Skip to Content
Discovering the causes of cancer and the means of prevention

Scientific Highlights

July 2012 - Linkage Newsletter


All Cancers

Clonal Mosaicism, Aging, and Cancer

In an analysis of 31,717 cancer cases and 26,136 cancer-free controls from 13 GWAS, the authors observed large chromosomal abnormalities in a subset of clones in DNA obtained from blood or buccal samples. Mosaic abnormalities, either aneuploidy (an abnormal number of chromosomes) or loss of heterozygosity (diversity in genes and possible loss of the normal function of a gene due to mutations), larger than 2 megabases were observed in non-sex chromosomes of 517 individuals (0.89%). In cancer-free individuals, frequency increased with age (see Figure 1), from 0.23% at under 50 years to 1.91% between 75 and 79 years (p = 4.8 × 10−8). Mosaic abnormalities were more common in individuals with solid tumors (0.97% versus 0.74% in cancer-free individuals; OR = 1.25). Detectable mosaicism also was more common in individuals from whom DNA was collected at least one year before leukemia diagnosis than in cancer-free individuals (OR = 35.4). These findings underscore the time-dependent nature of somatic events in the etiology of cancer and potentially other late-onset diseases. (Jacobs KB, Yeager M, Zhou W, et al. Detectable clonal mosaicism and its relationship to aging and cancer. Nat Genet 2012; May 6 [E-pub ahead of print])

Figure 1 shows the frequency of detectable clonal mosaic events by age and cancer status. The analysis excluded 1,000 individuals with unknown age at DNA collection.

Back to Topics

All-Cause Mortality

Coffee Drinking

The investigators examined the association of coffee drinking with subsequent total and cause-specific mortality among 229,119 men and 173,141 women in the NIH-AARP Diet and Health Study who were 50–71 years of age at baseline. Participants with cancer, heart disease, and stroke were excluded. During follow-up between 1995 and 2008, a total of 33,731 men and 18,784 women died. In age-adjusted models, the risk of death was increased among coffee drinkers. However, coffee drinkers were more likely to smoke, and after adjustment for tobacco-smoking status and other potential confounders, there was a significant inverse association between coffee consumption and mortality. Adjusted HRs for death among men who drank coffee compared with those who did not were as follows: 0.99 for drinking less than one cup per day, 0.94 for one cup, 0.90 for two or three cups, 0.88 for four or five cups, and 0.90 for six or more cups of coffee per day; the respective HRs among women were 1.01, 0.95, 0.87, 0.84, and 0.85. Inverse associations were observed for deaths due to heart disease, respiratory disease, stroke, injuries and accidents, diabetes, and infections, but not for deaths due to cancer. (Freedman ND, Park Y, Abnet CC, et al. Association of coffee drinking with total and cause-specific mortality. N Engl J Med 2012;366:1891–1904)

Sedentary Behaviors

In a follow-up study of 240,819 adults in the NIH-AARP Diet and Health Study, the authors found that higher amounts of sitting time and television viewing were associated with mortality, even after adjustment for age, sex, education, smoking, diet, race, and moderate-vigorous physical activity (MVPA). Compared with participants who reported viewing television for less than one hour daily, participants who reported viewing television seven or more hours daily were at greater risk of all-cause (HR = 1.61), cardiovascular (HR = 1.85), and cancer (HR = 1.22) mortality after adjustment for MVPA. Overall sitting time was associated with all-cause mortality. Even for adults reporting more than seven hours per week of MVPA, high television viewing time remained associated with increased risk of all-cause (HR = 1.47) and cardiovascular (HR = 2.00) mortality. (Matthews CE, George SM, Moore SC, et al. Amount of time spent in sedentary behaviors and cause-specific mortality in US adults. Am J Clin Nutr 2012;95:437–445)

