Advanced Epidemiologic Study Designs

what are the 4 advanced study designs?

Nested Case-Control Study

case-cohort sudy

case-crossover

N-of 1 trial

Case-control stidy

• Start with who has the disease → move back in time to see exposure

• create odds ratio

• sampling from a cohort.

• provides valid estimate of rate ratios or risk ratios

What is the TROHOC fallacy?

The misconception that case-control studies cannot provide valid estimates of risk or rate ratios. Properly designed and sampled case-control studies can estimate these measures.

How does a cohort study differ from a case-control study?

A cohort study starts with exposed and unexposed individuals and follows them to determine who develops the disease. A case-control study starts with individuals who have and do not have the disease, then looks backward to compare their previous exposures.

What measure of association does a cohort study directly estimate?

Risk (incidence) and relative risk (RR).

What measure of association is commonly estimated in a case-control study?

The odds ratio (OR), which can approximate the relative risk under certain conditions.

What is incidence density sampling?	

• Selecting controls from the cohort's person-time at risk. This approach allows case-control studies to estimate valid rate ratios.

What is a nested case-control study?

• A case-control study conducted within an existing cohort. Cases are compared with a selected sample of non-cases from that same cohort.

Why conduct a nested case-control study instead of analyzing the entire cohort?

To reduce the cost and time required when exposure measurements are expensive or difficult to obtain. Researchers only need to analyze exposure data for the cases and selected controls rather than the entire cohort.

How is a nested case-control study conducted?

1. Establish a cohort. 2. Follow participants over time. 3. Identify cases. 4. Select controls from the same cohort. 5. Compare exposures between cases and controls.

In the DDT example, why could researchers study breast cancer decades after DDT was banned?

DDT is persistent, and its byproduct DDE can be measured in frozen blood samples collected years earlier. This allows researchers to assess past exposure even decades after DDT was banned in 1969.

In the DDE example, what was the full cohort and outcome?

The cohort included 100,000 female teachers aged 45–65. Researchers followed them for outcomes including breast cancer; 2,000 developed breast cancer during follow-up.

In the DDE cohort example, how was the relative risk calculated?

RR = (risk in high-DDE group) ÷ (risk in low-DDE group) = (500/15,000) ÷ (1,500/85,000) = 1.89.

In the DDE example, what were the case and control counts in the nested case-control table?

The table included 2,000 cases: 500 with high DDE and 1,500 with low DDE. It included 4,000 controls: 600 with high DDE and 3,400 with low DDE.

How was the odds ratio calculated in the DDE nested case-control example?

OR = (500 × 3,400) ÷ (1,500 × 600) = 1.89.

Why did the nested case-control example produce the same OR and RR?

The sampling approach preserved the exposure odds comparison. The example illustrates how a properly designed case-control sample can reproduce the cohort association.

What are the five strengths of nested case-control studies?

1. Efficient and cost-effective because only a subset of the cohort is analyzed. 2. Exposure precedes disease, supporting temporal clarity. 3. Reduced selection bias because cases and controls come from the same cohort. 4. Useful for studying rare outcomes. 5. Reduced recall bias when biological samples or recorded exposure data are used.

What are the five limitations of nested case-control studies?

1. Limited to data and samples available from the original cohort. 2. Controls are generally only controls for one disease. 3. Matching cases and controls can be challenging. 4. Generalizability depends on how representative the original cohort is. 5. Time lags can make findings less relevant as exposures and diagnostic criteria change.

What is a case-cohort study?

An observational study nested within a larger cohort that compares all identified cases with a random subcohort selected at baseline.

What distinguishes case-cohort from nested case-control sampling?

A nested case-control study selects controls for individual cases. A case-cohort study selects one random subcohort at baseline and uses it as the comparison group for cases.

How is a case-cohort study conducted?

1. Establish a cohort. 2. Randomly select a subcohort at baseline, before outcomes occur. 3. Identify all cases during follow-up. 4. Compare exposures between the cases and the subcohort.

When is a case-cohort study particularly useful?

When outcomes are rare or exposure measurements are expensive or difficult to obtain.

Can the same case-cohort subcohort be used to study multiple outcomes?

Yes. The same randomly selected subcohort can serve as the comparison group for multiple outcomes, making the design efficient.

What happens if a member of the case-cohort subcohort develops the disease during follow-up?

They are included both as a case and as a member of the original subcohort.

Why is it acceptable for a case to remain in the case-cohort subcohort?

It preserves the original random sampling, avoids the need to find replacement controls, and specialized statistical methods account for the overlap.

What statistical method is mentioned for handling case-subcohort overlap?

Prentice's method. The lecture also mentions weighted Cox regression as a specialized analytical approach.

What are the four strengths of case-cohort studies?	

1. Efficient. 2. The same subcohort can be used for multiple outcomes. 3. Random sampling helps minimize selection bias. 4. The design maintains the temporal sequence between exposure and outcome.

What are the four limitations of case-cohort studies?

1. Complex analysis requiring specialized methods, such as weighted Cox regression. 2. Possible exposure misclassification if data are not updated or accurate. 3. Case-subcohort overlap complicates analysis. 4. Generalizability depends on how representative the subcohort is.

