LECTURE 0: How do we identify causes of disease?

(Part 1)

●​ Define epidemiology, and explain what is meant by exposure and outcome

Epidemiology
The study of the distribution and determinants of health-related states or events in specified

✅
populations, and the application of this study to control of health problems”

✅
to help resource allocation
Comparing disease around the world can helps hypothesise different risk factors between
different communities

Purpose:
●​ Measure the burden of disease in population
●​ Health care policy & planning
●​ Are there patterns / trends in disease occurrence
●​ Identify causes or factors that sustain disease
●​ Evaluation interventions

Public Health: The science and art of preventing disease, prolonging life and promoting
health through the organised efforts of society.

Exposure is a risk factor, biomarker and phenotype.

Health Outcome: health related state we are interested in

(e.g. pregnancy, death, cure, exacerbation of asthma, smoking cessation)

●​ Explain the meaning of a risk factor or an association

Risk factors: increase disease, can be beneficial, it’s something that alters the risk
helps to compare across population to investigate why disease is happening at a higher rate

Risk Factors of stroke:

Modifiable (possible to alter levels)
-​ Diabetes
-​ Smoking
-​ Medications e.g. HRT , hormone altering
-​ High BP
-​ Obesity
-​ Lipid levels
-​ Alcohol consumption = higher BP
-​ Physical activity is a beneficial risk factor
Non modifiable:
-​ Age/Sex

Association: apparent association between exposure and health outcome may be due to:
chance, bias, confounding variable, cause & effect
Alcohol relation to lung cancer can be due to the positive correlation with also smoking and
therefore causing disease
Cause of disease: factor that is associated with the disease so that when the intensity or
frequency of the factor changes in the population, the frequency of the disease also
changes.

Cause & effect is NOT inevitable, necessary, sufficient

Eg: Smoking causes cancer (smoking is cause, cancer is effect)
• inevitable: not all smokers will get lung cancer
• necessary: not all lung cancers caused by smoking (can’t get disease without it)
• sufficient: not the single cause

We can very rarely say that a factor caused the disease for an individual but we can say it
caused a disease in a POPULATION and increased the risk in an individual.

Can't say a factor caused disease in an individual, it increased risk in the individual. Or
caused disease in a population

Bradford Hill causality criteria determines evidence of a causal relationship between a

presumed cause and an observed effect

LECTURE 1: Measuring health and disease

●​ Contrast the difference between ecological and cross-sectional studies and
understand the advantages and disadvantages of these two study designs​

Descriptive study: Differences over time, variations in diff places, diff characteristics
-​ Ecological and cross sectional

Analytical study: Do people have the disease? Develop the disease more frequently?
Compare characteristics
-​ Cross sectional
Cross sectional: SNAPSHOT IN TIME (Who’s vegan?), looking at individuals

Ecological: things are measured as a group not individual level

Design: Researchers examine the relationship between exposure and outcome variables by
comparing different populations or groups rather than individuals within those populations.

Example: Studying the relationship between air pollution levels (exposure) and respiratory
diseases (outcome) by comparing the rates of respiratory diseases in cities with varying
levels of air pollution.

●​ Outline how routine sources of data and observational data are used to obtain
mortality and morbidity rates​
Health data: census, primary care data

People only seek care after experiencing symptoms

Then this can lead to GP or hospital records
→ registration data

Cohort studies help health outcomes

HEALTH survey for England: monitors trends in nations health care, adults aged 16+ ,
children 0-15, living in private household in England, Interview followed by nurse who comes
to take measurements.
Measures individual prevalence of exposure (aka risk factor)

-​ Descriptive if outcome not reported in relation to exposure, or if exposure not

reported in relation to outcome
-​ Analytical if disease prevalence compared by exposure / risk factor status

Health survey for England is a cross sectional study

Population Health and Evidence Based Practice

Showed the prevalence of obesity in men and outlines the obesity epidemic!

●​ Define incidence, prevalence, risk and rate and be able to interpret examples of each​

PE IN the gym
Prevalence = number existing cases / total population
✅
(^OVER SPECIFIC PERIOD OF TIME^)
useful for short term and episodic diseases

Proportion of people with the disease is how you compare prevalence

population size matters to understand incidence!

