EBVM Week 1: Evidence for risk factors and prognosis Flashcards
What type of studies are used to investigate risk factors?
Analytic –> Observational –> Cross-sectional studies
What is a cross-sectional study?
Exposure and outcome assessed at the same time
Assess prevalence in population
Easy
Inexpensive
Weak evidence for association
Temporal sequence not demonstrated (except time invariant factors)
What is a case-control study?
Advantages?
Disadvantages?
Identify cases and then compare with controls (non-diseased patients)
Look back at exposures/risk factors
Then identify risk factors
ADVANTAGES: Excellent for RARE diseases, good for evaluating multiple risk factors, good for disease with long latency, cost efficient
DISADVANTAGES: inefficient for rare exposures, susceptible to bias (selection and information), temporal relationship exposure may make outcome unclear, intermediate strength of evidence
How can the validity of a case-control study be assessed? 5
whether there are clearly define groups of patients
if risk factors are measured in same way in both grops
if potential confounders are identified and adjusted for
if patients are followed for sufficient time
if conditions of causation are satisfied
Define recall bias (a type of information bias)
difference in recall of risk factor status based on outcome status (often cases are more likely to remember exposure than control)
Define observer bias
investigators assessing risk factor status do so with differing accuracy for cases and controls
Define confounder
effect of an extraneous variable wholly or partially accounting for apparent association between exposure and disease
How can conditions of causation be satisfied?
Not just statistical association but biological? Strength of association consistency with other work specificity of association temporality of relaitonship biological gradient/dose - response biologically plausible supporting experimental evidence
Define Odds Ratio (OR)
Odds of exposure in cases/ odds exposure in controls
OR>1 = increased odds
OR< = decreased odds
OR=1 no association
Explain confidence intervals (CI)
how precise the estimates of an effect are (i.e. the level of uncertainty in the measured results)
CI = the range in which the true treatment effect is likely to lie within a given degree of confidence. “95% sure that the true value lies within the interval”
Define prognosis
the prediction of the patient’s clinical course over time and anticipated complications
What kind of study is used to investigate prognosis?
Analytical study –> observational –> cohort study (preferred method).
Also: systematic reviews (frequently unavailable), randomised control trials (strong evidence, rarely available), case-control (rarely appropriate, may identify factors associated with improved prognosis)
What is a cohort study?
Types? 2
group of subjects with a common/defined characteristic - i.e. exposure/risk factor status. EITHER:
PROSPECTIVE (preferred) or RETROSPECTIVE
Advantages - cohort study - 6
good for rare exposures and common outcomes
time sequence exposure followed by outcome
can assess multiple outcomes
less susceptible to selection bias
measure incidence and relative risk
strong evidence for causal link
Disadvantages - cohort study
Prone to bias (losses to follow up) time consuming, slow if long incubation exposure status can change over time inefficient for rare outcomes expensive
How are cohort studies critically appraised?
Validity:
representative sample?
clear definitions (exposure/outcome)
adequate follow-up
objective outcome applied in blinded manner
adjustment for prognostic factors and confounders
validation of an independent test set
Importance of results:
magnitude of effect?
precision of the estimates of the effect?
Where can the unexposed control group come from?
Internal source - preferred
External source - from separate population, less ideal
What % losses to follow up should you aim for?
<5-10% losses to follow-up (bear in mind that certain patients/groups of patients may be more likely to drop out, so identify these and think how it may bias results, adjust if necessary/possible)
Give some examples of confounding factors
patient age/gender/disease type/location
concurrent disease
confounding variables
How do you calculate proportion?
numerator included in denominator
How do you calculate odds?
numerator not included in the denominator
Define prevalence
Number of EXISTING cases/ total number of animals at a specific time point
Define incidence
Number of NEW cases in a defined population within a specified PERIOD of time
Define Incidence risk (‘cumulative incidence’)
new cases/number of patients at risk over a specified period of time
Define incidence rate
Rate - the denominator has patient-time units at risk. new cases per animal-time unit (e.g. 10 new cases of lameness/ 100 cow-years)
Define risk ratio (RR)
Ho many TIMES more likely with exposure. Risk exposed/risk unexposed RR>1 = increased risk with exposure RR=1 = no association RR<1 = reduced risk with exposure
Define median survival time
Time at which 50% if patients have developed the outcome
What are survival curves?
each point on the curve represents the proportion of patients that have NOT developed the outcome. Probability of surviving. Kaplan Meier curves show how quickly patients are dying.
How can the importance of valid evidence in cohort studies be assessed?
precision of estimates - Confidence interval provides range within which true value is likely to lie:
survival curve
survival times
relative risks
When is a result clinically important? 3
will it affect our advice to the client?
will it influence the treatment?
will it inform us of the likely future for our patient?
