Trends in 21st Century Epidemiology: From Scientific Discoveries to Population Health Impact

Robert N. Hoover Presentation: Historical Perspectives on the Evolution of Cancer Epidemiology

Slide 1 of 29

Robert N. Hoover, M.D., Sc.D.
Trends in 21st Century Epidemiology: From Scientific Discoveries to Population Health Impact
December 12, 2012


Slide 2 of 29: Formal Cancer Epidemiology: The Early Years

[Table] showing historical studies and the relative risks found. Smoking and lung cancer found to have a relative risk of 14 in study by Richard Doll & A. Bradford Hill published in BMJ 1950;221(ii):739-45. In another study of smoking and lung cancer by Earnest L. Wynder and Everts A. Graham published in JAMA 1950:143:329-36, a relative risk of 13 was found. For studies linking alcohol to upper GI cancers, radiation to leukemia, tobacco to other cancers, and 18 specific chemical or industrial processes (IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, Supplement 1, 1979), relative risks ranged from 4 to several hundred.


Slide 3 of 29: Epidemiology - Then

  • Small, simple studies
  • Small study teams
  • PI did virtually everything

Slide 4 of 29: Epidemiology - Now

  • Large, complex studies
  • Large, multidisciplinary teams
  • Specialization

Slide 5 of 29: Why the Differences?

  • Major changes in the goals of Classical Epidemiology
  • Introduction of, and major shift to, Molecular Epidemiology

Slide 6 of 29: Classical Epidemiology

Then

  • Large Risks
  • Evident Exposures
  • Main Effects

Now

  • Low-level Risks
  • Difficult to measure exposures
  • Effect modification

Slide 7 of 29: Molecular Epidemiology: Opportunities

Overcome some weaknesses of Classical Approaches

  • Measure exposures
  • Measure outcomes
  • Assess susceptibility
  • Mechanistic studies
  • Assess larger numbers of markers simultaneously

Slide 8 of 29: Hormone Therapy (HT) for Menopause and Cancer

[Table] showing results of a pooled analysis of 4 follow-up studies of all cancers and hormone therapy published in Lancet 1971;1:135-6. Among 1130 subjects exposed to hormone therapy, 7 cases of cancer were observed, and 74 cases of cancer were expected.

[Table] showing results of another study published in the New England Journal of Medicine 1976:295:401-5. The left column consists of 5 categories of hormone therapy use duration, ranging from <1 year to 15+ years. The middle column describes the relative risk for each duration of hormone therapy. The right column lists the related 95% confidence interval for each. Results indicate that as the duration of HT use increases, the relative risks and corresponding confidence intervals increase. The p-value for trend is significant at 0.02. The number of exposed cases is 49.


Slide 9 of 29: Relative Risk (RR) of Breast cancer: Never Users, Recent Users, and Past Users

[Table] showing relative risks for recent users of hormone therapy that have used for at least 5 years. In the left column, results are stratified by weight. For users weighing 65 kg or less, the relative risk is 1.65. For users weighing 65 kg or more, the relative risk is 1.06. The p-value for trend is 4x10-3. In the right column, results are stratified by BMI. For users with a BMI of 25.0 or less, the relative risk of 1.52. For users with a BMI of 25.0 or more, the relative risk is 1.02. The p-value for trend is 1x10-4. Source is Lancet 1997; 350:1047.


Slide 10 of 29: Genetic Epidemiology

1980s onward: Mendelian Inheritance

  • Genome-wide linkage
  • High-risk families

1990s onward: Susceptibility Genes

  • RFLP and other technologies
  • Candidate genes

Slide 11 of 29: Genomics - A Lost Decade

Thousands of candidate genes

Pursued "to extinction" in tens of thousands of studies

Tiny fraction of reported associations ever replicated

  • Even fewer GxE interactions

Slide 12 of 29: Genetic Epidemiology

1980s onward: Mendelian Inheritance

  • Genome-wide linkage
  • High-risk families

1990s onward: Susceptibility Genes

  • RFLP and other technologies
  • Candidate genes

2006 onward: Susceptibility Genes

  • Database and SNP chip
  • Agnostic search

Slide 13 of 29: Published Cancer GWAS Etiology Hits: 10.18.12

[Image] showing GWAS etiology hits in cancer.

