Background: Germline variation in DNA damage response may explain variable treatment outcomes in squamous cell carcinoma of the head and neck (SCCHN). By grouping patients according to stage and radiation treatment, we compared SCCHN survival with regard to ERCC2 A35931C (Lys751Gln, rs13181) and CCND1 G870A (Pro241Pro, rs9344) genotypes.

Methods: In a hospital-based SCCHN case series (all white, 24.7% female, mean age 58.4 years), this treatment-outcome cohort study genotyped 275 stage III–IV cases that were initially treated with radiation (with or without chemotherapy) and 80 stage III–IV and 130 stage I–II cases that were initially treated without radiation or chemotherapy and used Kaplan–Meier and Cox regression analyses to compare genotype groups on the basis of overall, disease-specific, progression-free, and recurrence-free survival rates.

Results:ERCC2 35931 AA predicted worse survival in stage III–IV cases treated with radiation [multiply-adjusted HR = 1.66, 95% confidence interval (CI), 1.15–2.40; HR over the first 3 follow-up years = 1.92; 95% CI, 1.28–2.88] and better survival in stage III–IV cases not treated with radiation (HR = 0.26; 95% CI, 0.11–0.62). Although not associated with survival in stage III–IV cancers treated with radiation (HR = 1.00; 95% CI, 0.67–1.51), CCND1-870 GG predicted better survival in stage III–IV cancers not treated with radiation (HR = 0.14; 95% CI, 0.04–0.50). Survival in stage I–II did not depend on ERCC2 A35931C or CCND1 G870A genotype.

Conclusions: Although promoting tumor progression in untreated patients, germline differences in DNA-repair or cell-cycle control may improve treatment outcome in patients treated with DNA-damaging agents.

Impact:ERCC2 A35931C may help distinguish advanced stage SCCHN with better outcomes from radiation treatment. Cancer Epidemiol Biomarkers Prev; 20(11); 2429–37. ©2011 AACR.

The American Cancer Society expected nearly 50,000 oral cavity, pharynx, and larynx cancer cases in 2010, representing 3.2% of all cancer expected in the United States (1). During 2001–2007, 61% of oral cavity and pharynx cancer and 25% of larynx cancer presented at stages (regional or distant) that were associated with less than 60% 5-year relative survival rate (2). Primary treatment of these later stage cancers generally includes surgery and/or radiation, sometimes combined with chemotherapy (3). The less-than-desired results from these standard treatments motivate the search for predictors of treatment outcome. Based on the notion that tumor cells, as a result of germline genetic differences, may exhibit a different therapeutic response to the DNA damage caused by radiation and chemotherapy, a line of research aims to differentiate disease outcomes according to variation in DNA-repair and cell-cycle control genes.

In a hospital-based case series comprising 485 white patients with squamous cell carcinoma of the head and neck (SCCHN), we examined survival outcomes with regard to stage, treatment, and 2 polymorphisms, ERCC2 A35931C (Lys751Gln, rs13181) and CCND1 G870A (Pro241Pro, rs9344). ERCC2 A35931C located on exon 23 of ERCC2, a gene that codes for a DNA helicase involved in nucleotide excision repair, produces an amino acid change (LysGln) associated with DNA–repair-phenotype differences (4–6), with ERCC2 35931C generally associated with lower ERCC2 mRNA levels (6), lower DNA-repair efficiency (7), and higher DNA adduct levels (7). CCND1, the gene for a cyclin (cyclin D1) that controls the G1 to S-phase cell-cycle transition, contains the exon 4 synonymous G870A (ProPro) polymorphism. The CCND1 870A variant promotes alternative splicing of the CCND1 transcript, producing a truncated protein (cyclin D1b) with absent exon 5 (8–10). Cyclin D1b lacks 2 regulatory motifs, a PEST sequence and threonine 286, associated with protein degradation and nuclear export, respectively (10). In addition, cyclin D1b may possess intrinsic cancer-promoting properties (10). Current interest in the survival effects of these 2 polymorphisms includes not only cancer of the head and neck (11–17) but also cancers at various other sites (18–24).

Study population

We used a cohort study design to test genetic associations with SCCHN treatment outcome. Over a 7-year period (February 14, 2000, through February 14, 2008), the University Ear Nose and Throat Specialists practice at the University of Pittsburgh Medical Center recruited 686 white patients in the age range 20 to 81 years within 1 year of the diagnosis of a biopsy-verified primary SCC of the head and neck. A standardized interviewer-administered questionnaire was used to collect risk factor information (25) and blood samples were obtained for genotyping. Every case signed an informed consent (IRB #981041) form approved by the University of Pittsburgh Institutional Review Board. This case series constituted approximately one half of all 20- to 81-year-old white persons treated for a primary SCCHN by the University Ear Nose and Throat Specialists practice. Our analysis excluded 201 cases for the following 8 reasons: (i) DNA unavailable for analysis (n = 24); (ii) ERCC2 and/or CCND1 genotype assay failure (n = 21); (iii) outcome information unavailable (n = 70); (iv) enrollment after second or later primary head and neck cancer (n = 24); (v) visceral metastasis at diagnosis (n = 3); (vi) unknown stage at diagnosis (n = 6); (vii) chemotherapy as primary treatment without radiation therapy (n = 6); and (viii) radiation (with or without surgery) as primary treatment of stage I or II disease (n = 47). Compared with cases enrolled in the current study (n = 485), cases that were excluded (n = 201) were more frequently enrolled between 2006 and 2008 and less frequently enrolled between 2003 and 2005 (P = 0.01; Supplementary Table S1). In other regards, included and excluded cases were similar with respect to sex, age at diagnosis, education, cigarette smoking and alcohol-use history, and ERCC2 A35931C and CCND1 G870A genotypes (P > 0.15; Supplementary Table S1).

