Objective To estimate risk of colorectal cancer (CRC) for first-degree relatives of CRC cases based on CRC molecular subtypes and tumour pathology features.
Design We studied a cohort of 33 496 first-degree relatives of 4853 incident invasive CRC cases (probands) who were recruited to the Colon Cancer Family Registry through population cancer registries in the USA, Canada and Australia. We categorised the first-degree relatives into four groups: 28 156 of 4095 mismatch repair (MMR)-proficient probands, 2302 of 301 MMR-deficient non-Lynch syndrome probands, 1799 of 271 suspected Lynch syndrome probands and 1239 of 186 Lynch syndrome probands. We compared CRC risk for first-degree relatives stratified by the absence or presence of specific tumour molecular pathology features in probands across each of these four groups and for all groups combined.
Results Compared with first-degree relatives of MMR-proficient CRC cases, a higher risk of CRC was estimated for first-degree relatives of CRC cases with suspected Lynch syndrome (HR 2.06, 95% CI 1.59 to 2.67) and with Lynch syndrome (HR 5.37, 95% CI 4.16 to 6.94), but not with MMR-deficient non-Lynch syndrome (HR 1.04, 95% CI 0.82 to 1.31). A greater risk of CRC was estimated for first-degree relatives if CRC cases were diagnosed before age 50 years, had proximal colon cancer or if their tumours had any of the following: expanding tumour margin, peritumoral lymphocytes, tumour-infiltrating lymphocytes or synchronous CRC.
Conclusions Molecular pathology features are potentially useful to refine screening recommendations for first-degree relatives of CRC cases and to identify which cases are more likely to be caused by genetic or other familial factors.
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Family history of colorectal cancer (CRC) is a well-established risk factor for the disease, but even for specific family histories there is a wide variation in CRC risk limiting the effectiveness of CRC screening based on family history alone.
We hypothesised that molecular and pathology features of a CRC may be useful to estimate CRC risk for their relatives.
What are the new findings?
Compared with first-degree relatives of mismatch repair (MMR)-proficient CRC cases, a higher risk of CRC was estimated for first-degree relatives of CRC cases with suspected Lynch syndrome (HR 2.06, 95% CI 1.59 to 2.67) and with Lynch syndrome (HR 5.37, 95% CI 4.16 to 6.94), but not with MMR-deficient non-Lynch syndrome (HR 1.04, 95% CI 0.82 to 1.31).
A greater risk of CRC was estimated for first-degree relatives if CRC cases were diagnosed before age 50 years, had proximal colon cancer or if their tumours had any of the following: expanding tumour margin, peritumoral lymphocytes, tumour-infiltrating lymphocytes or synchronous CRC.
How might it impact on clinical practice in the foreseeable future?
Molecular and pathology characterisation of CRCs can help define the risk of CRC for first-degree relatives and therefore may be useful to refine screening recommendations for first-degree relatives of CRC cases and also to identify which cases are more likely to be caused by genetic or other familial factors.
Having a family history of colorectal cancer (CRC) is a well-established risk factor for the disease; people with one first-degree relative (parent, offspring, sibling) with CRC are, on average, twice as likely to be diagnosed with CRC than those without a family history.1 ,2 This has clinical implications for prevention as more intensive screening can be targeted to those who have a relative with CRC; many guidelines for CRC screening reflect this approach.3–6 In addition, CRC diagnosed in a setting of a family history is more likely to be due to an inherited mutation and, when identified through genetic screening, has significant and quantifiable risks for CRC and other cancers.7–9 However, there is substantial variation around this average risk of family history.10 For example, the younger the age at diagnosis of the affected relative and the more closely related the affected relative, the greater the risk.11
As for breast cancer, recent analyses of population-based family studies have shown that, at least and perhaps more so for cancers diagnosed before the age of 40 years, tumour morphology can be used to further subcategorise familial risks. For example, trabecular growth pattern and high mitotic index are strong predictors for having a germline mutation in BRCA112 and hence predict risk for first-degree relatives. After excluding known BRCA1 and BRCA2 mutation carriers, absence of extensive sclerosis, circumscribed growth, extensive intraductal carcinoma and a lobular growth pattern of breast cancers were independent predictors associated with a twofold to threefold increased risk of breast cancer for first-degree relatives.13 The first-degree relatives of the one-third of early-onset cases who had none of these morphology features showed no evidence for an increased (familial) risk of breast cancer.
