types of experimental knolwedge

anecdotal

observational

tested knowledge

observational studies

treatement effect is **immediate**

impact must be large

comparisons are grounded in __historical memory__

larged biased by **placebo effect** and **recall bias**

no information about sid effects

Kaplan Meier Curve

survival curves

need to note how many patients are represented at each juncture

percentages represent how many patietns are at risk at each time

hash marks are sensored patients because of some reason

Why isn't observational and anecdotal success enough?

just because a substance may be noxious, does nto mean that all people exposed to it will have an untoward event

How do we know the outcome is valid?

an outcome is valid when chance and bias are taken into account and deemed to not have a major impact on the result

minimizing chance

randomization balances important variables

the mor variables being examined, the greater the number of people needed to balance the variables

alpha/type I error

probability that a study found a treatment difference when, in truth, no association exsists

a study is statistically significant if the probability of an alpha error is <0.5

beta/Type II error

the probability that a study found no difference in treatment when, in truth, an assciation exists

1-beta = power

power is analogous to sensitivity of a dignaosti ctest

benefit of randomization

more likely to balance unknown and unmeasured confounders

number needed to harm

to calciulate, determine the number of people needed to accrue one dath

if 3.3% die with each person treated, then for one death you would need x people or 1=3.3%*x

possible biases

selection biase (where do patients come from?)

spectrum bias (is the severity of disease reasonable)

blinding bias (do the patients or the assessors of their disease know what treatment they are receiving?)

intention to treat analysis

maintains all of the patietns in each of the groups established with randomization even if the patients do not follow their assigned group's pathway

a patient in the placebo group can bet active therapy and a patient in the active treatment group can stop medication or intervention

preserves effects of randomization

all subjects should be analyzed in the group that they were randomly assigned

corss-over trial

patients serve as their own controls

makes randomization unnecessary

condition must be chronic and similar in spectrum when treatment is stopped

what to look for in a trial

an outcome that matters

a practical intervention

similar groups

unbiased results

evaluation for chance

if all are present then you may we decide to use the new intervention