EPID 410 Exam 2
Review of cumulative incidence
- A measure of risk - All individuals in denominator are followed for the entire time period - Numerator is new cases of a condition
Confidence interval
- A range of values surrounding the point estimate that is likely to contain the true value in the parent population - The point estimate is always included within the bounds of CI - The level of confidence is chosen by the investigator (95% is most common) - Indicates how reliable study estimates are
Significance testing
- A test of credibility of the null hypothesis - Attempts to evaluate chance as a possible explanation of study findings (ex: assuming that the null hypothesis is true, how unusual are the observed data) - Can use: confidence interval, P-value. These are both closely related
Descriptive studies
- Describe a disease or characteristic by person, place, and time - Not used to determine causation
Types of analytic studies
- Experimental: Investigator manipulates the exposure, Gold standard of study designs - Observational: Investigator does not manipulate the exposure, observes what is occurring naturally
Analytic studies can be
- Experimental: Randomized control trial - Observational: Cohort, Case-control, Cross-sectional, and Ecologic
Factors that influence choice of study design
- Features of exposure and outcome - Current state of knowledge - Logistical considerations (ex. time, money, and other resources)
Ways to measure exposure and outcome
- Present/absent - Categorical - Ordinal - Continous
Measure of association
- Quantifies the relationship between an exposure and an outcome - Measures the strength and the direction of the relationship
Bias
- Systemic, nonrandom error that results in an inaccurate estimation of the effect of an exposure on an outcome - A lack of internal validity - Leads to the wrong conclusion
Analytic studies
- Test hypotheses about exposure-outcome relationships - Use a comparison group to determine exposures that are more common among people with a specific outcome than those without the outcome - Quantify the associations between exposures and outcomes (measure of association)
Point estimate
- The numerical result obtained when calculating a measure of association for a study - Single best estimate of the effect that can be determined from a study sample - The farther away from the null value, the larger the estimated effect
Hypothesis
- You can never prove a hypothesis - Either build evidence in support of the hypothesis or find evidence with which it is inconsistent, thereby eliminating the hypothesis
What is a cohort?
A well-defined group of individuals who share a common characteristic -General characteristic -Experience a common exposure associated with a specific setting
What other type of epidemiologic study could be susceptible to recall bias?
A.Retrospective cohort study
Directionality
Describes how participants are grouped at the start of the study. - Forward: participants are grouped by exposure status. (E --> O, which is going from the risk factor to the outcome) - Backward: participants are grouped by outcome status (O --> E) - Non-directional: Exposure and outcome observed simultaneously
Timing
Describes whether the outcome has occurred before or after the study begins
Analytic studies
Determines what causes a disease or condition.
Informed consent
Federal law requires participants to be notified of: - Participation in research - Procedures - Risks and discomfort - Potential benefits - Can withdrawal at any time without penalty - Possible advantageous alternative procedures (for randomized control trials)
When analyzing data from an RCT, participants who are non-compliant should be removed from the analysis in order to prevent bias. A.True B.False
Flase
The direction of a random control trial is
Forward. From exposure to outcome.
Case-control study null hypothesis
H0: The proportion with the exposure among those with the outcome equals the proportion with the exposure among those without the outcome.
Cohort study null hypothesis
H0: The risk of developing the outcome is the same among the exposed and the unexposed.
Risk ratio vs. attributable risk (cohort studies)
Risk ratio •Measures the strength of the association between an exposure and outcome • •Important in establishing etiologic relationships -What causes disease? Attributable risk •Measures how much of the disease risk is attributable to a specific exposure • •Important in clinical practice and public health -How much disease can be prevented if exposure is eliminated? If you are asked to calculate the measure of association for a cohort study in our class, calculate a risk ratio.
