Brainscape's Knowledge GenomeTM

Browse over 1 million classes created by top students, professors, publishers, and experts.


See full index

Evidence Based Medicine > Clinical Trials > Flashcards

Clinical Trials Flashcards

1
Q

How to assess if a paper is worth your time?

A
  • look at the title
  • read the abstract/summary
    1. does the paper address your clinical Q? (consider the site of the study… is it similar enough to apply the results to your practice?)
    2. is the study design appropriate for the questions being asked? (methods section)
    3. if steps 1 & 2 are appropriate - critically appraise the study
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What Qs should you ask about the study subjects (P in PICO)?

A
  • how were the subjects for the clinical trial selected?
  • what population did they come from?
  • are the animals more or less ill than the animals you see in your practice (tertiary care setting vs primary care)
  • did the animals receive more attention than you could ever possibly give (was this a research herd or kennel?)?
  • were the subjects studied in real life circumstances?
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

How do we select our sample?

A

A sample population is taken from the eligible population in the general population

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What is external validity?

A
  • external validity of a trial refers to how applicable the results are to the general population of interest
  • that is, are the animals, herds, or patients in the study similar to the animals, herds, or patients that you are going to apply the results of the study to?
  • subject selection obviously has a major impact on external validity
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What Qs should you answer before you go too far looking at a paper and where should you look for these answers?

A
  • what specific intervention was being considered & what was it being compared w/? (the “I & C” in PICO)
  • is it a placebo (negative control) or a “positive control”?
  • look at the materials & methods section or abstract
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What is a historical vs a concurrent control group?

A
  • concurrent control grps are formed @ the same time as the treatment grp (parallel arm trials)
  • historical controls are before and after (these make nice stories but are probably not often of much use - too many other factors can change over time which might produce the differences btwn treatment & control grps)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What is a cross-over trial?

A
  • utilize the same animals as treatment & control grps (order of treatments must be randomized)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What are 3 key elements of clinical trial design?

A
  1. outcome measures (O in PICO)
  2. bias
  3. chance
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

what are Qs to ask about outcome measures?

A
  • what is the most clinically relevant?
  • what is the easiest to measure reliably?
  • which is the most objective?
    > blinding is more important w/ subjective outcomes
    > case definitions become more important w/ subjective outcomes
  • which outcome is the most specific?
    > decrease the noise or measurement of unrelated effects (ex: Histophilus vx trial - outcome could be all mortalities or hemophilus specific mortalities)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Why should you ask how many outcomes?

A
  • beware of the multiple outcomes paper!
  • if they look @ enough outcomes, 1 just might be significant
  • @ the v least the author should prioritize them
  • want either 1 or 2 primary outcome measures
  • if you monitor 6 outcome measures, you will have 1 chance in 4 of incorrectly concluding the treatment or vaccine is effective by chance alone
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What are possible outcomes for a BRD trial?

A
  • mortality (most objective)
  • morbidity - 1st pulls (case definition of “sick”)
  • serological conversion (interpretation?)
  • avg daily gain
  • feed efficiency (what lvl can we measure these at?)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What is the experimental unit?

A

smallest independent unit to which treatment is allocated
- leg of the animal
- udder quarter
- individual animal
- pen of animals
- herd/flock/kennel
important to analyze results on the basis of what was the experimental unit (pen not individual animal, etc.)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What is “herd immunity” in a study?

A
  • vaccinated/treated animals may protect/reduce the challenge to all individs in the herd/grp
  • if vaccinates & non-vaccinates are commingled in pen (ex: BVD vaccine trials)
  • anthelmintic trials
  • minimizes differences in outcome
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

three reasons for an outcome variable difference btwn treatments?

A
  1. meloxicam is effective @ reducing pain thus affecting the outcome variable
  2. bias is present in the trial (something other than the treatment status - meloxicam vs placebo - accounts for the outcome variable difference)
  3. outcome variable difference could occur simply by chance
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

What is bias?

A
  • did some factor other than the treatment itself “cause” a difference in the outcome btwn the treatment & the control grp
  • bias = “systematic” difference btwn the treatment & the control grp which could affect the outcome measure
  • if there are important biases we may conclude the treatment works, when it really does not work or vice versa
  • potential for bias may cause us to Q the validity of the trial! (are it’s conclusions correct?)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

4 key times when bias might occur:

A
  1. selection/randomization
    - systemic difference in treatment & control grps attributable to lack of randomization (1st half given meloxicam & 2nd half saline, what is oldest calves hang out in front?)
  2. performance bias (cointervention; systematic differences in care provided apart from the intervention being evaluated)
    - castration vs not; person castrating should be blinded to control grps so cant change aftercare
  3. exclusion bias
    - systematic differences in w/drawals from trial (ex: bathing dogs every 3 days vs Bravecto - Os may drop out on one side only)
  4. detection bias
    - systematic differences in outcome assessment or follow-up
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

What are the target populations for each type of bias?

A
  • Selection bias: intervention grp/control grp
  • performance bias (co-intervention): exposed to intervention/not exposed to intervention
  • exclusion bias: follow up
  • detection bias: outcomes
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

Why should the animals in the treatment & control grps be treated as uniformly as possible?