For CRC, it is well established that tumour molecular features can predict germline mutation status with greater sensitivity and specificity and in a more cost-effective way than using family history alone, particularly in early-onset disease.14 Tumour-infiltrating lymphocytes, proximal colonic location, mucinous histology, poor differentiation and Crohn's-like reaction are now well-recognised pathology predictors for mismatch repair (MMR)-deficiency resulting from a germline mutation in one of the MMR genes.15 ,16 In this study, we estimated CRC risk for first-degree relatives of CRC cases based on tumour molecular subtypes and tumour pathology features.
Materials and methods
We studied a cohort of the first-degree relatives of people with incident invasive CRC (probands) recruited by the Colon Cancer Family Registry between 1997 and 2007 through state or regional population cancer registries in the USA (Washington, California, Arizona, Minnesota, Colorado, New Hampshire, North Carolina and Hawaii), Australia (Victoria) and Canada (Ontario). Study designs and recruitment for the Colon Cancer Family Registry can be found at http://coloncfr.org/ and have been previously published in detail.17 Written informed consent was obtained from all study participants, and the study protocol was approved by the institutional research ethics review board at each study centre.
Information on demographics, personal characteristics, personal and family history of cancer, cancer-screening history and history of polyps, polypectomy and other surgeries was obtained by questionnaires from all probands and participating relatives. Participants were followed up approximately every 5 years after baseline to update information across all the study centres. The present study was based on all available baseline and follow-up data until 2012. Reported cancer diagnoses and age at diagnosis were confirmed, where possible, using pathology reports, medical records, cancer registry reports and death certificates. The tumour anatomic location and histology were coded and stored using International Classification of Diseases for Oncology, third edition (ICD-O-3).18 Blood samples and permission to access tumour tissue were requested from all participants.
CRC tumour pathology review
Probands’ CRCs were reviewed by pathologists from each study centre of the Colon Cancer Family Registry and assessed for the following features: histologic type (adenocarcinoma, mucinous, signet ring cell), histologic grade (low or high grade), tumour margin (expanding or infiltrative) and peritumoral lymphocytes, Crohn's-like reaction, tumour-infiltrating lymphocytes, venous invasion and synchronous CRC (present or absent). Tumours were classified as mucinous carcinoma or signet-ring-cell carcinoma if at least 50% of the tumour demonstrated the specific morphology. Low grade was defined as adenocarcinoma with ≥50% gland formation and high grade as adenocarcinoma with <50% gland formation. Tumour margin describes the advanced edge of the tumour into the bowel wall and was categorised as expanding margin for well-circumscribed edge pushing into the deep portion of the wall and infiltrative margin for widespread dissection of the bowel wall by the cancer cells.19 The presence of peritumoral lymphocytes was based on the finding of a cap of lymphocytes at the deepest part of the invasive tumour front.20 Crohn's-like reaction was recorded as present for tumours demonstrating at least four nodular lymphoid aggregates within a single ×4 field, deep to the advancing edge of the tumour.21 Tumour-infiltrating lymphocytes were regarded as present when ≥5 intratumoral lymphocytes were identified within a single ×40 field of a haematoxylin and eosin-stained section.16 Colon cancer was defined as any diagnosis of cancer in the proximal colon (caecum to transverse colon; ICD-O-3 codes C18.0, C18.2–C18.4), distal colon (splenic flexure to sigmoid colon; C18.5–C18.7) or an unspecified location of the colon (C18.8, C18.9 and C26.0). Rectal cancer was defined as any diagnosis of cancer in the rectosigmoid junction (C19.9) and rectum (C20.9 and C21.8).
