Clinical Epidemiology Flashcards
Clinical Epidemiology
The study of the patterns, causes and effects of health and disease in patient populations and the relationships between exposures or treatments and health outcomes.
(bioscience= animals models, genes/cells.
clinical sciences= patients
epidemiology= populations
health services research= heath care systems)
What does clinical epidemiology do?
Seeks to answer clinical questions and to guide clinical decision making.
Uses epidemiological approaches applied to groups of individuals seen by health care providers.
Provides the foundation for evidence-based medicine.
What are its clinical applications?
- research question
-study population
-exposure
-outcome
-bias/confounding
-risk estimation
-impact
Types of research questions
Risk:
What factors are associated with increased risk of disease?
Screening/prevention:
What interventions prevent people getting a disease?
Will early detection change the course of the disease?
Prognosis:
What is the natural history of the disease?
What factors determine disease outcomes?
Treatment:
Which treatment is more effective?
What are the unintended consequences of treatment?
How are study population groups defined?
Specific illness or disease
Symptoms/ disease characteristics
At risk population
Diagnostic procedure
Treatment
Clinical epi exposures
interventions (i.e prevention, screening)
patient characteristics (i.e age, ethnicity, clinical features)
treatments or procedures (i.e type, intensity, dose)
Outcome in clinical epi
mortality (i.e survival)
disease progression (i.e recurrence)
morbidity/ complications
Outcomes are usually more frequent in clinical epidemiology compared with population studies
Data sources in clinical epi
Patient registries
Electronic health records
Population health surveys
Routinely collected administrative data
Primary patient-level data collection (prospective or retrospective)
Data collection can be:
study specific (protocol driven)
routine (practice driven i.e hospital records)
combination (i.e clinical registry or patient cohort)
Bias in clinical studies
Selection bias
- study population differs from broader population who would receive intervention in terms of relationship between exposure and outcome (i.e self-selection of participants)
Information bias (measurement error, mis-classification)
- methods of measurement of outcome differ between groups (closer monitoring of one treatment group)
Confounding:
- factors associated with both exposure and outcome may distort the main effect if not taken into account.
Methods to control for bias
Randomisation: selection bias/confounding
Restricted sample: selection bias/confounding
Blinding: measurement bias
Matching: selection bias/ confounding
Stratification: selection bias/ confounding
Multivariate analysis: confounding
Sensitivity analyses: all potential biases
Validity of clinical epi study findings
Internal validity
- Are the findings of the study correct?
- Depends on study design and appropriate analysis
External validity
- Are the findings generalisable to the broader population?
- Depends on study sample & setting
Types of clinical epi study designs:
Observational (non-randomised) studies
- Case-control studies
- Cohort studies
- Ecological studies (aggregated population data)
- Cross-sectional
Randomised studies
- RCTs –> Gold standard (but not always feasible)
Meta-analyses
- Combining results of multiple similar studies
Levels of clinical epi evidence
Meta-analysis RCTs
- All evidence from trials
- Combines data from multiple trials
Level: high
Randomised controlled trials (RCTs)
- Least biased: comparison groups should be equivalent (in all other ways) die to randomised assignment
Level: high
Meta-analyses of observational studies
- Combines data from multiple studies of same type
- Increases statistical power
Level: medium
Observational studies
(Cohort & Case-control)
- Data at the individual level on exposures and outcomes
- Biased if comparison groups differ in underlying risks that affect outcome
Level: medium
Observational studies (Ecological/trend analyses/cross-sectional)
- Data are at an aggregate level so findings may not translate to individuals
- Indicative rather than definitive
Level: low
RCTS
Patients are randomly assigned to one of two or more groups being offered different therapeutic measures
Chance alone indicated whether a particular patient will be assigned to a particular group
Patient outcomes in each group are monitored
- The occurrence of event(s) that the intervention seeks to prevent, and/or
-The occurrence of side effects/adverse effects
Non-randomised (observational studies)
Treatment not assigned as part of study protocol but rather occurs as part of clinical practice
Observe outcomes (prospective or retrospective) following treatment
Observed alternative therapies may be the result of underlying differences between the groups that affect the chance of progression or complications (Bias)