Measuring disease occurrence:

Incidence: number of new cases of a disorder that develop in specified time interval per
recruited (defined) study population (HOW FAST/RATE A DISEASE IS DEVELOPING)

●​ Interpret standardised rates and ratios from a plot or table and explain why they are
used

As paracetamol sales go up, wheeze prevalence in 13-14 yos goes up

BUT
-​ Country-level exposure data, but centre-level prevalence data
-​ Exposure data for all ages, but prevalence data for one specific age-group
-​ No information on who consumes paracetamol, nor on how frequently they take it
Mortality

May be for a population overall

Expressed in categories: sex, age groups, regions, countries and compare death rates
overtime

Standardisation allows comparing death/disease rates in two or more populations

Account for differences between populations in characteristics that affect mortality /disease
occurrence
-​ Age distribution is most important
-​ Many rates differ between men and women

1.​ Direct method (typical for age standardised rates).

Obtain age specific mortality rates from the populations of interest
(e.g. 5 year age groups from each country)

Then repeat process for all populations and compare age standardised mortality rates
Answer: B (140/40= 3.5)

2.​ Indirect Method (age adjusted rate RATIOS)

Obtain age specific mortality rates from standard/reference population
we can calculate a reference number of deaths in the country

Standardised mortality rates (SMRs)

1.​ Cross-Sectional Studies:

○​ Definition: In cross-sectional studies, data is collected at a single point in time,
providing a snapshot of a population at that particular moment.
○​ Design: Researchers gather data from a sample of individuals or units from a
population and examine the relationship between variables at that specific time.
○​ Example: A survey conducted to assess the prevalence of smoking among different
age groups in a community at a particular time.
 ○​ Strengths: They are relatively quick and inexpensive to conduct, providing
information on the prevalence of a condition or behaviour in a population.
○​ Limitations: Since data is collected at a single time point, it can only show
associations between variables at that time and cannot establish causality or assess
changes over time.
2.​ Ecological Studies:
○​ Definition: Ecological studies analyze data at the group or population level rather
than individual level.
○​ Design: Researchers examine the relationship between exposure and outcome
variables by comparing different populations or groups rather than individuals within
those populations.
○​ Example: Studying the relationship between air pollution levels (exposure) and
respiratory diseases (outcome) by comparing the rates of respiratory diseases in
cities with varying levels of air pollution.
○​ Strengths: They can provide insights into the population-level associations between
variables and are useful for generating hypotheses.
○​ Limitations: Ecological fallacy can occur, where associations observed at the group
level may not apply to individuals within those groups. Also, since individual-level
data is not collected, confounding variables at the individual level cannot be
controlled for.

LECTURE 2: Evidence on risk factors for

cardiovascular disease
Cohort studies:
●​ Describe the design of a cohort study and define exposure and outcome in
observational epidemiology
Cohort study: An observational study where a population is studied for the presence of a
fixed or modifiable exposure, which is thought to be an aetiology of a condition (e.g. CVD),
prior to the onset of a condition. The entire population is followed up in time, and the
incidence of disease in exposed individuals is compared with the incidence in those not
exposed.
-​ Group of people with a statistic in common
-​ Looking at incidence (new cases) of disease over time AKA prospective/longitudinal

Cohort Study Design:

-​ Start with population without disease outcome
-​ General population, geographically/occupationally defined
-​ Can use a highly exposed group eg assessing workers w occupational
exposures like asbestos or radiation. Compare with workers in same factory
without exposure
-​ Measure those that are exposed to something and those that are not
-​ Then look at the development of disease and compare in both

Compare = Risk in exposed/risk in unexposed

How do we define exposure?
●​ DEFINE:
○​ industrial (rare); vinyl chloride
○​ radiation – infectious; hepatitis B, HIV
○​ social; stress, social class, community
○​ non-biological and biological
●​ DEGREE of exposure: type of exposure
●​ DURATION of exposure: intermittent, constant; changes in exposure over time

Cohort studies allow examination of multiple outcomes

Outcome data from various sources:
-​ routine surveillance systems; death and cancer registers
-​ medical records
-​ questionnaire data
-​ physical examination
Outcome data collection must not be influenced by knowledge of exposure to prevent
measurement bias

●​ Describe methods used for quantifying disease or mortality rates between exposure
groups or by level of risk factor exposure
Isn’t this referring to excess deaths and risk rate?
Excess death = death in exposed - death in unexposed