Genotyping

We used kits procured from Gentra Systems Inc. to isolate DNA from whole blood, a Thermo Scientific Nanodrop 1000 full-spectrum UV/visible spectrophotometer to evaluate DNA quantity and quality, and commercial TaqMan allele discrimination assays run on the ABI 7700 Sequence Detection System (Applied Biosystems) to determine ERCC2 A35931C (rs13181) and CCND1 G870A (rs9344) genotypes. Each amplification reaction included negative assay controls and positive control samples with known genotype confirmed by sequencing. Two laboratory technicians independently interpreted each assay. Assays of blind duplicate samples and repeated analysis of a 10% sample set produced completely concordant genotypic results.

Tumor and outcome information

Using inpatient and outpatient medical records, the Social Security Death Index, and a previously described computer system to control data quality (26), a dedicated and certified tumor registrar abstracted and recorded anatomic subsite, TNM stage, primary treatment, and clinical outcomes. Clinical outcome variables that were ascertained, classified, and dated included death, death due to head and neck cancer, treatments for head and neck cancer recurrence, progression, or metastasis, and treatments for new primary SCCHN. Recording disease activity [no evidence of disease (NED) vs. alive with disease (AOD)], the tumor registrar also maintained a catalog of outpatient visits to the University Ear Nose and Throat Specialists.

Study endpoints

Analyses considered 1 primary endpoint (overall survival rate) and 3 secondary endpoints (disease-specific, progression-free, and recurrence-free survival rates). The overall survival endpoint calculated survival as time between the treatment start date (date of definitive surgery or date chemotherapy or radiation therapy started for cases treated without surgery) and date of death. The disease-specific survival rate endpoint calculated survival as time between the treatment start date and date of death due to head and neck cancer. The progression-free and recurrence-free survival rate endpoints calculated survival as time between the treatment start date and the earliest (i) date of first AOD visit occurring after an NED visit; (ii) date of first treatment of head and neck cancer recurrence, progression, or metastasis; (iii) date of first treatment of new primary SCCHN; or (iv) date of death due to head and neck cancer. For cases not experiencing an endpoint, overall survival rate analyses censored follow-up on the date of last live contact. Disease-specific, progression-free, and recurrence-free survival rate analyses censored follow-up on either the date of death due to known causes unrelated to head and neck cancer or otherwise on the date of last alive contact. Analyses for recurrence-free survival included only cases with a posttreatment NED visit.

Other variables

Analyses used sex, age on treatment start date, primary head and neck tumor subsite, American Joint Committee on Cancer (AJCC) diagnostic stage group (27), and treatment category (no radiation or chemotherapy, radiation therapy combined with platinum chemotherapy, and radiation therapy with or without nonplatinum chemotherapy) to define analytic cohorts and to adjust genotype associations with clinical outcome. Analyses grouped subsites into 4 categories: (i) lip and oral cavity; (ii) oropharynx (including soft palate, base of tongue, vallecula, and tonsil); (iii) larynx (ICD-O-3 C32.0 through C32.9); and (iv) other (including nasal cavity and sinus, nasopharynx, hypopharynx, and undetermined primary subsite). A previous publication defines cigarette smoking and alcohol-use history and describes procedures used to calculate lifetime cigarette and alcohol dose exposures (25).

Statistical analysis

We based inferences on 3 cohorts: 1 cohort of primary interest (a stage III–IV cohort exposed to radiation, with or without chemotherapy, as part of initial treatment) and 2 cohorts of secondary interest (stage I–II and stage III–IV cohorts, exposed neither to radiation nor to chemotherapy as part of initial treatment). We used the product-limit (Kaplan–Meier) method to estimate survival functions and the log-rank test with a Wilcoxon weight (the number at risk ni at failure time ti) to compare survival differences according to genotype. We evaluated overall survival differences under dominant and recessive genetic models, selected the model that produced the most statistically significant survival difference in our cohort of primary interest, and based statistical inferences on a Bonferroni-corrected P value = 0.0125 = 0.05/4 (for 2 polymorphisms evaluated under 2 models). Using SAS 9.2 (PROC PHREG), we fit proportional hazards (Cox) models to control genotype-outcome associations for potential confounders and applied a log-likelihood ratio test to genotype by cofactor interaction terms to screen for effect modification. To detect departures from the proportional hazards assumption, we examined Schoenfeld residual plots and tested the statistical significance of terms representing the time-dependent interaction between genotype and the natural logarithm of survival time. If results indicated nonproportional hazards, we examined Wald-type test statistics from single-variable and multiple-variable piecewise-constant time-varying coefficients (Gray's) models (28, 29) to confirm the statistical significance of outcome differences according to genotype.

Description of study cohorts

Table 1 summarizes the 3 study cohorts, which represent SCCHN categorized according to stage at diagnosis and use of radiation as part of initial treatment. Among all cohorts, key characteristics include 24.7% female, mean age 58.4 years (SD = 10.7 years), 13% positive history of cancer, 76.7% history of cigarette smoking, mean 44.3 pack-years among ever smokers (SD = 30.0 pack-years), 86.2% history of alcohol use, and median 16,400 lifetime total drinks among ever drinkers [interquartile range (IQR) = 3,500–39,900], an amount corresponding to 1.5 drinks per day for 30 years. Male sex, younger age, and an oropharyngeal subsite primary tumor location characterized stage III–IV radiation-treated cases, the cohort of main interest.

Table 1.