Back to Topics

Bladder Cancer

Genetic Variants in the PSCA Gene

GWAS have identified a SNP, rs2294008, on 8q24.3 within the prostate stem cell antigen (PSCA) gene as a risk factor for bladder cancer. To fine map this region, the authors imputed 642 SNPs within 100 kilobases of rs2294008 and 33 markers genotyped in one of the reported GWAS in 8,652 subjects. A multivariable logistic regression model adjusted for rs2294008 revealed a unique signal, rs2978974. In the combined analysis of 5,393 cases and 7,324 controls, the authors detected a per-allele OR of 1.11 for rs2294008 and 1.07 for rs2978974. The effect was stronger in carriers of both risk variants (OR = 1.24), and there was a significant multiplicative interaction (p = 0.035) between these two SNPs, which requires replication in future studies. The T risk allele of rs2294008 was associated with increased PSCA mRNA expression in two sets of bladder tumor samples and in normal bladder samples (see Figure 2), but rs2978974 was not associated with PSCA expression. A joint effect of two PSCA SNPs, rs2294008 and rs2978974, suggests that both variants may be important for bladder cancer susceptibility, possibly through different mechanisms that influence the control of mRNA expression and interaction with regulatory factors. (Fu YP, Kohaar I, Rothman N, et al. Common genetic variants in the PSCA gene influence gene expression and bladder cancer risk. Proc Natl Acad Sci USA 2012;109:4974–4979)

Figure 2 shows PSCA mRNA expression for rs2294008. Total PSCA expression is shown in bladder tumor tissues measured with microarrays, in bladder tumor tissues measured with an expression assay PSCA-123, and in normal bladder tissues measured with an expression assay PSCA-123.

Mapping of the UGT1A Locus Identifies Protective Variant

A recent GWAS of bladder cancer identified the UGT1A gene cluster on chromosome 2q37.1 as a novel susceptibility locus. The association within the UGT1A locus was detected by the SNP rs11892031. The authors performed detailed resequencing, imputation, and genotyping in this region, clarifying the original genetic association detected by rs11892031 and identifying an uncommon SNP, rs17863783, that explained and strengthened the association in this region. The authors found the protective T allele of rs17863783 to be associated with increased mRNA expression of UGT1A6.1 both in vitro and in human liver tissue samples. Rs17863783 may protect from bladder cancer by increasing the removal of carcinogens from bladder epithelium by the UGT1A6.1 protein. (Tang W, Fu YP, Figueroa JD, et al. Mapping of the UGT1A locus identifies an uncommon coding variant that affects mRNA expression and protects from bladder cancer. Hum Mol Genet 2012;21:1918–1930)

Back to Topics

Brain Cancer

Mobile Phone Use and Glioma Risk

In view of mobile phone exposure’s classification as a possible human carcinogen by the International Agency for Research on Cancer, the authors compared the observed U.S. incidence trends for glioma (a form of brain cancer) with expected rates based on risks reported in two European studies. Projected rates were estimated by combining relative risks reported in the 2010 Interphone study and a 2011 Swedish study with U.S. population data from the Surveillance, Epidemiology and End Results program. Age-specific incidence rates of glioma in the U.S. remained generally constant from 1992 to 2008 (−0.02% change per year), a period coinciding with a substantial increase in mobile phone use. Based on relative risks of glioma by tumor latency and cumulative hours of phone use in the Swedish study, predicted U.S. rates should have been at least 40% higher than observed rates in 2008 (see Figure 3). The authors determined that increased risks of glioma with mobile phone use, as reported by the Swedish study, are inconsistent with observed incidence trends in U.S. population data, although the U.S. data could be consistent with the modest excess risks reported in the Interphone study. (Little MP, Rajaraman P, Curtis RE, et al. Mobile phone use and glioma risk: Comparison of epidemiological study results with incidence trends in the United States. BMJ 2012;344:e1147)

Figure 3 shows observed and projected rates of malignant glioma in non-Hispanic white people in the United States using the relative risks, periods of latency, and cumulative hours of phone use from the Swedish study and the Interphone study.

Back to Topics

Breast Cancer

Estrogen Metabolism

The authors evaluated the influence of estrogen metabolism in a case-control study of breast cancer nested in the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial. Participants included 277 women who developed invasive breast cancer and 423 matched control subjects; all were aged 55–74 years, postmenopausal, and not using hormone therapy at PLCO baseline. Liquid chromatography-tandem mass spectrometry was used to measure baseline serum concentrations of 15 estrogens and metabolites. Most estrogens, estrogen metabolites, and metabolic pathway groups were associated with increased breast cancer risk; the serum concentration of unconjugated estradiol was strongly associated (HR = 2.07). No estrogen, estrogen metabolite, or metabolic pathway group remained significantly associated with breast cancer risk after adjusting for unconjugated estradiol. The ratio of the 2-hydroxylation pathway to parent estrogens (HR = 0.66) and the ratio of 4-hydroxylation pathway catechols to 4-hydroxylation pathway methylated catechols (HR = 1.34) were associated with breast cancer risk. More extensive 2-hydroxylation of parent estrogens was associated with lower risk, and less extensive methylation of potentially genotoxic 4-hydroxylation pathway catechols was associated with higher risk. (Fuhrman BJ, Schairer C, Gail MH, et al. Estrogen metabolism and risk of breast cancer in postmenopausal women. J Natl Cancer Inst 2012;104:326–339)