●​ Calculate a relative risk and interpret it along with the associated 95% confidence
interval You
Risk: number of outcomes per recruited study population
Rate: number of outcomes per person time of those in study interval
Relative risk is measured as rate or risk ratio
Calculation of incidence rate = rate in exposed cohort / rate in unexposed = rate ratio
Incidence risk = risk in exposed cohort / risk in unexposed = risk ratio
Absolute excess risk: risk in exposed minus risk in unexposed
Q1 answer is C (the units cancel out)

Q2 answer is D (exposed - non exposed)

 ●​ Recall the main advantages and disadvantages of cohort studies and explain how
bias or confounding might influence the finding
ADVANTAGES of cohort study:
●​ Evidence of exposure does not depend on memory (except in historical cohorts)
●​ Exposure is measured before disease onset, reducing potential for bias, and allowing
causality to be assumed
●​ Multiple outcomes (diseases) can be studied for any one exposure
●​ Incidence of disease can be measured in the exposed and unexposed groups
●​ Potential for nested case control studies
○​ Nested case control study within cohort study: people who develop disease
become a ‘case’, people who don't are controls. (based on exposure)

DISADVANTAGES of cohort study:

●​ Slow, expensive, administratively difficult and complex
●​ Needs large numbers and a long time
●​ Collection of data may alter behaviour
●​ Problems keeping standards
●​ Changes in exposure and diagnostic criteria with time
●​ Ascertainment of outcome status may be influenced by knowledge of exposure
●​ Losses to follow up may introduce selection bias (i.e. those left different health
behaviour

Confounding variables: associated with the exposure being investigated but also
independently associated with the risk of developing the outcome of interest

What do we do when we have a confounder we want to look at in a cohort study?

example:
 1.​ Collect data on smoking status
2.​ Analyse stratified by smoking status: is the association similar in population without
exposure ie is coffee association with CHD the same in non smokers?
3.​ Statistically adjust for differences in smoking between levels of coffee drinking
4.​ Historical cohorts may lack data on confounders: limitation
5.​ Prospective studies allow new confounders to be studied

What are the biases in cohort studies?

●​ Losses to follow-up occur in all cohort studies
●​ Reduce power and dilute results (due to drop outs)
●​ Problematic if more drop-outs in one exposure group compared with another
●​ Very important if drop-out is due to development of disease
●​ Beware of other types of bias, e.g., ascertainment, interviewer, non-response,
selection bias

Drop outs can affect data^^

Aims:
●​ Consider risk factors for CVD
●​ How cohort studies have been used to examine modifiable risk factors for CVD
●​ Introduce some well known cohort studies
Frambergs Study
British Regional Heart Study: 8000 men 1970s, blood samples taken from back then
still used to look into different potential exposures.

Another example:
●​ How measures of effect are obtained
QRISK - calculate risk of vascular event over next 10 years
 ●​ How evidence is assimilated to examine effect of reducing BP, cholesterol and BMI
on vascular related deaths

LOBs in blue weren’t really answered

Eeeeek

LECTURE 3: Case-control studies in cancer

research
●​ Describe the basic design of a case-control study noting its strengths and
weaknesses
Case control: observational & data from individuals
Data collected on past exposure to risk factors, collect info on risk factors and disease at
the same time.

●​ Cases: people with the disease

○​ Find them with Hospital records, Cancer registries
●​ Controls: subjects who do not have the disease but are otherwise comparable to
cases - same opportunity to be exposed to risk factor
○​ Find them in:
■​ general population: Random sample from the population that gave rise
to the cases; Not always practical; Subjects non-co-operative
■​ hospital controls: Readily available, time to spare, far more
co-operative; Must not have a condition related to exposure(s) of
interest with study disease

Strengths:
●​ Cheap and quick to do
●​ Relatively low cost
●​ Efficient for rare disease or disease with low incidence
●​ Can investigate a range of risk factors/exposures in one disease
●​ Can work out odds ratio ~ even more accurate in rare disease

Weaknesses:
●​ Prone to selection and information bias
●​ Recall bias: Retrospective nature may lead to random misclassification error
●​ Cannot calculate risk ratio as there are no cases of incidence
●​ Not suitable for exposures that are rare
●​ Not suitable for diseases with several main exposures
●​ Need to ensure data collection is not influenced by knowledge of exposure to prevent
measurement bias
●​ The disease may affect the exposure ~ reverse causality
●​ Confounding
Nested case control: from a cohort study, people who develop disease become a ‘case’,
people who don't are controls - based on exposure

●​ Construct a 2x2 table of exposure by outcome and use it to estimate an odds ratio

COHORT

CASE-CONTROL
 ●​ Interpret an odds ratio, its 95% confidence interval and associated p value
Odds ratio: odds among cases/odds among control
If the disease is rare, the odds ratio provides an estimate of the relative risk for the
association between exposure and disease. In a case-control study we CANNOT estimate
incidence or prevalence.