Probands’ CRCs were characterised for MMR-deficiency by microsatellite instability (MSI) using a ten-marker panel and/or by immunohistochemistry (IHC) for the four MMR proteins.22 Tumours were classified as MMR-deficient if they were MSI-high (≥30% or more of the markers show instability) and/or showed loss of expression of one or more of the MMR proteins by IHC; and MMR-proficient if they were microsatellite stable (no unstable markers) or MSI-low (<30% unstable markers) and/or showed normal expression of all four MMR proteins by IHC. Of the 4843 proband CRCs with a MMR IHC or MSI result, 3325 (68.7%) had both tests performed, 1508 (31.1%) had MMR IHC only and 10 (0.2%) had MSI only (see online supplementary table S1). Of the 3325 proband CRCs that had both tests performed, 3202 were concordant between MSI and MMR IHC (96.3%, 95% CI 95.6% to 96.9%). Probands with an MMR-deficient CRC underwent germline mutation testing (Sanger sequencing and MLPA) as previously described.17 ,23–25
A fluorescent allele-specific PCR assay was used to detect the somatic T>A mutation at nucleotide 1799 in exon 15 of the BRAF gene (BRAF V600E) as has been previously described.26 ,27 Methylation of the MLH1 gene promoter was measured in all MSI-high and MSI-low cases with sufficient tumour DNA and a random sample of microsatellite-stable cases. MLH1 methylation was measured using the MethyLight MLH1-M2Methylight reaction using an ALU control reaction to normalise for bisulfite-converted input DNA,28 with the modifications described in Poynter et al.29 We classified samples with a proportion of methylated reference greater than or equal to 10 as positive for MLH1 methylation. An ALU control reaction cycle threshold [C(t)] value of <25 was used to retain the largest sample size possible for the analysis while minimising the potential for false negatives.
We categorised the probands into four groups:
‘MMR-proficient’: probands with an MMR-proficient CRC.
‘MMR-deficient non-Lynch syndrome’: probands with a CRC that had loss of MLH1 and PMS2 protein expression with evidence of MLH1 methylation and/or had the BRAF V600E mutation.
‘Suspected Lynch syndrome’: probands with a CRC that had MLH1/PMS2 loss with no evidence of MLH1 methylation and/or BRAF V600E mutation or had MSH2/MSH6 loss or solitary loss of PMS2 or MSH6 or were MSI-H, for which no MMR germline mutation had been identified.
‘Lynch syndrome’: probands known to carry a pathogenic MMR germline mutation.
Observation time for each member of the cohort (a first-degree relative of a CRC-proband) started at birth and ended at first diagnosis of invasive cancer, first polypectomy, last contact or death, whichever occurred earliest. Of the 35 567 first-degree relatives, unaffected relatives with no information on age were censored at birth, that is, excluded from analysis (n=2071; 6%) and the remaining 33 496 entered into the analysis. For cancer-affected relatives with missing age at diagnosis (n=120; 0.4%), age at diagnosis was assumed to be 1 year prior to the last known age (n=102) or, if last known age was not available, the median age at diagnosis of the specific cancer in the general population (n=18). The sex-specific median age for each cancer was obtained from the Surveillance, Epidemiology, and End Results (SEER) Cancer Statistics Review (1975–2000).30
We estimated the standardised incidence ratio (SIR) of CRC for the first-degree relatives of CRC cases compared with the general population as a function of measured covariates by using the data on first-degree relatives of case-probands. Each SIR was calculated by dividing the observed number of CRCs first-degree relatives by the number expected based on the age-specific, sex-specific and country-specific incidence in the general population.31 The Jackknife method was used to estimate 95% CIs to adjust for correlation of risk between relatives from the same family.32
In this study, exposure is the tumour pathology feature of CRC of the proband and outcome is the incidence of CRC in the first-degree relatives. We estimated CRC HRs for first-degree relatives of CRC cases depending on the absence or presence of specific tumour pathology features. The HRs and robust estimates of corresponding 95% CIs were calculated from Cox proportional hazard regression models by taking into account clustering by family membership to allow for correlation of risk between relatives from the same family.33 To investigate independent associations of the tumour pathology features with familial risk, we fitted variables that showed a nominally statistically significant association in the univariable analyses into a multivariable model. We estimated CRC risk for first-degree relatives associated with a total number of proband's features that were shown to be statistically significant in the multivariable model, as a categorical (≥3 vs <3) as well as continuous (per feature) factor.