Measure of association in cross-sectional studies: Prevalence ratio (PR)
Two approaches: •Prevalence of outcome* -Compare prevalence of the outcome among those exposed to prevalence of the outcome among those unexposed •Prevalence of exposure -Compare prevalence of exposure among those with the outcome to prevalence of exposure among those without the outcome
Definition of an RCT (random control trial)
•An epidemiologic experiment in which participants are randomly assigned into 2 or more groups to receive or not to receive an intervention and then followed to see if they develop a specified outcome • •Exposure/intervention is determined prior to the development of the outcome - Temporality • •Gold standard of study designs
Intention to treat rule
•Analyze what you randomize -During analysis, all participants who were randomized are categorized according to their treatment assignment, regardless of whether they complied with treatment or were lost to follow-up - •Preserves integrity of randomization and prevents bias
Advantages of cross-sectional studies
•Can characterize the prevalence of a number of exposures and outcomes in a population • •Can be carried out quickly • •Useful for generating hypotheses about exposure-outcome relationships
Disadvantages of case-control studies
•Cannot calculate risk directly •Not efficient for the study of rare exposures •Susceptible to selection bias •Potential for recall bias •Temporality cannot always be established •Can only study one outcome
Disadvantages of cross-sectional studies
•Cannot determine temporality (whether the exposure preceded the outcome) • •Prevalent cases may not be representative of all cases • •Not useful for rare exposures, rare outcomes, or for diseases with short durations • •Cannot calculate risk
Measures of association in ecologic studies
•Cannot fill the cells of a 2x2 table from the data available •Summary measures are usually correlations between population averages
Review
•Case series - describes a small group of patients with a similar disease •Ecologic studies - unit of observation is group • •Cross-sectional studies - exposure and outcome measured at the same time -Prevalence ratio
Hospital-based case-control studies
•Cases admitted for outcome of interest •Controls admitted for another condition •Advantages: -Convenient way to identify participants -Participants are easily accessible -Often get lower refusal rates •Limitations: -Prevalence of exposure in controls may not be indicative of that in the source population that produced the cases -Illnesses of cases and controls may have etiologic similarities
Advantages of ecologic studies
•Cheap and easy to carry out • •Data often already available • •May suggest new hypotheses
Features of an RCT
•Closest approximation to a lab experiment •Investigator randomly assigns participants to intervention groups •Can provide strongest evidence of a relationship between an exposure and an outcome -Temporality -Fewer biases
Risk ratio (relative risk) (cohort studies)
•Compares the risk of the outcome in exposed participants to the risk of the outcome in the unexposed participants •Can be calculated directly in cohort studies
Selection of cases: Practical scenario
•Convenience sample -e.g., a hospital (source population = ?) - • Use prevalent cases -If an exposure-outcome relationship is found, is the exposure associated with development of disease or with survival or duration of outcome?
Objectives (of case-control studies)
•Describe the design of a case-control study •Calculate an odds ratio •Interpret results of a case-control study •List strengths and limitations of case-control studies •Compare and contrast the different epidemiologic study designs
Case report/case series
•Describe the experience of a single patient or a small group of patients with the same diagnosis • •Not actually a study -No comparison group -No measure of association
Case report/Case series
•Describe the experience of a single patient or a small group of patients with the same, often unusual, diagnosis •If a clinician identifies an unusual feature of disease, it may lead to new hypotheses
Study designs and measure of association Objectives
•Differentiate among the epidemiologic study designs •Calculate and interpret appropriate measures of association for study designs •Determine statistical significance given 95% CIs and p-values
Randomization
•Distinguishes randomized controlled trials from other types of studies • •Assigns participants to treatment and control groups by chance -Does not depend on patient's, physician's, or researcher's preference -Coin toss, random number generator, etc.
Selection of controls
•Do not have the outcome •Ideally chosen from the source population that produced the cases -e.g., community, hospitalized patients •Exposure criteria should not be used to select controls Selecting appropriate controls is one of the most challenging problems in epidemiology.
