Biostats- Study design

What are the two categories of study design ?

Observational and experimental studies

What are experimental studies ?

Exposure/treatment is controlled by researcher

• Replication: same treatments assigned to diff samplings units to asses variation in responses

• Control (placebo): experimental unitsg iven no treatment or a standard treatment for comparison purposes.

• Blinding: participants or researchers are unaware of treatment assignments to reduce bias

  • Desirable but not always possible

• Randomization: random assignment of participants to different groups to ensure comparability between them to asses casuality

*strong case for causation

What are observational studies >

Exposure/treatment not controlled by researcher, rather collected data from existing situation (does not interfere w processes/system)

-no randomization needed

-imbalances btwn exposure/treatment groups done through data analysis

*strong case for confounding

What is confouder ?

A variable that is related to both the exposure and outcome, potentially leading to a spurious association in observational studies.

Intervening variable: only associated w one or the other

What are variables needed to take into account when determining causality ?

1. association due to chance (sampling variability)

2. bias

3. confounding 

Bradford hill for inferring causality

o Strength of association: too strong to be explained otherwise?

o Consistency: is it consistent with other studies?

o Specificity: Is exposure associated with specific, expected outcomes versus numerous outcomes?

o Temporality: Is the time sequence consistent with causality?

o Biological gradient: Is there evidence of a dose-response relationship?

o Plausibility: Is it biologically plausible?

o Coherence: agreement between epidemiological and laboratory findings

o Analogy: agrees with observed effects of similar risk factors

o Reversibility: does removing exposure remove the disease?

Rank study designs from highest to lowest level of evidence ?

1. Systematic reviews and meta-analyses

2. Randomized controlled trials (RCTs)

3. Cohort studies

4. Case-control studies

5. Cross-sectional studies

6. Ecologic studies

7. Case reports and series

Case report studies

Detailed reports of the characteristics (symptoms, diagnosis, treatment, and follow-ups) of individual patients. Used to highlight novel diseases, unique cases, or unexpected outcomes.

CONS: no controls, no replicates, cannot make causal inference, small numbers means insignificant patterns

PROS: described new disease/phenomenon, first step to a better designed study

Ecologic studies

Studies that investigate relationships bwtn exposure & outcome at the population level (NOT individual level) using aggregate data.

CONS: relates data about distinct populations, can’t compare individuals so we cannot accurately reflect correlations

PROS: compare larger units, uses existing data to evaluate hypothesis, can find links btwn risk factors and health outcomes in advance to laboratory approaches

Cross-sectional studies

Analyzes data from a specific population at a single point in time. These studies assess the prevalence of outcomes and compares exposures but do not establish cause-and-effect relationships.

CONS: shows association, not causation, outcome & exposure measured at same time so hard to tell whether exposure preceded outcome

PROS: relatively quick and inexpensive to conduct, useful for generating hypotheses, and can assess multiple outcomes at once

Case control studies

Observational studies that compare individuals with a specific outcome (cases) to those without it (controls) to identify risk factors or causes.

CONS: cannot estimate prevalence/risk, often has more controls than cases, possible bias in terms of evaluating exposure

PROS: minimizes confounding, allows for exploration of multiple exposures, efficient for rare outcomes, low cost

What can/cannot be estimated from case control studies

Cannot: prevalence or incidence rates

Can: risk exposure associated btwn exposures and outcomes, odds of exposure associated btwn exposures and outcomes, A VALID ODDS RATIO

What are odds ? Statistical odds ?

Chance of winning/ chance of losing

Statistics: odds: p / (1-p)

Chance of probability of event = p 

ODDS RATIO: odds of exposure among those w disease/ odds of exposure among those without disease

Odds ration EX 

• Odds of exposure among those with disease = 𝐴/(𝐴+𝐵) / 1− (𝐴/(𝐴+𝐵))= A/B

• Odds of exposure among those without disease = 𝐶/ (𝐶+𝐷)/ 1 − (𝐶/(𝐶+𝐷)) = C/D

• Odds ratio of exposure, comparing disease to no disease = (𝐴/𝐵)/ (𝐶/𝐷) = (𝐴/𝐵) / (𝐶/𝐷) = 𝐴/𝐵×𝐷/𝐶

• Odds ratio (of exposure) = A×D/B×C (odds ratio is also called the cross-product ratio)

Case control odds ration example

Odds ration vs relative risk

o OR < 1 if and only if RR < 1

o OR > 1 if and only if RR > 1

o OR = 1 if and only if RR = 1

• if RR < 1, then OR ≤ RR < 1

• if RR > 1, then 1 < RR ≤ OR

*if outcome is rare, odds ration approximates relative risk

Cohort studies

Subjects are followed(follow-ups) over time to determine the outcome based on exposure status. Participants are grouped based on whether they were exposed to a particular risk factor.

PROS: reduced bias in exposure, time of exposure/disease is clear

CONS: more time-consuming, expensive to conduct, potential for loss to follow-up, inefficient for rare diseases, potential for confounding

Randomized clinical trials

A study design where participants are randomly assigned to receive either the treatment or a placebo/control. This method reduces bias and allows for direct comparisons of treatment effects.

PROS: more complex analyses can be done(IR & CI), protects against confounding being associated w exposure, can examine multiple outcomes, allows for casual inference, timing of exposure and outcome is clear

CONS: high cost, ethical concerns, participant bias/recruitment may limit, cannot do an RCT to show harm

Why have randomized clinical trials failed before ?

Poorly concealed randomization, lack of blinding, poor deliver of intervention, loss to follow-up

Meta analysis studies

A statistical technique that combines results from multiple studies to identify patterns, trends, or effects. It provides a more comprehensive understanding of an effect or outcome by aggregating data across various studies.

PROS: provides more precise estimates, understands sources of variation in study results

CONS: possible publication bias, subjectivity in decided which studies to include

What is a forest plot ?

A graphical representation used in meta-analysis to display the point estimates and confidence intervals of individual studies, as well as the overall pooled estimate.