To account for the missing values that occurred for some pathology features, we devised a multiple imputation model based on previous recommendations of van Buuren et al34 and Ali et al.35 The missing data were assumed to be ‘missing at random’ so that the reason for missingness can be explained by the observed data.36 The model included all predictor variables, the outcome variable (diagnosis of CRC) and additional variables that we considered may increase the plausibility of the missing at random assumption in order to improve the imputation process. While some researchers claim that as few as 3–5 imputed data sets are sufficient for imputation,37 we chose 10 sets based on recommendations that the number of sets should approximate the percentage of subjects with some missing data.38 Location of CRC was imputed using polytomous logistic regression while the other pathology features were imputed using logistic regression. Missing values were sampled and replaced with a set of plausible values randomly drawn from their predicted distribution based on the other observed variables, thus creating 10 completed data sets. Cox proportional hazard regression models were run separately for each imputed data set and estimates of the predictor variables were combined using the programmes written by Carlin et al.39 As a sensitivity analysis, we also compared the estimates of association using the imputed missing data with the estimates from a complete case analysis.
Some centres of the Colon Cancer Family Registry used stratified sampling based on family history for recruitment, and we did take this into account when combining and analysing the data. To adjust for this stratified sampling, we gave each relative a probability weight equal to the reciprocal of the family sampling fraction of their proband used by each study centre.17 All statistical tests were two-sided, and p<0.05 was regarded as nominally statistically significant. All statistical analyses were performed using Stata V.12.1.40
Of the 4853 probands with an invasive CRC from the population-based resources of the Colon Cancer Family Registry, 4095 (84%) were MMR-proficient; 301 (6%) were MMR-deficient non-Lynch syndrome; 271 (6%) were suspected Lynch syndrome (104 MLH1/PMS2 loss, 45 MSH2/MSH6 loss, 17 PMS2 loss, 27 MSH6 loss and 78 MSI-high) and 186 (4%) were Lynch syndrome with confirmed pathogenic mutations in MMR gene mutations (59 MLH1, 70 MSH2, 25 MSH6, 30 PMS2, 2 EPCAM). The probands were diagnosed with a CRC at a mean age of 54.2 (SD 11.6) years, with approximately one-third of tumours diagnosed in each of the proximal colon, distal colon and rectum (table 1). Compared with MMR-proficient probands, the mean age at diagnosis of CRC was 10.3 (95% CI 9.02 to 11.6) years older for MMR-deficient non-Lynch syndrome probands, 5.28 (95% CI 3.95 to 6.61) years younger for suspected Lynch syndrome probands and 9.33 (95% CI 7.74 to 10.9) years younger for Lynch syndrome probands (see online supplementary table S2).
We identified 33 496 first-degree relatives of the participating probands: 28 156 (84%) relatives of MMR-proficient probands; 2302 (7%) of MMR-deficient non-Lynch syndrome probands; 1799 (5%) of suspected Lynch syndrome probands and 1239 (4%) of Lynch syndrome probands (table 2). Of all first-degree relatives, 1404 (4%) were diagnosed with a CRC at a mean age of 62.1 (SD 14.1) years. Compared with first-degree relatives of MMR-proficient probands, mean age at diagnosis of CRC was 3.10 (95% CI 0.57 to 5.63) years older for first-degree relatives of MMR-deficient non-Lynch syndrome probands, 6.01 (95% CI 3.54 to 8.47) years younger for first-degree relatives of suspected Lynch syndrome probands and 14.8 (95% CI 12.5 to 17.0) years younger for first-degree relatives of Lynch syndrome probands (see online supplementary table S2). Verification was sought for all reported CRCs in the relatives. Of the 1404 CRCs reported in the first-degree relatives, 605 (43%) had been verified by pathology reviews, medical records, death certificates or cancer registries (see online supplementary table S3). For cancers that we could not obtain verification, we did not solely rely on the proband to report family history of cancer. Rather, we interviewed multiple family members to ascertain all diagnoses of cancer in family. For apparently conflicting reports within the family, we gave priority to the report provided by the closest family member.
Using all the data, we observed 1404 CRCs in the 33 496 first-degree relatives of CRC cases over the 1 792 526 person-years. We estimated that first-degree relatives of CRC cases had an approximately twofold higher risk of CRC compared with the general population (SIR 1.79, 95% CI 1.66 to 1.94). Compared with first-degree relatives of MMR-proficient CRC cases, a higher risk of CRC was estimated for first-degree relatives of suspected Lynch syndrome CRC cases (HR 2.06, 95% CI 1.59 to 2.67) and for first-degree relatives of Lynch syndrome CRC cases (HR 5.37, 95% CI 4.16 to 6.94), but not for relatives of MMR-deficient non-Lynch syndrome CRC cases (HR 1.04, 95% CI 0.82 to 1.31) (table 3).