“The odds of exposure were 2.5 times higher in the diseased group compared with
the non-diseased group” (for odds ratio table above)

Odds are a good estimation for rare diseases.

Interpreting the odds ratio:

ORs are interpreted as follows:
=1 no difference in odds of exposure between disease & control groups
>1 increased odds of exposure in the disease group vs. control group
<1 reduced odds of exposure in the disease group vs. control group
To account for uncertainty in our study we estimate a 95% confidence interval
1 is within 95% CI => Not enough evidence of an association
1 is NOT in 95% CI => Statistical evidence of association (p<0.05)

Another example:

Dose response: increasing levels of exposure are associated with either an increasing or a
decreasing risk of the outcome
AKA Graded association
 ●​ Recognise that estimated odds ratios from case-control studies are particularly prone
to confounding and bias and describe common sources of such bias

Confounding variables:
In designing and analysing a case-control study we aim to ensure that our comparisons:
-​ Are not explained by confounding.
-​ Are not explained by bias (systematic error)

Both can happen in case-control and cohort studies but are particularly likely in a
case-control study

How can you control for confounders?

1.​ Adjusting analysis is: collect info on potential confounders.
2.​ Matching (Each case is matched to 1 (or more) controls on basis of the potential
confounder) can reduce the impact of confounding variables, can be based on the
individual or the stratum.

Sources of Bias:
●​ In study design:
 ○​ Selection bias
●​ In methods/data collection:
○​ Recall Bias
○​ Information bias
○​ Interviewer bias
○​ Random error

Collecting exposure data:

In many case-control studies exposure information is obtained from historical records
(e.g., studies of occupational exposure). – These are sometimes referred to as historical
case-control studies

Advantages:
-​ Information recorded prior to disease onset, so may reduce the possibility of reverse
causation
-​ Reduce recall bias

Disadvantages:
-​ Records may be incomplete or missing (loss of cases and controls from analysis).
-​ Those extracting the data may introduce bias unless ‘blind’ to case-control status

LECTURE 4: What is evidence synthesis?

●​ List the pre-requisites for conducting a systematic review and explain in broad terms
how and why they are carried out
Systematic review (overview) of the literature represents ‘a systematic assembly, critical
appraisal and synthesis of all relevant studies on a specific topic’
-​ Can be quantitative (meta analysis) or qualitative
WHY?
Purpose:
-​ To summarise and simplify a complex and vast literature
-​ Increase precision of effect estimates and improve statistical power
 -​ Assess the consistency of the evidence and thereby the generalisability of the
findings
-​ Establish possible explanations for any inconsistencies
-​ Highlight areas of relative certainty or deficiencies in evidence

HOW?
How do you design a systematic review?
-​ Define the primary exposure and outcome
-​ Define the measures of exposure or intervention, outcome or effect
-​ Exclusion and inclusion criteria for the systematic retrieval of relevant studies
-​ Detail the method of extracting the relevant quantitative and / or qualitative
information
-​ Summarise the evidence using the appropriate statistical methods
-​ Interpret the results acknowledging systematic error (bias) and random variation
(chance)

How to choose/define a health outcome?

●​ Depends upon the purpose of the review
●​ Defined outcomes
○​ Mortality, birthweight etc.
●​ Range of outcomes
○​ Non-fatal illness, indices of clinical severity, symptom reports, measures of
health service usage, quality of life indices
●​ Choice of outcome often depends on what is published

Retrieval of studies:
-​ Searches of electronic databases (e.g. Medline, Embase) using text words and
subject headings
-​ Retrospective citations in recent reviews
-​ Prospective citations of key methodological or classic papers (Science Citation
Index)
-​ Hand searches of relevant journals
 -​ Increasing use of data from unpublished studies – so called “grey literature”

Extracting relevant info:

