Lesson 3 Epidemiology, Causes, Study design

• Epidemiology

  The study of the occurrence and determinants of health and disease in populations
• Sources of data

  • Australian Institute of Health and Welfare 2019. Cancer in Australia 2019.
  • Australian Institute of Health and Welfare 2021. Cancer in Australia 2021.
  • http://www.insidecancer.org

  
• Incidence rates don't need to be remembered but the concept should be understood
• Cancer incidence rises with age

  
• The rate of cancer incidence has remained fairly stable

  
• Most common cancers

  • Incidence by cancer type

  
• Incidence

  The number of diagnoses (often per population), reflects occurrence of the disease and our ability to diagnose it
• Cancers that kill the most Australians

  • Mortality by cancer type

  
• Mortality

  The number of deaths, it reflects the incidence and the lethality of the disease

  
• Cancer deaths have risen over time, but the rate of cancer deaths fell

  
• Fewer people diagnosed with cancer die in Australia than elsewhere

  
• MIR

  Mortality to Incidence Ratio, 1 means all diagnosed die, 0.5 means half die

  
• Survival rates vary enormously between cancer types

  
• Cancers overtook cardiovascular diseases as the leading cause of death in Australia in 2010

  
• Globally, cancer accounts for the second largest number of deaths

  
• Cancer accounts for more premature death than any other disease, and poses the highest overall burden of disease

  
• Causes of cancer

  • Inheritance
  • Environment
  • Mutations in dividing stem cells
• Environmental carcinogens and/or inherited deficiencies in DNA repair pathways can increase mutation rates
• Environmental factors (eg diet) and/or genetic factors (eg hormone levels) can promote replication of stem cells, rendering them vulnerable to mutation
• Clues from epidemiological data about cancer causes

  • Changes over time in rates of cancers
  • Geography
  • Age
  • Effects of treatment

  
• Changes over time in cancer rates can indicate changes in exposure to environmental factors or shifts in the genetic makeup of the population
• Geography can hint at environmental or genetic factors influencing cancer
• Cells accumulate changes so most cancers increase in incidence with age, deviations from this can provide clues to other factors

  
• Defining cancer risks

  
• Alterations in survival rates can correlate with changes in clinical practice, but clinical trials provide much better quality information
• Participants

  • 375,000 people accepted a mailed invitation to participate in a detailed survey about many health-related factors
  • 35,372 were women aged 35-69 years: subjects in this study
• Would these women be representative of the general population? No: they were more health conscious (eg only 11% were smokers). If meat-eating tends to cause breast cancer but only in smokers, this study may not detect this association.

  
• Data acquisition


  • Between 1995 and 1998, survey asked 217 diet-related questions
  • Subjects kept a 4-day diet diary
  • In 2004, the incidence of cancer in each participant was investigated using the National Health Register
• The investigators accessed health records for cancer incidence data, rather than asking the women themselves about their cancer status. Avoided the "lost-to-follow-up" problem (trouble tracking people down). Objective, uniform diagnoses.

  
• This was a prospective study (cancer-free women were recruited then the investigators waited to see who got cancer). No issue of recall bias (accurate diet information). A retrospective design would have involved recruiting women in 2004, finding out whether they had breast cancer or not and asking what they ate 8 years ago. Faulty memories, possibly substantial recall bias.

  
• Hazard ratio

  Adjusted for years in study, age, energy intake, menopausal status BMI, physical activity, smoking status, HRT use, OCP use, parity, total fruit and vegetable intake
• Relative risk


  Risk in exposed people (meat-eaters) relative to risk in controls (veges)
• If not accounted for: confounders. What else might explain these differences?
• Cohort study


  • Advantages: Can be useful for rare exposures. Prospective studies avoid faulty memory and recall bias and because exposure always precedes disease.
  • Disadvantages: Confounding!! Prospective studies can require long follow-up periods, are expensive and prone to "loss-to-follow up". Recall bias can be a problem for retrospective studies.
• Confounders


  • Characteristics that differ between meat-eaters and vegetarians which may influence breast cancer risk
  • If it were possible to randomly allocate study subjects to meat-eating or vegetarian diets, researchers could perform a randomised controlled trial and confounding would be eliminated.

  
• HRT clinical trial


  • Study population was 16,608 post-menopausal women
  • Random, blind, allocation of subjects to receive HRT or not
  • Gave 8506 women daily oestrogen plus progesterone, Gave 8102 women daily placebo
  • When trial stopped, 166 breast cancer in HRT group, 124 breast cancer in placebo group

  
• Relative risk (RR)


  % risk in treated group / % risk in untreated group

  Example:

  % risk in HRT group / % risk in placebo group
• The 95% confidence interval was 1.00-1.59. This means that we can be 95% sure that women in this age range, taking this form of HRT, will be 1.0-1.59 times more likely to develop breast cancer than those not taking these drugs. We are pretty sure HRT very slightly increases the chance of breast cancer.

  
• Disadvantage: non-compliance will lead to underestimate of effect. Significance = "Statistical significance", not "substantial" or "warranting change in behaviour".
• Randomised controlled trial


  • Advantages: Least opportunity for bias and confounding
  • Disadvantages: Often not practical. Can be expensive and take a long time. Non-compliance to assigned group can lead to underestimates of associations.