A higher risk of CRC was estimated for first-degree relatives if the CRC case's tumour had diagnosis before age 50 years (HR 1.64, 95% CI 1.43 to 1.88); a proximal colonic location (HR 1.33, 95% CI 1.16 to 1.53); an expanding tumour margin (HR 1.66, 95% CI 1.35 to 2.05); peritumoral lymphocytes (HR 1.26, 95% CI 1.06 to 1.49); tumour-infiltrating lymphocytes (HR 1.33, 95% CI 1.15 to 1.55) and a synchronous tumour (HR 1.69, 95% CI 1.23 to 2.32) (table 4). There was marginal evidence that presence of Crohn's-like reaction in CRC case was associated with a higher CRC risk for first-degree relatives (HR 1.17, 95% CI 0.99 to 1.49). There was no evidence that CRC risks for first-degree relatives depend on the CRC case's sex (p=0.42), tumour histologic subtype (p=0.68), grade (p=0.72) and venous invasion (p=0.91). (See detail in online supplementary table S4.)
We therefore fitted proband's age at diagnosis, tumour location, tumour margin, peritumoral lymphocytes, tumour-infiltrating lymphocytes and synchronous CRC in the multivariable model (table 5). When all groups were combined, we found that the CRC case's age at diagnosis <50 years (HR 1.61, 95% CI 1.40 to 1.87), tumour proximal colonic location (HR 1.29, 95% CI 1.12 to 1.49), expanding tumour margin (HR 1.48, 95% CI 1.19 to 1.83), presence of peritumoral lymphocytes (HR 1.21, 95% CI 1.01 to 1.44), presence of tumour-infiltrating lymphocytes (HR 1.21, 95% CI 1.03 to 1.43) and presence of synchronous CRC (HR 1.61, 95% CI 1.19 to 2.18) were all independently associated with an increased CRC risk for first-degree relatives. When we stratified by MMR-deficiency status of the CRC case, we estimated that a higher risk of CRC for first-degree relatives was associated with age at diagnosis <50 years (p=0.008) and an expanding tumour margin (p<0.001) when the CRC case was MMR-proficient; age at diagnosis <50 years (p=0.006) and presence of synchronous CRC (p=0.004) when the CRC case was MMR-deficient non-Lynch syndrome; and age at diagnosis <50 years (p=0.001) and presence of peritumoral lymphocytes (p<0.001) when the CRC case was Lynch syndrome (table 5).
When we combined the proband's age at diagnosis and tumour pathology features that were independently associated with CRC risk for first-degree relatives, we estimated that there was an approximately twofold increased risk of CRC for first-degree relatives (HR 1.91, 95% CI 1.65 to 2.22) if probands had three or more of the following six features: age at diagnosis <50 years, tumour's proximal colonic location, presence of expanding tumour margin, peritumoral lymphocytes, tumour-infiltrating lymphocytes and synchronous CRC. When we stratified by MMR-deficiency status of the CRC case, the approximate twofold increase in risk was still present for first-degree relatives of CRC cases with three or more features compared with first-degree relatives of CRC cases with less than three features in each group (table 6).
When we fitted these six features as a continuous factor, we estimated the HR per feature of the CRC case to be 1.33 (95% CI 1.26 to 1.42) for first-degree relatives when all groups were combined, 1.21 (95% CI 1.13 to 1.30) for first-degree relatives of MMR-proficient CRC cases, 1.54 (95% CI 1.23 to 1.92) for first-degree relatives of MMR-deficient non-Lynch syndrome CRC cases, 1.28 (95% CI 1.06 to 1.55) for first-degree relatives of suspected Lynch syndrome CRC cases and 1.33 (95% CI 1.11 to 1.58) for first-degree relatives of Lynch syndrome CRC cases.
