the two groups. In addition, MSI was not done on all patients in the Iceland study, so it is possible that there is an under-representation of unexplained MSI-H/intact IHC patients, as MSI was not assessed uniformly in that cohort.
In conclusion, we have shown that patients with DS MMR PVs do not appear to have features consistent with an inherited cancer syndrome, and assessing clinical characteristics such as family history, personal cancer history, PREMM5 scores and IHC staining patterns can help distinguish them from patients with LS.
The authors are grateful to the patients, families and OCCPI network of 51 participating hospitals. The authors also thank OSU undergraduate student interns Angela Onorato, James Miner, Jessica Purnell, Emily McDowell, Hannah Datz, Chloe Kent and Meghan Bartos, who assisted with database entry, as well as OSU graduate student Abigail Schaber who assisted with pedigree analysis. We thank all the staff in the University of Washington Genetics and Solid Tumors Laboratory who assisted with sequencing studies.
RP and SH contributed equally.
Contributors RP: study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; statistical analysis; study supervision. SH: study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; statistical analysis; obtained funding; study supervision. SL: critical revision of the manuscript for important intellectual content; statistical analysis. WLF, TR, CCP: analysis and interpretation of data; critical revision of the manuscript for important intellectual content; administrative, technical or material support. KS, JGJ: critical revision of the manuscript for important intellectual content; administrative, technical or material support. AdlC: study concept and design; critical revision of the manuscript for important intellectual content; study supervision. HH: study concept and design; acquisition of data; analysis and interpretation of data; critical revision of the manuscript for important intellectual content; statistical analysis; obtained funding; study supervision.
Funding The data reported here were derived from the Ohio Colorectal Cancer Prevention Initiative, which is supported by a grant from Pelotonia, an annual cycling event in Columbus, Ohio, that supports cancer research at The Ohio State University Comprehensive Cancer Center–James Cancer Hospital and Solove Research Institute. This study was supported in part by grant P30 CA016058, National Cancer Institute, Bethesda, MD.
Competing interests HH is on the scientific advisory board for InVitae Genetics and Genome Medical, has conducted collaborative research with Myriad Genetics Laboratories, Ambry Genetics and InVitae Genetics, and has stock in Genome Medical. RP has done collaborative research with Myriad Genetics Laboratories and InVitae Genetics. TR and KS are employees of deCODE Genetics/Amgen. SH, AdlC, JGJ, WLF, CCP and SL have no conflicts to disclose.
Ethics approval Institutional Review Board (IRB) approval for the Ohio Colorectal Cancer Prevention Initiative was obtained by the individual participating hospitals, Community Oncology Programs or by ceding review to the Ohio State University (OSU) IRB (2012C0123; IRB of record). The Iceland study was approved by the Icelandic National Bioethics Committee (VSNb2013010033/03.15), the Icelandic Data Protection Authority (2013010109TS) and the OSU IRB (2013C0144).
Provenance and peer review Not commissioned; externally peer reviewed.
Author note Rachel Pearlman and Sigurdis Haraldsdottir are co-first authors.
Patient consent for publication Not required.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.