Fine Mapping of 14q24.1

In the Cancer Genetic Markers of Susceptibility (CGEMS) GWAS, a SNP marker, rs999737, in the 14q24.1 interval was associated with breast cancer risk. To fine map this region, the authors imputed a 3.93 megabase region flanking rs999737 for stages 1 and 2 of the CGEMS study (5,692 cases, 5,576 controls) using the combined reference panels of the HapMap 3 and the 1000 Genomes Project. They performed single-marker association testing and variable-sized sliding window haplotype analysis. For both analyses, the initial tagging SNP rs999737 retained the strongest association with breast cancer risk. Investigation of contiguous regions did not reveal evidence for an additional independent signal. The authors concluded that rs999737 is an optimal tag SNP for common variants in the 14q24.1 region, thereby reducing the number of candidate variants that should be investigated in follow-up evaluation. (Lee P, Fu YP, Figueroa JD, et al. Fine mapping of 14q24.1 breast cancer susceptibility locus. Hum Genet 2012;131:479–490)

Risk for Hispanic Women

The authors evaluated the use of the Breast Cancer Risk Assessment Tool (BCRAT) for Hispanic women. The model, which estimates a woman’s risk of developing invasive breast cancer over a defined period of time given her age and risk factor profile, combined 1990–1996 breast cancer incidence for Hispanic women with relative risks for breast cancer risk factors from non-Hispanic white women. The authors compared the relative risks and calibration of BCRAT risk projections of 6,353 Hispanic and 128,976 non-Hispanic white postmenopausal participants aged 50 and older in the Women’s Health Initiative (WHI). Calibration was assessed by the ratio of the number of breast cancers observed (O) to the number expected (E) by the BCRAT (O/E). The authors re-evaluated calibration for an updated BCRAT that combined BCRAT relative risks with 1993–2007 breast cancer incidence, which is contemporaneous with the WHI. In the WHI Main Study, the BCRAT underestimated the number of breast cancers by 18% in both Hispanics (O/E = 1.18) and non-Hispanic whites (O/E = 1.18). Updating the BCRAT improved calibration for Hispanic women (O/E = 1.08) and non-Hispanic white women (O/E = 0.98). For Hispanic women, relative risks for number of breast biopsies (1.71 vs. 1.27) and age at first birth (0.97 vs. 1.24) differed between the WHI and BCRAT. The modest discriminatory accuracy of the BCRAT for Hispanic women might be improved by using risk factor relative risks specific to Hispanic women. (Banegas MP, Gail MH, LaCroix A, et al. Evaluating breast cancer risk projections for Hispanic women. Breast Cancer Res Treat 2012;132:347–353)

Back to Topics

Cervical Cancer

HPV Infection, Cytologic Abnormalities, and Co-testing

The authors assessed the longitudinal relationship of abnormal cytology and HPV positivity in a seven-year prospective study of 2,500 women in Guanacaste, Costa Rica. At each semiannual or annual visit, cervical specimens were screened using liquid-based cytology and tested for more than 40 HPV types using MY09/MY11 L1 degenerate primer polymerase chain reaction–based methods. Based on previous work, the authors separated prevalent and newly detected infections in younger and older women. Among newly detected HPV-positive or cytology-positive events, HPV and cytology appeared together 58.2% of the time; when discordant, HPV tended to appear before cytology in younger and older women. Combining newly detected and prevalent events, HPV and cytology disappeared at the same time more than 70% of the time. When discordant, HPV tended to disappear after cytology in younger and older women. Detection of HPV DNA and associated cytological abnormalities tend to begin and end together. However, when discordant, detection of HPV DNA tends to precede and/or last longer than associated cytologic abnormalities. (Markt SC, Rodriguez AC, Burk RD, et al. Longitudinal analysis of carcinogenic human papillomavirus infection and associated cytologic abnormalities in the Guanacaste Natural History Study: Looking ahead to cotesting. J Infect Dis 2012;205:498–505)