In a complete case analysis without any missing data, we found that use of the imputed missing data had no considerable effect on estimates of association (details not shown). We also estimated CRC risk for first-degree relatives by fitting the specific tumour location of the proband as an ordinal (continuous) variable in the model. We found a HR of 1.05 (95% CI 1.02 to 1.07) showing that the more proximally located CRC of the proband was associated with a higher risk of CRC for first-degree relatives when all groups were combined (see online supplementary table S5). When we investigated the association between BRAF V600E mutation of proband's CRC tumour and CRC risk for first-degree relatives, we found no evidence of association even after stratification by MMR status (see online supplementary table S6).
Using a large family cohort study, we estimated that first-degree relatives of suspected Lynch syndrome CRC cases had a higher risk of CRC compared with first-degree relatives of MMR-proficient CRC cases and a lower risk of CRC compared with first-degree relatives of Lynch syndrome CRC cases. We found that first-degree relatives had a higher risk of CRC if the CRC case was diagnosed <50 years, had proximal colon cancer, an expanding tumour margin, peritumoral lymphocytes, tumour-infiltrating lymphocytes or a synchronous CRC. These CRC features could be useful in identifying the first-degree relatives of CRC cases who are at increased CRC risk, even within a specific molecular subtype of CRC.
The clinical management of CRC cases with suspected but genetically unproven to have Lynch syndrome (ie, MMR-deficient, no MMR germline mutation and not MHL1-methylated or BRAF-mutated), and of their relatives, can be particularly challenging as it is not possible to distinguish which if any relatives are at increased risk because they carry a mutation in an MMR gene. Little is known about the cancer risks for members of these families and therefore screening and treatment guidelines are not well defined. The average age at CRC diagnosis for first-degree relatives of CRC cases with suspected Lynch syndrome was between that for first-degree relatives of MMR-proficient cases and that for first-degree relatives of confirmed Lynch syndrome cases (58 vs 49 and 64 years, respectively). We estimated that CRC risk for first-degree relatives of suspected Lynch syndrome CRC cases was between that for first-degree relatives of MMR-proficient cases and that for first-degree relatives of confirmed Lynch syndrome cases, and this is consistent with the observation by Rodriguez-Soler et al.41
It is possible that a proportion of these suspected Lynch syndrome CRC cases might be due to the existence of complex or cryptic mutations in MMR genes that are not readily identified by current Sanger sequencing and MLPA techniques.42–44 It could also be argued that the suspected Lynch syndrome CRC cases carrying undetected mutations have associated with them a more moderate CRC penetrance compared with that for the exonic and splice site mutations that are more readily detected.45 With the increasing use of next generation sequencing in clinical diagnostics, further investigation of more complex and cryptic mutational mechanisms will be possible. This will be accompanied by an increasing need to assess the pathogenicity of variants of uncertain clinical significance (VUS) in the MMR genes and quantify the risk of CRC for carriers of VUS and their relatives, and build upon the significant progress achieved to date.46 ,47 Alternatively, a proportion of suspected Lynch syndrome cases have been shown to result from somatic inactivation within the tumour through biallelic mutations.48 ,49 Though an intermediate screening recommendation has been recommended for suspected Lynch syndrome families,41 optimal screening strategies are yet to be defined given this group is likely to be heterogeneous with regards to family history and to the mechanism of MMR inactivation.
First-degree relatives of CRC cases with confirmed Lynch syndrome have a greater CRC risk than first-degree relatives of MMR-proficient CRC cases. Current screening recommendations for untested first-degree relatives of Lynch syndrome CRC cases are typically annual colonoscopy starting at age 20–25 years or 10 years younger than the youngest diagnosed CRC in the family, whichever is earliest.50 Our finding that the average age at CRC diagnosis for first-degree relatives of Lynch syndrome cases is approximately 15 years earlier than for first-degree relatives of MMR-proficient cases supports the rationale for starting CRC screening at an earlier age in first-degree relatives of Lynch syndrome cases until they have been genetically tested and shown not to be MMR gene mutation carriers. Similarly, our finding that younger age at diagnosis of CRC for Lynch syndrome cases is a risk factor for CRC in their first-degree relatives supports the rationale for varying the recommendations for age of starting CRC screening in first-degree relatives based on the age of the youngest CRC in the family.
