|Home | About | Journals | Submit | Contact Us | Français|
Whether people who inherit a mutation in MUTYH from only one parent (monoallelic mutation) are at increased risk of colorectal cancer (CRC) remains controversial. Most previous studies and meta-analyses have not found statistically significant associations but, given carriers are relatively rare, may be underpowered to detect small increased risks. We have conducted a systematic review and meta-regression analysis of previously published case–control studies to estimate the strength of association for monoallelic MUTYH mutation and CRC risk. Potential sources of heterogeneity were evaluated. We have compared the carrier frequency in cases with a family history of CRC to that of controls, as a novel and powerful design, to measure statistical evidence of an association but not the strength of association. The magnitude of the genotype-disease association, estimated from a pooled odds ratio comparing cases unselected for family history with controls, was 1.15 (95% CI = 0.98–1.36) and not substantially altered by adjustment for potential sources of heterogeneity. Monoallelic mutation carrier frequency was greater for cases ascertained due to a family history (3.3%; SE 0.9%) than for controls (1.4%; SE 0.3%) (P = 0.02). Monoallelic MUTYH mutation carriers are at increased risk of CRC but the average increase is small.
People with a family history of colorectal cancer (CRC) are, on average, about twice likely to be diagnosed with CRC than those with no family history [1, 2]. All the causes for this excess risk are unknown as less than half of it is explained by rare mutations in the known major CRC susceptibility genes (mismatch repair genes  and adenomatous polyposis coli gene ) and the common, but not necessarily causal, variants associated with small increments in risks recently identified by genome wide association studies .
Human MutY homologue (MUTYH) gene is a base excision repair gene which repairs oxidative damage of DNA by excising adenines incorporated opposite 8-oxo-7,8-dihydro-2′-deoxyguanosine (8-oxoG) . A number of studies [7-20] have confirmed that biallelic mutation carriers are at high risk of CRC. It remains uncertain, however, as to whether, and to what degree carriers of a mutation inherited from only one parent (monoallelic mutations) are at increased risk of CRC despite many case–control studies [10-24] and several meta-analyses [15-18,20, 25] that have been conducted to date. This lack of certainty is in part due to the rarity of monoallelic mutation carriers in the population and likely small increased risk. For example, assuming a population carrier frequency of 1.7% in Caucasian populations, a study of at least 34,000 cases and 34,000 controls is needed to have 90% power to detect an odds ratio of 1.2 or greater at the 0.05 level of significance.
Justifications for a further meta-analysis are: (1) the most recently published meta-analysis  did not include two unselected case–control studies (including the second largest study to date ); and (2) there has been no meta-analysis of familial case–control studies which is a powerful design to measure the statistical evidence of an association. Comparison of the genotype frequency for cases ascertained because they have a family history of disease with that for controls is a more powerful but seldom used design to address the evidence (i.e. P value) for an association [26, 27]. If monoallelic mutations are more common in CRC cases with a family history of disease compared to controls (with or without a family history of disease), then there is evidence that the monoallelic mutation is associated with disease risk. This is a more powerful design than the traditional case–control design, as the cases with a family history of CRC will be enriched for monoallelic mutation carriers if monoallelic mutations increase the risk of CRC. Note, however, that this “familial cases versus controls” design cannot be used to estimate the magnitude (i.e. odds ratio) of the association in the population.
One aspect of monoallelic MUTYH mutations under-studied to date is the role of the less commonly reported mutations. Almost all previous studies [11, 12, 14-16, 20,21] have restricted their analyses to the variants Y179C and G396D (previously known as Y165C and G382D) which account for more than 80% of all reported mutant alleles of MUTYH in Caucasian populations [9, 28, 29]. Previous meta-analyses have not included MUTYH sequence variants which are common in different ethnic groups; for example, E480X in Indian, Y104X in Pakistani [8, 28, 30], c.1437_1439delGGA in Italian , c.1228_1229insGG in Portuguese , Q498H in German , and G25D and P18L in Chinese populations . In this analysis we have assessed the combined role of all previously identified pathogenic variants in MUTYH mutations on the risk of CRC, and separately for Y179C and G396D.