Advantages of case-control studies
•Efficient for the study of rare outcomes •Useful for the study of chronic diseases •Tend to require a smaller sample size than cohort studies •Relatively inexpensive •May be completed fairly quickly •Loss to follow-up not an issue
Uses of ecologic studies
•Examine trends in population-level outcomes •Uncover potential relationships between exposures and outcomes that warrant further study -Data can be used to develop hypotheses
Randomized controlled trials summary
•Experimental design -Participants are randomly assigned to intervention groups and then followed to see if the outcome occurs - •Measure of association - risk ratio (relative risk)
Review of RCTs
•Experimental design - participants are randomly assigned to intervention groups and then followed to see if the outcome occurs -Reduces bias -Increases likelihood that intervention groups are comparable - •None of the treatments should be known to be inferior • •Ideally, blind participants and study personnel to treatment assignment • •Analyze according to treatment assignment (intention to treat rule) • •Measure of association - risk ratio (relative risk)
Ascertainment of outcome to prevent detection bias
•Exposed and unexposed should have: -Similar degree of surveillance for outcome -Same standard diagnostic protocol •Ideally, outcome is determined by someone who is not aware of exposure status of participants blinding
Ecologic study
•Exposure and outcome are measured at the group level •Data not available for individuals -No measure of association to calculate in this class
Cross-sectional study
•Exposure and outcome are measured simultaneously -Directionality: non-directional •Captures prevalent cases of disease -Timing: retrospective •Participants not chosen based on exposure or outcome status -Selected because they are members of a certain population subset at a certain time •Measure of association - prevalence ratio
Cross-sectional study
•Exposure and outcome measured simultaneously •Non-directional •No temporality •Measure of association: prevalence ratio
Detection bias (cohort studies)
•If exposed participants are followed more closely than the unexposed, the outcome is more likely to be diagnosed in the exposed group •Can lead to incorrect conclusion about exposure-outcome relationship
Disadvantages of ecologic studies
•Individual risk factors not identified • •Exposure and outcome cannot be linked on an individual basis • •Exposure tends to be crudely assessed
Unexposed participants (cohort studies)
•Internal comparison group -For population-based sampling -Identify a single cohort and divide into exposed and unexposed •External comparison group -For exposure-based sampling -Find unexposed group that is as similar as possible to exposed with respect to all other factors that may be related to disease
Validity in RCTs
•Internal validity is usually strong -Randomization reduces bias •External validity is usually poor -Participants in RCTs do not typically represent the general population
Dose-response (cohort studies)
•Is increasing level of exposure associated with increasing (or decreasing) risk of the outcome? •If exposure is ordinal or continuous, dose-response can be assessed •If dose-response is present, stronger argument that the exposure is associated with the outcome
Disadvantages of RCTs
•Limited generalizability of study results •A long period of time is often required to reach a conclusion •A large number of participants may be required •Financial costs are typically high •Ethical concerns may arise •Participants may not comply with treatment assignments
Blinding (RCT's)
•Masking of the treatment assignment -Single blind trial - patient unaware of which treatment is being received • -Double blind trial - patient and study personnel are both unaware of which treatment is assigned - •Eliminates bias introduced when patients and/or study personnel know which treatment is being received
Odds ratio (OR)
•Measure of association in case-control studies • •Compares the odds of the exposure among cases to the odds of exposure among controls •Cannot calculate risk directly in a case-control study
Potential issues in RCTs
•Non-compliance -Participants do not follow treatment regimen to which they were assigned •Loss to follow-up -Participants move away and cannot be contacted, withdraw from study, etc.
Ethical concerns (of RCT's)
•None of the treatments in a trial should be known to be inferior • •Trial should be monitored and have a protocol in place for when to stop early -Once one treatment is shown to be superior, can no longer justify withholding it from patients
Ecologic study
•Observational study where variables are measured at the group level -Unit of observation is group, not individual §e.g., countries, states, universities, hospitals -Exposure and outcome data is not available for the individuals in the population •Also referred to as ecological, correlational, or aggregate studies
Case-control study
•Participants are chosen based on outcome status -Backward directionality, from O E -Usually retrospective timing -Cases have the outcome, controls do not have the outcome •Assess exposure from the past -Through interviews or record reviews •No data on incidence •Temporality is not always clear
Review: Case-control studies
•Participants are chosen based on outcome status -Exposure is measured from the past •Prone to selection and recall biases •Incidence cannot be calculated •Compare the proportion with exposure among cases to the proportion with exposure among controls •Measure of association: odds ratio