Methylation of HPV 16 Genome

Previous studies have suggested an association between HPV 16 genome methylation and both cervical intraepithelial neoplasia (CIN) 3 and cancer, but the results have been inconsistent. The authors designed a nested case-control study within a large prospective cohort of women who underwent multiple screenings for cervical cancer in Guanacaste, Costa Rica. The authors collected diagnostic specimens at the time of CIN 3 diagnosis and persistent HPV 16 infection, prediagnostic specimens at the first HPV 16–positive screening visit, and control specimens from women with infection clearance within two years. Increased methylation in diagnostic vs. control specimens at nine CpG sites, or regions of DNA where a cytosine nucleotide occurs next to a guanine nucleotide, was associated with an increased risk of CIN 3 and persistence. High methylation at three of these CpG sites was associated with a much higher risk when combined compared with low methylation at these sites (OR = 52). In this HPV 16–infected cohort, increased methylation of CpG sites within the HPV 16 genome before diagnosis and at the time of diagnosis was associated with cervical precancer. (Mirabello L, Sun C, Ghosh A, et al. Methylation of human papillomavirus type 16 genome and risk of cervical precancer in a Costa Rican population. J Natl Cancer Inst 2012;104:556–565)

Back to Topics

Esophageal Cancer

Genotypic Variants at 2q33

The authors conducted a meta-analysis of susceptibility loci that showed nominally significant p values in two previously published genome-wide scans of esophageal squamous cell carcinoma. The meta-analysis revealed five SNPs at 2q33 with p < 5 × 10−8, and the strongest signal was for rs13016963 (combined OR = 1.29, p = 7.63 × 10−10). An imputation analysis of 4,304 SNPs at 2q33 suggested a single association signal, and the strongest imputed SNP associations were similar to those from the genotyped SNPs. The authors conducted an ancestral recombination graph analysis to identify haplotypes that harbor the variants directly responsible for the detected association signal. This showed that the 5 SNPs exist in a single haplotype along with 45 imputed SNPs in strong linkage disequilibrium, and the strongest candidate was rs10201587, one of the genotyped SNPs. The meta-analysis found genome-wide significant SNPs at 2q33 that map to the CASP8/ALS2CR12/TRAK2 gene region. Future studies of esophageal and other cancers should focus on comprehensive sequencing of this 2q33 locus and functional analysis of rs13016963, rs10201587, and other strongly correlated variants. (Abnet CC, Wang Z, Song X, et al. Genotypic variants at 2q33 and risk of esophageal squamous cell carcinoma in China: A meta-analysis of genome-wide association studies. Hum Mol Genet 2012;21:2132–2141)

Non-steroidal Anti-inflammatory Drug Use

The authors pooled data from six population-based studies within the Barrett’s and Esophageal Adenocarcinoma Consortium (BEACON) to evaluate the association between non-steroidal anti-inflammatory drug (NSAID) use and risks of esophageal adenocarcinoma (EAC) and esophagogastric junctional adenocarcinoma (EGJA). Compared with nonusers, individuals who had used NSAIDs had a reduced risk of EAC (OR = 0.68) and possibly of EGJA (OR = 0.83). Similar reductions were observed among individuals who took aspirin and those who took non-aspirin NSAIDs. Daily or more frequent NSAID use and 10 or more years of use showed ORs of 0.56 and 0.63, respectively, for EAC risk. Although reverse causation could partly explain the inverse association between NSAID use and EAC risk, these results suggest a possible role for NSAIDs in prevention of EAC and EGJA. (Liao LM, Vaughan TL, Corley DA, et al. Nonsteroidal anti-inflammatory drug use reduces risk of adenocarcinomas of the esophagus and esophagogastric junction in a pooled analysis. Gastroenterology 2012;142:442–452)

