Lecture Flashcards
What is equipoise?
the assumption that there is not one better intervention present during the design of a RCT
what are the types of evidence?
trials and studies
what are trials?
involves an intervention -> measure an outcome
gold standard = randomised double blinded controlled trial
what are studies?
observational, no intervention
Diagnostic - how good a diagnostic test is
Prognostic - look at rf on outcomes often of cohort study
Case control - retrospective, two groups of different outcomes
Cohort study - group of people. prospective
what levels of evidence are 1a/1b?
Ia: systematic review/meta-analysis of RCTs
Ib: at least one rct
what level of evidence are IIa/IIb?
IIa: at least one well-designed controlled study without randomisation
IIb: at least one well-designed quasi-experimental study such as a cohort study
what is a quasi experiment?
subject to concerns regarding internal validity, because the treatment and control groups may not be comparable at baseline. With random assignment, study participants have the same chance of being assigned to the intervention group or the comparison group
what level of evidence is III?
III: well designed, non-experimental descriptive study eg comparative study, correlation, case-control study, case series
what level of evidence is IV?
opinions and/or clinical experience
What are the levels of evidence?
Ia (highest)
IV (lowest
What is phase 1 of trials?
safety testing of new med on few (often) healthy volunteers - establish side effect profiles
What is phase 2 of trials?
testing on larger groups of disease patients - establish idea of efficacy and further side effects
what is phase 3 of trials?
large comparison trials on disease patients - for robust efficacy data in comparison to existing treatment
what is efficacy?
the ability to produce a desired or intended result
what is a null hypothesis? (Ho)
there is no difference between the two groups, any difference between these two groups is due to chance
what is the alternative hypothesis?
there is a difference between the two groups
what is the p value?
probability of detecting a difference with a priori assumption that the null hypothesis is true
what is the p value set at?
5% or 0.05
what does it mean if p value <0.05?
consider such a low probability that there must be an error in the logical argument - the error is considered to be in the priori assumption, therefore the null hypothesis cannot be true so the alternative hypothesis must be true
what are the limitations of hypothesis testing?
the amount of difference is not stated simple binary (yes/no) answer
what is the estimation?
we can never know true difference between two groups in a population so we take a sample - estimate the difference between the two groups using the data from the sample
what is the absolute difference?
the difference found in a study = estimation of the whole population
what is the true difference?
the actual difference in population
how can the estimation be presented?
- as a number
- as an absolute difference
- as a ratio
- as a proportion
what are confidence intervals?
is the range around the estimation
normal 95% CI = the range between which there is 95% chance that the true value will lie
what are the units of no effect in confidence intervals?
with absolute difference - no difference would be 0
with a ratio - no difference would be 1
what is good about confidence intervals?
width of interval indicates PRECISION of estimate
can be used to estimate type 2 statistical error
indicates lack of statistical significance if interval crosses unit of no effect
gives info of size and direction of difference
what is a type 2 statistical error?
the study has shown that there is no statistically significant difference in the groups but in the real world there might be (false negative)
what is the primary outcome?
the variable outcome of most interest
any difference in outcome between the groups is believe to be due to the intervention
what is the minimum clinically important difference?
the least difference outcome that is clinically important -
looking at the clinical significance with statistical significance
what is the intervention?
what we do to the patient that can change things
what are time points to measure?
when the outcome is measured
what is the power of a study?
the probability that a trial will detect a difference that truly exists
what does the power calculation look at?
type 1 error: finding a difference when there isn’t one in real life (false positive) = ALPHA
type 2 error: not finding a difference when there is one (false negative) = BETA
power is dependent on what four things?
- probability of T1 error (0.05) [ALPHA]
- sample size
- standard deviation of sample
- minimal clinically significant difference
what does power equal (calc)
1-beta
beta = probability of getting a type 2 error
what is a type 1 error?
false positive
incorrectly rejecting null hypothesis
experiment shows significant difference when the null hypothesis is true (no difference in actual population)
probability of type 1 error = alpha
what is a type 2 error?
false negative
incorrectly accepting null hypothesis
experiment does not show significant difference when the alternative hypothesis is true (there is a difference in actual population)
probability of type 2 error = beta
what is type 1 error due to?
