nov 11 williamtigbe ureca quantitative research by xV3WmZE


									The role of different types of
  evidence – Quantitative

        William Tigbe
  Learning objectives:
At the end of this session you should be able to:
recognise and understand different types of
  quantitative research design
understand the role of different study designs in
  terms of evidence
group work: identify which study designs are
  appropriate to answer specific research questions
understand criteria for causality
What is quantitative evidence?
 investigate disease aetiology
    exposure : outcome relationship/ association
    Interested in causal understanding
 Identify risk factors or protective effects
    quantified in some way
    absolute / relative risk (or protective effect)
 evaluate health / disease / treatment and
  prevention strategies
 What kinds of study? what
 kinds of research design?

primary vs. secondary research studies
experimental vs. observational methods
descriptive vs. analytic
prospective vs. retrospective
types of study design - features, advantages
 and disadvantages, strengths and weaknesses
    Primary studies:
primary studies
  collect & analyse original data
  routinely collected data
types of study designs
  RCTs, cohort, case control, cross-sectional studies
  vast majority of published studies in journals,
    general and specialised
Secondary studies:
use data collected for primary studies
  review the results (review validity of methods)
  appraise, combine results, synthesise evidence
  re-analyse data
review - choose papers and re-appraise
systematic review
meta-analysis – systematic review plus
 quantitative methods → pooled estimates
  Study designs
 Routine data
 Observational
  Case Control
 Experimental
  Randomised controlled clinical trials
Epidemiological Study

 Observational Studies           Experimental Studies

Prevalence Surveys
                         Analytical Studies

Cohort Studies                     Case-Control Studies
    Study designs
    Is there an Association?
 Routine data (studies of group characteristics)
    Time-trends

    Ecological (fallacy)

 Observational studies (of individual characteristics)
    Cross-sectional (exposure/outcome simultaneously assessed)

    Useful to raise hypothesis, do not prove causality!!!
    Study designs
    Is the Association Causal?

 Observational studies (of individual characteristics)

   Cohort (prospective, longitudinal, incidence of outcome)

   Case-Control (retrospective assessment of exposure)

   May suggest causality/provide the foundation for RCT
Study designs
Assessing the efficacy of
preventive/ therapeutic measures

 Experimental
   Randomised controlled trials (RCT)
   Testing new drugs
   Testing changes in lifestyle (primary prevention)
   Testing public health interventions (primary prevention)

     Considered the gold standard to prove causality
Observational studies:
features of observational studies:
  no researcher controlled intervention
  simply observe and record what happens
  observe disease / condition status
  measure exposure / determinant / attribute
  follow people over time, or look back in time
possible biases
   Types of Observational
cohort (group of people)
  identify a cohort, defined by exposure
  follow over time, compare outcome
case-control study
  identify cases (with outcome of interest), identify
   controls (no outcome)
  look back to compare exposure(s)
cross-sectional study
  identify cohort, collect all data at same time
Are observational studies
can’t always do experiments
  ethical, practical reasons, inappropriate
can investigate long-term effects
  e.g. cohort studies (Whitehall Civil Servants,
   Doctors and smoking, USA Framingham study)
retrospective investigations possible
  identify cohort retrospectively and follow up to
   present time
  case-control study - identify events and look back
potential for bias in observational studies
  difficulty of identifying good comparison groups
  impossibility of ensuring random distribution of
   known (and unknown) factors
  blinding often not possible, no control over
   exposures, lack of scientific control
philosophical objection
  inference is only possible with controlled
   experimental study designs
  evidence from observational studies is too weak to
   be of any use
Cohort studies:
define cohort
  comparison groups identified by exposure
  follow over time (prospective)
  record events of interest (outcomes)
usually interested in one exposure only
  different doses of same exposure
can investigate more than one outcome
  environmental exposure may be associated with
   more than one outcome
  e.g. radiation exposure, smoking and cancers
Cohort studies advantages:
temporal relationship clear - causation
 studies (exposure --- outcome)
can get direct RR estimate, and incidence
multiple outcomes
rare exposures - define cohort by exposure
common diseases - numbers of cases
unbiased measure of exposure
prospective studies long term, good for
Cohort studies disadvantages:
need for long follow-up periods
  loss of cohort members to follow-up
  endurance test for researchers & cohort members
poor for rare diseases
  since cohort exposure based, recruitment slow
poor for multiple exposures
possibility of biases
  selection bias
  misclassification of outcome
Case-control studies:
essential feature is retrospective nature
  identify cases of disease, disease specific
   complications, death due to a specified cause
  identify control group without disease /outcome
  look back over time for exposures
  investigate causality
investigate one outcome (defines groups)
investigate many exposures / causes
rare diseases / outcomes - defines the cases
Case-control studies advantages:

good for rare diseases - groups identified by disease
investigate multiple exposures
cheap (and quick) studies
retrospective, data may be routinely available
minimum outcome bias
Case-control studies disadvantages

can’t measure incidence, or RR directly
weaker evidence than cohort studies
difficult to select acceptable control group
exposure assessment bias (recall bias)
chance of reverse causation (if exposure is a
Cohort studies
             Count events and p-yrs


