3. Cohort Studies Flashcards
What is a cohort study?
Identify individuals (the corhort), measure ‘exposures’ in each individual, follow-up individuals to determine disease/disorder occurance. Relate information on disease occurance to exposure.
- Objective*
- •To assess the impact of breast feeding on the risk of being overweight in children at the time of entry to school.*
- Methods*
- •Mothers asked about breast feeding shortly after delivery and at 6 months.*
- •At 5 years children were assessed physically.*
a) What is the disease/disorder?
b) What is the exposure
c) Is this a cohort study?
a) Child being overweight
b) Child being breast fed
c) Yes - cohort = mums and children
- Objective*
- •To assess the impact of breast feeding on the risk of being overweight in children at the time of entry to school.*
- Methods*
- •A group of overweight children were identified when they started school and the same number of normal weight children at the same school were identified.*
- •Their mothers were asked about breastfeeding.*
Is this a cohort study?
No, this is a case control study
List some of the bias in cohort studies.
List some of the disadvantages of cohort studies.
Look at the data (pic). If you were a non-smoker what was your risk of dying from lung cancer in 1 year ?
Loss to follow-up, exposure usually measured at just one time point, selection of cohort
Take a long time, need a lot of people, very expensive
17 per 100,000
What is incidence?
Look at the data (pic). If you were a smoker how much greater risk of dying from lung cancer was there relative to the risk for non-smokers?
What is relative risk?
INCIDENCE is usually taken to be a measure of RISK. Incidence = number of new cases (or deaths) of a disease per 100,000 people per year.
417/17 = 24.5
RR = Risk in exposed population / Risk in unexposed population
Look at the data. What is the relative risk of heart disease amongst heavy smokers compared to non smokers?
What is the relative risk of lung cancer amongst heavy smokers compared to non smokers?
What should you look at to investigate if the results could have arisen due to chance alone?
1111/619 = 1.8
417/17 = 24.5
p-value and CIs
How would you describe the probability of risk of lung cancer (pic) - does smoking increase the risk of lung cancer?
What is the 95% CI of a sample mean?
What is the 95% CI of a sample relative risk?
If the risk of lung cancer for smokers is the same as the risk for non-smokers what is the relative risk ?
The probability of getting a risk at least 24.5 times greater in smokers than non-smokers is less than 1 in 1000 assuming smoking has no effect. (P-value definition). This is evidence that smoking does increase the risk of lung cancer.
The 95% CI of a sample mean contains the population mean with a probability of 95%.
The 95% CI of a sample relative risk contains the population relative risk with a probability of 95%.
RR = 1
Do we have enough evidence here that smoking increases risk of lung cancer and heart disease?
What number are you looking for in the 95% CI when looking at:
a) Differences in means
b) Ratios (e.g. RR)
95% CI does not include 1. Evidence that smoking does increase the risk of lung cancer and heart disease.
Diffs in means = look for 0 in 95% CI (0 = no diff). Ratios = look for 1 in 95% CI (1 = no diff)
What is a confounder?
A confounder is a factor that is associated both with the exposure and also with the disease. E.g. In a cohort study heavy drinkers were found to be more likely to develop lung cancer. Does drinking cause lung cancer? - not actually but more likely to drink and smoke, which causes LC
In a cohort study heavy drinkers were found to be more likely to develop lung cancer. Does drinking cause lung cancer?
How can you adjust for confounders?
- Analyse alcohol consumption and risk of lung cancer in non-smokers
- Analyse alcohol consumption and risk of lung cancer in smokers
- Might expect these risks to be similar – in which case you could combine them*
- [PIC]*
- NB: always go for adjusted RR*
What is survival analysis used for?
What kind of curve plots the proportion of people surviving over time?
Do you calculate relative risk in survival analysis?
If, after adjusting for age, sex and tumour stage, the Hazards Ratio for exercisers vs non-exercisers = 0.73 (95%CI:0.54-0.99), what would you conclude about the survival rate? What percentage is the risk of death?
Often want to look at how long someone lives – particularly with cancer
Kaplan Meier
No - you calculate hazard ratio. They take time to death/event into account. They are similar to RR and give the risk of dying at any point in one group compared to another.
People who exercised prior to developing colorectal cancer survived longer after the colorectal cancer was diagnosed, and risk of death is 27% less.
What are the 2 measures of importance?
If you stop smokers from smoking will you prevent more lung cancer deaths or IHD deaths? (Lung cancer RR = 24.5, IHD RR = 1.8)
Absolute excess risk. Attributable portion.
IHD b/c it’s much more common (to die from) than LC. Use absolute excess risk (pic)
How do you calculate absolute excess risk?
What is the absolute excess risk of lung cancer amongst heavy smokers compared to non smokers? (Data in pic)
AER = Risk in Exposed – Risk in Unexposed
417 - 17
How do you calculate the attributable portion?
What is the attributable portion for lung cancer and IHD. Assume 20% of the population are heavy smokers. Use the data below.
= incidence in population attributable to exposure / incidence in population
= p (RR - 1) / 1 + p (RR - 1)
p = proportion exposed in population
82% of LC deaths and 14% of IHD deaths are attributable to smoking. (Working below)