●​ Study sample
○​ Sample size, location, age, gender, ethnicity
●​ Study design
○​ Experimental; parallel, cross-over, factorial
○​ Observational; cross sectional, cohort, case controlled
●​ Potential for bias
○​ Experimental; intention-to-treat, complete case analysis, blinding
○​ Observational; method of exposure measurement, with or without knowledge
of outcome
○​ Can use Risk of Bias tools - to understand reliability of underlying evidence
■​ Feeds into overall GRADE of evidence (high to very low)
●​ Confounding
○​ Large RCT limit the role of confounding
○​ Observational studies prone to confounding
●​ Measures of effect
○​ Differ for qualitative and quantitative reviews
○​ Direction and magnitude of effect
○​ Dichotomous health outcomes
■​ Relative risk, odds ratio, hazard ratio (SE, 95%CI)
○​ Continuous health outcomes
■​ Mean difference and associated variance (SE, 95%CI)

hazard ratio = hazard in exposed/hazard in unexposed

Hazard ratio takes into account person YEARs, looking at length of time
Risk ratio just looks at risk in exposed/risk in unexposed

●​ Outline the main statistical methods for pooling different studies or estimates, and
interpret their meaning
Pooling = data sets coming from different sources are combined

2 methods of pooling data from meta analyses:

●​ ‘Fixed effects’ models:
○​ Pools data when studies are very similar ~ homogeneity between study
estimates
○​ Gives greater relative weight to larger studies
○​ It's more conservative with small studies having sensational results - assumes
publication bias
●​ ‘Random effects’ models
○​ Heterogeneous studies
○​ Greater relative weight to smaller studies
○​ Causes/sources of heterogeneity are examined - why are there differences?
○​ Publication bias contraindicates this method of pooling

●​ Interpret a forest plot for both dichotomous and continuous health outcomes
Table and forest plot (right) are showing exact same data

Characteristics of a forest plot:

●​ Horizontal lines represent 95% CI
●​ Boxes represent risk ratio (point estimates)
○​ Box size proportional to inverse of variants; larger study = bigger box + tighter
95% CI
●​ Pooling (combined) estimate gives assimilated value (+associated CI)
○​ Including test for heterogeneity
■​ I-squared value: low value = little heterogeneity, higher value = lots of
heterogeneity (Low < 25%, Moderate < 50%, High < 75%)
■​ P-value: Chi squared test for estimates, statistical significance
-​ p value now statistically significant, i-squared increased = more heterogenous

-​ No longer statistically significant bc CI bridges over the null line (1)

Answer: C
Big CI + small study, = less precision, less consistent evidence
Describe data above:
Graded / dose prospective associations

Describe data results above:

1st & 2nd pooled estimates have high heterogeneity and statistical significance
3rd & 4th pooled estimates have less heterogeneity and less/no statistical significance
The mean BPs are reducing as intervention increases (reducing alcohol intake) which shows
causal effect & dose response.
Also shows consi stency and reversibility
As you get more data = more certainty

●​ Explain what is meant by heterogeneity in a meta-analysis, and be able to recognize

and examine potential sources of bias
Estimates from individual studies esp observational studies can be biassed. If you do a
systematic review of biassed studies, you’ll be more certain of a biassed estimate

Heterogeneity = differences

Variations based on clinical differences is due to the way the study was carried out
Could also have statistical heterogeneity from estimations because of differences between
people in general (ie if you did the same study on different people, you might not necessarily
get the same result).
Idk if you wanna summarise parts of it bc its a bit chunky
Sorry for being so late will do the lecs now insha’Allah

- is this relevant??? Or just an example

LECTURE 5: Quantifying benefits of disease
prevention
●​ Recognise the importance and scope for prevention of coronary heart disease and
other chronic diseases
What are the top 5 causes of death for men and women?
-​ CHD
-​ Dementia
-​ Chronic lung disease
-​ Stroke
-​ Lung cancer

How can disease be prevented?

●​ Removing the cause of disease
●​ Strengthening host resistance
●​ Interfering with disease pathogenesis
●​ Identifying and treating disease early

How does the scope for disease prevention vary between diseases?
●​ Lung cancer has a high preventability at the early stages but not at the later stages
(i.e. make people stop smoking, they won't get the cancer. If they stop after they have
the disease stopping won't make a big difference)
●​ But CHD has a long development time therefore there is a high preventability scope
at all stages (lots of opportunities for prevention throughout all stages)
We know about high BP effects on disease risks because there have been many cohort
studies carried out to show this

●​ Define and compare different types of prevention (primary, secondary, tertiary)

Disease prevention can be classified into Primary, Secondary and Tertiary

Primary Secondary Tertiary

Disease No disease is present Early disease is present Established or late

Status disease present

Detail - Remove the causal - Encourages early Ensures effective

exposure identification of disease treatment
- Enhance the host ~ early patient-led
resistance presentation and = good medical
- Interfere with the undertake screening practice
disease pathogenesis + early preventive and
treatment measures

Aim To prevent the To ensure treatment is To prevent

development of given in early stage of complications, limit
disease (ie reduce the disease to cure it or disability and distress
incidence of disease) delay the progression

Why is it difficult to distinguish between primary and secondary prevention?

