Week 6 Flashcards
what is the ranking of the hierarchy of study designs based on their contribution to causal inference from least to most?
case series -> cross-sectional -> case-control -> cohort -> RCT
which forms of bias should be considered in a RCT?
selection bias, performance bias, attrition bias, and detection bias
how do we control for selection bias?
randomized selection
allocation concealing
how do we control for performance bias?
blinding participants
how do we control for detection bias?
blinding outcome assessors
how do we control for attrition bias?
how was missing data dealt with? hard to control for, but should be considered when interpreting data
how do we control for reporting bias?
determine the factors that could result in reporting bias. reflecting on how under or over reporting may occur
identify the 2 ways standardized reporting was mandated
The Uniform Requirements for Manuscripts Submitted to Biomedical Journals
Consolidated Standards of Reporting Trials (CONSORT)
what are the downsides to simple randomization?
it is less efficient and there’s potential for group imbalance in small trials
what factors are considered inappropriate ways to randomize participants?
-test result
-alternating presentation
-case record number
-date of presentation
-age or birthday
what is stratified randomization?
study participants are stratified by prognostic factors (gender, age, etc) to reduce group imbalance in smaller trials
what is block randomization?
population is split up to account for the effects of a particular variable
ex. group 1 are smokers, they are evenly distributed into treatment group and control group. group 2 are nonsmokers and they are also evenly distributed into treatment group and control group
define what minimization means in a study
a complex computer-based method of allocating subjects into balanced groups based on key characteristics
T/F allocation concealment is different than blinding
true
when does blinding of a study occur? what bias does it prevent?
after subject allocation has been completed and is intended to occur throughout the entire duration of the study to prevent performance and detection bias
when does allocation concealment of a study occur? what bias does it prevent?
during allocation to prevent selection bias
T/F allocation concealment is always achievable
true
T/F blinding is not always achievable
true
what form of bias does triple-blinding help prevent?
reporting bias
define selection bias
when can it occur during a trial?
Systematic differences between baseline characteristics in the groups being compared
during allocation
define performance bias
when can it occur during a trial?
Systematic differences in the care provided to study participants apart from the intervention
during treatment intervention
define attrition bias
when can it occur during a trial?
how can it occur?
systematic differences in withdrawals from trial
during followup
drop-out participants or missing data (loss to followup)
define detection bias
when can it occur during a trial?
how can it occur?
systematic differences in outcome assessment
during outcome collection (end of trial)
if someone knows or figures out the intervention (treatment or placebo) which can consciously or unconsciously affect the trial outcome
an example of attrition bias is incomplete outcome data, what might cause this outcome?
trial stopped early
participants missing outcome assessment appointment
incomplete documentation from participants
an example of attrition bias is missing outcome data, what might cause this outcome?
patient data or records lost
patients excluded by investigators
non-adherence to medication regimen
how does the intention-to-treat analysis mitigate attrition bias?
what does ITT analysis mean towards the study?
all randomized patients are analyzed according to randomization allocation regardless of protocol deviation or study withdrawal
ITT means the investigators analyze all data without disregarding any data that can skew the outcome results such as lack of medication adherence
why is the intention-to-treat analysis good for superiority trials but not for non-inferiority trials?
it may bias non-inferiority trials into showing non-inferiority by making treatment and control arms look similar
how is intention-to-treat different from “per protocol”?
if the study only uses data using the “per protocol” method, then they can exclude any significant events that occur prior to the treatment intervention which may affect the analysis of the treatment. using the intention-to-treat method, you include all events during the entirety of the trial regardless of when the treatment was introduced.
For example, if 10 people suffered a stroke before having surgery (the treatment intervention), the per protocol method would not include those 10 people before the surgery and would only analyze participants who have a stroke after receiving surgery. the intention-to-treat method would include both the 10 people who had a stroke before surgery and those after surgery, providing a more accurate representation of if the treatment intervention was significant or not.
what percentage of patient loss is considered concerning in regards to determining attrition bias?
20%
define what multiplicity is in regards to a study
potential inflation of the probability of making a Type I error due to multiple testing (observing a difference when it does not exist)
what is Type 1 error? (a error)
rejecting the null hypothesis when it is true (observing a difference when it does not exist, false positive)
in order to adjust for multiplicity, we can use what’s called the Bonferroni method, what is it?
a statistical adjustment method which assumes the null hypothesis is the same for all outcomes as an attempt to correct the inflation of type I error from multiple testing
why should calculating the relative risk reduction not be used for primary outcomes?
may exaggerate the importance of the effect and shows a much greater risk difference compared to calculating the absolute risk reduction
define what the number needed to treat is
how is it calculated?
what is the ideal NNT?
it is an estimation of how many pts who have received a treatment over a certain amount of time is required before one person would experience a beneficial outcome
1/ARR where ARR is the percent difference between the two study variables (20% risk in placebo - 10% risk in treatment= 10% ARR)
ideal NNT=1 (one person needs to be treated to prevent one bad outcome over a specific period)
define what the number needed to harm is
an estimation of the number of patients who would have to receive the treatment over a specific time before one person would experience a negative outcome
*only calculate NNT and NNH for statistically significant results
