Flashcards in Clinical Studies Deck (33):
P-values & Confidence intervals?
P-value is a measure of statistical significance
Confidence intervals are made-up intervals used to show the likelihood that a certain result will fall into this range
Significance and Key components in clinical trial design
Phase 1 - test safety of treatment, small number of volunteers
Phase 2 - See if treatment works and safety; a few hundred people (efficacy)
Phase 3 - compare new with old and see which is better, see how well it works; several thousands of patients (effectiveness)
Phase 4 - once released into circulation, look for side effects
Identify potential biases and limitations
Allocation biases - Bias in which group someone is appointed; sorted by randomisation
Measurement biases - Bias in measurement, sorted by blinding
reporting bias - Positive traits published, Negative and neutral traits not
Sample size - must be large to properly represent the population
How do you interpret findings from clinical trials?
Outcomes presented in terms of efficacy or effectiveness
Outputs of trial:
Experimental event rate(EER): incidence in cases
Control even rate(CER): Incidence in control cases
Relative risk: EER/CER
Relative reduction: (CER-EER)/CER
Absolute risk reduction(ARR): CER - EER
Number needed to treat(NNT): 1/ARR
Tell me strengths of case control studies
Good for rare diseases
Investigate lots of exposures at the same time
Good for long-lasting diseases
Tell me weaknesses of case-control studies
Possible bias in selection
Recall bias during investigation
Unsure of temporal relationship
Poor for rare exposure
Cannot calculate incidence
Tell me about case-control studies
It is retrospective and observational
Uses odds ratio to calculate relative risk
Only occurs post diagnosis
Tell me about cohort studies
Starts with a group of exposed or unexposed and follows up their lives
Uses Risk Ratio to calculate it
Incidence in exposed/Incidence in unexposed
Advantages of cohort studies
Able to see multiple outcomes
Follow through natural life of disease
Able to look at rare exposures
Incidence can be calculated
Minimise bias in estimating exposure if prospective
Disadvantages of cohort studies
Inefficient for rare disease
Loss to follow-up may introduce bias
Healthy worker effect may cause bias in occupational population
What is the healthy worker effect?
The idea that those who are healthy enough to work are not representative of the whole population
Major sources of health data in the UK
Advantages of routine data
Standardised collection procedure
Wide range of recorded data
Available for past years
Experience in use and interpretation
Disadvantages of routine data
May not answer the question
Validity may vary
Disease labelling may vary over time or by area
Coding changes may create artefactual increases or decreases in rate
Need careful interpretation
Purpose of systematic reviews
It is a review of all sources using systematic and explicit methods to: Identify, Select, Critically appraise relevant research, collect data from studies, analyse data from the studies that are included in the review.
E.g. Debate on statins
What is involved in a systematic review
Efficient searching of data
Applying formal rules for critical appraisal of the sources
Stages of a systematic review
Stages of a systematic review
Stage 1: Plan the review
Stage 2a: Identification of research
Stage 2b: Selection of studies
Stage 2c: Quality assessment
Stage 3a: Data analysis
Stage 3b: Data visualisation
Stage 3c: Reporting and dissemination
What's happening at each stage?
Stage 1: Planning, definition of study by PICOS (Population, interventions/comparison, outcomes, study design)
Stage 2a: Defined search criteria and a thorough research of all publicised material
Stage 2b: Eligibility and inclusion criteria based on: Study design, year of study, Publication language, Sample-size/precision, Specific exposure/intervention, Specific outcome, Completeness of information
Stage 2c: may be assessed according to recognized or user-defined criteria
Stage 3: Study details need to be abstracted from each eligible study along with the effect estimate
What is a forest plot?
Forest plot is a graphical representation of the results from each study included in the meta-analysis, combined with the meta-analysis result
Each study is shown by a box with whiskers showing confidence intervals
Overall estimate shown by diamond at bottom of forest plot, centre of diamond shows pooled point estimate, width is 95% confidence intervals
What is a meta-analysis?
The use of statistical techniques in a systematic review
to integrate the results of included studies
Limitations of systematic reviews and meta-analyses?
1. Publication Bias
2. Inconsistency of results (heterogeneity)
3. Low study quality
Only the subset of the relevant data is available
Non-significant findings are less likely to be mentioned
Published studies not representative of all valid studies
How is it measured?
Publication bias measured by a funnel plot, each point meets one study and is plotted on odds ratio and sample size
Studies differ with respect to:
Unknown study characteristics
Measurements of heterogeneity
Tau2: estimate of between-study variance based on random-effect model
Chi2: test of statistical significance for heterogeneity (low power to detect existing inconsistency )
I2: measure or index of heterogeneity
Low study quality
Studies are of a low quality
Advantages of systematic review and meta-analysis
Generates a pooled overall estimate
Produce a more reliable and precise estimate of effect
Explore differences between published studies
Identify publication bias
Critically appraising a systematic review
Was the question addressed?
Was a comprehensive search for relevant literature carried out
Was the quality of each study assessed properly
Was heterogeneity explored?
How credible is the evidence?
Check guidelines for reporting
Critically appraising a meta-analysis
Was heterogeneity explored?
Was publication bias an issue?
Was it appropriate to pool the studies?
Was the appropriate model used to pool effect estimates? (fixed vs effect model)
Did different sub groups of study give similar results?
Define: Incidence, Prevalence, Mortality and Morbidity
Incidence: Number of new cases of a disease in a specific time interval (new cases/original population)
Prevalence: Frequency of a disease in a population at a given point in time (no. cases/population)
Mortality: Number of deaths associated with a condition in a given time period (Deaths/population)
Morbidity: Number of cases of ill health, complications, side effects attributed to a specific condition over a particular time period
Fixed effects model?
Single true underlying effect
Used when effect of exposure is not heterogeneous