690 likes | 700 Views
Genetic susceptibility to head and neck cancer and to lung cancer. Paolo Boffetta. International Agency for Research on Cancer Lyon, France. Avoidable cases of lung cancer (%) France, 2000 - Totals. Assumption of no interaction between risk factors.
E N D
Genetic susceptibility to head and neck cancer and to lung cancer Paolo Boffetta International Agency for Research on Cancer Lyon, France
Avoidable cases of lung cancer (%)France, 2000 - Totals Assumption of no interaction between risk factors
Avoidable cases of oral and pharyngeal cancer (%) - France, 2000 Assumption of no interaction between risk factors
Avoidable cases of laryngeal cancer (%) France, 2000 Assumption of no interaction between risk factors
Cancer deaths (x1,000) attributable to known risk factors – France, 2000 Women Men
Setting the scene • Why to study genetic susceptibility to cancers whose etiology is dominated by exogenous (preventable) agents? • Quantification of individual risk, especially among those formerly exposed to the risk factors • Cancers with strong exogenous risk factors as model to study interactions between genetic and exogenous factors in carcinogenesis.
Genetic Susceptibility to Lung and H&N Cancers • High-penetrance susceptibility genes do not seem to account for a substantial proportion of lung and H&N cancer cases. • Family history of cancer has been reported as a risk factor
Familial relative riskMeta-analysis of three registry-based studies from Utah, Iceland and Sweden Cancer RR 95% CI Lung 2.1 1.8, 2.6 Larynx 4.8 2.2, 10
Explanations for high FRR • Shared exogenous risk factors • Shared genes with a role in pathways relevant to exogenous risk factors • Combination of the two above • Shared genes not interacting with exogenous risk factors
Genetic Association Studies on Lung and H&N Cancers • 1st generation • Very small studies (<100 cases) • Usually not epidemiologic study design; 1-2 SNPs • 2nd generation • Small studies (100-500 cases) • More epi focus; a few SNPs • 3rd generation • Large molecular epi studies (>500 cases) • Proper epi design; pathways • 4th generation • Consortium-based pooled analyses (>2000 cases) • GxE analyses • 5th generation • Post-GWS studies
Sample size needs as a function of genotype prevalence and OR for main effects
N cases in the 47 studies of GSTM1 polymorphism and lung cancer published until 2001
Alcohol Dehydrogenases (ADH), Aldehyde Dehydrogenases (ALDH) & H&N Cancer
ADH1C and H&N Cancer • ADH1C Ile350Val (rs698) in exon 8 • ADH1C*1 allele increases ethanol oxidation by 2.5-fold compared to *2 allele • Suspected to increase risk of H&N cancer • We conducted a pooled analysis of 7 published case-control studies. Brennan et al., Pooled analysis of alcohol dehydrogenase genotypes and head and neck cancer - a HuGE review. Am J Epidemiol 2004.
Inconsistent results may be due to: • Moderate prior probabilities that each SNP individually alters gene function or confers substantial increase in risk • Difficulty in detecting associations with modest risk sequence variants in studies of small sample size • False positives • Publication bias Meta-Analysis limitations • Publication Bias • Unable to adjust for potential confounders • Low power in stratified analysis • Unable to examine gene-gene or gene-environment interactions Pooled-Analysis limitation • Publication and participation bias
Central Europe Multicenter StudyUpper Aerodigestive Tract Cancers • Centers ROMANIA - Bucharest HUNGARY – Budapest POLAND - Lodz & Warsaw RUSSIA - Moscow SLOVAKIA - Banska Bystrica CZECH REPUBLIC - Prague & Olomouc • Incident cancer cases of the UADT SCC, diagnosed at designated hospitals • Controls were randomly chosen from in-patients or out-patients in the same hospitals • Of the 906 eligible cases, DNA was available for 811 (and for 1083 controls)
Genotyping • Genotyping was by Taqman • Designs of genotyping assays for SNP were taken from the SNP500 website • ~100 SNP were selected based on a combination of prior knowledge (functional data, reports of associations, important pathways) and allele frequency observed in Europeans. • Results available on ~50 SNP • DNA repair • Cell cycle • Carcinogen metabolism
SNP in Alcohol Dehydrogenases (ADH) & Aldehyde Dehydrogenases (ALDH) • ADH1BArg48His (rs1229984) in exon 3 • His = *2 allele = ‘fast’ allele • ADH1C Ile350Val (rs698) in exon 8 • Val = *1 allele = ‘fast’ allele • ADH1C Arg272Gln (rs1693482) in exon 6 • ALDH2 355 A>G (rs886205) in the 5’ UTR • ALDH2 348 T>C (rs440) in intron 6 • ALDH2 483 T>C (rs441) in intron 6 • Linkage disequilibrium demonstrated for ADH1B*2 (fast) and ADH1C*1 (fast) in Asians
ADH1B and ADH1C results Hashibe et al., submitted.
ALDH2 results Hashibe et al., submitted.
