Ali Khalid

Co-President of the cardiology and

cardiothoracic surgery society
§ The Studies pyramid including summarised !

§ Interpretations of statements with IRR, RR– Commons !

§ All formulas you need to know

§ Funnel plots and Forest Plots

§ Important Definitions
§ Calculations

§ Levels of prevention
§ Questions Time
SURVEYS – CROSS Causality - Bradford Hill’s Criteria Advantages of Disadvantages of
– SOME STUPID CHILDREN CAN’T BE survey survey
SECTIONAL
Asked TO DO REVISION CHEAP AND Can not determine
§ WHAT MAKES A GOOD • Some: Specificity association QUICK temporality
STUDY : • Stupid: Strength association
● Large sample size • Can’t: Consistency of finding DEFINE HEALTH Susceptible to bias
● Blind • Children: Coherence theory NEEDS IN Selection bias - over or
● Random allocated • Be: Biological plausibility POPULATION under-representation
● Control • Assed: Analogy of people within a
● Low follow up bias • To: *Temporal sequence* sample
● Replicated in multiple • Do: Dose response
studies • Revision: Reversibility
● Prospective study
Measurement error: accuracy RANDOM Sampling
and precision types :

● Assess accuracy - validation tests

● Assess precision - repeating
CASE CONTROL STUDY
Odds of Exposure in Cases
-----------------------
§ Case control study : A study that identifies Odds of Exposure in
individuals with an established disease and Controls
compares their exposures retrospectively.
Diseae (case ) group and non-diseased
(control) group. uses an odds ratio
§ Cohort studies are good for rare exposures

§ Case control studies are good for rare

diseases (rare outcomes) The odds of (exposed ) having (disease) is (odds
ratio ) times that of ( non diseased –control ) +
95% confidence interval ( x-y) tells you that the
odds will lie between X, Y . +
If null hypothesis= 1 is not included in the CI ,
EVIDENCE IS STATISTICALLY SIGNIFICANT ,
There is enough evidence to reject the null
hypothesis
 IN EXAM : Null hypothesis : the intervention doe
FIRST IDENTIFY EXPOSURE , not have any effect on the selected
§ Observational study , link exposure to OUTCOME sample
outcomes , to compare incidence and SECOND : CALCULATE INCIDENCE The data risk ratio can indicate
calculate relative (between exposed RATE OF BOTH EXPOSED AND higher risk associated with an
and exposed ) , risk 2 types: UNEXPOSED ( Diseased exposed / exposure BUT if the null
total exposed ) etc
§ prospective : link current exposure to hypothesis value (=1) lies within
Third calculate IRR using the formula
development of outcomes (disease ) in Forth ;use the template the CI e.g.( 0.5-2) , the data is not
future patients starts disease free people (exposed ) are IRR times as statistically significant
ALSO p>0.05 : results are NOT statistically
§ Retrospective : link past exposure and likely to develop OUTCOME than
significant :can not accept a link between the
to current outcome (disease ) those UNEXPOSED E.G people who two variables in the research question
smoke are 17 times as likely to develop lung AND VICE VERSA
cancer than those who don’t smoke. REMEMBER THAT CI tells you that we are
Relative risk /Risk Ratio RR indicates Fifth As the 95% confidence interval does 95% confident that the result lie between CI.
not overlap/include the null value of 1 it is ETA STYLE QUESTION
the likelihood of developing the disease in the statistically
DO NOT significant,
COMPAREvice
THEversa
RISK RATIO
exposed group relative to those who are not
AND CONFIDENCE INTERVALS
exposed NAME 2 OF EACH
● >1 exposure increases risk of disease THE ONLY WAY TO ASSES
● <1 exposure decreases risk of disease - WHETHER THE RESULTS ARE
protective STATICALLY SIGNIFICANT IS
● If 1 – no difference between exposed and BY USING NULL HYPOTHESIS
unexposed VALUE & CI OR P-VALUE
● calculate risk of new disease as incidence rate REMEMBER : THE SMALER
ratio THE P-VALUE , THE MORE
CONFIDENT WE ARE IN THE
RESULTS
§ Experimental , 2 groups exist: intervention
and control group , Randomisation and
blinding eliminates bias
§ Methods of randomisation : Simple randomization
, Block randomization , Stratified randomization ,
Minimized randomization
§ Allocation concealment is where the person
randomizing does not know what the next
treatment allocation will be. It happens during
the design phase of the RCT - when
participants are split into the two groups.
Blinding occurs after the design phase when
the drug/placebo is given to the individual in
a masked manner , can be single , double ,
triple depending on how many group are
blinded ( participants , clinicians ,
researchers)
§ RCT : is used to prove that intervention
(prophylactic , drug , etc) has some benifit
over the existing drug or comparison group
 NULL VALUE AND CONFIDENCE
INTERVAL GLODEN RULE :
§ STEPS • IF THE CI INCLUDES NULL VALUE
ASK(PICO) à ACQUIRE (systematic =1 , THE EVIDENCE IS NOT
reviews & meta-analysis . Using Pubmed , SIGNIFICANT ,
medline , embase ) • IF THE CI EXCLUDES THE NULL
à APPRAISE (CASP) à APPLY VALUE=1 , THE EVIDENCE IS
àASSES SIGNIFICANT
Ho = (i.e. the ratio between the risks are the same)
§ P - Patient/ Problem “Does smoking 95% CI : 95 percent of the time your results will match the results
increase the risk you get from a population
§ I - Intervention
of women
§ C - Comparison having strokes The p-value states how likely the results in the study would have
occurred by chance if the null hypothesis was true à small P
§ O – Outcome
Value (p<0.05) = strong evidence against the null hypothesis =
Bias : • Selection bias ( Sampling bias • Non- Reject the null hypothesis = SIGNIFICANT EVIDENCE
responder bias )
• Information bias • Instrument bias • Inter-observer
bias. confounding factor : An Adjusting by :
external environmental @ design stage:
Remove Bias by: Blinding , variables factor that is matching , restriction
Randomisation , Calibration of independently associated , stratification
equipment with the exposure and the @ analysis stage :
outcome but is not on the standardisation ,
causal pathway e.g. AGE , regression ,
SEX , DRINKING , stratification
SMOKING ETC
SYSTEMATIC REVIEW ADVANTAGES Systematic review is a study of
AND DISADVANTAGES primary data explicit ,
transparent and reproducible

