Evidence based dentistry Flashcards
risk
What are the chances…?
Good (positive) or bad (negative)
- …having a heart attack
- …surviving prostate cancer
outcome
“something” that might happen, what you are measuring
- Death, heart attack, cancer diagnosis (bad)
- Tooth decay/ periodontitis/ TMJD (bad)
- Caries free (good)
what are statistics?
Numbers that summarize information
- Based on observations of large numbers people
- Useful in predicting what is likely to happen in the future
what are risk statistics?
The chance that an outcome will happen – risk of an outcome
- Fractions or differences between 2 numbers
numerator in risk statistics
top number in fraction)
= number of people who actually experience the outcome
denominator in risk statistics
(bottom number in fraction)
= the number of people who could potentially experience the outcome
risks and odds in binary events
express the chance of being in one of the two states.
binary events
one or another outcomes
risk =
number of events of interest /
total number of observations
odds =
number of events of interest /
number without the event
ways of expressing risk
can be asked in exam, numbers and words
- The chances of falling were one in four,
- 25%.
ways of expressing odds
can be asked in exam, numbers and words
- The chances of falling were one third of the chances of not falling;
- One person fell for every three that didn’t fall;
- The chances of falling were 3 to 1 against.
questions to ask when interpreting risk (4)
Risk of what?
- What is the outcome?
(Getting a disease? Dying from a disease? Developing a symptom? Surviving a disease?)
How big is the risk?
- What are the chances of experiencing the outcome?
(Out of how many? Need baseline/total)
- What is the timeframe? (Next year? Next 10 years? Lifetime?)
e.g. Out of 1000 people in the last 10 years
Does the risk information reasonably apply to me or my patient?
- Age/sex/lifestyle similar enough to apply results, need to dissect paper
How does this risk compare with other risks?
- Perspective- which risk should I do something about? Personal decision
questions to ask when you see messages about risk reduction (5)
Reduced risk of what?
- What outcome?
- How much do you care about it?
How big is the risk reduction?
- What are my chances if I don’t get treatment?
- Starting and modified risks
Does the risk reduction information reasonably apply to me?
- Is the study based on people like you (or your patients)?
Any downsides?
- Life threatening side effects? (Mental health. Need to think of other side for patient)
- Time/cost/hassle
Is the benefit (risk reduction) worth the downsides?
starting and modified risks in drug studies
are the chances of the outcome in the untreated and treated groups
those who did not take the drug and those that did
2 by 2 - contingency tables
Putting numbers into table
Rows – groups
- Both groups got a placebo pill
- group 1 had normal brushing
- group 2 had perio treatment
Modified risk – is given treatment, what we are interested in (group 2)
need total for each group
issue with relative risk reductions
Makes even small risk reductions sound big
- We then overestimate the benefit
- Relative Risks that seem large might not mean much if starting risk is small
absolute risk reduction is more representative
FRAMING
i. e. 4.9% down to 0.8% = 84% reduction in risk of premature birth
4. 9% - 0.8% = 4.1% is the absolute risk reduction (difference)
(absolute) Risk difference
Difference in risk between groups
good outcome measure
- Risk in paracetamol group= 40/63= 63%
- Risk in placebo group=5/27= 18%
- Risk Difference = 63% - 18% = 45% more patients experienced pain relief in the paracetamol group
(absolute) Risk difference when there is no difference in modified and control group
0
- No benefit between the 2 conditions (take away from each other – no difference)
- cannot straddle as unsatisfied of evidence
Number Needed to Treat (NNT)
The number of patients you would need to treat to prevent one patient from developing the disease/ condition/ outcome
- Numerically: 1/ Absolute Risk Difference
e. g. NNT for paracetamol = 1/ 0.45 = 2.22 (round up, measured in people)
- Would need to treat 3 people with paracetamol post-operatively to have one person experience pain relief of >50% in 4 hours
risk ratio/relative risk
How many times more likely is a participant in the paracetamol group to experience pain relief than a participant in the placebo group?
