Clinical trials Flashcards
why don’t we like before and after studies
due to regression to the mean
regression to the mean
take most extreme populate with intervention and will always tend to regress to the mean
regression…. extended
reason people have joined the study [predisposed to an upswing
i.e. usually both arms of the trial will improve
e.g. someone with bad migraines- migraines ill tend to get better e.g. just by talkingg and being reassured by the doctor
due to regression to the mean
we must be sceptical
example of variable dependent variables
e. g. did have high bp, then goes down to normal at T2
- however someone from T1 with normal BP may have high BP at T2
-population stays the same, but people move in the population
due to upswings and downswings
after new findings how long doe it take for clear, widespread practice to start in hospitals
17 years on average e.g. cortisone
however only 2 years for artificial surfactant- money and companies
why do we randomise
best way of making differences in baseline in groups only occur due to change
- want differences at baseline to be due to chance rather than systematic on one way or another
randomisation means that
it is the manipulation of treatment which manipulates the data and not confounding variables
statistics help us with
chance differences
statistics do not help us with
systematic differences
why manipulate data
- a desire for a positive outcome
- stop money feeling wasted
- much more likely to get published in big journal
- get famous
why is there a need for allocation concealment
if clinicians are part of the trial they may want to give their patients the active treatment- can be conscious or unconscious
allocation concealment
around 30% diff in effect
getting the question right
i.e. if the study has addressed a male population, the results shouldn’t be extrapolate for females
questions to answer
is the research valid?
is it applicable to the problem
clinical problem vs study question
PICOs should match
example PICO
P people who have snakebites in rural india, treated with anti-venom
I adrenaline, anti-venom
C antihistamine , steroid. Anti-venom along
O deaths, morbidity, mortality, cost
randomisation involves
how allocation of groups occured
what can be used to randomise
can be computer generated- no bas
randomisation ensures that
differences in demographic groups are just due to chance and reduces ALLOCATION BIAS
in clinical trial papers it is important to show
the extent where baseline characteristics int he two groups are similar
e.g. sex, age, species of snake, delay in admission to hospital, first aid, worst outcmes
important to show baseline characterisitcs
so countries can see ow relevant data is for their populations- allow additional studies
Follow up- complete and long-enough?
- neurological evaluations?
- Kept in hospital?
- adverse reactions with graded criteria
- monitored for 96h
- long term check up?
why are follow ups important to record
allows future researchers to know what possible adverse effects could be
blinding
ensure both groups are treated the same
why should we blind
- clinicians- way change the way they manage patients
- patients- placebo effect
- researchers- outcome measurers- clearly objective
stats cannot be used for
baseline variation
what does a p value of 0.04 mean
if we did an infinite number of studies of this size 4% of the time we would get results like this more more extreme, if there was no difference
what drives p value
sample size and variation
