Colorectal Cancer Risk in Monoallelic Carriers of MYH Variants

摘要: To the Editor: Since low-frequency variants in base-excision-repair gene MYH (MIM 604933) were first demonstrated to confer a recessive colorectal cancer (CRC) risk,1,2 there has been speculation of an additional dominant effect.3,4 In recent article, Farrington et al.5 described results screening series 2,239 CRC cases and 1,845 controls for germline human homolog Escherichia coli muty (MYH). whites, Y165C G382D are principal disease-causing MYH. Among cases, detected 46 monoallelic carriers these (14 heterozygotes 32 heterozygotes) 11 biallelic (8 homozygotes 3 compound heterozygotes), with corresponding frequencies among 28 0 carriers. They employed method Hugot al.6 estimate genotype relative risk (GRR) associated variant Their study confirmed well-established increased (GRR=92.65; 95% CI 41.60–213.20). Although they did not find statistically significant (GRR=1.35; 0.92–2.07), authors suggested that heterozygosity might be elevated risk—primarily later life—after arbitrarily restricted their analysis diagnosed after age 55 years (GRR=1.68; 1.07–2.95). We believe such conclusion may premature. First, GRR calculation approach is necessary assessment heterozygote Moreover, other analytical approaches, including standard asymptotic exact yield CIs do exclude odds ratio (OR) 1. Second, type stratification raises issue post hoc analysis. Third, we have similarly determined large case-control no evidence status influences risk. Our was based on 2,561 patients histologically adenocarcinomas (1,474 males 1,087 females; mean ± SD at diagnosis 61 11.4 years) ascertained through ongoing initiative Institute Cancer Research/Royal Marsden Hospital National Health Service Trust. We previously reported subset 358 cases.4 A total 2,695 control individuals (836 1,859 59 10.9 spouses malignancies, recruited as part Research Network Trial (1999–2002), Royal Trust/Institute Family History DNA Registry (1999–2004), Study Colorectal Genetics (2004), or UK Breast Genetics, all established within United Kingdom. None had personal history malignancy time ascertainment. All British obvious differences demography terms place residence Blood samples obtained informed consent ethics review board approval, accordance tenets Declaration Helsinki. Genotyping performed using customized Illumina Sentrix Bead Arrays manufacturer’s protocols. Assay validation conducted TaqMan by direct sequencing samples. Among CRC, identified 4 (1 homozygote 53 (38 15 heterozygotes). controls, identified, but 57 (40 17 Y165C). For each SNP, distributions deviate significantly from Hardy-Weinberg equilibrium (exact P=1.0). These comparable those documented populations—specifically Kingdom–based (P=.17 controls; P=.97 cases) (table 1)—and translate risks 5.57 (95% 0.69–∞) 0.98 0.66–1.46) carrier status, respectively. Stratification data 10-year bands provided influenced (P=.13). Furthermore, partitioned, al.,5 early- late-onset disease (for onset ⩽55 years, OR 0.83; 0.42–1.51; >55 1.03; 0.66–1.58). Table 1.  Summary Published Case-Control Studies Relationship between Variants Risk CRC To further explore possibility affect applied kin-cohort compare 14,668 first-degree relatives noncarriers. Data any cancer, well vital current death, collected parents, siblings, offspring validated questionnaire. Fourteen (4.3%) 324 431 (3.0%) 14,344 noncarriers received CRC. Age-specific cumulative estimated marginal-likelihood approach,7 bootstrap estimates hazard ratios (HRs) used calculate CIs. The HR generated this 1.74 0.62–3.60). To date, seven published studies frequency controls3,5,8–12 1 fig. 1). Collectively, provide information 8,546 7,949 controls. quantify pooled our data. ORs calculated use logistic regression, since five contained <5 single category. Meta-analysis methods combining weighted sum log estimates, inverse variance weight. An conditional-likelihood approach13 obtain OR. There heterogeneity (Cochran’s Q=3.74; P=.81); however, both fixed- random-effects models combine results. Under fixed-effects model, 1.26 0.99–1.60), whereas, under 1.24 0.98–1.59). finite because representation yields lower confidence bound 7.39 estimate. Alternatively, naive convention adding 0.5 empty cell generates 6.06 2.02–18.19). provides robust indicate excess compared Our is, fact, likely inflated, pathogenic some heterozygous carry variants. Hence, it additional, apparently will, reality, heterozygotes. example, Croitoru al.3 carried rare (three one case, respectively). With adjusted observations, 1.23 0.96–1.58). Figure 1.  Funnel plot variants, model. plotted order decreasing log(OR). Horizontal lines represent Each box represents point ... Our quantifying illustrates inherent problem studying low-penetrance By definition, alleles risks. If population at-risk low (i.e., <2%), then exceptionally required precisely detect comprehensively 1.2 would require 22,000 conclusion, assertion confers unsupported basis