Evidence Based Medicine Flashcards
Decision making
- Using the best evidence that we can to make decisions for & w/ patients is incredibly important.
Clinical practice
- Virtually e/t done in clinical practice has evidence for it.
Scientific journals
- Just because it is in a scientific journal doesn’t mean that it is of good quality or true.
Clinical trials
- A control group identical to the experimental patients.
- Have to be identical to prevent another/ other variables from having an effect (Confounding bias).
- Show cause precedes the effects = shows causation.
- Fewer biases
- Gold standard of evidence= Particularly true for meds
- Clinical trails can also be used for different types of care or health care professionals i.e diff types of surgery.
- Clinical trials are NOT GOOD for finding the harm of meds. You won’t find v/ rare side effects in clinical trials.
Efficacy
Efficacy = Whether a treatment gives the desired results in ideal experimental settings ( e.g w/i a randomised controlled trial).
- Clinical trials & what was found w/i the study
- i.e If s/t reduces the risk of death in heart failure by 30% in a a clinical trial, that the effigy of the drug
Effectiveness
- The degree of beneficial effect in “real world” clinical settings.
- What is it like in the real world?
- Drugs/ meds tend to be less effective in the real world WHY? Clinical trials are the ideal situation.
CON of EBM
- A problem w/ EBM is that when you have evidence from clinical trials it can take people a long time for people to change their actual practice.
Bad decisions
- Making bad decisions in EBM can be quite serious & can lead to quite long advocate problems.
Criticism of EBM
- It just looks at science it doesn’t take into account patients.
- This is not true
- If you do a proper EBM you have to merge the evidence with/ the values of the person you’re treating w/ the population values.
Values
- Is s/t worth doing in the long run
Definition of EBM
- Evidence - Based Medicine is the integration of best research evidence with/ clinical expertise & patient values- Dave Sackett.
Factors of EBM
- Patient Concerns
- Best Research evidence
- Clinical Expertise
- Evidence-based medicine should sit at the intersection of these 3 elements
The Evidence Hierarchy
- Systematic reviews = Strongest
- Randomised Controlled Trials
- Cohort Studies
- Case- Control Studies
- Case Series, Case Reports
- Editorials, Expert Opinion = WEAKEST
Systematic Reviews
- Gold standard ( for whether a drug works or not)
- Gathering info from different trials & putting them together
- In cases of drug interactions, you can’t run clinical trials = Ethically this is awkward
Randomised Controlled trials
- Look @ them in the case of evidence for a drug working
- V/ hard to find adverse effects in RCTs
- You wouldn’t know that you were looking for it
- V/ V/ rare i.e 1 in 200,00 people
Cohort studies
- Observational studies
- Looking @ real world data to look for trends/ evidence for an effect
Case - Control Studies
- S/t this is the v/ 1st evidence of the adverse effects of drug interactions
Case Series, Case Reports
- Look @ case reports in the case of looking @ adverse drug reactions, particularly ones that haven’t been found prior to the drug being marketed.
- Observational studies looking @ real world data —> Full of confounding bias
- Can look @ adverse reactions that haven’t been reported yet—> v/ rare
- Can’t user randomised & controlled tests
Editorials, Expert Opinion
- Expert is no good w/o evidence
- Essay written by an expert.= Useful to read, but they are not gold standard for whether s/t is true or not
- When it comes to drug therapy isn’t always the best one
- Can be wrong
Types of research
- Hypothesis Generation
-Hypothesis testing
What does Hypothesis Generation include?
- Qualitative research
- Ecological
- Survey
- Observational studies (Case-control studies & Cohort studies)
Qualitative research
- Not representative
- Usually based around interviewing people or having a focus group
- Focus group= just tells you what people think & s/t people involved keep their preferences a secret.
- Purely about looking for more info that can help you make decision about what sort of thing you might want to look into.
Ecological
- Studies looking @ things happening at a Nationwide level
Survey
- You can’t test the hypothesis
- No control group
Observational studies
- Look @ / Take real world data & try to find things out from them.
- Case- control
- Cohort study = Collect data in future & Retrospectively = use data from the past. Follow cohorts prospectively ( in the future) through time. Prospectively = you collect the data in the future.
S/ cohort studies can be like RCTs, you can set up a group like a control group & an experimental group. Though may not have the same random allocation of individuals in the 2 groups —-> enables you to test the hypothesis.
Hypothesis Testing
- Quantitative Research
- Intervention Studies = RCT’S
- RCT’S = Randomised Controlled trails
- Cohort studies
Quantitative Research
- Things dealing with/ numbers , which you can do statistical tests on.
- can do s/ observational studies, but if your data is rubbish (regardless of stats) you can’t prove anything/ results don’t mean anything.
Intervention studies (RCTs)
- W/ RCTs you can find out whether s/t works or doesn’t.
Case studies or Case series
- Describe experience of one or more people w/ disease or a drug.