What is a case-crossover study?

A self-controlled observational design comparing exposure immediately before a sudden event with exposure during another control period in the same individual.

What is the defining feature of a case-crossover study?	

Each participant serves as their own control. Exposure at different times in the same person's life is compared.

What types of exposures and outcomes are studied using case-crossover designs?

Transient, short-term exposures that may trigger sudden events. Examples include vigorous exertion or alcohol consumption before a myocardial infarction.

What is the main research question addressed by a case-crossover study?

Whether a transient exposure is more likely to occur shortly before an event than during another comparable period when the event did not occur.

Why is a case-crossover design described as self-controlled?	

The same individual contributes both the hazard period and the control period. This allows researchers to compare the person's exposure at different times without comparing them with a separate control group.

What is a hazard window in a case-crossover study?	

The relevant period immediately before the event during which the exposure is hypothesized to trigger the outcome.

In the vigorous-exertion and myocardial infarction example, what could the hazard window be?

If heavy exertion is thought to trigger a myocardial infarction within two hours, the hazard window could be the two hours immediately before the event. For example, if the MI occurs at 4 pm, the hazard window could be 2–4 pm.

What is a referent window in a case-crossover study?

A comparison period during which the event did not occur. Exposure during this period is compared with exposure during the hazard window.

How are hazard and referent windows compared in the vigorous-exertion example?

Researchers could compare exertion between 2–4 pm on the day of the myocardial infarction (hazard window) and 2–4 pm on another day when the individual did not experience an MI (referent window).

Why is choosing the correct hazard window important?

The plausible induction time between the exposure and outcome may be unknown or may vary. The window must be long enough to capture relevant exposures but not so long that it includes exposures unrelated to the event.

What happens if the hazard window is too short?

It can reduce statistical power because relevant exposure events may be excluded.

What happens if the hazard window is too long?

It can bias the association toward the null by including exposure periods that are less relevant to the event.

What factors should be considered when choosing a referent window?

1. It should be far enough from the event that the event does not affect exposure. 2. It should be sufficiently recent that the underlying exposure rate is comparable to the hazard period. 3. It should represent an appropriate period when the individual was at risk.

Why might five minutes before a car crash be a poor referent window when studying mobile phone use?

A phone call may have started before that period and continued into the crash. The referent period could therefore overlap with the exposure around the event.

Why might one year before a car crash be a poor referent window when studying mobile phone use?

Exposure patterns may have changed substantially over a year. For example, COVID-19 lockdowns could have changed phone-use patterns, making the earlier period less comparable.

What did the mobile phone use and motor vehicle crash example demonstrate?

The odds of mobile phone use in the minutes immediately before a crash were several times greater than during a similar time earlier in the week.

What did the case-crossover example involving medication prescriptions and inpatient falls examine?

It examined new antihypertensive or hypnotic prescriptions in the three days before inpatient falls, compared with earlier referent periods.

What did the recent hospital discharge and opioid overdose example find?

People were four times more likely to die from an opioid overdose shortly after discharge from inpatient treatment than during the previous two years.

What is the purpose of using the same individual as their own control in a case-crossover study?

It controls for characteristics that do not change over time within the individual, including factors such as age, sex, and stable lifestyle characteristics.

What types of confounding can a case-crossover design help eliminate?

Time-invariant confounding, including measured and unmeasured characteristics that remain constant within the individual over the comparison periods.

How is the matched odds ratio calculated in a case-crossover study?

The matched OR is calculated using discordant pairs: OR = b/c. Here, b is the number of pairs exposed during the hazard window but not the referent window, and c is the number exposed during the referent window but not the hazard window.

In the lecture's matched case-crossover table, what were the four pair counts?

Hazard exposed and referent exposed = 4; hazard exposed and referent not exposed = 50; hazard not exposed and referent exposed = 9; hazard not exposed and referent not exposed = 1,165.

Using the lecture's matched-pair table, what was the matched odds ratio?

OR = 50/9 ≈ 5.56. Only the discordant pairs (50 and 9) are used in the matched odds ratio calculation.

What are the strengths of case-crossover studies?

1. They control for time-invariant confounders, including unmeasured factors such as age and sex. 2. They are useful when finding an appropriate separate control group for a traditional case-control study is difficult.

What are the limitations of case-crossover studies?

1. Co-occurring acute exposures can be difficult to separate. 2. They are limited to studying short-term effects of transient exposures. 3. They may not capture the long-term effects of an exposure. 4. Exposure status must vary over time within individuals for the design to be informative.

Why can co-occurring acute exposures be a problem in case-crossover studies?

If multiple exposures occur close together before an event, it can be difficult to determine which exposure contributed to the outcome.

Why might a case-crossover study of running and myocardial infarction give a misleading impression if interpreted without considering long-term effects?

A single bout of running may temporarily increase the immediate risk of myocardial infarction, while regular running may reduce long-term risk. The short-term trigger and long-term effects are different.

Why do case-crossover studies use information from cases only?

The design compares each case's exposure during the hazard period with their exposure during referent periods. It is informative when the individual's exposure varies over time.

