PHRM3031 interventions RCTS I/II Flashcards
Study Types
Experimental - RCT
Observational - cohort (longtudinal), case-control, cross-sectional, ecological (population-based)
Study Types & their questions
RCT
interventions
Study Types & their questions
Cohort
Aetiology
Prognosis
Diagnosis
Risks
Study Types & their questions
Case Control
Prognosis
Diagnosis
Risks
Study Types & their questions
Cross Section
Aetiology
Diagnosis
Frequency
Study Types & their questions
Case Reports
Next Steps
Which Study is best for:
Intervention/Therapy
RCT
Which Study is best for:
Diagnosis
Cohort>Case-Control>cross-sectional
Which Study is best for:
Risk, prognosis
cohort>case-control
Which Study is best for:
aetiology
cohort>cross-sectional
Which Study is best for:
frequency
cross-sectional
Which Study is best for:
observational
qualitative
Cross Sectional Study
- the study population is ascertained at one point in time - a snap shot
- participants are asked about their current disease status (outcome) and their current and past treatments (exposure)
Case Control Study
-compares disease outcomes in cases and controls
Cases: patients with a particular disease outcome
Controls: have similar attributes (age, sex, ses) to the cases but do not have the disease
Cohort Study
Cohort: a group of individuals with a common characteristic
Cohort Study: cohrts are followed up (observed) over time
Aim –> is exposure to a risk factor associated with new disease outcome in the future
Prospective Study
observe outcomes or events that occur after identification of subjects
-RCTs, cohort or longitudinal studies
Retrospective Study
review records from the past or obtain information about past events
-case-controls studies
Randomised Controlled Trial (RCT)
- a sample of patients are selected from the population with a particular condition
- patients are randomised into the treatment group or the control group
- the two groups are treated in exactly the same ways in everything apart from the treatment
- clinical measures are made over time and any differences are attributed to the intervention
Critical Appraisal -definition
- structured summary
- are the results of the study valid?
- what are the clinically relevant results?
- will the results help care for my patient?
Critical Appraisal - Design
Parallel - common, patients are randomised to either group A or group B at the start of the study and the patient receives the assigned medication for the duration of the trial
Cross-over - fairly rare, patients are randomised to either group A or group B at the start the study and midway through the study, are swapped to the other group
Factorial - rare, sometimes used when interactions between various drugs are to be investigated
Assessing Bias: Seven Questions
- were patients randomised?
- was group allocation concealed?
- were groups similar at baseline (start of trial)?
- were participants, health professional and study personnel blinded to group allocation?
- were all participants who entered the trial properly accounted for at it conclusion and how complete was follow-up?
- Aside fro the experimental intervention, were the groups treated equally?
- were patients analysed in the groups to which they were first allocated?
Randomisation: defintion
- process by which patients in a clinical trial are allocated to a group
- assignment is according to chance rather than any systematic approach
- randomisation tables, often computer generated
- patients have an equal chance of being assigned to either group (treatment or control)
- does not mean that the treatments are evenly allocated to all patients
why randomise?
- aims to ensure that the control and intervention groups are as similar as possible
- known prognostic factors (such as age, weight, gender)
- unknown factors (cofounder eg. adherence, genetics, socioeconomic status)
- -> reduces the chance of representation of any one characteristic with the study groups
randomisation levels
simple: at level of patient
cluster - intact clusters of individuals
stratified - participants are grouped into ‘blocks’ and the allocated to treatment groups; often uses stratification
baseline characteristics
the treatment and the control group should be similar for all prognostic characteristics except for treatement
blinding
-concealing the treatment status from the patient, investigators or analysts
-used to guard against bias on the part of investigators or subjects when assessing or reporting the outcome of treatment
Four Types:
-single
-double
-triple
-open label (non blinded)
single blind
- the investigator, but not the patient, knows the identity of the assigned treatment
- can be subject to investigator bias
double blind
neither the investigator nor the patient knows the identity of the assigned treatment
the ultimate triple blind
no one knowns what is going on
-most rigorous but least commonly performed blinding
non blinded studies
- open label design
- both the investigator and the patients know the identity of the assigned treatment
- used when blinding is not possible (eg surgery, radiotherapy, diet)
- subject to bias by patients and investigators and so there is a greater chance of seeing a positive result
power of a trial
power is the ability to detect a difference between two groups when a difference actually exists
- type I (alpha) error: is the detection of a difference when a difference does not exists (false positive)
- Type II (beta) error: is the inability to detect a difference when a difference actually exists (false negative)
data analysis
may be either:
- according to the treatment to which patients were randomised
- according to the treatment actually received
analysis by randomisation
treatment the patient ws randomised to –> ‘intention to treat’ analysis
evaluates medicine effectiveness: how the medicine work when people take it in an (almost) ordinary setting
analysis by actual treatment
analysis is according to the treatment actually receive
evaluates medicine efficacy: doe the medicine work when it taken exactly as prescribed?