The risk of 50% pain relief after 4 h in placebo group = 5/27 = 18%
The risk of 50% pain relief after 4h in paracetamol group = 40/63 = 63%
Risk in paracetamol group / Risk in placebo group
- 63/ 18 = 3.42
A participant in the paracetamol group is 3.42 times more likely to experience pain relief than a participant in the placebo group.
risk ratio/relative risk value is the risks in both groups were equal (no benefit)
1
- value of no difference (1:1
- cannot straddle as unsatisfied of evidence
odds ratio
Ratio of odds of pain relief in both groups (usually intervention/control)
Odds of pain relief in paracetamol group = 40 (pain relief)/23 = 1.74
Odds of pain relief in placebo group = 5 (pain relief)/ 22 = 0.23
Odds ratio = 1.74/0.23 = 7.56
odds ratio value were there would be no benefit between the 2 groups
1
- value of no difference (1:1)
- cannot straddle as unsatisfied of evidence
difference between odds and risk ratio
OR only approximates RR when the outcome is rare
- OR tends to overestimate
- the RR as in this case
confidence intervals
Confidence intervals quantify the level of uncertainty for the population of interest
Sampling introduces uncertainty, don’t know how will affect rest of population
confidence intervals for risk ratio
A confidence interval tells us the range of values that a true population treatment effect (e.g. RR) is likely to lie (subject to a number of assumptions)
what value can a CI not straddle for risk ratio
1
“value of no difference” between treatments indicates that
- there is insufficient evidence for a difference between the treatment and control group in the population
things to query once read the published study
- Much research is based on weak science
- Many results are disseminated too early – trial not fully completed
- Is there any science behind the numbers?
- How good is the science?
- Some research only makes a weak case for the message – be sceptic
- Some research makes a strong case
Treatments shown to fight diseases in test tubes won’t necessarily translate to humans (bench science in petri dishes, isn’t just a building block to clinical trial which will be benefit to patient)
Treatments that work in animals often don’t have the same results in people – Dolly the Sheep – positive benefit in animals may not translate to humans
be sceptical about treatments that have only been proven in animal or lab studies
observational uncontrolled studies
Researchers watch what happens to a group of people
- A group of patients has disease A and is treated with drug X
- No control group
The researchers observe how many get better
- No intervention
- Has the treatment caused them to survive or something else attributed?
controlled studies
Cohort or case control
- Treatment better than control group – shows treatment may be beneficial
Researchers observe what happens to people in different situations-without intervening
randomised controlled trials
patients randomly split into 2 groups-one gets intervention the other placebo (TAU)
splitting into 2 groups randomly – equal chance of being in one group or the other
Any differences at follow-up caused by intervention
evidence levels
- systematic reviews and meta analysis
- randomised controlled trials
- cohort studies
- case-control studies
- cross-sectional surveys
- ecological studies
- case series and case reports
- ideas editorials and opinions
higher: experimental interventions
lower: observational, non-interventional
case report/ case series
A report on a single patient or series of patients with an outcome of interest
- No control group is involved
case report/case series disadvantages
- Cannot demonstrate valid statistical associations
- Lack of control group
cross sectional study
The observation of a defined population at a single point in time (or time interval) snapshot
Exposure and outcome are determined at the same time
- Can see relationship which may be incorrect as collected at one point e.g. asthma and smoking (cannot determine what came first)
cross sectional study used for
- Estimating prevalence of a disease e.g. caries in Scottish schools
- Investigate potential risk factors – can’t say actual link as cannot measure prospectively
case report/ case series used for
- Identify new disease outcome e.g. when e-cigarettes issue arise, drug for morning sickness in 70s
- Hypothesis generation a hunch/feeling of a link, lead onto a trial
cross sectional study disadvantages
- Causality
- Confounding – measure one thing but other factors contribute e.g. healthy diet, less likely to smoke, more exercise etc
- Recall bias - inaccurate
case-control study
The study of people with a disease and a suitable control group of people without the disease
- Looks back in time at exposure to a particular risk factor in both groups
- retrospective is most common, look back at previous exposures
See if a higher prevalence between case and control group of risk factor
case-control study used for
looking at potential causes of disease
case-control study disadvantages
- Confounding
- Recall /selection bias
- Selection of controls differences between clinicians decisions
= Time relationships (did exposure occur before disease? Didn’t can’t have caused it)
See if a higher prevalence between case and control group of risk factor
- Well conducted is excellent but hard to carry out
Rare disease is hard to recruit (need a huge population to identify cases)
- Recruit and look back retrospectively
cohort study
Establish a group of individuals in population
- Measure exposures (through questionnaires, examinations etc)
- Follow up over a period of time