We performed meta-analysis of case–control studies to estimate the strength of association between monoallelic mutations and CRC risk. We have compared the frequency of monoallelic mutation carriers in cases with a family history of CRC to that of controls to estimate the statistical evidence of the association.
MEDLINE, PUBMED (http://www.ncbi.nlm.nih.gov/) was searched for all relevant case–control studies of the association between germline monoallelic mutations in MUTYH and risk of CRC using the following combinations of key words: “colon or colorectal”, “cancer or neoplasm or tumour” and “MUTYH or MYH or hMYH or human MutY or base excision repair gene”. Search engines, such as “Google Scholar” and “SuperSearch” of the University of Melbourne library, were also used to search for related articles. No language restrictions were imposed. References from relevant articles, letters, reviews and previous meta-analyses were reviewed to identify any additional studies that were not indexed by the electronic database. Studies were reviewed initially on the basis of title and abstracts, and then all full manuscripts for those that appeared relevant were obtained and checked for eligibility.
case–control studies were eligible if they reported analyses of the associations between MUTYH mutations and CRC risk. Eligible studies needed to have presented the genotype data for cases and controls, the number of cases and controls tested, and to have described whether or not case ascertainment was independent of family history of CRC. Studies were excluded if they (1) defined cases by presence of multiple adenomas or polyposis, and not CRC, (2) did not state how CRC diagnoses were confirmed, (3) ascertained cases only because they had no family history, (4) recruited cases under more than one category of family history but did not provide numbers for each ascertainment method, (5) only reported on variants of MUTYH previously described as non-pathogenic, or (6) reported data included in a later publication.
Exposure was defined as carrying a monoallelic germline mutation in MUTYH (as evidenced by testing DNA extracted from a blood sample). CRC cases were MUTYH genotyped people with a histologically confirmed colorectal adenocarcinoma and controls were MUTYH genotyped people without a previous diagnosis of CRC. In terms of groups, familial cases were defined as CRC cases ascertained because they had a family history of CRC. Unselected cases were defined as CRC cases who were ascertained irrespective of their family history.
For each study, the following data were extracted: first author’s name, year of publication, country in which the study was performed, name of the study if given, study design, case ascertainment method, recruitment and selection criteria for cases and controls, number of individuals and their age and sex distributions for both cases and controls, and the number of monoallelic and biallelic carriers of MUTYH mutations by variant tested for both cases and controls in each study. Corresponding authors were contacted to confirm and request data if required. The names of MUTYH variants were provided using nomenclature according to the Leiden Open Variation Database of the MUTYH gene .
For familial case–control studies, we compared the pooled frequency of monoallelic mutations in cases to that of controls. For unselected case–control studies, the pooled odds ratio (OR) and its 95% confidence interval (CI) was estimated to compare the frequency of monoallelic MUTYH mutation carriers between cases and controls. Odds ratios were estimated for each genotype, and forest plots of the odds ratios were generated. We fitted linear meta-regression models to the log-transformed individual study odds ratios to create pooled estimates and to evaluate the role of several potential sources of heterogeneity; carrier frequency in controls, study site (Australia, Europe and North America), ascertainment method of controls (population-based or hospital-based), and the year of publication. To test the influence of individual studies on the pooled estimate, we omitted data from each study one at a time and repeated the analysis. Random and fixed effects models were fitted, and heterogeneity was tested using Cochran’s Q statistic. Funnel plots and statistical tests for funnel plot asymmetry were performed to test the evidence of publication bias [36, 37].
All statistical tests were two-sided and, following convention, statistical significance for testing a predetermined null hypothesis was assessed by P < 0.05. All statistical analyses were conducted using Stata 10.0 .