●​ Difficult if the disease has a gradual developmental phase
●​ Whether or not the risk factor is considered a disease in its own right (e.g. high BP)

Why is it difficult to distinguish between secondary and tertiary prevention?

●​ Difficult to separate if the disease is a continuum
●​ It's hard to see when its early or established
(Key separation is between primary and secondary prevention, before/after disease
develops)

●​ Discriminate between the concepts of population and high-risk strategies for the
primary prevention (the ‘prevention paradox’) of coronary heart disease and be able
to list their strengths and limitations

2 methods of primary prevention:

●​ High risk strategy
○​ Identify people with high levels of risk factors
○​ Then offer advice/treatment or preventative measures to reduce the chance of
them getting disease
●​ Population strategy
○​ No identification
○​ Reduces whole population risk profile by a population wide change

These strategies start from the distribution of causal factors (e.g. blood pressure) in the
population.

I love making tables now omg

 High risk strategy Population strategy

✅ ADVANT
AGES
-​ Intervention/treatment is tailored to
the individual ~ appropriate and
-​ Radical and aims to reduce disease
incidence
effective -​ Everyone benefits
-​ Intervention may be cost effective -​ Behaviourally appropriate ~
-​ Motivation of subject and doctor is everyone acts the same
usually high -​ Does not require screening or
-​ Avoids interference with low risk identification of a high risk group
population -​ Does not depend on medical care
services which is more expensive

❌ DISADV
ANTAGES
-​ Involves risk factor screening -
imperfect prediction + costly
-​ Only offer a small benefit so there
may be poor motivation
-​ Limited efforts to interfere with the -​ Measures may need to be complex
incidence of disease - not a radical -​ Risk ratio needs to be assessed if
strategy, more of a palliative one its worth it
-​ Often behaviourally inappropriate - -​ If the strategy involved adding in a
singling out a small subset of the new factor with potential harmful
pop who have to behave differently effects, the balance of benefit and
from others harm would be narrow
-​ Not useful in disease widely
distributed and a significant
proportion of people with low risk
factors still get the disease

Which primary prevention method is better and more effective?

●​ The choice of preferred method depends on how disease risk is distributed at
different levels of risk factor
●​ So if most cases of CHD are occurring at extremely high BPs then a high risk
strategy
●​ But if most are occurring at lower or average BP then you would favour the
population strategy

Rose paradox:
It suggests that the majority of the cases who develop a disease, come from a population
who had a low or moderate risk factor rather than a high one (because of the population
distribution)
E.g. most CHD cases occur at moderate BP levels despite highest BP is associated with the
highest risk of CHD

Implication: In a situation where most disease cases (CHD) occur at moderate risk factor
levels, population strategy often more effective than a high-risk strategy
 ●​ Interpret individual and population risks when assessing the scope for prevention

To assess all of these measures, first need to define a 'normal’ (low risk) group which is not
exposed to cause.

Relative risk: strength of causality

Attributable risk (/excess risk): impact of cause on individual
Population attributable risk:overall amount of disease risk in a population associated with
a particular cause
●​ Attributable risk x proportion of population exposed
●​ It gives you the attributable risk for the whole population in a specific community but
cannot be used for other populations unless the factors are the same

Population attributable risk fraction: proportion of all disease in the population associated
with that cause
●​ PARF = (population attributable risk/risk in whole population) x 1000
●​ Risk in whole population = (risk in exposed x proportion exposed) + (risk in
unexposed x proportion unexposed)
TUTORIAL: How do we communicate and
interpret statistical risks to patients
●​ Recognise the importance and scope for prevention of coronary heart
disease and other chronic diseases
Context of risk comm:
1.​ Timing
2.​ Seriousness of situation (e.g. cancer Tx or screening)
3.​ Complexity of situation
 4.​ Consequences of intervention i.e. size of beneficial consequences vs. risk of
harmful consequences
5.​ Availability of evidence