Case-control studies summary
•Participants are chosen based on outcome status; exposure is measured from the past • •Prone to selection and recall biases • •Measure of association: odds ratio
Cohort studies summary
•Participants are classified by exposure status and followed over time for development of outcome -Temporality - •Provides incidence data used to identify possible causes of disease -Compare the incidence of outcome among the exposed vs. the incidence in the unexposed - •Epidemiologic measures -Risk ratio - measure of association -Attributable risk - amount of disease among the exposed group that can be attributed to the specific exposure
Review: Cohort studies
•Participants are grouped by exposure status and followed over time for outcome (forward directionality: E O) -Can be prospective (outcome in the future) or retrospective (outcome in the past) - •Incidence data, temporality • •Risk ratio, attributable risk • •Good for studying rare exposures • •Can be expensive and time-consuming
Randomized controlled trial
•Participants are randomly assigned into 2 or more intervention groups •Follow over time to see if they develop the outcome •Measure of association: risk ratio
Case-control study
•Participants chosen based on outcome status •Exposure assessed from the past •Measure of association: odds ratio
Cohort study
•Participants in a disease-free, at-risk cohort are grouped by exposure status •Follow over time for development of outcome -Can be prospective or retrospective •Measure of association: risk ratio
Loss to follow-up (cohort studies)
•Participants moving, dropping out of the study, or unable to be contacted during follow-up •Leads to missing outcome data •Potential source of bias -If those who are lost to follow-up are not evenly distributed between outcome groups and they differ from those who complete the study
Sampling (cohort studies)
•Population-based -Sample from the general population -When exposure is fairly common •Exposure-based -Sample special population with a common exposure -When exposure is rare •e.g., chemical used in a certain industry, World Trade Center attack, Gulf War exposure -Must identify appropriate unexposed group
Main drawbacks of cross-sectional studies
•Prevalent cases might not be representative of all cases -Tend to identify prevalent cases of long duration •Individuals who die quickly or recover quickly are less likely to be identified • •Unable to determine the temporal relationship between exposure and outcome -Unless the exposure of interest never changes •e.g., race, blood type, genetics
Why randomize?
•Prevent biases of investigators from influencing the treatment assignment of each participant •Increase the likelihood that each treatment group is comparable on all factors except for exposure -Controls for factors, both known and unknown, that could affect the outcome (bias)
Cohort study timing: Can be prospective or retrospective
•Prospective - exposure status is determined at baseline, outcome occurs in the future •Retrospective - exposure status is obtained from historical records or participant recall, outcome has already occurred by the start of the study •Can combine retrospective and prospective designs
Advantages of RCTs
•Randomization makes groups comparable •Results provide strong evidence for causality •Detailed information on participants can be collected at baseline and follow-up •Blinding can reduce bias in assessment of outcomes •Direct clinical implications
Epidemiologic measures of note in cohort studies
•Risk ratio (relative risk, cumulative incidence ratio) •Attributable risk
Potential biases of note in case-control studies
•Selection bias -Error due to the manner in which participants (either cases or controls) are selected for a study •Recall bias -Differential recollection of exposure between cases and controls •Berkson's bias -In hospital-based studies, frequency of exposure might not reflect that of the general population
Cohort study
•Starts with a disease-free, at-risk cohort •Participants are grouped by exposure status -Forward directionality, from E O •Follow cohort over time for development of outcome -Incidence Exposure status is determined PRIOR to observation of outcome status temporality
Uses of cross-sectional studies
•Study trends over time • •Understand characteristics of a population • •Estimate prevalence of an exposure or outcome • •Generate hypotheses about the relationships between exposures and outcomes
Advantages of cohort studies
•Temporal sequence between exposure and outcome is clear •Direct calculation of risk ratio •Can study rare exposures •Can examine multiple outcomes of a single exposure •Can study multiple exposures •Minimizes bias •Strongest observational study design
Disadvantages of cohort studies
•Time-consuming •Costly •Often require large sample size •Not efficient for the study of rare diseases •Loss to follow-up may cause bias •Exposure status may change during the course of the study
Example: Sampling and unexposed participants
•To study the effects of a hazardous chemical spill in a specific factory -Population-based or exposure-based sampling? -Internal or external comparison group? - •To study the effects of HPV vaccination among college students -Population-based or exposure-based sampling? -Internal or external comparison group?