A
  • treating all of the animals from 1 source & leaving all animals from another source as controls could result in substantial differences in the outcome measure merely due to “source effect”
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

What are some example sources of selection bias?

A
  1. non-randomized clinical trials
    - just comparing 2 gps of patients that for various reasons (O or vet bases) that were treated differently
    - there are inherent self-selected differences in these 2 grps before the intervention is applied!
  2. vaccinate the 1st half of the truck load
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

How do you virtually eliminate selection bias?

A

RANDOMIZATION

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

what is randomization?

A
  • absolutely necessary to avoid bias
  • esp eliminates trial entry biases
  • legitimizes the use of subsequent statistical analyses
  • protects against systematic differences in control & treatment grps
  • baseline comparison of known covariate helps to evaluate effectiveness of randomization
22
Q

How do you randomize?

A
  1. simple randomization
    - toss a coin, D & D dice
    - random number tables (most statistical texts)
    - random number generator (spreadsheets)
  2. systematic randomization
    - not true randomization
    - randomizing 1st animal & then alternating assignment btwn treatment & control grps
23
Q

What are confounding variables?

A
  • factors that may be related to the outcome measure that are not equally distributed btwn treatment & control grps
  • proper randomization should balance most confounding variables
  • sm trials, randomization may not actually balance these factors
  • result could be that the difference in outcome is due to the confounding factor & not the treatment
24
Q

how do you overcome selection bias?

A

Design
1. randomization
- tends to produce study grps comparable w/ respect to known & unknown risk factors
- removes investigator bias in the allocation of participants
- guarantees statistical tests will have valid significance levels
2. other additional options (esp if sample sizes are sm)
- stratification or blocking
- matching

25
Q

what is stratified randomization or blocking?

A
  • simple randomization may not equally balance important confounding variables
  • divide the population into different strata according to key confounding variables
  • ex: breed, age, sex, etc
  • then use simple randomization w/in each subgrp
  • esp important if sample sizes are sm
26
Q

what is matching?

A
  • @ the individual lvl means that a pr of treatment & control subjects are chosen to be as similar as possible in terms of certain key variables (age, sex, breed, time of admission)
27
Q

What is performance and exclusion bias?

A
  • treatment & control patients may receive unequal observation, follow-ups, or ancillary treatments
  • esp true if researchers or clinicians or animal caretakers are aware of treatment status
  • some patients will almost always be lost to follow-up (are those that are lost equivalent among treatment grps & controls?)
28
Q

what are problems associated w/ ancillary treatment?

A
  • contamination: control patients accidentally receive the experimental treatment
  • co-intervention: performance of additional dx or therapeutic acts on the experimental but not control patients
29
Q

What is detection bias?

A
  • bias may occur b/c dx decisions can be affected by knowledge of treatment status
  • this becomes more important in outcomes that are subjective measurements
  • process to measure the outcome may actually differ btwn treatment & control grps which could potentially bias the trial
30
Q

What is blinding?

A
  • subject & possibly the observer are unaware of the treatment grp to which subjects have been allocated
  • single blinding: only subjects (clients) are unaware of which treatment grp they belong to (requires use of placebo)
  • double blinding: both subjects & observers are unaware of treatment grps
  • triple blinding: statistical analysis is blinded
31
Q

What 3 biases are eliminated by blinding?

A

performance, exclusion, & detection bias!

32
Q

What is the fourth key element of clinical trial design?

A
  1. outcome measures
  2. bias
  3. chance
  4. TREATMENT EFFECT
33
Q

What is hypothesis testing?

A
  • remember we started by examining the CLINICAL Q or “hypothesis” of the clinical trial paper we are reading
  • this is often stated in the abstract or in the materials & methods section
  • hopefully the objective or “hypothesis” is stated v clearly
34
Q

What is chance?

A
  • clinical trial is attempting to infer something about the general population from their study population
  • clinical trial will simply give us an ESTIMATE of the treatment difference in the population
  • if the authors had the ability to do the same trial again w/ a different sample population the results would not be exactly the same each time
  • amt of uncertainty about the clinical effect of their treatments needs to accompany their results
35
Q

What is the null hypothesis (H”0”)?

A
  • researchers have a theory/hypothesis which they are investigating about the population (clinical Q or PICO)
  • in trials or experimental designs, this hypothesis is usually comparative in nature
  • most statistical techniques start w/ the hypothesis that there is NO DIFFERENCE btwn the treatments (H”0” = no treatment effect in the population)
  • researchers then examine the data from their trial to ascertain if they can reject the null hypothesis
  • rejecting the null hypothesis would mean that there actually is a treatment effect & 1 treatment is “significantly” different than the other
36
Q

What is the alternative hypothesis?