Where do we get information from?

Variation in studies = heterogeneity. , • Primary Care reporting systems e.g. GPs,
Meta-analysis requires low Pharmacists
heterogeneity , all studies should be • Hospital information systems
similar in methodology , study design , • Laboratories
exposures & outcomes
Featueres of Meta-analysis
– Quantitative synthesis of primary data
– Summarises effect sizes and their
uncertainty –Displayed as a Forest Plot
INTERPRE
T THE The incidence rate ratio for lung cancer comparing smokers

FOLLOWI
with non-smokers is 13.

NG
The IRR for dental caries in those living in a fluoridated area compared with
a non-fluoridated area is 0.33.

In a study, if patients were followed up for 10 years and the cumulative

incidence of stroke in those who took a statin was 1.5% compared to a
cumulative incidence of stroke in those who took only placebo, 3.5%.what
is The relative risk (or risk ratio) ?
randomised 99 participants to receive falls prevention service and 102 to
receive control treatment. The incidence rates of falls per year were 3.46 in
the intervention group and 7.68 in the control group. IRR 0.45 (95% CI 0.35
to 0.58) Interpret the incidence rate ratio • Interpret the 95% confidence
interval

study found that current users of HRT were more likely than never users to
develop breast cancer (adjusted relative risk 1.66 (95% CI 1.58-1.75)
p<0.0001 , Interpret the relative risk and the 95% CI , and the P value

joinmyquiz.com
992260
Odds of Exposure in Cases
------------------------------------
Odds of Exposure in Controls
Funnel plot : looking @studies involved in publication (systeuetic review or meta analysis )
used to asses : publication bias
Each blob = study involved in systematic review .
X- axis = effect of study, vertical line = null hypothesis value (value of effects )
Y –axis =study size :the size of study included in systematic review

1- About 10-12 dots in this systematic review -----→ 10- 12 studies

included.
2- Smallest studies are @ the bottom ,largest studies @the top
3- studies are equally represented @ both sides of effects (of treatment)
either treatment works or doesn’t work .included all studies even the
smalles with insignificant finding
---→ NO publication bias = even distribution of large ,small, negative
&positive
1- there are some missing studies ,systematic review didn’t include
studies which have shown a negative effect (treatment doesn’t work)
2- if there is no bias towards publishing small trials without just having
positive effect then you should have Even distribution across the plot.
---→ Publication Bias
If there is no study saying that treatment doesn’t work ,they still need to
explain
Tutorial:
How to
read a
forest plot -
Students 4
Best
Evidence
(cochrane.
org)
 Random error - Validity - Used to
Bias - systematic error
probability that a result measure
in sampling or
occurred due to measurement
measurement
chance performance.