~265 Disease Loci marked by SNPS
1 Locus marked by a CNV

Another ~90 coming soon...
Breast, bladder, esophageal, kidney, lung, osteosarcoma, ovary, prostate, testicular...


Slide 14 of 29: Early Established Susceptibility Loci for Breast Cancer

[Table] listing various susceptibility loci and their associated properties. For CASP8, the minor allele frequency in Europeans was 0.13, in Asians was 0.00, and in Africans 0.21. The odds ratio for heterozygotes was 0.89, and for homozygotes 0.74. The population attributable risk was 20. For FGFR2, the minor allele frequency in Europeans was 0.38, in Asians 0.30, and in Africans 0.50. The odds ratio for heterozygotes was 1.23, and for homozygotes 1.63. The population attributable risk was 19. For TNRC9, the minor allele frequency was 0.25 in Europeans, 0.60 in Asians, 0.53 in Africans. The odds ratio for heterozygotes was 1.23, and 1.39 for homozygotes. The population attributable risk was 10. For MAP3K1, the minor allele frequency in Europeans was 0.28 in Europeans, 0.54 in Asians, 0.35 in Africans. The odds ratio for heterozygotes was 1.13, and 1.27 for homozygotes. The population attributable risk was 7. For 8q24, the minor allele frequency for Europeans was 0.40, 0.56 for Asians, and 0.58 for Africans. The odds ratio for heterozygotes was 1.06, and 1.17 for homozygotes. The population attributable risk was 6. For LSP1/H19, the minor allele frequency was 0.31 for Europeans, 0.14 for Asians, and 0.12 for Africans. The odds ratio for heterozygotes was 1.06, and 1.17 for homozygotes. The population attributable risk was 4. For the 2q35 loci, the minor allele frequency for Europeans was 0.50, 0.15 for Asians, and 0.69 for Africans. The odds ratio for heterozygotes was 1.20, and 1.40 for homozygotes. The population attributable risk was 19.

Source: Pharoah P. N Engl J Med 2008;358:2796-803.


Slide 15 of 29: Cigarette Smoking, NAT2 Phenotype, and Breast Cancer Risk in Two Large Consortial Analyses

[Table] showing risk of breast cancer based on NAT acetylator status and smoking. In a study by Ambrosone et al published in Cancer Epidemiology Biomarkers and Prevention 2008;17(1), rapid acetylator phenotypes showed a 1.07 risk among smokers who smoked 20 or less pack years compared to never smokers, and 1.04 risk among smokers who smoked more than 20 pack years (Confidence interval 0.9-1.3). The p-value for interaction was 0.03. Among slow acetylator phenotypes, smokers who smoked 20 or less pack years had a risk of 1.21 compared to never smokers, and smokers who smoked more than 20 pack years had 1.33 risk (confidence interval 1.2-1.9).

[Table] showing risk of breast cancer based on NAT acetylator status and smoking. In a study by Cox et al in the American Journal of Epidemiology 2011;174(11), rapid acetylator phenotypes who smoked 20 or less pack years showed a 1.13 risk of developing breast cancer compared to nevers, and a 1.24 risk for smokers who smoked than 20 pack years compared to never smokers (Confidence interval 1.1-1.4). The p-value for interaction was 0.87. For slow acetylator phenotypes, smokers who smoked 20 or less pack years had a 1.08 risk compared to never smokers, and smokers who smoked more than 20 pack years had a 1.25 risk (Confidence interval 1.1-1.4).