●​ Define and compare different types of prevention (primary, secondary,

tertiary)
●​ Discriminate between the concepts of population and high-risk strategies
for the primary prevention (the ‘prevention paradox’) of coronary heart
disease and be able to list their strengths and limitations
●​ Interpret individual and population risks when assessing the scope for
prevention

LECTURE 6: Screening and secondary

prevention
●​ Describe the levels of prevention and outline the principles of screening
Screening: The presumptive identification of unrecognised disease or defect by the
application of tests, examinations or other procedures which can be applied rapidly.
-​ Sorts out apparently well persons who probably have a disease from those who
probably do not
-​ A screening test is NOT intended to be diagnostic.

✅
Prevention: modifying or removing risk factors that are causally related to a disease

❌
Early detection
Individuals without disease have to go with further diagnostic procedures if they have a
false positive

Primary prevention
Aim: remove cause of disease
Examples: avoiding smoking = reduces lung cancer incidence, improve nutrition/vaccination
= prevention of infectious disease, removal of harmful agent (lead in paint) = no deaths from
it

Secondary prevention
Aim: early detection of preclinical disease, identifies high risk individuals
Example: prediabetes (pts do a blood test and if they are on the border then lifestyle advice
given/small amount of med to prevent diabetes)
 ●​ Define measures of screening test performance and demonstrate how they are
determined from a 2x2 table

How can you interpret a 2x2 table of a screen test performance

●​ If person has the disease and are screened with a positive screen result = TRUE
POSITIVE
●​ If person has the disease and are screened with a negative screen result = FALSE
NEGATIVE
●​ If person does not have disease but screen result says positive = FALSE POSITIVE
●​ If person does not have disease and screen result says negative = TRUE NEGATIVE

Odds of being affected with a positive result: No affected individuals / No unaffected

individuals that did screen

Detection rate (AKA sensitivity)

= prob of ppl WITH THE DISEASE, test positive

Calculation:
Number of affected positive screen ppl/ no. of affected +ve screen ppl + no. of affected -ve
screen ppl
False Positive Rate

= prob of ppl without the disease, testing positive

Calculations: No. of unaffected +ve screen ppl/ no. of unaffected +ve screen ppl + no. of
unaffected -ve screen ppl

False Positive Rate = No. false positive / Total No. unaffected

Specificity

= 100% - false positive rate OR (no of unaffected -ve screen pts / total no of unaffected pts)
x 100
Expressing consequences:

Positive predictive value

= No. of affected ppl with +ve screen / total no. of ppl with +ve screen results

Negative predictive value is just the complete opposite [d/(c+d)]

interpret the test characteristics of screen test performance assessment

●​ Positive predictive value

●​ Odds of being affected given a positive result
●​ Negative predictive value

●​ Recognise the requirements for a worthwhile screening programme

Using example of cervical screening
-​ Well defined disorder (well defined medically): Cervical cancer
-​ Prevalence (known and of public health importance): Very common cancer in women
 -​ Natural History (possible to identify early disease from healthy): HPV positive but no
cancerous cells
-​ Treatment: removal of these cells early on
-​ Test: Smear test at GPs
-​ Test performance: looks at HPV presence and also cell cytology to check for
abnormal cells BRUV
-​ Access (all ppl who could benefit should have access): GP surgery
-​ Financial (cost effective): LOL, smears are COSTLY but yh i guess its good if it
prevents ppl from getting cancer as management of that is even more costly

●​ Interpret measures of test performance and explain why knowing the prevalence of
the condition being screened for is important
Knowing prevalence is important as it indicates what type of prevention should be used

Thank you supa star ⭐️🌟💫✨⭐️🌟💫✨

LECTURE 7: How do we identify causes of

disease? (Part 2)
Kind of a recap lecture

●​ Give examples of Bradford-Hill’s criteria for causation

Bradford-Hill
●​ Suggested a set of criteria to determine evidence of a causal relationship between a
presumed cause and an observed effect
●​ 9 original criteria, they have been developed and refined
●​ Guidelines rather than a complete checklist

Modified version:
Criteria of strength and association:
●​ The larger an association between exposure and disease, the more likely it is to be
causal
●​ Less likely due to chance, bias, or confounding

Criteria of consistency:
●​ Do multiple epidemiological studies using a variety of locations, populations, and
methods show a consistent association? If they do the association is more likely to
be causal

Criteria of specificity:
●​ Associations more likely to be causal when they are specific - meaning the exposure
causes only one disease

Criteria of independence:
●​ Not original part of Bradford Hill criteria
●​ Is the association independent of known confounding variables?