Characteristics of all study subjects and of study subjects subdivided into 3 cohorts defined by stage at diagnosis and use of radiation as part of initial treatment

Stage III–IV
CharacteristicALL (n = 485)XRT (n = 275)No XRT (n = 80)Stage I–II no XRT (n = 130)Pa
Year enrolled     0.083 
2000–2002 30.7 26.9 40.0 33.1  
2003–2005 30.5 29.5 30.0 33.1  
2006–2008 38.8 43.6 30.0 33.8  
Sex     0.028 
Men 75.3 79.6 72.5 67.7  
Women 24.7 20.4 27.5 32.3  
Age, y     0.006 
<50 19.6 21.8 20.0 14.6  
50–59 33.8 38.5 25.0 29.2  
60–69 30.5 28.4 36.3 31.5  
≥70 16.1 11.3 18.8 24.6  
History of cancer     0.896 
Yes 13.0 12.4 13.8 13.8  
No 87.0 87.6 86.3 86.2  
Cigarette smoker     0.157 
Ever 76.7 74.9 85.0 75.4  
Never 23.3 25.1 15.0 24.6  
Pack-years (2)  (2)  0.283 
Never smoker 23.4 25.1 15.4 24.6  
<15 11.8 10.5 12.8 13.8  
15–39 25.5 27.6 21.8 23.1  
≥40 39.3 36.7 50.0 38.5  
Alcohol drinker (1) (1)   0.826 
Ever 86.2 85.8 85.0 87.7  
Never 13.8 14.2 15.0 12.3  
Lifetime total drinks (4) (3) (1)  0.076 
0–499 20.0 18.4 22.8 21.5  
500–4,999 19.3 15.8 25.3 23.1  
5,000–19,999 23.1 27.2 12.7 20.8  
≥20,000 37.6 38.6 39.2 34.6  
Subsite     <0.0001 
Lip and oral cavity 49.5 36.7 56.3 72.3  
Oropharynx 26.4 41.8 8.8 4.6  
Larynx 17.9 13.5 28.8 20.8  
Other 6.2 8.0 6.3 2.3  
Stage group      
19.6   73.1  
II 7.2   26.9  
III 23.7 24.7 58.8   
IVa 46.8 71.3 38.8   
IVb 2.7 4.0 2.5   
Chemotherapy (4) (4)    
None 55.5 21.0 100.0 100.0  
Platinum 36.0 63.8    
Nonplatinum 8.5 15.1    
ERCC2 A35931C (Lys751Gln)     0.566 
AA 39.2 37.1 40.0 43.1  
AC 44.7 47.6 40.0 41.5  
CC 16.1 15.3 20.0 15.4  
CCND1 G870A (Pro241Pro)     0.040 
GG 27.8 28.4 23.8 29.2  
GA 49.3 44.4 55.0 56.2  
AA 22.9 27.3 21.3 14.6  
Stage III–IV
CharacteristicALL (n = 485)XRT (n = 275)No XRT (n = 80)Stage I–II no XRT (n = 130)Pa
Year enrolled     0.083 
2000–2002 30.7 26.9 40.0 33.1  
2003–2005 30.5 29.5 30.0 33.1  
2006–2008 38.8 43.6 30.0 33.8  
Sex     0.028 
Men 75.3 79.6 72.5 67.7  
Women 24.7 20.4 27.5 32.3  
Age, y     0.006 
<50 19.6 21.8 20.0 14.6  
50–59 33.8 38.5 25.0 29.2  
60–69 30.5 28.4 36.3 31.5  
≥70 16.1 11.3 18.8 24.6  
History of cancer     0.896 
Yes 13.0 12.4 13.8 13.8  
No 87.0 87.6 86.3 86.2  
Cigarette smoker     0.157 
Ever 76.7 74.9 85.0 75.4  
Never 23.3 25.1 15.0 24.6  
Pack-years (2)  (2)  0.283 
Never smoker 23.4 25.1 15.4 24.6  
<15 11.8 10.5 12.8 13.8  
15–39 25.5 27.6 21.8 23.1  
≥40 39.3 36.7 50.0 38.5  
Alcohol drinker (1) (1)   0.826 
Ever 86.2 85.8 85.0 87.7  
Never 13.8 14.2 15.0 12.3  
Lifetime total drinks (4) (3) (1)  0.076 
0–499 20.0 18.4 22.8 21.5  
500–4,999 19.3 15.8 25.3 23.1  
5,000–19,999 23.1 27.2 12.7 20.8  
≥20,000 37.6 38.6 39.2 34.6  
Subsite     <0.0001 
Lip and oral cavity 49.5 36.7 56.3 72.3  
Oropharynx 26.4 41.8 8.8 4.6  
Larynx 17.9 13.5 28.8 20.8  
Other 6.2 8.0 6.3 2.3  
Stage group      
19.6   73.1  
II 7.2   26.9  
III 23.7 24.7 58.8   
IVa 46.8 71.3 38.8   
IVb 2.7 4.0 2.5   
Chemotherapy (4) (4)    
None 55.5 21.0 100.0 100.0  
Platinum 36.0 63.8    
Nonplatinum 8.5 15.1    
ERCC2 A35931C (Lys751Gln)     0.566 
AA 39.2 37.1 40.0 43.1  
AC 44.7 47.6 40.0 41.5  
CC 16.1 15.3 20.0 15.4  
CCND1 G870A (Pro241Pro)     0.040 
GG 27.8 28.4 23.8 29.2  
GA 49.3 44.4 55.0 56.2  
AA 22.9 27.3 21.3 14.6  

NOTE: Values given are percentages. Parentheses contain the number of observations with missing data for the characteristic tabulated.

Abbreviation: XRT, radiation treatment.

aχ2 test.