Slide 16 of 29: Genomics History as Lesson for Future

Two major caveats:

  • Importance of high quality epidemiologic methods
  • Assay development

Slide 17 of 29: Lessons for the Future - #1

We are not as smart as we wish we were

  • Less a priori, more listening to data
    • Mandatory Corollary: Replication, replication, replication

Slide 18 of 29: Lessons for the Future - #2

Remarkable opportunities from new science and technologies

  • Classical Epidemiology: Internet, environment, and lifestyle monitoring tools, linked datasets
  • Molecular Epidemiology: All of the "omics"
    Mandatory Corollaries:
    • Work with lab to bring to "primetime"
    • Best epidemiologic methods

Slide 19 of 29: Lessons for the Future - #3

Bigger, Better, Sooner

  • Many of the important, contemporary questions in biology and public health can only be addressed by aggregating large amounts of high quality epidemiologic data.

Slide 20 of 29: Lessons for the Future

  1. Listen to the Data
  2. Remarkable opportunities from new science and technologies
  3. Bigger, Better, Sooner
  4. Much faster and better at adapting methods to meet scientific needs and opportunities as they emerge

Slide 21 of 29: Formidable, but surmountable, obstacles to implementing "Lessons for the Future"

  • Appropriate "credit" for participating in team science and consortial efforts
  • Role for junior investigators
  • Relative value and timing of individual vs. pooled analyses
  • Cultural differences between disciplines
  • Rapid changes in state-of-the-art technologies
  • Study subject participation, cooperation, and consent
  • Rapid and broad data-sharing
  • Funding for necessary infrastructure
  • Inadequacy of traditional grant mechanisms for funding broad "discovery" efforts
  • ETC, ETC, ETC...

Slide 22 of 29: General Trends Over Time, NOT Dogma

Then

"Big Science" studies did exist

  • CPS1, Dorn, British physician cohorts
  • International Breast Cancer and National Bladder Cancer Case-control Studies

Interdisciplinary studies did exist

  • Hepatitis B and liver cancer

Slide 23 of 29: General Trends Over Time, NOT Dogma

Now

  • Still an important role for relatively small, innovative studies
  • Still will be high-risk risk factors
  • Many things will not be well-assessed by biomarkers

Slide 24 of 29: Ever Use of Artificial Sweeteners and Bladder Cancer Risk in 632 Cases and 632 Controls

[Table] showing data from a study by Howe GR et al., Lancet 1977 Sept: 17(8038): 578. Men who had ever used artificial sweetners had a relative risk of 1.6 for developing bladder cancer, with a p-value of 0.018. Women had a relative risk of 0.6 with N.S. p-value.


Slide 25 of 29: Bladder Cancer and Ever Use of Artificial Sweeteners in 3,000 Cases and 5,766 Controls

[Table] showing results from a study by Hoover RN, et al., Lancet 1980:1:837-40. Men exhibited a relative risk of 0.99, with a 95% confidence interval of 0.89-1.10. Women exhibited a relative risk of 1.07, with a 95% confidence interval of 0.89-1.29. When both sexes were taken into consideration, they had a 1.01 relative risk, with a 95% confidence interval of 0.92-1.11.


Slide 26 of 29

"As a general rule of thumb, we are looking for a relative risk of three or more [before accepting a paper for publication]."

Marcia Angell
Editor, New Engl J Med
1995


Slide 27 of 29

Advances will be accelerated by "Collective Intelligence"

"I not only use all of the brains I have, but all I can borrow" - Woodrow Wilson


Slide 28 of 29: Breast Cancer and Candidate Genes

Search Study:
170 SNPs in 120 Candidate Genes in 4400 cases and 4400 controls
None significant after control for population stratification and multiple testing

Source: Pharoah PDP et al. PLOS Genet 2007; 3:401-406


Slide 29 of 29: Cigarette Smoking, Genotype, and Breast Cancer

Since 1995, 50 studies have examined this relationship in relation to a total of 11 susceptibility genes

"Literature is complicated by methodologic limitations, ...which likely contributed to the inconsistent findings. These methodologic issues should be addressed in future studies."

Source: Terry PD et al. Cancer Epidemiol Biomarkers Prev 2006.

Return to Top