Back to Topics

Eye Cancer

Family History and Second Cancer Risk

Investigators evaluated the risk of second cancer in a retrospective cohort of 1,852 retinoblastoma (RB) survivors of at least one year, classifying the RB1 germline mutation, inherited or de novo, by inference from laterality of RB and positive family history. In bilateral survivors, the investigators found a relative risk (RR) of 1.37 for second cancers associated with a family history of RB, adjusted for treatment, age, and length of follow-up. The risk for melanoma was significantly elevated for survivors with a family history of RB (RR = 3.08), but the risk for soft tissue and bone sarcomas was not. The cumulative incidence of second cancers 50 years after diagnosis of bilateral RB, with adjustment for competing risk of death, was significantly higher for survivors with a family history (47%) than survivors without a family history (38%). (Kleinerman RA, Yu CL, Little MP, et al. Variation of second cancer risk by family history of retinoblastoma among long-term survivors. J Clin Oncol 2012;30:950–957)

Back to Topics

Gastric Cancer

Effects of Helicobacter pylori, Garlic, and Vitamins

In an extension of the Shandong Intervention Trial to 14.7 years of follow-up, two weeks of antibiotic treatment for Helicobacter pylori (H. pylori) reduced the prevalence of precancerous gastric lesions, whereas 7.3 years of oral supplementation with garlic extract and oil (garlic treatment) or vitamin C, vitamin E, and selenium (vitamin treatment) did not. In this masked factorial placebo-controlled trial of 3,365 randomly assigned subjects, gastric cancer was diagnosed in 3.0% of subjects who received H. pylori treatment and in 4.6% of those who received placebo (OR = 0.61). Gastric cancer deaths occurred among 1.5% of subjects assigned H. pylori treatment and among 2.1% of those assigned placebo (HR of death = 0.67). Garlic and vitamin treatments were associated with non-significant reductions in gastric cancer incidence and mortality. In this high-risk population, short-term antibiotic treatment significantly reduced the incidence of gastric cancer by 39% over 15 years of follow-up. (Ma JL, Zhang L, Brown LM, et al. Fifteen-year effects of Helicobacter pylori, garlic, and vitamin treatments on gastric cancer incidence and mortality. J Natl Cancer Inst 2012;104:488–492)

Back to Topics

Lung Cancer

Body Mass Index and Smoking

The authors prospectively examined the association between body mass index (BMI) and the risk of lung cancer among 448,732 men and women in the NIH-AARP Diet and Health Study. BMI was inversely associated with the risk of lung cancer among both men and women (BMI ≥ 35 vs. 22.5–24.99 kg/m2: HR = 0.81 and 0.73, respectively). The inverse association was restricted to current and former smokers and was stronger after adjustment for smoking. Among smokers, the inverse association persisted even after finely stratifying on smoking status, time since quitting smoking, and number of cigarettes smoked per day. Sensitivity analyses did not support the possibility that the inverse association was due to prevalent undiagnosed disease. Results suggest that a higher BMI is associated with a reduced risk of lung cancer among current and former smokers. (Smith L, Brinton LA, Spitz MR, et al. Body mass index and risk of lung cancer among never, former, and current smokers. J Natl Cancer Inst 2012;104:778–789)

Diesel Exhaust in Miners Case-control Study

See article on pages 1–3 of this issue of Linkage. (Silverman DT, Samanic CM, Lubin JH, et al. The Diesel Exhaust in Miners Study: A nested case-control study of lung cancer and diesel exhaust. J Natl Cancer Inst 2012; March 5 [E-pub ahead of print])

Diesel Exhaust in Miners Cohort Study

See article on pages 1–3 of this issue of Linkage. (Attfield MD, Schleiff PL, Lubin JH, et al. The Diesel Exhaust in Miners Study: A cohort mortality study with emphasis on lung cancer. J Natl Cancer Inst 2012; March 5 [E-pub ahead of print])

Back to Topics


Bayesian Model for Studying Gene-environment Interaction

Investigators developed an approach for analyzing how a gene or chromosomal region associated with a disease interacts with an established environment risk factor to influence the disease risk. The approach was based on a Bayesian model that uses a latent genetic profile variable to capture all of the genetic variation in the entire targeted region and allows the environment effect to vary across different genetic profile categories, rather than analyzing one genetic marker at a time. Using data collected in the Environment and Genetics in Lung Cancer Etiology (EAGLE) study, investigators applied the Bayesian model to evaluate the joint effect of smoking intensity and genetic variants in the 15q25.1 region, which contains a cluster of nicotinic acetylcholine receptor genes and has been shown to be associated with both lung cancer and smoking behavior. Investigators found evidence for gene-environment interaction (p = 0.016), and the smoking effect appeared to be stronger in subjects with a genetic profile associated with a higher lung cancer risk. The conventional test of gene-environment interaction based on the single-marker approach is far from significant. (Yu K, Wacholder S, Wang Z, et al. A flexible Bayesian model for studying gene-environment interaction. PLoS Genet 2012;8:e1002482)