Genetics of ERCC2 A35931C (Lys751Gln) and CCND1 G870A (Pro241Pro)

In an aggregate group of all 3 cohorts, ERCC2 35931 AA, AC, and CC genotypes appeared in 39.2%, 44.7%, and 16.1% (C minor allele frequency = 0.384, SE = 0.016) and CCND1-870 GG, GA, and AA genotypes in 27.8%, 49.3%, and 22.9% (A minor allele frequency = 0.475, SE = 0.016), respectively. Neither genotype distribution violated the Hardy–Weinberg equilibrium (P > 0.2). In a white control group described elsewhere (25), but not otherwise mentioned in this article, ERCC2 35931 AA, AC, and CC genotypes appeared in 36.4%, 46.5%, and 17% (C minor allele frequency = 0.403, SE = 0.012; n = 834) of cases and CCND1-870 GG, GA, and AA genotypes in 28.6%, 48.2%, and 23.2% (A minor allele frequency = 0.473; SE = 0.012; n = 838) cases, respectively. Whether evaluated under additive, dominant, or recessive genetic models, ERCC2 A35931C and CCND1 G870A case and control genotype differences were not statistically significant (P > 0.3).

Among all cases and among stage III–IV radiation-treated cases only, ERCC2 A35931C and CCND1 G870A genotypes were statistically independent of factors such as enrollment year, sex, age, prior history of cancer, smoking and alcohol histories, pack-years, and lifetime total drinks (P > 0.05; Supplementary Table S2). As shown in Table 1, ERCC2 A35931C genotype distributions in the 3 study cohorts were statistically indistinguishable from each other. However, CCND1 G870A genotype distributions differed (P = 0.04; Table 1), with relative paucity of 870A homozygotes among stage I–II cases. This latter finding corresponds to the lower proportion of stage I–II disease observed among all CCND1 870A homozygous cases (Supplementary Table S3).

Survival outcomes

Figure 1 shows overall survival, separately for each cohort, according to ERCC2 A35931C and CCND1 G870A genotypes. The radiation-treated stage III–IV cohort comprised 275 cases, including 155 censored before death at a median interval of 4.3 years (IQR = 3.2–6.0 years) after the start of treatment. The stage III–IV cohort untreated with radiation comprised 80 cases, including 46 censored before death at a median interval of 4.3 years (IQR = 2.6–6.8 years) after the start of treatment. The stage I–II cohort comprised 130 cases, including 103 censored before death at a median interval of 3.8 years (IQR = 2.1–5.6 years) after the start of treatment.

In stage III–IV cases, the effect of genotype depended on treatment. In cases treated with radiation, overall survival was worse in ERCC2 35931A homozygotes (AA) than in cases with at least 1 35931C allele (AC + CC; P = 0.0013) and statistically indistinguishable between CCND1 870G homozygotes (GG) and cases with at least 1 870A allele (GA + AA; P = 0.49). In cases not treated with radiation, however, overall survival was better in ERCC2 35931A homozygotes than in cases with at least 1 35931C allele (P = 0.0013) as well as in CCND1 870G homozygotes than in cases with at least 1 870A allele (P = 0.0089). In stage I–II cases, the overall survival rate was independent of genotype. Although levels of statistical significance varied with the number of events, as shown in Table 2, analyses of genotype associations with disease-specific, progression-free, and recurrence-free survival were consistent in direction with the associations observed with overall survival.

Table 2.

Outcome summary, by cohort—number at risk, number of overall survival, disease-specific survival, and progression-free survival events, number with no evidence of disease activity posttreatment, number of recurrence-free survival events, and statistical significance (log-rank P values) of outcome differences according to ERCC2 A35931C and CCND1 G870A genotype

ERCC2 A35931C P valuesCCND1 G870A P value
CohortnOSDSSPFSNEDRFSOSDSSPFSRFSOSDSSPFSRFS
Stage III–IV XRT 275 120 62 95 235 63 <0.01 0.02 0.03 0.16 0.49 0.21 0.80 0.83 
Stage III–IV no XRT 80 34 18 32 71 26 <0.01 0.05 0.02 0.18 <0.01 0.17 0.02 0.01 
Stage I–II no XRT 130 27 10 40 126 38 0.78 0.98 0.79 0.59 0.71 0.41 0.58 0.54 
ERCC2 A35931C P valuesCCND1 G870A P value
CohortnOSDSSPFSNEDRFSOSDSSPFSRFSOSDSSPFSRFS
Stage III–IV XRT 275 120 62 95 235 63 <0.01 0.02 0.03 0.16 0.49 0.21 0.80 0.83 
Stage III–IV no XRT 80 34 18 32 71 26 <0.01 0.05 0.02 0.18 <0.01 0.17 0.02 0.01 
Stage I–II no XRT 130 27 10 40 126 38 0.78 0.98 0.79 0.59 0.71 0.41 0.58 0.54 

Abbreviations: DSS, disease-specific survival; n, number at risk; NED, number with no evidence of disease activity; OS, overall survival; PFS, progression-free survival; RFS, recurrence-free survival; XRT, radiation treatment.

Table 3 shows results from multiple variable models of overall survival. Controlling for sex, age, subsite, stage, platinum treatment, and CCND1 G870A genotype, ERCC2 35931A homozygosity remained a statistically significant predictor of poor overall survival in radiation-treated stage III–IV SCCHN (HR = 1.66; 95% confidence interval (CI), 1.15–2.40). Controlling for sex, age, subsite, and stage, ERCC2 35931A and CCND1 870G homozygosity both remained statistically significant predictors of better overall survival in stage III–IV SCCHN not treated with radiation (ERCC2 A35931C AA vs. AC + CC: HR = 0.26, 95% CI, 0.11–0.62; CCND1 G870A GG vs. GA + AA: HR = 0.14; 95% CI, 0.04–0.50). Screening procedures suggested a nonproportional ERCC2 A35931C hazard over time in relation to overall survival rate for radiation-treated stage III–IV SCCHN. However, the ERCC2 A35931C genotype remained a statistically significant predictor in a multiple-variable piecewise-constant time-varying coefficients model (Table 3) and in a model that truncated follow-up 3 years following entry (HR = 1.92, 95% CI, 1.28–2.88; Supplementary Table S4). In stage III–IV cases treated with radiation, interaction tests did not identify statistically significant differences in the ERCC2 A35931C polymorphism survival effect in relation to head and neck cancer subsite, stage, platinum chemotherapy, or CCND1 G870A genotype.