Treatment and control groups (RCT's)
•Treatment group receives the new treatment or intervention •Control group receives either: -Standard treatment that is currently in use -Placebo §A substance having no pharmacological effect but which gives participants the perception that they are receiving treatment
Interventions assessed through RCTs
•Treatments • •Preventive measures • •Screening methods • •Diagnostic procedures • •Behavioral interventions • •Healthcare delivery services
Population-based case-control studies
•Use cases from the general population •Controls sampled from an appropriate source population •Advantage: -Prevalence of exposure among controls is likely to be indicative of that in the population that produced cases •Limitations: -Expensive and time-consuming to locate controls -Tend to have low cooperation levels
Selection of cases: Ideal scenario (case-control study)
•Well-defined source population • •Outcome is a reportable condition or from a registry • •Use only newly diagnosed cases
When to conduct a cohort study
•When the exposure is rare •To learn about multiple outcomes due to a single exposure •To study the outcomes of multiple exposures
When to conduct a case-control study
•When the outcome is rare • •Useful when studying a disease with a long latency period (e.g., many chronic diseases) • •To evaluate the relationships between multiple exposures and a single outcome • •Often done early in the investigation of a suspected exposure-outcome relationship
Case-control studies are good for rare outcomes.
.
Cohort studies are good for rare exposures
.
If You see an Odds Ratio, It is a case-control study
.
If you see Risk Ratio, It is a randomized control trial or cohort study.
.
If you see a prevalence ratio, it is a cross-sectional study.
.
In RANDOMIZED CONTROL TRIALS, THE EXPOSURE IS RANDOMLY ASSIGNED
.
Retrospective design of a case-control study (most common) Incidence cannot be calculated.
.
You CANNOT get incidence rate from case-control studies.
.
Hypothesis
An educated guess about an association that is testable in a scientific investigation
External validity
Are results applicable to other populations?
Random control trials
Are the gold standard of study designs
Inferential statistics
Because epidemiologic studies use only a sample of individuals from a larger parent population, we can only estimate the true value of an association between an exposure and an outcome. - Then generalize the result from the sample to the entire population - must take into account random variation in sampling
Participants (cohort studies)
Both exposed and unexposed must be: -Free of outcome of interest at the start of the study -At risk for the outcome
Descriptive studies can be
Case report/Case series, Ecologic, and Cross-sectional
Internal validity
Do study results reflect the true situation of the study population?
Prevalance
Existing cases/total population
Dose-response
Is increasing or decreasing exposure associated with increasing or decreasing risk of the outcome? - If yes, stronger argument for for causality
Example of alternative hypothesis
MSM who always use condoms have a decreased risk of HIV when compared to those who don't
When to do a randomized controlled trial?
Must have good reason to believe that the new intervention will be as effective or better than the current, standard intervention
Additional concerns (of RCT's)
Must take into account: -Side effects/complications -Compliance -Cost
Careful! Randomly selected vs. randomly assigned
Randomly selected •Choosing a random sample of people to participate in your study •Should be done for all study types Randomly assigned •Assigning exposure status to participants in a random manner •Only applicable to RCTs
Retrospective study
Outcome occurred prior to baseline data collection
Prospective study
Outcome occurs after the baseline data collection
Measures of association in epid 410
Prevalence ratio, risk ratio, and odds ratio
Attributable risk (cohort studies)
The amount of disease incidence among the exposed that can be attributed to that specific exposure How much of the disease could be prevented among the exposed group if the exposure were eliminated
Validity
The degree to which a study reaches a correct conclusion
Ecologic fallacy
The error made when associations observed between an exposure and outcome at the aggregate level are attributed to associations at the individual level -Results cannot be assumed to apply to individuals
Null value
The number corresponding to no association between an exposure and an outcome
If a null hypothesis is true with prevalance ratio, this means
The prevalance among the outcome exposed is equal to the prevalance of the outcome among the unexposed. Answer will be 1.0
Example of null hypothesis
The risk of contracting HIV is the same among MSM who always use condoms and those who don't
Example of alternative hypothesis
The risk of death from lung cancer is different (increased) for smokers vs. non-smokers
Example of null hypothesis
The risk of death from lung cancer is the same for smokers and non-smokers
If the null hypothesis is true with risk ratio, this means
The risk of the outcome among the exposed is equal to the risk of the outcome among the unexposed. Risk ratio will be 1.0
Alternative hypothesis
There is an association between the exposure and the outcome
Null hypothesis
There is no association between the exposure and the outcome