- often first stat alerting to new disease/ condition/adverse effect or syndrome.
SPECIFIC LIMITATIONS: - No denominator (no comparison group), Don’t know how many people have been exposed to the drug from 1 or 2 cases.
- No comparison= Only happening in a small number.
- Study population selected = Only those prescribed the drug.
- Sampling variation
Confounding
- Does the exposure cause the effect.
- Exposure = to a drug, disease or particles from Road traffic.
- Mixing of the effects of the exposure with/ that of a 3rd factor.
- The distortion of the apparent effect (because of other factors that you are not aware of) of an exposure brought about by the association w/ other factors that can influence the outcome.
Confounding
- Limit of observational studies
-RCTs control for confounding. - i.e Exposure = taking paracetamol
Effect = it removes your headache
Case control studies (observational studies)
- Observational study
- ALWAYS work backwards in time = Look back @ old data
- Find diseased (adverse effect) & non-diseased individuals(no adverse effect).
- Discover /Find exposure in groups
-i.e you think that a drug is associated with/ blood cancer, so you find those w/ blood cancer & look back through their med records- to see what they have been exposed to . If the drug you concerned w/ is present, you could think that the drug & blood cancer are linked.
Case & controls
- Good for finding rare events.
- Cases = Have the disease or outcome.
- Controls= Do not have the disease or outcome.
- Case control studies identify subjects based on outcome status ( when you start the study you already know whether they have it or not).
How to select cases?
- Define your disease/outcome, that you’re looking for.
- Select incident cases ( new cases) = cases w/ disease or adverse drug reaction.
- Select controls from same underlying population as cases.
- Matched controls to increase similarity. I.e in height, weight, diet.
- Common to have 3-4 controls for each case to increase “power”= Control group is bigger than the case group = increases statistical power of the study.
MMR & development disorders
MMR= Measles Mumps & Rubella
Smeeth Lancet 2004
- Concern that autism was linked to MMR vaccine.
- 1294 development disorders CASES = w/ autism 78% had MMR
- Follow backwards to see exposure
- 4469 W/ no disorder, matched on age, sex, & GP practice CONTROLS= w/o autism , 82.1% had MMR
- shows that Kids w/ MMR vaccine are less likely to have autism.
Benefits
- Good for rare diseases w/ a long latency (long time response/ effect to happen after you have been exposed).
- Allow for multiple exposures- you might find more than one influence.
Limitations of case-control
- Observational study = subject to confounding (might find s/t else going on or an association but it’s on a 3rd factor).
- Recall bias = people forget things, memory is diff when asked to recall events.
Cohort studies ( observational study)
-Find people free of disease.
- Find Cohorts of people exposed & unexposed.
- Follow the cohorts through time.
- Compare the risks of incident outcomes in each cohort (adverse drug reactions or the benefit that you’re looking for).
- Opposite of case-control, because you find the outcome 1st & then look backwards for cohort studies and find people who have been exposed to the drug and those w/o the disease & follow these people through time. Then you compare the risk of incident outcomes.
Cohort studies
- similar to RCT’S you have control group and experimental group but NO RANDOMISATION.
- Can do retrospective = look at the past
- Can do prospective = Look at the future
- Look @ exposed and unexposed and follow through time and loom at the differences.
When are cohort studies useful?
- To find a temporal relationship.
- Can be used for multiple outcomes
- Cohort studies identify subjects based on their exposure status ( To a drug or a pollutant).
Limitations of cohort studies
- Observational – Can’t prove cause and effect – just an association. Other factors may have caused the change.
- Inefficient for rare diseases( use a case-controlled study) or effects with long latency.
Intervention studies
- Control Group Data collection.
- Intervention Group Data collection. ( intervention like medication).
- Collect data as you go along.
- you are interfering in the real world.
Randomised Controlled Trials= Intervention study
When to use?
- Uncertainty about a treatment or an area of medicine (exposure)
- When exposure can be modified ( w/ tablets or injections, 1 group gets it, 1 group doesn’t).
Why
- Remove confounding (differences in the characteristics of
exposed and unexposed individuals) so you can prove causality.
- Equipoise (can’t assign people to a treatment KNOWN to be
worse) – needs uncertainty. If you already know that the treatment you are going to use is worse than the pre-existing treatment you can’t do the clinical trial= Unethical.
Population Sample
- Randomly allocate to either intervention or control and then you get your outcomes.
Comparison Groups
- No therapy
- Similar therapy or drug
- Placebo ( So people don’t know what groups they are in/ secret)
- Pre-existing standard ( goal standard treatment)
- Depends on the question and EQUIPOISE
- How serious the illness is i.e for HIV a placebo =unethical.
Randomisation
Carried out to balance characteristics (i.e. gender) of control and treatment group.
*Small studies are less likely to achieve balance. You get imbalance.