Table 3.

Results from multiple variable regression—multiply adjusteda risk factor associations with overall survival, estimated by proportional HRs, and 95% CIs

Stage III–IV XRT (n = 275)Stage III–IV no XRT (n = 80)
Risk factor comparisonHR95% CIPHTVCHR95% CIPHTVC
ERCC2b 
AA vs. AC + CC 1.66 1.15–2.40 <0.01 <0.01 0.26 0.11–0.62 <0.01 0.01 
CCND1c 
GG vs. GA + AA 1.00 0.67–1.51 0.99 0.26 0.14 0.04–0.50 <0.01 0.03 
Sex 
Female vs. male 1.12 0.72–1.75 0.61 0.99 0.60 0.25–1.44 0.25 0.24 
Age 
Per year of age 1.03 1.01–1.05 <0.01 <0.01 1.04 1.00–1.07 0.06 0.08 
Subsite 
Oropharynx vs. oral cavity 0.46 0.30–0.70 <0.01 <0.01 1.01 0.32–3.23 0.99 0.59 
Larynx vs. oral cavity 0.51 0.28–0.92 0.03 0.06 0.49 0.21–1.15 0.10 0.39 
Other vs. oral cavity 0.76 0.37–1.56 0.45 0.67 0.33 0.04–2.57 0.29 0.62 
Stage 
IV vs. III 1.90 1.17–3.10 <0.01 <0.01 3.49 1.53–7.95 <0.01 0.03 
Treatment 
Platinum vs. no platinum 0.91 0.63–1.32 0.62 0.70     
Stage III–IV XRT (n = 275)Stage III–IV no XRT (n = 80)
Risk factor comparisonHR95% CIPHTVCHR95% CIPHTVC
ERCC2b 
AA vs. AC + CC 1.66 1.15–2.40 <0.01 <0.01 0.26 0.11–0.62 <0.01 0.01 
CCND1c 
GG vs. GA + AA 1.00 0.67–1.51 0.99 0.26 0.14 0.04–0.50 <0.01 0.03 
Sex 
Female vs. male 1.12 0.72–1.75 0.61 0.99 0.60 0.25–1.44 0.25 0.24 
Age 
Per year of age 1.03 1.01–1.05 <0.01 <0.01 1.04 1.00–1.07 0.06 0.08 
Subsite 
Oropharynx vs. oral cavity 0.46 0.30–0.70 <0.01 <0.01 1.01 0.32–3.23 0.99 0.59 
Larynx vs. oral cavity 0.51 0.28–0.92 0.03 0.06 0.49 0.21–1.15 0.10 0.39 
Other vs. oral cavity 0.76 0.37–1.56 0.45 0.67 0.33 0.04–2.57 0.29 0.62 
Stage 
IV vs. III 1.90 1.17–3.10 <0.01 <0.01 3.49 1.53–7.95 <0.01 0.03 
Treatment 
Platinum vs. no platinum 0.91 0.63–1.32 0.62 0.70     

NOTE: Results from 2 models, one for stage III–IV tumors treated with radiation (XRT) and a second for stage III–IV not treated with radiation.

Abbreviations: PH, P value from proportional hazards model; TVC, P value from piecewise-constant time-varying coefficients model; XRT, radiation treatment.

aHRs mutually adjusted for ERCC2 A35931C genotype, CCND1 G870A genotype, sex, age, subsite, stage, and treatment.

bERCC2 A35931C genotype.

cCCND1 G870A genotype.

Finally, in subgroup analyses restricted to persons without personal history of cancer, ERCC2 35931A homozygosity persisted as a statistically significant predictor of poor overall survival in stage III–IV SCCHN treated with radiation (n = 241, log-rank P = 0.0006) and better overall survival in stage III–IV SCCHN not treated with radiation (n = 69, log-rank P = 0.0072).

In our cohort of main interest, comprising radiation-treated stage III–IV SCCHN cases, we observed worse overall survival rate in 102 patients who were homozygous for ERCC2 35931A than in 173 patients with at least 1 ERCC2 35931C allele (P = 0.0013; Fig. 1). ERCC2 35931A homozygosity predicted not only worse overall survival, but also worse disease-specific (P = 0.02) and progression-free survival (P = 0.03; Table 2). The ERCC2 A35931C genotype overall survival effect appeared limited to the first 3 years of follow-up, after which time the product-limit estimated risks of death were 48% (SE = 5%) in ERCC2 35931A homozygotes and 30% (SE = 4%) in persons with at least 1 ERCC2 35931C allele, a 1.6-fold difference. Adjustment for CCND1 genotype, sex, age, head and neck cancer subsite, stage, and platinum chemotherapy did not change this association (unadjusted HR = 1.65, 95% CI, 1.15–2.38; adjusted HR = 1.66, 95% CI, 1.15–2.40; Table 3). Finally, our statistical screens for effect modification in stage III–IV cases treated with radiation could not reject the null hypothesis of equivalent ERCC2 A35931C genotype survival effects in CCND1 870G homozygotes versus persons with at least 1 CCND1 870A allele or in oropharyngeal versus nonoropharyngeal, in stage III versus stage IV, or in platinum-treated versus nonplatinum-treated SCCHN.

Figure 1.

Overall survival (Kaplan–Meier plots) for each of 3 squamous cell carcinomas of head and neck cohorts: stage III–IV cases treated with radiation (with or without chemotherapy), stage III–IV cases not treated with radiation or chemotherapy, and stage I–II cases not treated with radiation or chemotherapy, according to ERCC2 A35931C (AA in gray) and CCND1 G870A (GG in gray). Plus (“+”) symbols signify censored observations.