size of effect
best estimate is the difference in means (or medians if reported) between the intervention and control group
clinical vs statistical significance
clinical significance: the size of the effect is real (i.e) statistically significant) and that it is of a magnitude sufficient to change practice or to be considered clinically important
statistical significance: the size of the effect and the 95% CI in relation to the null hypothesis (usually <5% error rate)
trials can be statistically significant but not clinically significant
P values
-measures of the probability that a result is purely due to chance
a low p-value suggests that chance is an unlikely cause of the observed difference between groups
if the p-value is love (p<0.05) then the probability that the result was due to chance is also low (<5%) –> a statistically significant effect
Confidence Intervals (CIs)
-generally more informative than p-values
-estimates of the range of values that are likely to include the real value
usually quotes as 95% CIs
-if the 95% CI for the difference between a treatment and control group is small and does not overlap the ‘no effect’ point(1 for a ratio an 0 for a difference) we are confident that the result is real
Absolute Risk (INE)
AR (NE) absolute risk (incidence) of events in control (not exposed) group (%)
Absolute Risk (IE)
AR (E) absolute risk (incidence, I) of events in treatment (exposed, E) group (%)
Relative Risk (RR)
- definition
- formula
- tell us how many times more (or less) likely an event will occur in the treatment (exposed) group relative to the control (non-exposed) group
- RR=ARe/ARne
- RR=1 means no difference between the two groups
- RR<1 means the treatment reduces the risk of the event
- RR>1 means the treatment increases the risk of the events
Relative Risk Difference (RRD)
RRI - increase
RRR - reduction
(formula)
RRD = (ARne-ARe)/ARne
or (1-RR)
Absolute Risk Difference (ARD)
- definition
- formula
-tells us the absolute difference in the rates of events between the two groups (treatment difference)
ARD=ARe-ARne
-ARD=0 means that there is no difference between groups (i.e treatment had no effect)
Number needed to treat (NNT)
- definition
- formula
-is the number of patients that need to be treated (compared to the control group) for a period of time to prevent one extra event of interest
NNT=(1/ARD)x100
Clinical outcome
Number needed to harm (NNH)
- definition
- formula
-is the number of patients that need to be treated (compared to the control group) for a period of time to cause one extra event of interest
NNH=(1/ARD)x100
safety outcome
Summary Ratings Green Yellow Red Black
Green - therapies rated green are best you can get- there is clear evidence of patient-important benefits and these benefits clearly out-weigh any associated harm
Yellow - these therapies require more study. the data is inconclusive or not substantial enough to be able to give a clear rating yet. these would benefit from a large, randomised controlled trial to give better sense of effect and harms
Red - benefits and harms may be equal or equivocal, while there may be some benefits the benefits dont outweigh the harms
Black - harms without benefits, these therapies have very clear harms to patients without any recognisable benefit
Time to event analysis
-survival analysis
-event= dichotomous clinical outcomes that can occur only once
-outcomes may be negative (death), positive (still alive) or neutral (start glucose monitoring etc)
-use the term ‘survival’ regardless of the type of outcome and whether it is related to survival
time periods can . be continuous (months) or categorical (0-6 months)
2 methods of analysing survival curves
- kaplan-meier
- compares two groups using log-rank test (expected vs observed) - cox-proportional hazards
- cuts period o observation into small intervals, in each interval, compare rate of events (taking into account numbers of events in the intervals and number of person at risk in the interval) –> a relative risk
- combines all the estimates of relative risk –> hazard ratio
- can consider effects of variables (eg. gender, age, smoking)