*1ST THING YOU SHOULD LOOK @ IN CLINICAL TRIALS=Check baseline characteristics ( Tells you whether there is a bias in the control & treatment group & what the 2 populations in the groups are).
Baseline characteristics
- Can include
- Age
- Male/ female ratio
Extra info= women tend to be under represented in clinical trials.
RCT limitations
- Generalisability of the study population ( Collected from particular groups of people).
- RCT populations are v/ particular as the study has v/ specific criteria.
- Generalisability of the study environment.
- You will be seen in a hospital environment routinely if you are apart of a clinical study.
- Limited questions & v/ specific
- Limited Clinical applicability = linked to the type of questions being asked.
Intention to treat analysis
- Patients allocated to control and intervention @ start of the study).
- Analysis carried out on the group (control or intervention) they were initially allocated to…
- Regardless of whether they dropped out, took the
medicines, started on another drug etc. - Assesses effectiveness – as mirrors REAL LIFE as similar problems may occur.
what is risk?
- risk= Probability of s/t happening
- Harm= Outcome of that risk.
- risk = number of events of interest / total number of observations
- risk = Number of people who had the outcome/ Total number of people in that specific group
- example
24 people climbing trees, and 6 fall - risk of a fall
= 6 falls/24 who could have fallen
= 6/24 = ¼ = 0.25 = 25% have fallen off out of the potential 100% who could have fallen.
What is odds?
- odds = number of events of interest/number without the event
- Odds= Number of people who had the outcome/ Number of people who did not have the outcome
- Add the numerator and denominator together you get the total population.
- example
24 people climbing trees, and 6 fall
-odds of a fall
= 6 falls/18 did not fall
= 6/18 = 1/3 = 0.33 (not usually as %)
Risk ratio
- Risk ratio=Relative risk
- Risk on treatment/risk on control
- Where risk ratio = 1, this implies no difference in
effect
Expressing risk & odds ratio
RISK
- What was the risk in %
- How much was the risk reduced by %
ODDS
- How much was the odds reduced by %
- comparison of control and treatment
Odds ratio
- Odds ratio = odds on treatment/odds on control
- Where odds ratio = 1, this implies no difference in effect.
Absolute Risk difference
- risk on treatment – risk on control
example for Blum et al
119/164 – 130/164 = 0.726 – 0.793
= -0.067
usually expressed as a %, -6.7%
If - = reduced
If + = increased - treatment reduced the risk of being dyspeptic by about 7
percentage points - Where risk difference = 0, this implies no difference in effect
Qualitative/Narrative Review
These are not good ways of reviewing evidence.
- Summary of all data
- More up to date than a textbook(?) tend to be 5-10 years out of date.
- Subjective – reflecting bias of author
Systematic Reviews and Meta-analysis
- Bring together similar RCTs & set a criteria
- Systematic collection of good quality RCTs
- Uses pre-defined explicit methodology (reduces bias) of which databases your going to search.
- Seeks to minimise bias
- May combine statistics (meta-analysis)
- Not merely a review – since criteria for inclusion of RCTs explicit (a scientific process)
Systematic reviews
- Not all have numbers you can meta analyse.
Meta-analyses
- Have to have a predefined question.
Pre-defined Question (essential)
- “Does aspirin increase the chance of survival to 6 months after an acute stroke?= Tells us what the drug is, the condition & the outcome of the condition
- The questions should be specific
extra info
- can’t get an average by adding the studies together as all the studies are diff.
- Effect measure = Risk ratio & Relative risk
Forest plots
- Line of no effect in the centre
- If data arms do not touch the line of no effect = statistically significant ( could show that the drug works).
- Black box = Effect size for the study.
- The bigger the box the more patients in the study.
- Arms either side = confidence intervals = can be 95% confident that the true effect size of the study is w/i those arms.
- On right side = drug works
- on left side =drug doesn’t work
- Small studies = give wider confidence intervals ( To reduce uncertainty about the results w/ the effect size you need lots of patients.
- Diamond = Combination of all studies = Summary effect measure.
- Meta- analysis puts all of the studies together & creates a summary effect measure.
When can we do a meta-analysis?
-When more than one study has estimated an effect ( Can’t do meta-analysis w/o having an effect size).
- When there are no differences in the study characteristics that
are likely to substantially affect outcome (Studies have to be similar)
- When the outcome has been measured in similar ways
- When the data are available (take care with interpretation when only some data are available), S/studies are not published & there is publication bias and bias in academics.
Disease Oriented Outcomes (DOOs)
- physiological or surrogate outcomes e.g. changes in blood pressure, blood lipid levels, etc.
- Measure things in your body that may be correlated to other health aspects but are not actually the thing that you are concerned about.
- Not as good as POOs.
Patient Oriented Outcomes (POOs)
- outcomes that matter directly to patients: e.g. morbidity, mortality, symptom improvement, cost reduction and/or quality of life.
- Measures things that matter to patients.