Figure 1.

Overall survival (Kaplan–Meier plots) for each of 3 squamous cell carcinomas of head and neck cohorts: stage III–IV cases treated with radiation (with or without chemotherapy), stage III–IV cases not treated with radiation or chemotherapy, and stage I–II cases not treated with radiation or chemotherapy, according to ERCC2 A35931C (AA in gray) and CCND1 G870A (GG in gray). Plus (“+”) symbols signify censored observations.

Close modal

Two hypotheses can explain these results. The presence of at least 1 ERCC2 35931C allele could result in the development of tumors that respond more favorably to treatment. Alternatively, the presence of 2 ERCC2 35931A alleles could simply result in the development of tumors with greater malignant potential. However, relatively poor survival with ERCC2 35931 AA does not appear to represent a feature of SCCHN more generally. Most notably, for stage III–IV cancer not treated with radiation or chemotherapy, overall survival was better, not worse, in 32 patients with ERCC2 35931 AA than in 48 patients with ERCC2 35931 AC or CC (P = 0.0024; Fig. 1). These results (the ERCC2 35931C allele associating with better survival in patients exposed, and worse survival in those unexposed, to DNA-damaging treatment) concur with observations in esophageal cancer either treated or not treated with cisplatin (18). A simple explanation for a treatment's influence follows. Ineffective DNA damage response due to germline polymorphisms in DNA-repair or cell-cycle control genes may promote the death of tumor cells that are unable to repair the DNA damage caused by radiation and, thereby, render these tumor cells more vulnerable to radiation therapy. Absent radiation, however, ineffective DNA damage response may actually promote tumor cell genetic instability and, consequently, tumor progression. The general notion that ERCC2 35931C signifies reduced DNA repair (6, 7) is consistent with the better survival rate we observed in radiation-treated stage III–IV cases with at least one ERCC2 35931C allele.

A CCND1-870 GG survival benefit also appeared to depend on treatment, with survival difference only observed in stage III–IV disease not treated with radiation (Fig. 1). Stage I–II disease occurred less often in cases with CCND1-870 AA than in cases with CCND1-870 GG or GA (Supplementary Table S3), an occurrence that explains the relatively low frequency of the CCND1-870 AA genotype in our stage I–II cohort (Table 1). To the extent that the CCND1 870A allele may contribute to malignant potential, the association observed between CCND1-870 AA and advanced stage among all cases complements the association observed between CCND1-870 GG and better survival in stage III–IV patients with cancers not treated with radiation.

A literature search located only 2 published studies, both included in a 2008 review (12), of ERCC2 A35931C and head and neck cancer outcome. Reporting in 2005, Gal and colleagues (11) determined the ERCC2 A35931C genotype for 276 18- to 65-year-old western Washington residents (71% male, mean age: 54 years, 93% white) with oral SCC (25% oropharynx, 60% treated with radiation, stage distribution not specified) incident between 1988 and 1995. Although the ERCC2 35931 CC genotype tended to predict lower risk of subsequent second primary SCCHN (adjusted HR = 0.17, 95% CI, 0.02–1.29), the ERCC2 A35931C genotype did not predict overall or disease-specific survival rates, with survival results said to be similar for persons who received radiotherapy as part of initial treatment. Reporting in 2006, Quintela-Fandino and colleagues (16) determined the ERCC2 A35931C genotype for 103 Spanish patients (94% male, median age: 60 years) treated between 1988 and 1995 for primary or relapsed stage IV SCCHN (48% oropharynx, 18% oral cavity, 16% larynx, and 18% other). Chemotherapy-naive patients received platinum-based induction chemotherapy, with radiation in 25% of cases. Patients carrying the 35931C allele experienced better survival than patients homozygous for the 35931A allele (P = 0.0012), a result that is in agreement with our observations in stage III–IV cases treated with radiation (with or without chemotherapy).

Four other publications mention CCND1 G870A-dependent outcomes in SCCHN, with reports from (i) Germany (n = 194; ref. 14); (ii) Portugal (n = 66; ref. 15); (iii) India (n = 165; ref. 17); and (iv) the United States (n = 698; ref. 13). Each study observed worse overall and/or disease-free survival rates in relation to CCND1 870G. These studies did not examine variation in effect with regard to both stage and treatment. In contrast, we observed better, not worse, overall, progression-free, and recurrence-free survival rates in relation to CCND1-870 GG, but only within a small subgroup, stage III–IV not treated with radiation.

Study limitations include noncomparable stage III–IV cohorts, one treated with radiation and another not treated. Radiation treatment cannot be regarded as a random exposure. In our study population, for example, more recent enrollment period, very late stage (stage IV), and involvement of oropharyngeal subsite characterized the radiation-treated cohort (Table 1). Although we can observe significantly better survival in radiation-treated than in nontreated patients with at least 1 ERCC2 35931C allele and significantly worse survival in radiation-treated than in nontreated patients homozygous for ERCC2 35931A (Supplementary Figure), the noncomparability of radiation-treated and nontreated cohorts precludes isolation and measurement of the specific effects of radiation treatment. Nevertheless, our results suggest that radiation, if effective, is either more effective or only effective in patients with at least 1 ERCC2 35931C allele.

