Case-control & cohort studies

What do you call a non-adjusted OR?

"crude" OR

If 41 people who ate >5 got hypertension and 108 didn't, and 199 people who ate <5servings per day got hypertension but 252 didn't, What is the crude odds ratio for consuming >5 servings fruit (cases) and vegetable compared to < 5 servings per day

(41/199)/(108/252)=0.481

A case-control study aimed to investigate the relationship between red meat consumption and cancer. The study included 160 cases with cancer and 609 control subjects. Each participant was classified as consuming red meat > 4 times per week (exposed) or ≤ 4 times per week (unexposed). The results were as follows: 56 cases ate red meat > 4 times per week and 104 cases ate red meat ≤4 times per week; whereas 109 controls ate red meat > 4 times per week. What is the odds ratio for eating red meat > 4 times per week among cases compared to controls?

2.47

If conducting a prospective cohort study on a disease that 1/75 people have, and you need 100 with outcome, how large of a sample size would be needed?

7500

What are the steps in interpreting an OR?

<1 or >1 - relationship CI include 1? - significance CI wide or small? - precision (generally larger n = smaller, more precise CI)

How would you interpret an OR>1? OR<1? OR=0

>1 exposure more likely in cases than controls (positive/direct relationship) <1 exposure more likely in controls (negative/inverse relationship - exposure protective against outcome =0 no association, null hypothesis

Name a few noteworthy nutrition-related cohort studies

Framingham heart study: CVD The nurse's health study: cohort of nurses, oral contraceptives CHILD study: allergies/asthma CanPath: cancer/chrinis illnesses

Describe ways to select cases and controls

From general population (registries, households, telephone) - can be costly, time consuming, low response rate. From hospital/clinic (more common) - often easier to identify and higher response rate

Define epidemiology terms including: incidence, prevalence, mortality, and morbidity

Incidence: new cases over a time period in a population Prevalence: total # cases of disease in a population at a given time Mortality: death rates Morbidity: incidence of ill health/disease

How is an odds ratio calculated

OR = Odds of exposure among cases/odds of exposure among controls

OR vs RR

OR: cohort, case-control, or experimental RR: cohort and experimental, NOT case-control

How are associations assessed in cohort studies?

Relative risk; likelihood that outcome occurs in the exposed group vs unexposed group

Differentiate between retrospective and prospective cohort studies

Retrospective: at study onset, outcomes have already occurred (looking from past to present) Prospective: at study onset, outcomes have not yet occurred (looking from present to future)

What's the null hypothesis for risk difference? What's absolute risk reduction?

Risk difference = 0 Only used when risk difference is negative (make it positive)

Differentiate between risk and odds. Give example with risk/odds of rolling a 2 on a di

Risk: probability of event occurring - chances of event occurring/total # possible outcomes (1:6) Odds: chances of event occurring/chances of event not occurring (1:5)

What's an adjusted OR? (aka multivariate OR)

Used when cases/controls differ on variables other than exposure, could act as confounders. Accounted for in analysis - gives a multivariate OR

Describe the advantages of case-control studies

Useful for studying rare conditions or with a long lag time between exposure/outcome. Generally requires fewer subjects than short/cross-sectional and generally quicker/cheaper than cohort studies

When is RR statistically significant?

When CI doesn't include 1

What are "person years"

a way to incorporate both time & the population at risk in risk calculations. Sum of years each person is in the study.

Define key features of cohort studies

assess exposure at baseline, then follow participants over time. Exposure of the who develop outcome compared to those who don't. Can demonstrate temporality and determine incidence

In a case control study, what is the independent and dependent variable?

exposure = independent, outcome = dependent

Explain when a case control design is appropriate to use

studying rare conditions or with a long time between exosure/outcome. Before doing a cohort study or RCT (usually fewer subjects and quicker)

Describe nested case-control studies

Case-control nested within a cohort study. Participants recruited for cohort study, after outcomes assessed, sample of data analyzed as a case control study

Describe a confidence interval. When is the CI of an OR statistically significant

Range of values within which the true population value truly lie. CI=1-alpha. Significant when CI excludes 1

How can RR be expressed? Use RR = 0.6 as an example

0.6 times as likely to have outcome 60% as likely to have outcome 1-0.6 = 40% less likely to get outcome (more likely if RR>1)

Differentiate between retrospective cohort and case control

Both look back at exposures. Cohort starts with exposure, then see if there's development of outcome. Case-control starts with outcome then asks about exposure

Describe the advantages of prospective cohort studies

Can establish timing/directionality of events (temporality), can calculate incidence, avoid recall bias, can assess many outcomes (and exposures) if well designed

What's a case? What's a control? How is potential confounding dealt with?

Cases have the outcome of interest. Need a defined method for case ascertainment (Ex: diagnosis) Controls don't have outcome of interest. May choose controls with another condition to control for healthcare system interaction. May match cases for age, sex, ethnicity, etc.

Describe the limitations of prospective cohort studies

Confounding, level of exposure may change over time, loss to follow up (withdrawal), large sample size needed (esp hard if outcome rare)

Describe the design of a case control study

Investigator selects cases and controls. Data about exposure to possible etiologic factors collected in both groups, frequency/amount of exposure is compared between groups. OR may be calculated to compare frequency of exposure

What happens in a case-control when there are multiple categories of exposure (tertiles)? How is significant determined?

Lowest group becomes reference/comparison group, may be indicated with OR of 1. All other OR calculated in comparison to this group. P<0.05 for significance.

What are the two possible types of misclassification in a case-control study>

Non-differential: both groups equally affected, may affect ability to see association Differential: error more likely in one group than the other, inter-rater reliability important - larger concern & may contribute to a false association

RR vs HR?

RR: Relative risk of developing the outcome among exposed vs unexposed participants HR: Relative risk of developing the outcome over time (ex: RR per year)

Distinguish between relative risk, absolute risk, and risk difference (absolute risk reduction)

RR= risk of outcome in exposed/risk of outcome in unexposed Absolute risk = risk of developing outcome = #with outcome/total sample size Risk difference = risk exposed - risk unexposed

How to interpret RR? >1? <1? =1?

RR>1 outcome more likely in exposed group -increased risk of getting outcome RR<1 outcome less likely in exposed group - decreased risk of getting outcome (negative association/protective) RR=1 null, no relationship between exposure and outcome (same risk in both groups)

Describe the limitations of case-control studies

Temporality difficult to establish. How far back to look for exposure (must determine appropriate time period using biological plausibility) Reliance on memory Recall bias (cases more likely to search memory/report past exposures) Misclassification (measurement errors) Cannot calculate disease incidence/prevalence (non-representative sample proportions)