We estimated that first-degree relatives of MMR-proficient CRC cases and MMR-deficient non-Lynch syndrome CRC cases had an approximately 1.6-fold higher risk of CRC compared with the general population. Current CRC screening recommendations for first-degree relatives of CRC cases vary, but in general the recommendation is to start screening earlier (at age 40 rather than age 50 for those without a family history of CRC) and to use colonoscopy if the CRC occurs in a relative before the age of 60 years. Our data that a younger age of the proband's CRC was associated with an increased CRC risk for first-degree relatives of the MMR-proficient and MMR-deficient non-Lynch syndrome CRC cases are consistent with this type of age-dependent screening recommendations.
We found that first-degree relatives of MMR-deficient CRC cases as a result of MLH1 methylation were at no greater risk of CRC compared with first-degree relatives of MMR-proficient CRC cases. However, we estimated that CRC risk for first-degree relatives of MMR-deficient non-Lynch syndrome cases is about 4.5-fold higher when the case's CRC was diagnosed before age 50 years and about twofold greater when the case had a synchronous CRC. It is possible that this increased CRC risk may be explained by shared lifestyle and environmental factors within families, such as smoking, which has been reported to be associated with MMR-deficient and BRAF-mutated CRC.51 These features are also comparable with familial serrated neoplasia (Jass syndrome),52 which has been linked to 2q3353 and characterised by early age at diagnosis, multiplicity of lesions, presence of BRAF V600E-mutated and MLH1-methylated MSI-H CRCs.54
We found no evidence of an association between BRAF V600E mutation status of proband's CRC tumour and an increased risk of CRC for their first-degree relatives, even after stratification by MMR status. A subset of this data has been previously reported highlighting an inverse association between BRAF-mutated CRC and family history of CRC.27 In contrast, three other studies support an association between a family history of CRC and an increased risk of the BRAF-mutated CRC.55–57 One potential explanation for the discrepancy in findings between this study and the previous reports might be the difference in the mean age at diagnosis of the CRC cases.
In this study, the molecular features analysed were those commonly used for the identification of Lynch syndrome. Additional molecular features may provide greater insights for CRC risk prediction for relatives. Two studies recently reported that a family history of CRC is associated with a high risk of MMR-proficient CRC with low levels of tumour LINE-1 methylation (LINE-1 hypomethylation).58 ,59 Furthermore, the recent comprehensive characterisation of CRCs from the Cancer Genome Atlas project has identified many key genetic and epigenetic mutations in CRC.60 Although these tumour molecular features were beyond the scope of this study, they require further investigation in large resources for their effectiveness in CRC risk prediction for relatives.
Currently anatomic and pathology features of CRCs are not used to guide CRC screening intensity or to estimate CRC risk for relatives. We found that proximal colonic location, synchronous CRCs, expanding tumour margin and inflammatory infiltrate assessed by the presence of peritumoral lymphocytes and tumour infiltrating lymphocytes in the proband's tumour are associated with an increased CRC risk for first-degree relatives. Many of these pathology features are strongly associated with MMR-deficiency secondary to Lynch syndrome or caused by somatic alterations in MMR genes.15 ,16 The mechanism accounting for the prominent inflammation in MMR-deficient CRC (peritumoral lymphocytes, tumour infiltrating lymphocytes and Crohn's-like reaction) is unclear and may be related to the host response to tumour cells demonstrating an altered phenotype. Interestingly, some of these pathology features were also found to be associated with a higher CRC risk for first-degree relatives of MMR-proficient CRC. This may be due to undetected MMR gene mutations within MMR-proficient CRC cases,61 or alternatively molecular alterations other than MMR-deficiency may trigger a similar immunological reaction in the tumour environment.