Other study limitations arose as a consequence of (i) case loss secondary to missing data; (ii) uncertainty regarding disease-specific, progression-free, and recurrence-free survival outcomes; (iii) incomplete knowledge of the extent or amount of radiation or chemotherapy delivered; (iv) unknown human papilloma virus (HPV) tumor status; (v) sample size; and (vi) polymorphism selection. Many factors contributed to case loss. However, case loss as a whole appeared to be random with regard to genotype and other prognostic variables (Supplementary Table S1). Moreover, our inclusion and exclusion criteria defined 3 distinct and relatively homogeneous cohorts that, together, encompass most of head and neck cancer clinically. Our primary analysis used a hard endpoint: death due to any cause. However, one would expect genetic factors interacting with treatment to affect, in particular, the cancer-relevant outcomes including disease-specific, progression-free, and recurrence-free survival rates. Results with these latter outcomes corroborated overall survival results (Table 2). However, determination of causes of death and ascertainment of cancer-progression events from medical records entail subjectivity and introduce variability. Our reliance on one research-dedicated and certified tumor registrar may have produced more reliable information about these difficult-to-measure secondary endpoints. More importantly, outcome determinations occurred in a manner blind to genotype. We documented radiation and type of chemotherapy started as part of primary treatment, but not the amount of radiation or doses of chemotherapy actually completed. This limitation prohibited study of genetic effect according to treatment intensity. We lacked information about HPV tumor status, a favorable prognostic factor usually observed in oropharyngeal cancer (30). However, the genotypic effects we observed on overall SCCHN survival were not only statistically equivalent in oropharyngeal and nonoropharyngeal radiation-treated stage III–IV SCCHN but also were observed in analyses restricted to oral cavity cancer (data not shown). Although our study included many cases (n = 485), sample size limited our ability to compare genotype effects between important subgroups, in particular radiation-treated stage III–IV cases subgrouped according to platinum chemotherapy. Finally, we studied only 2 polymorphisms in 2 genes. However, these genetic variants were chosen because of their known functional significance. Fully useful approaches to outcome prediction or treatment selection may require more complete genetic characterization of DNA-repair and cell-cycle control pathways (12, 16, 31).

Although genetic determinants of head and neck cancer outcome and treatment response remain poorly understood, we find that ERCC2 A35931C and, perhaps, CCND1 G870A when combined with information about stage and treatment may predict outcomes in advanced-stage SCCHN and help clinicians tailor treatment selection.

No potential conflicts of interest were disclosed.

