L19 Observational Studies

Question 1

List the type of observational study designs available.

Answer 1

1) Descriptive / Qualitative:
- Ecologic / Correlational Study
- Case Reports / Series

2) Analytical / Quantitative:
- Cross-Sectional Study (Outcome -> Exposure)
- Cohort Study (Exposure -> Outcome)
- Case-Control Study (Exposure & Outcome simultaneously)

Question 2

Describe the design of an ecologic / correlational study.

Answer 2

1) Unit of observation is populations, NOT individuals!
2) Disease rates & exposures are measured in populations & their relation is examined.
- Uses aggregated / group level data, NOT individual level data (e.g. country per-capita etc.)

Question 3

Explain what is an ecologic fallacy.

Answer 3

Used when data collected at a population / group level are analysed & results are assumed to apply to the associations at the individual level.
- Of particular concern when interpreting results from ecologic studies

Question 4

How does an investigator inform the readers of his/her paper of the potential for an ecologic fallacy?

Answer 4

Discussion section of paper:

• Data from this study are aggregated & what may apply at the population basis may NOT necessarily be observed on an individual basis.
• Aggregated data by country does NOT allow consideration or relationships w/in a country!

Question 5

Discuss the strengths & limitations of an ecologic study.

Answer 5

(+) Inexpensive & easy to conduct, usually using secondary data (e.g. published statistics)
(+) Hypothesis generating (e.g. to further investigate if correlation also applies at the individual level)

(-) Inability to link exposure to outcome in individuals (since population-level NOT individual-level data used)
(-) Likelihood of confounding by other variables (e.g. multicolinearity between independent variables)

Question 6

Describe the design of a case report or case series.

Answer 6

1) Most basic type of descriptive study of individuals
2) Careful, detailed report of the profile of a single patient (case report) or a series of patients (case series), with respect to factors that could be related to illness or outcome
- Usually on an unusual disease or association
3) Often the first alert by an observant healthcare provider that something might be going on

Question 7

Discuss the strengths & limitations of a case report or case series.

Answer 7

(+) Hypothesis generating (e.g. to further investigate if correlation also applies at the individual level)

(-) No comparison group to tell if it is by chance or not.

Question 8

How is the assessment of causality of an adverse drug event conducted in an individual patient?

Answer 8

Challenge (Administer) -> Dechallenge (Withdraw) -> Rechallenge (Re-administer)

Rechallenge ONLY if deemed safe!
- If life-threatening, do NOT rechallenge!!

Question 9

Describe the design of a cross-sectional study.

Answer 9

1) Results obtained at a snapshot in time
- i.e. information on presence & absence of exposure & outcome of individuals assessed simultaneously at one point in time

2) Provides information on prevalence
- i.e. proportion who have a specific characteristic at one point in time

Question 10

Discuss the strengths & limitations of a cross-sectional study.

Answer 10

(+) Efficient in terms of time and money
(+) Many outcomes & exposure factors can be assessed at one snapshot in time
(+) No loss to follow-up

(-) Unclear temporal relationship between exposure & outcome

• Difficult to establish causal relationship from data collected in a cross-sectional time-frame
• If prevalence is ONLY reported, it can actually be considered as a purely descriptive study; however if grouped upon analysis, it is considered as analytical study.

Question 11

Describe the design of a cohort study.

Answer 11

1) Follows two or more groups from exposure to outcome.
2) Selection into study on basis of exposure status
3) Can be further classified as prospective or retrospective cohort studies
- Prospective: Outcome has not yet occurred at initiation of study
- Retrospective: Outcome has already occurred at initiation of study

Question 12

How should the selection process of the comparison (unexposed) group be, when conducting a cohort study?

Answer 12

1) Baseline characteristics of exposed & non-exposed groups should be similar as possible wrt all factors, other than the factor under investigation.
- Control for confounders often required via multivariable regression analyses due to use of observational study designs!

2) Need to collect data on any potential baseline differences that could affect the outcome.

3) Selection of comparison groups include:
- Internal comparison (unexposed members of same cohort)
- Comparison cohort (another cohort from a similar population thought to be unexposed)
- General population data (pre-existing data from general population)
- Multiple comparison groups

Question 13

Discuss the strengths & limitations of a cohort study.