A
  • if we can prove the null hypothesis is wrong, we can then move on to the alternative hypothesis (Note: cant actually “prove” that the null hypothesis is true)
  • alternative hypothesis is that there is a difference btwn post-operative complications &/or pain scores in Ca that receive traditional OVH vs laparoscopic OVH
  • NOTICE THAT HAVENT STATED THE DIRECTION OF THE DIFFERENCE OR THE EFFECT (not presuming prior to the trial that laparoscopic OVH is better or worse than traditional OVH in terms of the outcomes measured)
  • this is called a TWO-TAILED TEST of the hypothesis!
37
Q

What are one-tailed tests vs two-tailed tests?

A
  • in almost every clinical trial, authors should use a two-tailed statistical test to evaluate the null hypothesis
  • this assumes that we dont know which direction the outcome will go if we reject the null hypothesis
  • it could be better than the control group or it could be worse!
  • be v suspicious of trials where the authors utilized one-tailed statistical tests
  • one-tailed tests allow the researcher to get a statistically significant result w/ fewer animals (more power)
  • however, there are v, v few situations where they can be used in a valid fashion
38
Q

How do you accept the alternative hypothesis?

A
  • cant be tested directly
  • accepted by default if null hypothesis is rejected
  • Note: cant ACCEPT OR PROVE the null hypothesis
39
Q

What does hypothesis testing involve?

A
  • conducting a test of statistical significance & quantifying the degree to which sampling variability may account for the results observed in a particular study
40
Q

What is the p value?

A
  • researchers will use the data from their clinical trial & they will compare the 2 (or more) grps using some form of statistical test
  • attached to the values of the test statistic is a probability or P-value (probability statement)
  • P-VALUE DESCRIBES THE CHANCE OF GETTING THE OBSERVED EFFECT IN THE OUTCOMES MEASURED, IF THE NULL HYPOTHESIS IS TRUE!
  • OR, WHAT IS THE LIKELIHOOD THAT WE WOULD SEE THIS RESULT BY CHANCE ALONE, EVEN IF THE TREATMENT HAD NO “REAL” EFFECT AT ALL
41
Q

How do you make a decision using the P-value?

A
  • if the P-value is v sm, then it is unlikely that we could have obtained the observed results if the null hypothesis was true, to therefore, we reject H”0”
  • if the P-value is v large, then there is a higher probability that we could have obtained the observed results if the null hypothesis were true & we do not reject H”0”
42
Q

How small of a p-value do we need?

A
  • researcher probably should have decided before they started the trial how sm of a p-value they would need before they would reject the null hypothesis
  • 5% is the standard lvl of “statistical significance” in most scientific journals but it’s completely arbitrary
  • p = 0.05
43
Q

What does statistical significance result in?

A

categorization of all results

44
Q

what gives more meaningful info than statistical significance?

A

confidence intervals

45
Q

What is the difference btwn statistical & clinical significance?

A
  • statistical significance simply defines the likelihood of achieving this treatment difference by chance alone
  • v trivial differences may be statistically significant if a large enough number of animals were included in the trial
  • in sm trails (low power), v lrg differences may not be statistically significant
  • if not significant, was the trial large enough to show a difference?
  • statistical significance is a necessary precondition for consideration of clinical importance but indicates NOTHING ABOUT THE ACTUAL MAGNITUDE of the effect
46
Q

What are two potential errors that can occur in regards to statistical significance?

A
  1. type I error: we can conclude that the drug or vaccine or new surgical technique actually is better than the control grp when in reality it is not
  2. type II error: we could also conclude that the new drug or vaccine or surgical technique is no better than the control grp when in reality it is
47
Q

What is a Type I error?

A
  • alpha error
  • type I error occurs if a study finds a treatment difference when, in fact, there was no difference
  • FALSE CLAIM
  • “alpha lvl” or “lvl of significance” is the lvl of type I error (or p value) that one is willing to accept (typically a < 0.05)
  • setting a lvl of 0.05 means that we will declare a difference statistically significant if p < 0.05
  • that means we will accept that potentially 5% of the time we will declare there is a treatment difference when, in fact, a difference does not exist
  • usually set @ a v sm lvl
  • 5% or 1%
  • PRIMUM NO NOCERE
  • alpha lvl is equivalent to the p-value where results will be declared significant
  • want to be 95% certain that a treatment difference is not just a random result
  • risk of a “false claim” due to chance
48
Q

what is a type II error?

A
  • (beta error)
  • occurs if a study fails to find a treatment difference, when in fact there was a difference
  • MISSED OPPORTUNITY
  • lrger the trial, less likely that a type II error will occur (the smaller the beta)
  • type II errors or beta errors are usually set at 0.20
  • willing to accept that a type II error will occur 20% of the time
  • that is, can accept that 20% of the time, will be unable to detect a treatment difference when in fact one exists
  • POWER = 1 - beta
49
Q

How are type II errors related to power?

A
  • type II error is only a possibility if there is no significant differences detected btwn treatments & controls
  • Q to ask: if there is no statistical difference, did the trial have enough power to detect a difference if 1 truly exists?
50
Q

If the vaccine didnt work in reality, but worked in the trial, what kind of error is this?

A

Type I

51
Q

if the vaccine worked in reality, but didnt work in the trial, what kind of error is this?

A

Type II

Evidence Based Medicine (8 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2024 Bold Learning Solutions. Terms and Conditions