This study received financial support through grants from NIH/NIDR P60 DE13059 and NIH/NCI 1P50CA097190.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Jemal
A
,
Siegel
R
,
Xu
J
,
Ward
E
. 
Cancer statistics, 2010
.
CA Cancer J Clin
2010
;
60
:
277
300
.
2.
National Cancer Institute. Cancer Stat Fact Sheets
.
[cited 2011 April 21]. Available from
: http://seer.cancer.gov/statfacts.
3.
National Cancer Institute. PDQ® cancer information summaries: adult treatment [cited 2011 Apr 21]; Available from
: http://www.cancer.gov/cancertopics/pdq/adulttreatment.
4.
Duell
EJ
,
Wiencke
JK
,
Cheng
TJ
,
Varkonyi
A
,
Zuo
ZF
,
Ashok
TD
, et al
Polymorphisms in the DNA repair genes XRCC1 and ERCC2 and biomarkers of DNA damage in human blood mononuclear cells
.
[Erratum appears in Carcinogenesis 2000;21:1457]
.
Carcinogenesis
2000
;
21
:
965
71
.
5.
Lunn
RM
,
Helzlsouer
KJ
,
Parshad
R
,
Umbach
DM
,
Harris
EL
,
Sanford
KK
, et al
XPD polymorphisms: effects on DNA repair proficiency
.
Carcinogenesis
2000
;
21
:
551
5
.
6.
Wolfe
KJ
,
Wickliffe
JK
,
Hill
CE
,
Paolini
M
,
Ammenheuser
MM
,
Abdel-Rahman
SZ
. 
Single nucleotide polymorphisms of the DNA repair gene XPD/ERCC2 alter mRNA expression
.
Pharmacogenet Genomics
2007
;
17
:
897
905
.
7.
Benhamou
S
,
Sarasin
A
. 
ERCC2/XPD gene polymorphisms and lung cancer: a HuGE review
.
Am J Epidemiol
2005
;
161
:
1
14
.
8.
Comstock
CES
,
Augello
MA
,
Benito
RP
,
Karch
J
,
Tran
TH
,
Utama
FE
, et al
Cyclin D1 splice variants: polymorphism, risk, and isoform-specific regulation in prostate cancer
.
Clin Cancer Res
2009
;
15
:
5338
49
.
9.
Pabalan
N
,
Bapat
B
,
Sung
L
,
Jarjanazi
H
,
Francisco-Pabalan
O
,
Ozcelik
H
. 
Cyclin D1 Pro241Pro (CCND1-G870A) polymorphism is associated with increased cancer risk in human populations: a meta-analysis
.
Cancer Epidemiol Biomarkers Prev
2008
;
17
:
2773
81
.
10.
Solomon
DA
,
Wang
Y
,
Fox
SR
,
Lambeck
TC
,
Giesting
S
,
Lan
Z
, et al
Cyclin D1 splice variants. Differential effects on localization, RB phosphorylation, and cellular transformation
.
[Erratum appears in J Biol Chem. 2004;279:30914]
.
J Biol Chem
2003
;
278
:
30339
47
.
11.
Gal
TJ
,
Huang
W-Y
,
Chen
C
,
Hayes
RB
,
Schwartz
SM
. 
DNA repair gene polymorphisms and risk of second primary neoplasms and mortality in oral cancer patients
.
[Erratum appears in Laryngoscope. 2006;116:507]
.
Laryngoscope
2005
;
115
:
2221
31
.
12.
Hopkins
J
,
Cescon
DW
,
Tse
D
,
Bradbury
P
,
Xu
W
,
Ma
C
, et al
Genetic polymorphisms and head and neck cancer outcomes: a review
.
Cancer Epidemiol Biomarkers Prev
2008
;
17
:
490
9
.
13.
Marsit
CJ
,
Black
CC
,
Posner
MR
,
Kelsey
KT
. 
A genotype-phenotype examination of cyclin D1 on risk and outcome of squamous cell carcinoma of the head and neck
.
Clin Cancer Res
2008
;
14
:
2371
7
.
14.
Matthias
C
,
Branigan
K
,
Jahnke
V
,
Leder
K
,
Haas
J
,
Heighway
J
, et al
Polymorphism within the cyclin D1 gene is associated with prognosis in patients with squamous cell carcinoma of the head and neck
.
[Erratum appears in Clin Cancer Res 1999;5:222]
.
Clin Cancer Res
1998
;
4
:
2411
8
.
15.
Monteiro
E
,
Varzim
G
,
Pires
AM
,
Teixeira
M
,
Lopes
C
. 
Cyclin D1 A870G polymorphism and amplification in laryngeal squamous cell carcinoma: implications of tumor localization and tobacco exposure
.
Cancer Detect Prev
2004
;
28
:
237
43
.
16.
Quintela-Fandino
M
,
Hitt
R
,
Medina
PP
,
Gamarra
S
,
Manso
L
,
Cortes-Funes
H
, et al
DNA-repair gene polymorphisms predict favorable clinical outcome among patients with advanced squamous cell carcinoma of the head and neck treated with cisplatin-based induction chemotherapy
.
J Clin Oncol
2006
;
24
:
4333
9
.
17.
Sathyan
KM
,
Nalinakumari
KR
,
Abraham
T
,
Kannan
S
. 
CCND1 polymorphisms (A870G and C1722G) modulate its protein expression and survival in oral carcinoma
.
Oral Oncol
2008
;
44
:
689
97
.
18.
Bradbury
PA
,
Kulke
MH
,
Heist
RS
,
Zhou
W
,
Ma
C
,
Xu
W
, et al
Cisplatin pharmacogenetics, DNA repair polymorphisms, and esophageal cancer outcomes
.
Pharmacogenet Genomics
2009
;
19
:
613
25
.
19.
Giachino
DF
,
Ghio
P
,
Regazzoni
S
,
Mandrile
G
,
Novello
S
,
Selvaggi
G
, et al
Prospective assessment of XPD Lys751Gln and XRCC1 Arg399Gln single nucleotide polymorphisms in lung cancer
.
Clin Cancer Res
2007
;
13
:
2876
81
.
20.
Horgan
AM
,
Yang
B
,
Azad
AK
,
Amir
E
,
John
T
,
Cescon
DW
, et al
Pharmacogenetic and germline prognostic markers of lung cancer
.
J Thorac Oncol
2011
;
6
:
296
304
.
21.
Le Morvan
V
,
Smith
D
,
Laurand
A
,
Brouste
V
,
Bellott
R
,
Soubeyran
I
, et al
Determination of ERCC2 Lys751Gln and GSTP1 Ile105Val gene polymorphisms in colorectal cancer patients: relationships with treatment outcome
.
Pharmacogenomics
2007
;
8
:
1693
703
.
22.
Monzo
M
,
Moreno
I
,
Navarro
A
,
Ibeas
R
,
Artells
R
,
Gel
B
, et al
Single nucleotide polymorphisms in nucleotide excision repair genes XPA, XPD, XPG and ERCC1 in advanced colorectal cancer patients treated with first-line oxaliplatin/fluoropyrimidine
.
Oncology
2007
;
72
:
364
70
.
23.
Tibaldi
C
,
Giovannetti
E
,
Vasile
E
,
Mey
V
,
Laan
AC
,
Nannizzi
S
, et al
Correlation of CDA, ERCC1, and XPD polymorphisms with response and survival in gemcitabine/cisplatin-treated advanced non-small cell lung cancer patients
.
Clin Cancer Res
2008
;
14
:
1797
803
.
24.
Bewick
MA
,
Lafrenie
RM
,
Conlon
MSC
. 
Nucleotide excision repair polymorphisms and survival outcome for patients with metastatic breast cancer
.
J Cancer Res Clin Oncol
2011
;
137
:
5430
50
.
25.
Buch
SC
,
Nazar-Stewart
V
,
Weissfeld
JL
,
Romkes
M
. 
Case-control study of oral and oropharyngeal cancer in whites and genetic variation in eight metabolic enzymes
.
Head Neck
2008
;
30
:
1139
47
.
26.
Amin
W
,
Kang
HP
,
Egloff
AM
,
Singh
H
,
Trent
K
,
Ridge-Hetrick
J
, et al
An informatics supported Web-based data annotation and query tool to expedite translational research for head and neck malignancies
.
BMC Cancer
2009
;
9
:
396
.
27.
Patel
SG
,
Shah
JP
. 
TNM staging of cancers of the head and neck: striving for uniformity among diversity
.
CA Cancer J Clin
2005
;
55
:
242
58
;
quiz 61–2
.
28.
Valenta
Z
,
Weissfeld
L
. 
Estimation of the survival function for Gray's piecewise-constant time-varying coefficients model
.
Stat Med
2002
;
21
:
717
27
.
29.
Gray
RJ
. 
Flexible methods for analyzing survival data using splines, with applications to breast cancer prognosis
.
J Am Stat Assoc
1992
;
87
:
942
51
.
30.
Marur
S
,
D'Souza
G
,
Westra
W
,
Forastiere
A
. 
HPV-associated head and neck cancer: a virus-related cancer epidemic
.
Lancet Oncol
2010
;
11
:
781
9
.
31.
Carles
J
,
Monzo
M
,
Amat
M
,
Jansa
S
,
Artells
R
,
Navarro
A
, et al
Single-nucleotide polymorphisms in base excision repair, nucleotide excision repair, and double strand break genes as markers for response to radiotherapy in patients with stage I to II head-and-neck cancer
.
Int J Radiat Oncol Biol Phys
2006
;
66
:
1022
30
.