Effect of Germ-Line Genetic Variation on Breast Cancer Survival in a Population-based Study

Study Population.

The ABC Study is an ongoing, population-based study of breast cancer in the region (19) currently covered by the East Anglian Cancer Registry in the United Kingdom (population = ∼2.2 million). Eligible cases are: (a) all cases of invasive breast cancer diagnosed at ≤65 years of age since the beginning of the study on July 1, 1996 (incident cases); and (b) all cases diagnosed at ≤55 years of age since January 1, 1991 who were still alive at the beginning of the study (prevalent cases). Cancer registry boundary changes have meant that some prevalent cases diagnosed before 1995 were identified through the North Thames Cancer Registry. Cases diagnosed after January 1, 1991 who did not survive until July 1, 1996 are not included in the study, and so the data are left-truncated. Approximately 70% of eligible cases have enrolled in the study. The study is approved by the relevant Local Research Ethics Committees. There were 2473 participants enrolled in the ABC Study at the time of this analysis; all of the patients participating provided informed consent, submitted a blood sample, and completed a comprehensive questionnaire.

Genotyping.

Genotyping was performed in the course of ongoing analyses of candidate low penetrance breast cancer susceptibility genes. Cases were genotyped for the common SNPs and the VNTR polymorphisms shown in Table 2 ⇓ . The number of cases genotyped varied for different polymorphisms because genotyping and case collection is ongoing, and fewer cases were available when the first polymorphisms were assayed >5 years ago [e.g., CYP19 (ttta)_n and CYP17 c−34t]. Most SNPs were genotyped using a fluorescent 5′ exonuclease assay (Taqman) and the ABI PRISM 7700 Sequence Detection System (PE Biosystems). In our laboratory, repeated genotyping of 14 different SNPs by this methods has been found to give >98% reproducibility. 5 Allele sizing on an ABI373 sequencer (PE Biosystems) was used to genotype VNTRs. Additional details of SNP identification and genotyping procedures are available in the references shown in Table 2 ⇓ or from the authors. BRCA1 and BRCA2 mutation analysis was carried out by heteroduplex analysis with sequencing confirmation as described previously (20) .

Survival Data.

Vital status was available for 2430 (98.2%) of the 2473 genotyped study participants through the cancer registries. Twenty (0.8%) patients could not be identified in either registry, and 23 (0.9%) were identified but had only provisional (recent) registration status with no vital status data. Both registries actively obtain vital status data of registered individuals through mail and telephone follow-up at 3 and 5 years after diagnosis (and subsequently at 5-year intervals). This active follow-up involves searches through hospital information systems for recent visits, and when no recent visit is observed then the general practitioner of the case is asked by letter to indicate vital status. In addition, the registries obtain notification of deaths through death certification flagging with the Office for National Statistics; lag time is a few weeks for cancer deaths and 2 months to a year for noncancer deaths.

Statistical Methods.

Cox regression analysis (21) allowing for left-truncated data were used to estimate genotype-specific HRs and 95% CIs in Stata version 6.0. The proportional hazards assumption was examined using log-log plots and tested formally by adding a time x genotype interaction term to the model. Time at risk began at date of blood sample receipt and ended at date of death from any cause, or, if death did not occur, September 30, 2000 (3 months before the start of this analysis). For each SNP, HRs were calculated with common allele homozygotes as the referent group (unless otherwise specified), and Ps presented indicate the significance of a test for heterogeneity among all three of the genotype groups (common allele homozygote, heterozygote, and rare allele homozygote).

Data on age at diagnosis, cancer stage, histological grade, and tumor type, was available from the registries for 1709 cases (70%). This enabled us to examine the significance of each polymorphism conditional on known prognostic factors. Factors were grouped as follows: age at diagnosis (< 45, 45–49, 50–53, and >53 years); tumor grade (well-differentiated, moderately differentiated, and poorly/undifferentiated); stage (Tumor-Node-Metastasis stages I-IV; Ref. 22 ); and histological subtype (ductal, lobular, tubular, medullary, mucinous, metaplastic, and other). Initial analyses were not controlled for stage, grade, tumor type, or age at diagnosis, because any effect of genotype may act through these prognostic variables. If an unadjusted HR was significant at the α = 0.05 level, likelihood ratio testing of the inclusion of covariates and interaction terms was performed to determine the best-fitting model. To avoid the proportional hazards assumption for other covariates, we then used “true” stratification for prognostic factors (as opposed to adjustment by inclusion of dummy covariates), so that baseline hazards could vary by strata (23) .

Protein-truncating Mutations in BRCA1 and BRCA2.

BRCA1 and BRCA2 mutation screening was performed on 1370 cases. Penetrance and prevalence estimates from this population have been reported previously (20) . Ten cases were observed to have a mutation in BRCA1: 1619deltaaat, 2379delg, 2774delc, 2800delaa, 3596delaaag, 3874deltgtc, 3879del4, 4158delag, 4184deltcaa, and 4912 + 2delt. Unadjusted survival analysis showed that BRCA1 mutation carriers had poorer survival compared with noncarriers and untested cases [HR, 4.14 (1.32 – 13.0); P = 0.047] (Fig. 1) ⇓ . Restricting analyses to the 1370 women screened produced similar results. BRCA1 mutation carriers were significantly younger at diagnosis (P = 0.003) and had higher grade tumors (P = 0.02), but there was no significant association with stage or tumor type (P = 0.17 and P = 0.08, respectively). After adjustment for grade and tumor type, the effect of BRCA1 status was reduced and no longer significantly associated with prognosis [HR_stratified, 1.99 (0.47–8.45)].