What is an N-of-1 trial?

A randomized crossover trial conducted with one participant to determine whether a treatment works for that specific individual.

What question does an N-of-1 trial aim to answer?

Does this treatment work for this particular patient?

Why might evidence from large clinical trials not predict an individual's response to treatment?

Large trials generally report average treatment effects across groups. An individual patient may respond differently from the average participant.

How is an N-of-1 trial related to a crossover randomized controlled trial?

It uses a crossover design with one participant. The same patient receives both the active treatment and the control or placebo during different periods.

In what clinical situation might an N-of-1 trial be useful?

It can be used in clinical practice to evaluate treatment for an individual with a chronic condition, such as a patient with a persistent rash.

How could an N-of-1 trial test whether a topical cream helps a patient with a chronic rash?

The patient alternates between an active dermatologic cream and an inactive placebo cream in a randomly allocated sequence. The patient's results during the active and placebo periods are compared.

Why should an N-of-1 trial be randomized?

Randomizing the order of treatment and placebo periods helps reduce bias from the order in which the treatments are received.

Why should an N-of-1 trial be blinded?

Blinding the patient and physician helps reduce the influence of expectations on symptom reporting and treatment assessment.

How are the results of an N-of-1 trial used to make a treatment decision?

If the patient improves during the active treatment periods compared with placebo periods, the treatment may be continued. If there is no benefit, the treatment may not be worthwhile for that patient.

What was the general treatment sequence in the lecture's donepezil N-of-1 example?

The participant randomly alternated between placebo and drug periods over 36 weeks. The sequence shown was PLACEBO → DRUG → PLACEBO → DRUG → PLACEBO → DRUG.

What dosing schedule was used during the treatment periods in the donepezil example?

The treatment schedule included 5 mg/day for four weeks, followed by 10 mg/day for two weeks.

What are the strengths of N-of-1 trials?

1. They provide personalized evidence about treatment effectiveness. 2. The patient serves as their own control, reducing between-person confounding. 3. Randomization and blinding improve internal validity. 4. They support patient engagement and shared decision-making.

Why do N-of-1 trials control for between-person confounding?

The same patient receives both the active treatment and control condition, so characteristics that differ between people are held constant within the trial.

For what types of conditions are N-of-1 trials most useful?

Chronic, relatively stable conditions where treatment effects begin quickly and wear off after treatment stops. Examples from the lecture include chronic pain, ADHD, and insomnia.

What are the limitations of N-of-1 trials?

1. Limited generalizability because the trial involves one person. 2. The condition needs to be stable. 3. Planning, randomization, and statistical analysis can be complex. 4. The trial can place a burden on the patient. 5. Carryover effects can occur. 6. Some interventions and outcomes are unsuitable.

Why does an N-of-1 trial require a relatively stable condition?

If the condition changes substantially over time on its own, it becomes difficult to determine whether changes in symptoms are due to treatment or to the natural course of the condition.

Why are rapid treatment onset and washout important in an N-of-1 trial?

The treatment should begin having an effect within a practical period, and its effects should wear off after stopping it so that the next trial period can be evaluated fairly.

What are carryover effects in an N-of-1 trial?

Carryover effects occur when the effects of a treatment persist into a later trial period, potentially influencing the results of the next treatment or placebo phase.

Why might an N-of-1 trial be unsuitable for surgery or mortality outcomes?

Surgery cannot generally be repeatedly switched on and off in crossover periods, and outcomes such as death cannot be repeatedly measured within one individual.

What were the feasibility results for the N-of-1 trials discussed in the lecture?

Of 73 planned trials, 70 began and 57 were completed. Fifty trials produced a definite clinical or statistical answer.

How often did N-of-1 trial results change a physician's prior treatment plan in the lecture's feasibility example?

In 15 trials, representing 39% of trials with appropriate data, the results changed the physician's prior plan.

What do the N-of-1 feasibility results suggest?

N-of-1 trials can be feasible and useful in clinical practice, and their findings can influence physicians' treatment decisions.

What is the key difference between nested case-control and case-cohort studies?

Nested case-control studies select controls for individual cases, whereas case-cohort studies select one random subcohort at baseline to compare with all cases.

What is the key difference between case-cohort and case-crossover studies?

Case-cohort studies compare cases with a randomly selected subcohort from a larger cohort. Case-crossover studies compare exposure during hazard and referent periods within the same individual.

What is the key difference between a case-crossover study and an N-of-1 trial?

A case-crossover study is observational and examines whether a transient exposure is associated with a sudden event. An N-of-1 trial is an experimental randomized crossover design that tests whether a treatment works for one patient.

Which study design is best suited to studying an expensive exposure measurement within an existing cohort?

A nested case-control study or a case-cohort study can be efficient because researchers do not need to measure exposure for the entire cohort.

Which study design uses one randomly selected comparison group for multiple outcomes?

A case-cohort study.

Which study design is best suited to investigating whether a short-term trigger occurred immediately before a sudden event?

A case-crossover study.

Which study design is best suited to determining whether a treatment works for one specific patient?

An N-of-1 trial.