Testing Gene-environment Interaction

Several methods for screening for gene-environment interaction that address the issue of assuming gene-environment independence have recently been proposed. In this report, the authors present a comparative simulation study of power and type I error properties of three classes of procedures: (1) the standard one-step case-control method; (2) the case-only method that requires an assumption of gene-environment independence; and (3) a variety of hybrid methods that aim to gain power by exploiting the assumption of gene-environment independence but can protect against false positives when the independence assumption is violated. These analyses suggest that although the case-only method generally has maximum power, it can create substantial false positives in large-scale studies even when a small fraction of markers are associated with the exposure in the population. All of the hybrid methods protect well against such false positives and can retain power advantages over standard case-control tests. For future genome-wide scans for gene-environment interactions, researchers can attain major power gain by using alternatives to standard case-control analysis. Whether a case-only scan or a hybrid method should be used depends on the strength and direction of gene-environment interaction and association, tolerance for false positives, and the nature of replication strategies. (Mukherjee B, Ahn J, Gruber SB, and Chatterjee N. Testing gene-environment interaction in large-scale case-control association studies: Possible choices and comparisons. Am J Epidemiol 2012;175:177–190)

Back to Topics

Oropharyngeal Cancer

Oral HPV Infection

HPV infection is the main cause of a subset of oropharyngeal squamous cell carcinoma. To determine the prevalence of oral HPV infection in the United States, a cross-sectional study was conducted as part of the National Health and Nutrition Examination Survey 2009–2010. Participants aged 14–69 provided a 30-second oral rinse and gargle with mouthwash. DNA purified from oral exfoliated cells was evaluated for detection of HPV types. Demographic and behavioral data were obtained by interview. The overall prevalence of oral HPV infection was 6.9%, and HPV 16 prevalence was 1.0%. Oral HPV infection followed a bimodal pattern for age with peak prevalence in individuals aged 30–34 years (7.3%) and 60–64 years (11.4%). Men had a higher prevalence than women for any oral HPV infection (10.1% vs. 3.6%; unadjusted prevalence ratio [PR] = 2.80). Infection was more common among those with a history of any type of sexual contact than those without (7.5% vs. 0.9%; PR = 8.69) and increased with number of sexual partners and cigarettes smoked per day. Associations with age, sex, number of sexual partners, and current cigarettes smoked per day were independently associated with oral HPV infection in multivariable models. (Gillison ML, Broutian T, Pickard RK, et al. Prevalence of oral HPV infection in the United States, 2009-2010. JAMA 2012;307:693–703)

Back to Topics

Ovarian Cancer

BRCA1 and BRCA2 Mutations

Approximately 10% of women with invasive epithelial ovarian cancer (EOC) carry deleterious germline mutations in BRCA1 or BRCA2. The authors conducted a pooled analysis of 26 observational studies of women with ovarian cancer to characterize the survival of BRCA carriers with EOC compared with noncarriers and to determine whether BRCA1 and BRCA2 carriers show similar survival patterns. The five-year overall survival was 36% for noncarriers, 44% for BRCA1 carriers, and 52% for BRCA2 carriers. After adjusting for study and year of diagnosis, BRCA1 and BRCA2 mutation carriers showed more favorable survival than noncarriers (for BRCA1: HR = 0.78; for BRCA2: HR = 0.61). These survival differences remained after additional adjustment for stage, grade, histology, and age at diagnosis (for BRCA1: HR = 0.73; for BRCA2: HR = 0.49). Among patients with invasive EOC, having a germline mutation in BRCA1 or BRCA2 was associated with improved five-year overall survival, and BRCA2 carriers had the best prognosis. (Bolton KL, Chenevix-Trench G, Goh C, et al. Association between BRCA1 and BRCA2 mutations and survival in women with invasive epithelial ovarian cancer. JAMA 2012;307:382–390)