             Count events and p-yrs
Case-control studies

Exposed?                 Case


   Population concept of a
   case-control study

• Find Cases ( CHD )
• Define Controls ( C )                                                C
• Compare and contrast
  case and control groups                   C         C
  on potential casual factors
                                    CHD     CHD




   Source: Bhopal R. Concepts of Epidemiology: an integrated introduction to
           the ideas, theories, principles and methods of epidemiology. 2002
  Cross-sectional studies:
both exposure and disease outcome are assessed
disease cases are prevalent cases (prevalence
cannot establish temporal relationships between
 exposure and disease
helpful to assess burden of disease
suggest associations (odds ratio)
possibility of several biases (selection, reverse
 causation, confounding effect)
Experimental study designs:

researcher manipulates environment
  to test an hypothesis, compare treatments
  control other factors, measure treatment effect
Randomised controlled trials (RCTs)
random allocation to treatment or intervention
  this is what distinguishes the RCT method
  neither clinicians, participants nor investigators
   should be able to interfere with this process
  randomisation gives all study participants the
   same chance to be in each study group
Theory – random distribution of measured
 and unmeasured factors (possible
 cofounders) in both groups – only difference
 is treatment

keeping randomisation status secret
  from patients (single blinded)
  from patients and clinician (double blinded)
  from patients, clinicians and researchers
is this important?
  psychological effects in patients
  allocating patients to “wrong” group (breaking
   randomisation) by those treating patients
  recording bias by researchers
RCTs are good for:
Comparison of effectiveness of alternative
 treatments/preventive interventions
  new vs. old drug
  new surgical procedure
  new preventive intervention
Health Services Research
  evaluation of new service delivery strategies
“Gold standard” for evaluation – high grade
Weaknesses / problems with
practicalities of RCTs
   recruiting patients to randomise
   preserving blinding (reduce researcher biases)
selected groups enter trials
   who actually gets randomised?
   often small percentage of total patients available
poor external validity (generalisability)
   results applicable to highly selected groups only
   transferring trial results into clinical practice (EBM)
Exercise – in your groups:

Quantitative Studies - Identification of
 appropriate designs

In the following research questions
  which designs would work?

   is there a “best” study type for each?
1. What is the risk of developing coronary
   heart disease in people exposed to
   second-hand smoke (SHS)?
2. Does vitamin D insufficiency increase the
   occurrence of seasonal affective
3. What are the risk factors for binge
   drinking among teenagers and young
4. In the UK, what is the prevalence of
   prostate cancer in men aged 55 – 65?
Different types of evidence –
Quantitative Research
number of different research designs
  different features
  advantages / disadvantages
  all → information
what about evidence?
  can we treat all information gathered equally as
  study designs, strengths and weaknesses
  do these features have an impact?
What are the threats to
validity of research results:
  minimise/eliminate bias in the study design
  can’t analyse out bias
play of chance
  can’t design this out
  use of appropriate statistical methods
  measure confounders, analyse effects out
  Synopsis of Different Study Designs

           Cross-sectional    Case-control     Cohort           Trial

Timing     short-term         short-term       long-term        variable

Exposure simultaneous         retrospective    starting point   intervention

Outcome    simultaneous       starting point   prospective      prospective

Costs      variable           usually low      high             high

Ethics     standard           standard         complex          complex

Strengths burden of disease rare disease       temporality      gold standard
                                                                for causality
Issues     confounding/      recall bias       selection bias   selection bias
           reverse causation
From Association to Causation:

Clinical / ecological / experimental observations

                 Available Data

             Case-Control Studies

                Cohort Studies

              Randomised Trials
Bradford Hill ‘Criteria’ for Causality:

• Temporal relationship
• Strength (relative risk, odds ratio)
• Dose-response relationship (some
• Replication of the findings
• Biologic Plausibility
• Reversibility (cessation of exposure)
Hierarchy of evidence:
Systematic reviews
1 well designed RCT
2 Cohort study
4 Case control study
5 Cross-sectional study
BUT – this is a guide only – need to critically
 appraise all – how valid were study methods?
Research designs
  primary vs. secondary
  experimental vs. observational
  cross-sectional, case-control, cohort studies
features, strengths and weaknesses
Study design books:

Pocock SJ. Clinical trials: a practical approach. Wiley
 (Chichester) 1983

Jadad A. Randomised controlled trials. BMJ Books
 (London) 1998

Gordis L. Epidemiology. W.B. Saunders Company

Bhopal RS. Concepts of Epidemiology. Oxford
 University Press 2002

Dr William Tigbe
HSRI – WMS – MSB room B146
Tel 02476150539

To top