Of the 221 studies identified by the literature search, 16 studies [11-15, 17-21, 23, 24, 33, 39-41] passed the inclusion and exclusion criteria (Fig. 1). Tables Tables11 and and22 show that there were five studies using familial cases [14,24, 33, 39, 40] comprising a total of 363 cases and 1,698 controls, and 12 studies using unselected cases [11-15, 17-21, 23, 41] comprising a total of 21,369 cases and 14,639 controls. One study  contained both familial and unselected cases. For this study we included the relevant cases for each analysis and used all controls for both analyses.
All five studies attempted to exclude known or suspected cases due to mutation in a mismatch repair gene (Table 1). All studies recruited controls from the general population irrespective of a family history of CRC [14, 24, 33, 39] except for one study where controls had no family history of CRC and no sign of neoplasia in colonoscopy .
Table 3 shows that all but one study found that the monoallelic MUTYH mutation carrier frequency for familial cases was greater than for controls, though the difference was not statistically significant in any study (all P ≥ 0.05). Overall, monoallelic MUTYH mutations were identified in 12 of 363 familial cases (3.3%; SE 0.9%) compared with 23 of 1,698 controls (1.4%; SE 0.3%) (P = 0.02).
Table 2 shows that eight of the 12 studies [11, 12, 14, 18-21, 23] found a higher frequency of monoallelic mutation carriers in CRC cases compared with controls, but for all except one , the difference was not statistically significant. Overall, carriers of monoallelic MUTYH mutation were found in 411 of 21,369 cases (1.9%; SE 0.9%) and 243 of 14,639 controls (1.7%; SE 0.1%).
The association between being a monoallelic MUTYH mutation carrier for any variant and CRC was estimated as a pooled odds ratio of 1.15 (95% CI = 0.98–1.36) (Fig. 2). There was no evidence to reject the homogeneity of studies (Cochran’s Q = 8.88, with 11 degrees of freedom; P = 0.6). The pooled odds ratio for the association between a monoallelic mutation and CRC was 1.35 (95% CI = 0.99–1.85) for Y179C; and 1.06 (95% CI = 0.88–1.28) for G396D.
From meta-regression analysis, association between monoallelic MUTYH mutation and CRC risk remained virtually unchanged after adjusting for: the monoallelic carrier frequency in controls (P = 0.15), the ascertainment method of controls (P = 0.34), study site (P = 0.40) or the year of publication (P = 0.84). There were no obvious differences in summary estimates when each study was omitted one at a time and analysis was repeated, consistent with there being no major influence of individual studies on the overall estimate.
Funnel plots did not reveal any evidence of publication bias of the studies included in either meta-analysis (Fig. 3). Begg’s statistical tests also showed no evidence of publication bias; P = 0.59 for studies using familial cases and P = 0.71 for studies using unselected cases.
We have shown that people who have inherited a mutation in the gene MUTYH from only one parent (monoallelic mutation carriers) are at increased risk of CRC but only by a small amount, about 15%. The novelty of this aspect of the analysis is that it includes the second largest casecontrol study to date, which was not included in the most recent meta-analysis. The statistical significance of the association is provided by our novel meta-analysis of studies using cases ascertained because they have a family history of the disease, which is more powerful in terms of sample size than unselected case–control studies for addressing this issue. This analysis allow us to reject the null hypothesis that the carrier frequency is the same for cases and controls (P = 0.02), as the cases with a family history of CRC will be enriched for monoallelic mutation carriers if monoallelic mutations increase the risk of CRC. The magnitude of the association is provided by the meta-analysis of standard case–control studies (OR = 1.15). The observed association is unlikely to be due to biallelic mutation carriers, as biallelic mutations were tested for both cases and controls by each study and we excluded them from our analysis.