Cohort study
May be retrospective or prospective
Participants selected based on exposure status (exposed and non-exposed)
Outcome does not occur prior to being part of the cohort
Followed up over time (longitudinal)
Time consuming and costly
Not efficient for rare outcomes
Case-control study
Study participants selected based on outcome
Some have outcome of interest (cases), some do not have outcome of interest (controls)
Exposure status is unknown at selection
Good for studying rare outcomes
Cross-sectional study
Outcome and exposure(s) are measured at the same time
Participants are selected based on the inclusion and exclusion criteria set for the study
After the entry into the study, the participants are measured for outcome and exposure at the same time
Relatively quick and inexpensive
Can give us information about the prevalence of outcomes or exposures
Only one time measurement so must be careful when looking at causal associations
see slide 21 for comparison of observational studies
Randomised trials VS observational studies
Advantages of RCTs:
Ensures that intervention is unrelated to the outcome
Comparison groups similar in all aspects other than the treatment/intervention
Can control for confounding by unknown factors
Disadvantages of RCTs:
Not always feasible/ethical to randomise
Often not timely
Not always generalisable due to restrictive eligibility criteria
Not designed to determine long term (rare) adverse effects
Advantages of observational studies
Feasibility: cheaper, convenient, timely
Better suited for rare outcomes and under-represented subgroups
May better reflect real world practice (effectiveness vs efficacy)
Disadvantages of observational studies
High risk of bias – treatment selection is based don clinical indications
Can only control for known measurable confounding factors
Systematic reviews and meta-analysis
Provide the “highest level of evidence” for various clinical questions
Collates information from all available published data on a particular research topic/question
Meta-analysis – “a quantitative summary of the results if the results are judged sufficiently similar or consistent to support such synthesis” (Porta 2014)
Helps us to understand the quality of the articles in literature or the type of studies that have been conducted and published
Example 1: studies of the natural history of illness
Measure health outcomes in ill persons who are not receiving a therapy that influences the presence or rate of these outcomes
Data from cohort studies are the major source of information on the natural history of a condition
Example 2: Studies of disease prognosis
Aim to predict a person’s risk of disease outcomes based on their clinical and non-clinical characteristics
Used to guide clinical decision making (i.e., suitability for treatment)
Cohort studies commonly used to develop prognostic tools
Usually assess multiple variables
Time to event analysis (i.e., death, disease progression)
- Kaplan-Meier survival analysis, Cox proportional hazards regression
Example 3: Studies in disease prevention
Test an intervention that aims reduces the risk or severity of disease:
- Primary prevention – reduces disease occurrence
- Secondary prevention- earlier detection or treatment
- Tertiary prevention – reducing complications
Primary/secondary prevention offered to ‘population’
Need to evaluate: Effectiveness, safety, cost-effectiveness
Variety of study designs
Example 4: Hormone Replacement Therapy for prevention of Coronary Heart Disease
Early observational studies showed reduced risk of CHD
Findings may be due to selection bias?
- Healthier women may be more likely to use HRT, less likely to develop CHD
RCTs to investigate potential of HRT chemo-prevention
RCTs no reduced risk, increased risk of CHD in 1st year
Why the contradictory results?
Debate about inferiority of observational methods
Results:
Differences mostly about timing of HRT use
How long after menopause women started HRT
Effect modification
Observational studies
- Compared current vs never users
- Women closer to menopausal age
RCT participants
- No prior HRT
- Older at enrolment (~63yr)
Reanalysis of data comparable groups
- Similar results
- No overall benefit of HRT for CHD risk
- Difference due to effect of timing or HRT
Example 5: Studies of diagnostic and screening tests
Tests performed in persons with a symptom or sign of an illness are termed diagnostic
Tests done in those with no signs or symptoms are referred to as screening
Variety of study designs can be applied
Mammography screening today
Screening technology is trial era was inferior to current day
Generalisability questionable outside of trial conditions
Relevance of mammography screening today given considerable advances/changes in BC treatment
There is therefore a potential role for observational studies in evaluating mammography
Advantages:
- Contemporary screening practice
- Real world setting