Answer 13

(+) Clear temporal sequence between exposure & outcome (since exposure -> outcome)
(+) Can study several outcomes associated with a single exposure
(+) Ideal & efficient for study of rare exposures (e.g. drugs uncommonly used, rare viral infections)
(+) Can directly measure incidence of outcome among exposed & unexposed subjects
- i.e. proportion who develop outcome in a specified period of time

(-) Inefficient in studying rare outcomes
- Very large sample size is required for studying rare outcomes
(-) Inefficient for study of outcomes that takes a long time to develop (as compared to case-control studies)
(-) Time-consuming (esp. for prospective cohort study)
(-) Costly (esp. for prospective cohort study)
(-) Potential for bias due to loss on follow-up (esp. for prospective cohort study)

Question 14

Compare the strengths & limitations between a prospective & a retrospective cohort study.

Answer 14

Prospective:
(+) More control of the quality & quantity of data (i.e. less potential for bias)
(-) More time-consuming
(-) More expensive

Retrospective:
(-) Less control of the quality & quantity of data (i.e. more potential for bias)
(+) Less time-consuming
(+) Less expensive

Question 15

Describe the design of a case-control study.

Answer 15

1) Groups of individuals are defined on basis of whether or not they have a given disease or outcome, and then compared with respect to their exposure histories.
2) Selection into study on basis of outcome status
- Cases: Outcome present
- Control: Outcome absent, BUT at risk of developing outcome

Question 16

How should the selection process of the cases & controls be, when conducting a case-control study?

Answer 16

Cases: Selected independently of exposure
- i.e. investigators are blinded to exposure details to prevent selection & information bias

Controls:

1) Ideally should be a direct, random sample of source population from which the cases originated
2) Represent population at risk of developing outcome (i.e. becoming cases)
- Challenge is defining source population & selecting controls
3) Must be sampled independently of exposure

Question 17

Compare the strengths & limitations of using hospitalised patients vs using general population as sources of controls for a case-control study.

Answer 17

1) Hospitalised patients:
(+) Convenient
(+) Relatively inexpensive to identify & interview
(+) Cases & controls are likely to be similar in their accuracy of recall
(+) Generally high level of cooperation of subjects
(-) Disease for which controls are hospitalised may be associated with risk factors under study (i.e. potential confounding)
- e.g. lung cancer as outcome, smoking as exposure, control as coronary heart disease patients but may be associated to smoking as risk factor
(-) Selection factors leading to hospitalisation may differ between cases & controls (i.e. referral patterns may not be representative of entire demographic/population)

2) General population:
(+) Generally ensures comparability
(-) Often difficult to enumerate all members of population as basis for selecting individuals
(-) Difficult to gain cooperation for participation
(-) Relatively expensive to identify & interview
(-) May not recall exposures with same degree of accuracy as cases

Question 18

Discuss the strengths & limitations of a case-control study.

Answer 18

(+) Ideal & efficient for study of rare outcomes
(+) Can study several exposures / risk factors associated with a single outcome
(+) Efficient for study of outcomes that takes a long time to develop (i.e long latency periods)
(+) Efficient in terms of time & money

(-) Inefficient in studying rare exposures
(-) Selection of appropriate control is difficult
(-) Appropriate data on exposure may be difficult to obtain (e.g. OTC, TCM, supplements -> reliant on recall of participants)
(-) Cannot estimate outcome rates directly, but can estimate relative measures via OR
(-) More potential for bias because outcome status known when assessing exposure status

Question 19

Describe the design of a nested case-control study.

Answer 19

Case-control study “nested” w/in a cohort study

Outline:

1) Identify a cohort w/ banked specimens, images or information.
2) Within the cohort, identify those participants who developed the outcome during follow-up (cases).
3) Within the same cohort, select a sample from the rest of the cohort to form controls.
4) Assess characteristics in the cases & controls that might predict the outcomes (i.e. exposures, risk factors).

Question 20

Discuss the strengths & limitations of a nested case-control study.

Answer 20

(+) Efficient in terms of time & money
- Useful for costly measurements that have been archived at the beginning of cohort study & preserved for later analysis for only cases & controls, not whole cohort
(+) Preserves the advantages of cohort studies that resulted from collecting predictor variables before outcomes happened
- Avoid interviewer’s bias to identify exposure variables before outcome occurs
(+) Avoids potential biases of conventional case-control studies that draws cases & controls from different populations

(-) Many research qn & circumstances are NOT amendable to strategy of storing materials for later analysis on a sample of study subjects
(-) When data are available for entire cohort at no additional costs, nothing is gained by studying only a sample of controls at reduced statistical power.
- Whole cohort data should be used instead!