Criteria of temporality:
●​ Exposure must precede the onset of disease
●​ Arguably an essential criteria

Criteria of biological gradient:

●​ If increased exposure results in increased incidence of disease (dose-response
relationship)

Criteria of plausibility:
●​ Is it biologically plausible?
●​ Are the mechanisms well understood?

Criteria of experiment (reversibility):

●​ Does disease risk decline following an intervention or cessation exposure?
●​ Is the risk reversible?

●​ Describe how we can assess epidemiologic evidence for potential causal risk factors
for disease

Fundamental variables in epidemiology: CORE

-​ Exposure: what we are interested in studying as a potential risk factor or cause of an
outcome
-​ Outcome: the variable (usually undesirable) we want to understand better so we can
(usually) reduce it in the population
-​ Confounder: not of primary interest here - we need to consider it because it is
associated with both the exposure and outcome
-​ Risk factors: Any attribute, characteristic or exposure of an individual that increases
the likelihood of developing a disease or injury
Difference between a risk factor and a causal factor:
●​ not all risk factors will have a causal effect
●​ e.g. age has a major effect on CVD but we cant say age causes CVD

What is a statistical association?

●​ Level of one variable is somehow related to the level of another

Causal factors: Causation is a key element of epidemiology BUT no consensus definition

exists

The "Causal Pie" model

But that's easier said than done
●​ Sometimes there is no single cause or
●​ Causes may interact
●​ Disentangling them almost impossible
●​ Causality may also be two-way
LECTURE 8: Benefits and Risks of
Postmenopausal hormone replacement
therapy (HRT)
Two main hormones used in HRT are:
●​ Oestrogen – types used include estradiol, estrone and estriol
●​ Progestogen – a synthetic version of the hormone progesterone, e.g.,
dydrogesterone, medroxyprogesterone, norethisterone and levonorgestrel

HRT involves either taking both of these hormones (combined HRT) or just taking oestrogen
(oestrogen-only HRT).
Most women take combined HRT because taking oestrogen on its own can increase your
risk of developing endometrial cancer. Taking progestogen alongside oestrogen minimises
this risk.
Oestrogen-only HRT is usually only recommended for women who have had hysterectomy

●​ Recall the principles and methods of evidence synthesis

Method: Design of systematic review (formal evidence synthesis)

●​ Define the primary exposure and outcome
●​ Define the measures of exposure or intervention, outcome or effect
●​ Exclusion and inclusion criteria for the systematic retrieval of relevant studies
●​ Detail the method of extracting the relevant quantitative and / or qualitative
information
●​ Summarise the evidence using the appropriate statistical methods
 ●​ Interpret the results acknowledging systematic error (bias) and random variation

Principles:

●​ To summarise and simplify a complex and vast literature

●​ Increase precision of effect estimates and improve statistical power
●​ Assess the consistency of the evidence and thereby the generalisability of the
findings
●​ Establish possible explanations for any inconsistencies
●​ Highlight areas of relative certainty or deficiencies in evidence

●​ Interpret published systematic reviews and meta-analyses

●​ Describe the potential health benefits and harms of postmenopausal hormone
therapy
Benefits:

●​ Decreased risk of fractures

●​ Increased BMD

Risks:

●​ Increased risk of thromboembolisms

●​ Increased risk of breast cancer
●​ Increased risk of endometrial cancer

●​ Contrast the consistency of evidence from reviews of observational vs

experimental studies
LECTURE 9: Population Interventions - The
Folate Flour Story

●​ Perform a critique of published research using a recommended checklist

●​ Be aware of the limitations of interpretation in relation to study design and conduct

●​ Recognise the difficulties in influencing public health policy

LECTURE 10: Early life exposures for disease

prevention
●​ Define life-course epidemiology and be aware of early life determinants that influence
health in later life
●​ Reflect on the difference between coeliac disease and gluten intolerance
●​ Recognise the association between age of introduction of wheat into the infant diet
and the likelihood of developing coeliac disease
●​ Critically assess and appraise the evidence around how the dose of gluten
consumed in infancy might contribute to the prevention coeliac disease