Identify those that develop disease (outcome of interest) –not know at start what you are going to capture e.g. 1990s baby cohort captured obesity crisis
Measure as exposure happening and measure at event so not relying on recall bias – good but time and labour expensive
cohort study used for
Estimating incidence of disease
– number of new cases of disease over a period of time that has developed
Investigating causes of disease
– when disease and exposure occurred is recorded
Determining prognosis
Timing and direction of events
– understand when things may have caused disease
cohort study disadvantages
- Controls difficult to identify
- Confounding
- Blinding difficult
- For rare diseases- need large samples
- Very expensive and time consuming
randomised controlled trials
Sometimes referred to as a Clinical Trial
RCTs considered the gold standard study design
- For effectiveness and efficacy
- Provides strongest evidence on effectiveness of treatments
Particularly useful for clinical studies – need to see for grounds on changing practice
- Measure baseline of groups before trial
Small groups as can’t afford large groups and wouldn’t be ethical
4 design elements of randomised controlled trials
Specification of participants (inclusion/ exclusion criteria)
Control/ Comparison groups
– what will be difference between groups
Randomisation
– allows to separate groups fairly and randomly that can be compared – difference are random, equality across groups so only treatment criteria
Blinding/ Masking
– administrator, patient and analyser don’t know the different groups
inclusion/exclusion criteria
Age
– very young or old have slightly different physiological factors so may not be appropriate to include
Disease severity/ diagnosis
Unambiguous – 100% clear in criteria
- Exact definitions
How you will know if the results of the trial are relevant to your patients
comparison/ control group
Existence of a comparison group
- Placebo
- Current standard treatment (TAU)
Why?
- People often get better on their own
Need control to compare – what would happen without treatment
randomisation
How participants/ subjects/ patients are allocated to treatments.
- To minimise Bias
Assign participant to one of the two groups
- Researcher nor participant has influence over allocation
Randomisation should be performed by computer, not up to people as easily manipulated (minimise bias and influence on results)
Only systematic difference between the arms of a trial are the treatments themselves
why would the researcher not carry out and allocate randomised groups?
because selectively choose to make drug look favourable
why would the patient not allocate their randomised group?
because want treatment
Random allocation of participant to treatment ensures
Each individual entered into trial has equal chance of being allocated to any treatment arm (paracetamol / placebo)
allocation concealment (selection bias)
A technique used to prevent selection bias by concealing the allocation sequence from those assigning participants to intervention groups, until the moment of assignment.
Allocation concealment prevents researchers from (unconsciously or otherwise) influencing which participants are assigned to a given intervention group.
what does allocation concealment in RCT prevent?
researchers from (unconsciously or otherwise) influencing which participants are assigned to a given intervention group.
blinding/MASKING
If the participant or researcher is aware of which treatment is given this may influence the results
- look, taste, smell, action be the same – no difference
needs to be blind to:
- Participant
- administrator of treatment
- Assessor of outcome
- Data analyst
advantages of RCTs
Provide strongest and most direct epidemiologic evidence for causality
4 disadvantages of RCTs
More difficult to design and conduct than observational studies
- ethical issues
- feasibility
- costs
Still some risk of bias and generalisibility often limited
Not suitable for all research questions – pregnant women, elderly generally excluded
Non-blinded RCTs may overestimate treatment effects e.g. estimates of effect from trials with inadequately concealed allocation have been 40% larger than clinical trials with adequately concealed random allocation
summary of cohort study
Participants are recruited to a study and followed up over time
Exposures and disease are measured prospectively
- No intervention – don’t manipulate end point
- Start with anyone and follow
summary of RCT
Participants are allocated by chance to different interventions and followed up and outcomes assessed
summary of case-series report (case study report)
Description of the medical history of one or more patients
summary of cross-sectional study
Observational study that analyses data collected from a population, or a representative subset, at a specific point in time
- Survey
- No longitudinal aspect
- No intervention/trial
summary of systematic review and/or meta analysis
All the evidence for RCTs looking at effectiveness of a particular treatment are synthesised
summary of case control study
People with a disease (cases) are matched to those without it (control) and earlier exposure to different factors are compared
- Start with outcomes – look back at exposures (get association)
- Quicker as already have outcomes
- Can’t assess prevalence in population
how to decide on which study design to use?
variety of study designs may be appropriate for a particular clinical question,
need to use the Evidence Levels pyramid / table
and question the feasibility of conducting such a study to determine what is most appropriate.
4 key study design features of RCTs
inclusion/exclusion criteria
comparison/control
randomisation
blinding