- bias
- confounding factors
- multiple hypothesis testing
what is type 2 error due to?
sample size
what is sample size?
number of participants required to detect a difference of a specified size
what is screening?
physically identifying potentially eligible participants
what criteria do you use when screening patients?
inclusion/exclusion
- every pt that fulfils inclusion criteria must be recorded
- everyone must have exclusion criteria applied -> if not, must record why
what is allocation concealment?
allocation sequence is concealed from all staff (reducing selection bias)
what is randomisation?
process of assigning to groups with known chance for each participant to enter any group (reduces allocation bias)
equally distribute confounders
what is blinding?
treatment assignment is concealed (reduces ascertainment bias)
concealment if whether in placebo group or not
how do you get ethics for your research?
submit protocols to REC/IRB for approval
what are the types of data?
continuous
categorical
what is reported in continuous data?
normally distributed - mean and standard deviation
non-normally distributed - median and inter-quartile range
what is categorical data?
nominal (without order) e.g. colour, gender
ordinal (with order) e.g. scale - slow, medium, fast
what is intention to treat analysis?
analysis includes every subject who is randomized according to randomized treatment assignment. It ignores noncompliance, protocol deviations, withdrawal, and anything that happens after randomization
what is imputation for missing data?
replacing missing data with substituted values
what are the measurements of central tendency?
mean
median
mode
what is the mean?
the average
used in normally distributed data
SD measured
what is the median?
central value
use if data is skewed
IQR measured
what is the mode?
most frequent value
rarely used
what are the techniques for comparing groups?
categorical: chi-squared (X2)
continuous: depends on distribution - T-test, ANOVA
does funding need to be stated?
yes - all funding needs to be clearly stated and role of funders
what is enrolment?
if they fit inclusion criteria - sign consent (NOK if don’t have capacity or PLR by nurse/doctor)
what is risk?
the probability of an event occurring
what is the absolute risk reduction? (ARR)
the difference between the event rate in the treatment group and the event rate in the control group
(event rate in control group) - (event rate in treatment group)
what is numbers needed to treat?
the number of patients that need to be exposed to intervention for one of them to sustain the effect
1/ARR
what is the risk ratio? aka relative risk
ratio of risk of outcome in treatment group to risk of outcome in control group
RR > 1 then the risk of the event in the treatment group occurring is great than the rate in the control group
RR<1 the risk of event in the treatment group occurring is less than the rate in the control group
(event rate in treatment group)/(event rate in control group)
What is the relative risk reduction?
the proportion by which the intervention reduces the outcome rate
watch out for high RRR in events that are rare
e.g. increased risk of breast cancer in young women due to OCP - high RRR, low ARR
proportion of ARR to the control event rate
(CER-EER)/CER
What are odds?
another way to express chance
the ratio of the number of times an event occurs to the number of times it does not occur
the odds of rolling a six on a dice = 1/5 =0.2
(NB risk of rolling a six = 1/6)
what is an odds ratio?
ratio of odds of event in treatment group to odds of an event in control group
expresses how much more likely the event will occur in treatment group in comparison to control
>1 = the event rate is greater in the treatment group <1 = the event rate is greater in the control group
what is accuracy?
inaccuracy is due to systematic error (bias)
an accurate study hits bullseye every time but have poor spread within bullseye
what is precision?
imprecision due to random error (chance)
precise study has closely clustered results but may not in the gold
what is accuracy in a test?
the proportion of patients correctly identified by the test (true positives and true negs)
what is the reference standard in 2x2 diagnostics?
definition of having the disease for the purposes of the paper
what is the gold standard?
definition of having the disease as recognised by international medical community
what is sensitivity?
proportion of patients with the disease that have a positive test
ability of a test to correctly classify patients as having disease
true positive/(true positive + false negative)
what is specificity?
proportion of patients without the disease that have a negative test
ability of test to correctly classify patients as not having disease
true neg/(false pos + true neg)
What is SpIn?
Rule in a disease
Test with high specificity (with few false positives) accurately rules in the diagnosis
what is SnOut?
Want to rule out disease
Test with high sensitivity (few false negatives) accurately rules out a diagnosis
does changing the population prevalence change the sens/specificity?
nope - about the test
what is the positive predictive value?
proportion of patients with positive test that have disease
true pos/(true pos + false pos)
what is the negative predictive value?
proportion of patients with neg test that do not have disease
true neg/(false neg + true neg)
will PPV and NPV change with population prev?
yes - about the patient
What happens to PPV and NPV when the prevalence increases?
PPV rises
NPV drops
Sens/spec stay the same
What kind of diagnostics do we use in ED?
Usually use tests to RULE OUT a diagnose more often than to rule in
so often have high SENSITIVITY (SnOut) and low specificity
What do you want in order to diagnose something?
high positive predictive value
but low prevalence gives low PPV so need to perform initial test on population that has a high prevalence -> aka screening
What is a likelihood ratio?
(probability of individual with condition having the test result)/(probability of an indiv without the condition having the test result)
How do you measure likelihood ratio for positive test?
sensitivity/(1-specificity)
How do you measure the likelihood ratio for negative test?
(1-sensitivity)/specificity
What does the likelihood ratio measure?
how much more likely (for positive) or less likely (negative) than the pre-test probability is it that a patient has the disease
what is a receiver operating characteristic curve?
another way of gauging the usefulness of a diagnostic test
x-axis: 1-specificity
y-axis: sensitivity
want each to be perfect = 1
what does the area under the curve represent?
the closer to 1 - the better the test
(between 0.5 and 1)
the bigger the area - the better the test