Haplotype Analysis E-M Algorithm used to estimate the expected haplotype/allele frequencies Unit of analysis is in chromosomes, not individual subjects
False positive report probabilities (FPRP) FPRP<20% are in bold
ADH/ALDH Summary of Results • ADH1B R48H variant appeared to be protective against UADT cancers • 2 ADH1C variants moderately increased the risk of UADT cancers • These associations were most apparent for esophageal cancer • Gene-alcohol interaction observed for ADH1B and ALDH2 variants • Gene-gene interactions were assessed but none were apparent • Replication of these results is necessary • Subsequent studies should focus not only on these variants, but other variants including haplotype tagging SNPs from Hapmap
International Head and Neck Cancer Epidemiology Consortium • Collaboration of research groups leading large molecular epidemiology studies of H&N cancer. • Established in 2004 http://inhance.iarc.fr
Case-control studies invited to participate in Inhance projects
SNP data pooling project • Studies on various SNPs and head and neck cancer have not been consistent due to limitations in sample size, publication bias, false positives • Greater statistical power in pooled analyses may help to identify modest risk SNPs • Assessed SNPs genotyped in common across the Inhance studies • 18 SNPs genotyped in at least 3 studies • Supported by the NIDCR R03 grant
Genotype all cases and controls for panel of 300k+ SNPs Identify most ‘interesting’ and replicate in second independent sample (and third etc) Advantages No prior hypothesis Disadvantages Sample size Funding Analysis Basics of whole genome scan
Proposal of GWS of H&N cancer • Phase 1 • 300k SNP • Central European study • 1000 cases and 2000 controls (already genotyped as part of a parallel GWS of lung cancer) • Phase 2 • 5k SNP • ARCAGE study (Western Europe) • 1300 case-control pairs • Latin American study • 1500 case-control pairs • Phase 3 • 30-50 SNP • INHANCE • 5-10K case-control pairs
Conclusions 1What have we learned about H&N cancer susceptibility? • Role of genes encoding for enzymes involved in alcohol metabolism and DNA repair • Improve marker selection • Select SNPs that are likely to alter gene function • Tag SNPs • Detailed analysis of genes having strong evidence of association • Quantify risk from genetic variants and interaction with alcohol drinking • Need to expand to other pathways
Conclusions 2What have we learned about search for cancer susceptibility factors? • Strengths of large-scale molecular epidemiological studies and consortia • role of local collaborators and junior investigators • need for novel funding mechanisms • Need for novel approaches • Examine multiple markers • multigenic model • combine genotype and phenotype data for a pathway driven approach • Statistical analyses • hierarchical modeling • How to put the evidence together?
Acknowledgements - 1 Central European Study David Zaridze Neonila Szeszenia-Dabrowska Dana Mates Vladimir Janout Eleonóra Fabiánová Peter Rudnai Vladimir Bencko Witold Zatonski NoN Muin Khoury John Ioannidis Julian Little • IARC • Mia Hashibe • Paul Brennan • Janet Hall • Federico Canzian • Rayjean Hung • Norman Moullan • Amélie Chabrier • Valérie Gaborieau • Julien Berthiller • CNG • Mark Lathrop
Acknowledgements - 2 INHANCE members - Chu Chen, Stephen Schwartz, Karl Kelsey, Silvia Franceschi, Simone Benhamou, Daniel Luce, Isabelle Stucker, Erich Sturgis, Qingyi Wei, Richard Hayes, Mark Purdue, Philip Lazarus, Joshua Muscat, Zuo-Feng Zhang, Andrew Olshan, Elaine Smith, Edward S. Peters, Ana Menezes, Alexander W. Daudt, Maria Paula Curado, Sergio Koifman, Victor Wünsch Filho, Jose Eluf Neto, Elena Matos, Vladimir Bencko, Eleonora Fabianova, Vladimir Janout, Witold Zatonski, Rolando Herrero, Xavier Castellsague, Renato Talamini, Luigino Dal Maso, Fabio Levi, Carlo La Vecchia, Pagona Lagiou, Antonio Agudo, Wolfgang Ahrens, Bernard E. McCartan, David Conway, Andres Metspalu, Gary J. Macfarlane, Ray Lowry, Kristina Kjaerheim, Lorenzo Simonato, Ivana Holcátová, Franco Merletti, Ariana Znaor
Possible explanations for protective OR observed for ADH1B*1 ‘fast’ allele • Linkage Disequilibrium • The SNP may be in LD with another SNP in another ADH that is protective • Ensemble of ADH activity may be lower even though this ADH1B allele is associated with higher activity • Residual confounding by drinking behavior • Though we did not observe an association between genotypes & drinking behavior in our data, perhaps subjects with the ADH1B slow genotype had different drinking behaviors such as binge drinking. • Complication by other substrates • ADHs can metabolize retinol. Protective effect may be due to dietary intake of vitamin A that is being metabolized more by ‘fast’ allele? • Functional studies not consistent • Perhaps ‘fast’ allele is not really fast • In vitro studies showed enzyme activity differences, but the ADH1B genotype did not affect blood acetaldehyde levels in individuals • Acetaldehyde Clearance • Enzyme activity of ADH1B*2 variant is 100-200 times higher. Perhaps a fast initial metabolism leads to a peak in acetaldehyde that induces clearance mechanisms • Enzyme activity of ADH1C*1 fast allele is 2.5 times, so it may not induce clearance?
How to move forward in the study of gene-environment interactions • Improve marker selection • Select SNPs that are likely to alter gene function • Tag SNPs • Detailed analysis of genes having strong evidence of association • Expand focus to other pathways • Examine multiple markers • Multigenic model • Combine genotype & phenotype data for a pathway driven approach • Employ novel statistical analyses • FPRP • Hierarchical modeling • Enhance statistical power • Limitations of previous studies include small sample size, publication bias, false positives • Meta-analysis, pooled analysis, consortia
Data distribution by study X=data available