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Survival curves for the common homozygotes are shown as a thick, solid line. The thin, broken line depicts the survival curve for the heterozygotes, and the thin, dashed line, the survival curve for the rare homozygotes. There were too few rare homozygotes to estimate survival probabilities for CHK2 and CYP1A1, and this category is not applicable for BRCA1/2 mutation carriers. Note that the curves do not begin at time = 0, because patients are not at risk until they enter the study (this is particularly noticeable for CYP17 and VDR, where cases from the prevalent series only were genotyped).

Nineteen cases were found to carry a BRCA2 mutation: 1215insa, 1537delaaga, 2023delttact, 253delc, 295 + 3a>g, 3034delaaac, 4538dela, 4867delg, 5849delttta, 6174delt, 6503deltt (n = 3), 6719deltg, 7061deltctta, 9303ins31 (n = 2), 983delacag, and 9276dela. BRCA2 mutation carriers were younger at diagnosis (P = 0.005) and had more advanced stage disease (P = 0.007) than noncarriers. Survival of mutation carriers did not differ from noncarriers (P = 0.47).

Polymorphisms in DNA DSB Repair Genes.

Ligase IV is involved with nonhomologous end joining repair of DSBs. A silent t>c polymorphism in codon 501 (D501D) of the ligase IV gene (LIG4) was significantly associated with survival after breast cancer diagnosis (P = 0.002). Rare allele homozygotes (cc; n = 36) showed a 4-fold risk of death compared with common allele homozygotes (tt; n = 1,440) [HR, 4.01 (2.09–7.68)]. LIG4 heterozygotes (tc; n = 538) had a risk similar to that of the common allele homozygotes (tt) [HR, 1.01 (0.70–1.45)]. Additional analyses were performed by comparing rare allele homozygotes with the other two groups combined. Variables for age at diagnosis, grade, stage, tumor type, and prevalent/incident status were then included in the Cox regression analysis. After excluding factors not significant at the 5% level, the best fitting model included grade, stage, and tumor type. With adjustment for these variables, the rare allele homozygotes (cc; n = 26) remained at significantly increased risk of death compared with other genotypes (tt and tc; n = 1444) [HR_stratified, 4.15 (1.85–9.30); P = 0.003]. There were no significant interactions between LIG4 genotype and grade, stage, or tumor type.

Like LIG4, KU70 is involved in the repair of DNA double-strand breaks through nonhomologous end joining. A silent g>t polymorphism at codon 593 did not affect breast cancer survival (P = 0.78). Six genes involved in homologous recombinational repair of DNA double-strand breaks were examined: ATM, BRCA1, BRCA2, RAD51, RAD52, and XRCC3. No polymorphism in these genes was associated with survival after breast cancer (Table 2) ⇓ . The gene for Nijmegen breakage syndrome (NBS1) is thought to be involved in DSB repair and cell cycle checkpoint functions. Four polymorphisms in NBS1 were examined; although rare homozygotes for all four of the polymorphisms appeared to have decreased risk of death compared with common homozygotes (HRs ranged from 0.58 to 0.79), no difference in risk was significant.

TP53 and CHK2 play roles in the cellular response to DSBs through the arrest of the cell cycle. We found a protective effect against death after breast cancer among cases that carried the proline allele of the TP53 R72P polymorphisms [heterozygote HR, 0.67 (0.48–0.94); rare allele homozygote HR, 0.77 (0.40–1.47)], although no significant heterogeneity existed among genotypes (P = 0.056). Considering heterozygotes and Pro-Pro homozygotes together, the estimated HR compared with Arg-Arg homozygotes was 0.68 (0.50–0.94; P = 0.018). Inclusion of known prognostic variables in the model reduced the apparent protective effect of this TP53 polymorphism [Pro carriers HR, 0.78 (0.52–1.15)]. The polymorphisms in CHK2 did not appear to affect survival after breast cancer.

Polymorphisms in Hormone Metabolism Genes.

There were 350 women genotyped at the intron 4 (ttta)_n repeat in CYP19. There was a significant improvement in survival with increasing repeat length considered as a continuous covariate [HR, 0.85 (0.75–0.96); P = 0.01]. However, stratification by grade, stage, and tumor type reduced the dataset to 154 cases, and the effect was no longer significant [HR, 0.87 (0.70–1.08); P = 0.20)]. Data on 1527 women were available for the CYP19 c>t polymorphism, which had no effect on survival.

The CYP17 promoter t-34c had a marginally significant effect on breast cancer survival (P = 0.05). Heterozygotes had significantly increased hazard of death [HR, 1.81 (1.10–2.97)] compared with CYP17 common homozygotes, but this was no longer significant after adjustment for stage, grade, and tumor type [HR, 1.70 (0.86–3.37)]. An increased hazard in rare homozygotes was not significant [HR, 1.38 (0.70–2.75)]. Nonsignificant increased risks were also observed for the VDR gene (Table 2) ⇓ . None of the four SNPs in COMT was associated with survival nor were the total number of repeats at the AR exon 1 repeat polymorphisms [poly(Gln)_n and poly(Gly)_n] associated with time to death.

Polymorphisms in Other Genes.

Four genes involved in the metabolism of carcinogens and other xenobiotics were analyzed: ADH3, which catalyzes the oxidation and reduction of alcohols and aldehydes; CYP1A1, which is involved with Phase I detoxification of polycyclic aromatic compounds; GSTP1, which is involved with Phase II detoxification of carcinogens and their intermediates; and NQO1, which detoxifies quinones derived from oxidation of phenolic metabolites of benzene. In addition, TGF-β and E-cadherin (CDH1) polymorphisms were assessed. No effect on survival by polymorphisms in any of these genes was observed (Table 2) ⇓ .