Repeatability –
Response rate - the Inverse Care Law :
measurements on a
percentage of the selected The availability of
different occasion
sample that takes part in good medical care
agrees. Used to
the study or survey. Used tends to vary inversely
measure to check how with the need for it in
measurement representative a sample is the population served
performance.
:Proportionate universalism
Notifiable addressing health inequalities,
Health
diseases: where universal provision of Inequalities
measles , services are combined with
mumps , targeting distribution to those
rubella most in need
 What public health action is required in
the management of an infection ?

§ Primary- Interventions designed q Cases are referred early to hospital.

to prevent the onset of a specific q Cases are reported promptly to CCDC.
Notification is a legal requirement notify the
health condition (eg) PHE OR LOCAL authorities in 3 days .
immunization and vaccination q Contacts are traced and given appropriate
chemoprophylaxis.
§ Secondary- An early intervention q Information is given to others including
that decreases the prevalence of primary care, schools/universities, and
a specific problem (eg) screening meningitis charities.
q Communication with the media is appropriate
§ Tertiary- Treatment to improve and efficient.
quality of life and reduce
symptoms after the disease has SURVEILLANCE Ongoing collection, compilation
developed (eg) rehabilitation and analysis of data to guide Public Health action. It
is the process by which we can obtain “information
for action.” The six steps are: Collect, collate,
analyse, interpret, disseminate, act.
§ Passive surveillance : A system by which a designated body (e.g. the local authority or
the PHE in England) receives reports of infectious diseases or other illnesses submitted
from hospitals, clinics, public health units, or other sources. Cheap, but with reduced
quality and completeness.
§ Active surveillance: A system whereby staff members of a designated body (e.g. the
HPA in England) regularly contact health care providers or the population to seek
information about health conditions. Relatively more expensive, but it provides more
accurate and timely information
§ Limitations of surveillance : under reporting , lack of representation , trends are difficult
to interpret , lack of denominators
§ Incident – one case of serious disease (e.g. Diphtheria, Botulism)

§ Index case – the first case to come to the attention of an investigator – not always primary
case
§ Secondary case – the case that contracted the infection from the primary case

§ Outbreak– observed number of cases greater than expected for a defined place and time
period or two or more cases with a common source
§ Pandemic - the worldwide spread of a new disease
 There have been 40 cases of
tuberculosis in Buckingham reported
in the past 5 years out of its population

CALCULATE of 1000 people. • How much is the

cumulative incidence over the 5
years? • How much is the incidence
A cross sectional study of 75-85 year
rate?
olds found that out of 397 individuals,
162 had cataract. Calculate prevalence
of cataract in this population.

Dr Smith, a GP in Bedford selected

4000 children at random from his
patient list. He found 200 of them had
asthma. What is the prevalence of
asthma in this study population?
joinmyquiz.com
992260
 REMEMBER : YOU CALCULATE INCIDENCE
RATE AND INCIDENCE RATIO IN
PREOSPECTIVE STUDIES (STARTING WITH NO
DISEASE )
YOU CALCULATE ODDS RATIO AND Odds for
§ Give examples of 3 non experimental case control studies (starting with a disease and
studies backwards )

select a sample of 27,281 healthy women, aged

between 40 and 45 and follow them up for 10 years.
Participants are asked to log how often they consume
§ soya milk, on an ongoing basis throughout the study.
What is they type of study and calculate the IR

week after a catered wedding party for 150

guests, many of the guests developed an acute
bloody diarrhoeal disease. suspect that the Duck
Liver Parfait has not been prepared according to
FSA guidelines. WHAT IS THEY TYPE OF STUDY
, Calculate OR ? Was the environmental health
correct to suspect Duck liver if CI ( 11.08-1052)
why is the CI very broad ?
joinmyquiz.com
992260
THANK YOU FOR
ATTENDING , ANY
QUESTIONS ?