Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
15 Cards in this Set
- Front
- Back
What is a cohort study?
|
Investigator:
Identifies a group of animals that have the ‘exposure’ of interest and that are free from the disease of interest Identifies a group of animals that do not have the ‘exposure’ of interest and that are free from the disease of interest. Follows both groups over time to determine how many animals in both groups develop the disease of interest Calculates incidence rates of disease in both groups and the risk ratio |
|
What are the advantages of a cohort study?
|
Incidence rates can be calculated directly from the study
Provides a complete description of the natural history of disease development Have opportunity to study multiple potential effects of exposure Allows for good quality control |
|
What are the disadvantages of a cohort study?
|
Large numbers required if studying a rare disease
Very long study required if the disease of interest has a long incubation/induction period Relatively expensive Maintaining follow-up often difficult Controlling extraneous variables can be difficult Little opportunity for detailed study of pathogenesis of disease |
|
What is a case control study?
|
Investigator:
From records, identifies a group of animals that have or has the disease of interest (the ‘cases’). Form records, indentifies a similar group of animals that do not have the disease of interest (the ‘controls’). The past history of both groups is investigated to determine how many animals in each group had the exposure of interest. Calculates the odds ratio to as an estimate of the risk ratio |
|
What are the advantages of a case control study?
|
No risk to subject in the trial - uses historical information
Well suited to the study of rare diseases or diseases with long incubation/induction periods Relatively easy and quick to conduct Require comparatively few subjects, generally inexpensive Can potentially estimate the effect of many exposures of interest at the same time |
|
What are the disadvantages of a case control study?
|
Relies on recall or records for information on past exposure
Validation of information is difficult or impossible Control of extraneous variables is difficult Selection of an appropriate control group is usually difficult and always subject to bias Incidence rates cannot be determined Method unfamiliar to most vets and hard to explain Detailed study of pathogenesis not possible |
|
What is odds ratio & relative risk?
|
Odds ratio and relative risk express how much more likely something is to occur if a certain condition is met, such as a patient being exposed to an illness or receiving a particular treatment.
✩ODDS RATIO -this divides the odds of an outcome given a condition by the odds of the same outcome when the condition is not met Odds ratio = (A/B)/(C/D) -studies that create a sample population based on outcome, such as case-control studies, must use odds ratio -studies that create a sample population based on exposure or treatment, such as a controlled trial or cohort study, can use relative risk ✩RELATIVE RISK -relative risk divides the likelihood (or percentage) of an outcome given a condition being met, by the likelihood of the same outcome if the condition is not met Relative risk is [A/(A + B) / C/(C+D)] |
|
How do we design good epidemiological studies?
|
Accuracy is the goal - freedom from errors
Only two sorts of error – random and systematic A study that is free from random error is a PRECISE study A study that is free form systematic error (bias) is a VALID study Both precision AND validity are required for accuracy |
|
What is precision? What is accuracy?
|
✩Precision - is the reproducibility of a measurement
✩Accuracy - is how close a measurement is to the true value |
|
What is a systematic error? What is a random error?
|
✩Systemaic error are errors that occur the same way every time a measurement is taken
-as a result, the measurements are wrong in the same way each time and thus are not accurate but precise ✩Random error - is unavoidable error that is different each time a measurement is taken -this reduces precision -it also reduces accuracy if the amount of error is large |
|
How do we improve precision?
|
Can be improved through two pathways
1. Increasing the number of subjects in the study 2. Increasing study efficiency ie. be smart about choices, ie. don't select young dogs for cancer study, probably won't get it, need older dogs Would be more precise if you had 60 disease and 60 control vs 20 cases and 100 control but that is still better than 20 cases and 20 control P= 0.12, 3 (95% CI 0.2-38) this drug more of a risk for causing cancer b:c of wide CI Vs p= 0.04, 0.075(95% CI 1.01-1.07) less risk Statistical significance is not the same as biological significance A precise estimate will have a narrow 95% confidence interval An imprecise estimate will favour acceptance of the null hypothesis - Statistical significance is different from biological significance |
|
What kinds of systematic error or bias can occur?
|
1. Selection bias
Non-responders ie. dropouts Poor choice of control Poor choice sampling frame ie. self selection 2. Information bias Information collected wrong or poorly interpreted -differential misclassification ie. those who are affected by abortion are more likely to remember details than non-aborted group -non-differential classification - just by history you make mistakes, if mistakes are equally likely for both groups, tends to smudge groups (gets you to say that there is less risk than there is) 3. Confounding -relationship between exposure and disease is distorted by the effect if an extraneous variable |
|
How do we asses confounding variables?
|
Description between exposure and disease is distorted by the effect of an extraneous variable
ie. in tennis player example, younger guy may have had more winning streak, both were bad at left handers and good with right handers, but older guy got less exposure to right handers so looked worse Look at the ‘crude’ estimate Then stratify on the suspected confounder Look at the ‘stratified’ estimate If an important difference exists between crude and stratified estimates, confounding exists REPORT STRATIFIED ESTIMATES DO NOT USE CRUDE ESTIMATES |
|
How do we prevent confounding variables by study design
|
Restriction
Prohibit the confounder from varying Effective, cheap, but poor basis for generalisation? -by how we select animals into the study ie. for testicular cancer the only animals we will allow into the study are 8 year olds -limits the animals you get into the study though -poor basis of generalization ie. 5 year olds could be different 2. Matching Selection of a comparison group (controls) that is identical to the case group with respect to the confounder Often costly, needs special analysis ie. every 5 year old German shepherd with undescended testicles, you find a five year old German shepherd with descended testicles |
|
How do we control of confounding variables in data analysis?
|
1. Stratified variable
If lots of categories, becomes inefficient 2. Mathematical modeling Logistic, log liner etc. Not easy to interpret Underlying assumption in the model may not be appropriate |