Stats + mock mistakes Flashcards
Cohort Study
What is it?
Follows a group (cohort) over time to compare the incidence of disease between exposed and unexposed individuals.
Direction: Forward (prospective) or backward (retrospective)
Best for:
- Studying causality and incidence
- When exposure is known and common
Key strength:
Establishes temporal relationship (exposure before outcome)
Example:
Follow smokers vs non-smokers for 10 years to see who develops Parkinson’s.
Case-Control Study
What is it?
Starts with people who already have the disease (cases) and compares them to those without it (controls), looking backward for exposures.
Direction: Retrospective
Best for:
- Studying rare diseases
- Quick and cost-effective research
Key weakness:
Cannot easily establish time-order or calculate incidence
Example:
Take 100 people with Parkinson’s and 100 without, and ask if they smoked in the past.
Cross-Sectional Study
What is it?
A snapshot of a population at one point in time to assess both exposure and outcome simultaneously.
Direction: None (one-time observation)
Best for:
- Measuring prevalence
- Hypothesis generation
Key limitation:
Cannot establish causality
Example:
Survey 5,000 people today on their smoking habits and Parkinson’s diagnosis status.
Cross-Over Randomised Trial
What is it?
A type of randomized trial where each participant receives both treatments in sequence, with a “washout” period in between.
Best for:
- Chronic, reversible conditions
- Comparing two treatments in the same person
Key limitation:
Not suitable for diseases with permanent effects
Example:
Compare two blood pressure drugs in the same patients — each tries both drugs.
Parallel Group Randomised Controlled Trial (RCT)
What is it?
Participants are randomly assigned to only one of the treatment groups — both groups are followed concurrently.
Best for:
- Testing interventions (e.g., new drugs or therapies)
Key limitation:
Not appropriate for harmful exposures (like smoking)
Example:
Randomly assign 200 people to a new drug vs placebo and monitor outcomes.
What is a t-test?
A t-test is a statistical test used to determine whether there is a significant difference between the means of two groups. It assumes that the data is:
- Continuous (e.g. height, blood pressure, weight)
- Normally distributed
- From independent or paired groups, depending on the type of t-test
What does a t-test compare?
The means of a continuous variable between two groups.
Independent t-test: 2 separate groups (different participants)
Paired t-test: same group, 2 time points (same participants)
When wld u use a t-test and when would u use a chi-squared test
If you’re comparing means → think t-test.
If you’re comparing proportions or categories → think chi-square test.
Phases of clinical trials
mnemonic: 0 Safety, 1 Safe, 2 Works?, 3 Prove it, 4 Watch it
Phase 0: First in humans, microdosing
Phase 1: Is it safe?
Phase 2: Does it work?
Phase 3: Prove it works — for licensing
Phase 4: Long-term monitoring after approval
What happens in Phase 0 of clinical trials?
Microdosing in a very small group to study how the drug behaves in the body (pharmacokinetics). No therapeutic intent.
What is the main purpose of Phase 1 trials?
To assess safety, tolerability, and dosage in a small group of healthy volunteers (or sometimes patients
What is tested during Phase 2 trials?
Preliminary efficacy and side effects in a larger group of patients with the condition
What is the goal of Phase 3 trials?
To confirm efficacy, monitor for adverse effects, and compare the drug with standard treatments. Forms the basis for licensing approval (so its licensed already but only newly)
What is the purpose of Phase 4 trials?
Post-marketing surveillance to detect rare or long-term side effects and assess ongoing safety and effectiveness in the general population
What are the Bradford Hill criteria for assessing a causal relationship in epidemiology?
The Bradford Hill criteria are nine principles used to evaluate whether an observed association is likely to be causal:
Strength of Association – A stronger association (e.g., high relative risk or odds ratio) is more likely to be causal.
Consistency – The association is observed repeatedly in different studies, settings, and populations.
Specificity – A specific exposure is linked to a specific outcome (less emphasized today).
Temporality – The cause must precede the effect. This is the only essential criterion.
Biological Gradient (Dose-Response) – Risk of disease increases with greater exposure.
Plausibility – The relationship is biologically or medically sensible based on current knowledge.
Coherence – The association does not conflict with existing theory or knowledge of the disease’s natural history.
Experiment (Reversibility) – Removal or reduction of exposure leads to a decrease in risk (e.g., quitting smoking lowers cancer risk).
Analogy – Similar factors are known to cause similar effects (e.g., asbestos and mesothelioma → silica and lung disease).
What is relative poverty, and how does it differ from absolute poverty in the UK?
Relative poverty (or relative low income):
When a household has less than 60% of the current median UK income, after housing costs (AHC) or before housing costs (BHC).
➤ It reflects inequality — being poor compared to others in society.
Absolute poverty (or absolute low income):
When a household earns less than 60% of the median income in 2010/11, adjusted for inflation (held constant in real terms).
➤ It reflects fixed hardship, not changing with societal wealth.
What statistical test should be performed to compare the means of a normally distributed variable between two groups?
2-sample t-test (also called Independent t-test)
When to use it:
Data follows a Gaussian (normal) distribution.
You are comparing the means of a continuous variable between two independent groups.
When should the Chi-squared test be used in statistical analysis?
Chi-squared test
When to use it:
For categorical data (nominal or ordinal variables).
To test if there is a significant association between two categorical variables.
Used to compare the observed vs. expected frequencies in categories (e.g., yes/no responses, disease/no disease).
When should Fisher’s Exact test be used instead of the Chi-squared test?
Fisher’s Exact test
When to use it:
Used for categorical data, particularly when dealing with small sample sizes or small expected frequencies (usually less than 5 in any cell).
Provides an exact p-value, unlike the Chi-squared test, which is an approximation.
When should the Paired t-test be used in statistical analysis?
Paired t-test
When to use it:
Used when comparing the means of a continuous variable in related or paired groups (e.g., same subjects measured before and after an intervention).
Assumes that the differences between pairs are normally distributed.
What are the following tests used for:
- 2-sample t-test
- paired t-test
- chi-squared test
- Fisher’s exact test
- Wilcoxon signed-rank test
2-sample t-test: For comparing means of normally distributed continuous data between two independent groups.
Paired t-test: For comparing means of related continuous data (e.g., before vs. after).
Chi-squared test: For testing relationships between categorical variables.
Fisher’s Exact test: For small sample sizes with categorical data.
Wilcoxon signed-rank test: For comparing paired data when the data is not normally distributed.
a high AST>ALT is considered >2 . What does this indicate?
aspartate aminotransferase (AST) to alanine aminotransferase (ALT) of >2 suggests;
- Alcohol-related liver disease
- Cirrhosis
- Advanced fibrosis
a AST/ALT ratio of <1 indicates what?
AST/ALT ratio of <1 i.e. ALT is higher
- Acute viral hepatitis
- NAFLD
n.b. Drug-induced liver injury and autoimmune hepatitis also generally have an AST/ALT ratio below 1, except in severe cases
What does a P value represent in statistical testing?
The P value is the probability of obtaining the observed results, or more extreme, assuming the null hypothesis is true. A low P value (< 0.05) suggests that the observed result is unlikely due to chance, and we may reject the null hypothesis
P < 0.05: Statistically significant
P > 0.05: Not statistically significant