When analysed for each of the molecular subtypes of CRCs, there was a complex relationship of these pathologic features to CRC risk for first-degree relatives. In an effort to simplify the pathology analysis, we investigated the role of all of the pathology factors that were related to CRC risk for first-degree relatives and found that having three or more features in CRC cases was associated with an increased CRC risk for first-degree relatives in the entire group as well as all of the molecular subtypes. When we considered the specific colorectal tumour location of the proband, we found an increased risk of CRC for first-degree relatives with a more proximal location of CRC in the proband. Further larger studies will be required to gain enough power to investigate the effect of specific colorectal tumour location in each specific molecular subtype of CRC. This site-specific study may be more informative to understand site-specific tumour molecular pathology feature differences beyond the simple proximal–distal divide.62 Future studies with computer-assisted morphology analysis may identify new pathology features associated with an increased cancer risk for relatives.63
This is the first study to investigate the association of CRC tumour molecular pathology features with CRC risk for relatives. We used a large population-based cohort of first-degree relatives of incident CRC cases, and therefore inference can be readily made, at least to the populations from which these families were sampled, if not more generally. Family history of cancer was collected in a systematic manner across the Colon Cancer Family Registry. Our application of weights to each relative depending on the different sampling strategies used by some centres of the Colon Cancer Family Registry would minimise any selection bias due to sampling based on family history.
For CRC cases with suspected Lynch syndrome and confirmed Lynch syndrome, we estimated risk for all first-degree relatives, even though on average only half will carry a mutation in an MMR gene. Therefore, if suspected Lynch syndrome is due to an autosomal dominantly inherited syndrome, the associations for both these groups are an average over carrier and non-carrier relatives. Our study families were predominantly Caucasians, and we therefore cannot infer that these results apply to other ethnic groups. Further, we did not report risks of cancers other than CRC for relatives given we emphasised only on associations between CRC tumour molecular pathology features and risk of CRC for their first-degree relatives in this study. Fifty-seven per cent of CRC diagnosis in the first-degree relatives were based on the proband's or relative's self-report, which, if a proportion are false positive, may lead to an underestimation of the true associations. However, the specificity of self-reported cancer diagnoses is generally more than 98%, suggesting that there are few false positives in such data.64 It is possible that people who are likely to have a genetic predisposition give a more complete family history of cancer compared with those who are not. If this is the case, there will be more complete cancer diagnosis data reported in Lynch syndrome or suspected Lynch syndrome families rather than non-Lynch syndrome families.
In conclusion, findings from this study provide evidence that molecular and pathology characterisation of CRCs can both help define the CRC risk for first-degree relatives, and therefore may be useful to refine screening recommendations for first-degree relatives of CRC cases and to identify which cases are more likely to be caused by genetic or other familial factors.
The authors thank all study participants of the Colon Cancer Family Registry and staff for their contributions to this project.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors AKW, JLH and MAJ developed study concept and design. All authors were involved in collecting and acquisition of required data for the study. AKW and RJM conducted statistical analysis, and all authors were involved in interpretation of data. AKW, DDB, CR, DJA and MAJ drafted the manuscript. All authors critically reviewed the manuscript for important intellectual content and approved the final version of the manuscript.
Funding This work was supported by the National Cancer Institute, National Institutes of Health under RFA #CA-95-011 and through cooperative agreements with members of the Colon Cancer Family Registry and Principal Investigators. Collaborating centres include Australasian Colorectal Cancer Family Registry (U01 CA097735), Familial Colorectal Neoplasia Collaborative Group (U01 CA074799) [USC], Mayo Clinic Cooperative Family Registry for Colon Cancer Studies (U01 CA074800), Ontario Registry for Studies of Familial Colorectal Cancer (U01 CA074783), Seattle Colorectal Cancer Family Registry (U01 CA074794) and University of Hawaii Colorectal Cancer Family Registry (U01 CA074806). This work was also supported by a Centre for Research Excellence grant from the National Health and Medical Research Council (NHMRC), Australia (APP1042021). AKW is supported by the Picchi Brothers Foundation Cancer Council Victoria Cancer Research Scholarship, Australia. MAJ is an NHMRC Senior Research Fellow. JLH is a NHMRC Senior Principal Research Fellow. CR is a Jass Pathology Fellow.
Disclaimer The content of this manuscript does not necessarily reflect the views or policies of the National Cancer Institute or any of the collaborating centres in the CFRs, nor does mention of trade names, commercial products or organisations imply endorsement by the US Government or the CFR. Authors had full responsibility for the design of the study, the collection of the data, the analysis and interpretation of the data, the decision to submit the manuscript for publication and the writing of the manuscript.
Competing interests None.
Ethics approval Written informed consent was obtained from all study participants, and the study protocol was approved by the institutional research ethics review board at each study centre.
Provenance and peer review Not commissioned; externally peer reviewed.
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.