The meta-analysis of studies in which cases were ascertained irrespective of a family history of CRC (unselected case–control studies) comprised a total pooled sample of 21,369 CRC cases and 14,639 controls, which is the largest study to date. We estimated the strength of association between monoallelic MUTYH mutations and CRC after adjusting for possible sources of heterogeneity including carrier frequency of controls, control ascertainment, study site and the year of publication, but the association was not changed. The estimate of the effect size for this analysis was similar to that of the recent meta-analysis of Lubbe et al.  (OR = 1.14). However, with the use of the familial cases we have been able to conclude that there was an statistically significant evidence that monoallelic carriers are at greater risk than non-carriers (P = 0.02), whereas the study of Lubbe et al. which did not utilize familial cases reported a P value of 0.12.
We did not find statistical significant evidence for an association between specific variants of MUTYH mutation and CRC risk separately for Y179C and G396D, but had little power to detect even moderate associations given the decrease in sample size from stratification by mutation type. Among previous case–control studies, the study of Clearly et al.  was the only one that screened for the nine most frequent variants of MUTYH mutations in cases and controls. Mutation testing in all other studies was restricted to the most common mutations Y179C and G396D in Caucasian populations, or only one or two additional variants. A comprehensive mutation screening approach might be required to observe a significant association with CRC risk, if present.
Houlston and Peto  have pointed out that standard case–control designs, in which cases are ascertained irrespective of a family history of the disease, have limited power to identify alleles with small associations with risk if the carrier frequency of the deleterious allele in population is low (e.g. MUTYH mutations). Antoniou and Easton  showed that the sample size required to detect a disease allele association was substantially reduced if the cases were selected because they had a family history of the disease. Therefore case–control studies using familial cases hold promise as more powerful to examine the question of the existence of an association between monoallelic MUTYH mutations and CRC risk, despite their typically low frequency in the population. Our analyses here demonstrate the importance of study design using cases selected for a family history to detect evidence for a role of genetic variants on disease risk.
Two family studies based on the relatives of monoallelic MUTYH carriers diagnosed with CRC did observe a significant association between monoallelic MUTYH mutation and CRC risk. A kin-cohort study  of the relatives of MUTYH mutations carriers observed that monoallelic mutation carriers who had a relative diagnosed with CRC were at threefold risk of CRC (Hazard Ratio 2.9; 95% CI 1.2–7.0; P = 0.02) in addition to that expected due to their family history. A retrospective cohort analysis  of obligate carriers of monoallelic mutations in MUTYH, being the parents of biallelic carriers, estimated monoallelic carriers had twice the CRC risk of general population (Standardized Incidence Ratio 2.12; 95% CI 1.30–3.28; P < 0.01). It is important to note that the estimates of association from these studies (two to threefold increased risk) are relevant to monoallelic carriers who have a relative with CRC, whereas the result we have reported here (1.15-fold increased risk) is relevant to monoallelic carriers from the general population (irrespective of family history of CRC).
We have included all previously published unselected and familial case–control study in our meta-analyses. To assist readers evaluate the quality of and differences between different study designs we have provided extracted relevant details in Table 1. Meta-regression models might be able to resolve some of the inconsistencies across studies, within the limitations of power, but considerable residual heterogeneity for unmeasured modifiers of risk might remain.
When combined, all previous case–control studies of monoallelic MUTYH mutations and CRC suggest that the risk of CRC is increased for carriers, but only to a small degree, on average. Given the rarity of monoallelic mutations they account for only a trivial proportion of CRC. Therefore the clinical significance of increased risk of CRC for monoallelic mutation carriers is minimal. This study demonstrates the value of using familial cases for detecting rare, “low penetrance” cancer susceptibility alleles.
The authors thank corresponding authors of the articles involved in this systematic review and meta-regression analysis for providing required information. This work was partly supported by the National Health and Medical Research Council of Australian Government (Grant No. 628836).
Conflict of interest None.
Electronic supplementary material The online version of this article (doi: 10.1007/s10689-010-9399-5) contains supplementary material, which is available to authorized users.