Trial methodology

1
Trial methodology

• Objectives
• The randomised, controlled trial
• characteristics, definitions, advantages,
  limitations
• randomisation, placebo, blinding
• baseline comparison and confounding
• non-compliance and intention-to-treat analysis
• effect modification and subgroup effects
• Special designs
• cross-over
• cluster (exercise in infectious disease
  epidemiology)
• factorial design (exercise in nutritional
  epidemiology)
• Cochrane Collaboration
• CONSORT Statement

2
Objectives

• Efter kurset skal du kunne
• redegøre for definitionen på et
  interventionsstudie, herunder det randomiserede,
  placebo-kontrollerede, dobbelt-blinde trial (RCT)
• redegøre for styrker og svagheder ved RCT
• definere randomisering, og redegøre for hvordan
  randomiseringen gennemføres og hvordan fordelene
  udnyttes i analysen
• angive undertyperne clinical og field trials, og
  specielle designs som cross-over, cluster, og
  factorial designs, og redegøre for deres
  anvendelse
• kende til Cochrane Collaboration og CONSORT
  Statement

• Rothman and Greenland er fåmælt mht RCT
• Grimes og Schulz artiklerne gode!

3
Algorithm for classification of study
Did investigator assign exposure?
Yes Experimental study
No Observational study
Random allocation?
Comparison group?
Yes Randomised, controlled trial
No Non-randomised, controlled trial
Yes Analytical study
No Descriptive study
Time sequence?
Exposure ? Outcome Case-control study
Exposure and Outcome Cross-sectional study
Exposure ? Outcome Cohort study
Grimes DA Schulz KF, 2002
4
Clinical and field trials

• Clinical trial
• therapeutic effect in patients
• Phase I pharmakinetics and safety in healthy
  individual
• Phase II initial clinical evaluation in few
  patients
• Phase III full clinical evaluation thru RCT
  in many patients
• Phase IV post-marketing surveillance
• Field trial
• preventive effect in healthy or high-risk
  individuals
• health-facility-based
• community-based

5
Efficacy and effectiveness

• Efficacy
• the extent to which a specific intervention
  produces a beneficial result under ideal
  condition
• Effectiveness
• the extent to which a specific intervention
  does what it is intented to do when deployed in
  the field
• Thus,
• efficacy depends on biology (e.g. vaccines,
  drugs, etc)
• effectiveness depends on efficacy and compliance

Last JM, A dictionary of epidemiology, 1988
6
Rating clinical evidence

• Quality of evidence
• I Evidence from at least one randomised,
  controlled trial
• II-1 Evidence from well-designed controlled
  trials without randomisation
• II-2 Evidence from well-designed cohort or
  case-control studies, preferably from more than
  one group
• II-3 Evidence from multiple time series with or
  without the intervention
• III Opinions of respected authorities, based on
  clinical experience, descriptive studies or
  reports of expert committees
• Strength of recommendations
• A Good evidence to support the intervention
• B Fair evidence to support the intervention
• C Insufficient evidence to recommend for or
  against
• D Fair evidence against the intervention
• E Good evidence against the intervention

US Preventive Services Task Force, 1996
7
(No Transcript)
8
Counterfactual definition of risk
risk if exposed risk if not exposed
RR

9
Intervention studyAdvantages

• As a cohort study exposure ? outcome
• Complete description of diease, incl incidence
  data
• Possible to study multiple effects
• The exposure allocated by the investigator
• Allow large contrasts in exposure
• The exposure compared to no, other or placebo
  intervention
• The groups to be compared will be similar
• The randomised, controlled trial
• Randomised eliminate confounding
• Placebo-controlled allow blinding
• Blinding reduce bias
• Strongest design to establish cause-effect
  relationships

risk among exposed risk among not
risk if exposed risk if not
RR

10
Intervention studyLimitations

• As a cohort study
• Expensive and time consuming
• Not useful for rare diseases and long induction
  time
• May not be ethically justifiable
• insufficient belief to justify intervention
• insufficient doubt to justify placebo
• intervention too expensive or difficult to be
  implemented
• intervention already widespread in population
• even if little evidence of efficacy or safety
• May not be feasible to have a true placebo group
• treatment already exist
• change in habits, surgery, zone therapy, etc

11
Does a confounder confound?

• Confounding
• En spurious association between exposure and
  outcome, due to the effect of a confounder

E- O-
C
E O

• Confounder
• A predictor of the outcome, which is associated
  with exposure

12
How do we avoid confounding?

• A confounder is
• a predictor of the outcome
• This we cannot do anything about
• associated with exposure
• We can make sure it is equally distributed
  between exposure-groups!

How do we do that?
13
How do we avoid confounding?

• How can we distribute the predictor equally
  between exposure-groups?
• In the design
• Restriction we only include one level of C
  either C- or C
• Randomization we distribute participants
  randomly to the exposure
• In the analysis
• Stratification we analyse our data separately
  for C- and C
• Multivariable analysis we let our statistical
  software do the trick!

14
Randomisation

• Interventions allocated based on random principle
  
• random does not mean haphazard or alternating
• optimal tool to avoid confounding, since
• both known and unknown potential confounders will
  be balanced
• Different types of randomisation
• Simple ABAAABABBAABABBAABBBABABABBBABBAABB
• Adequate with large sample sizes
• Blocked ABBA BBAA ABAB AABB BBAA BABA BAAB
• Using permuted blocks, e.g. of 4, 10 or more,
  varying block size
• Ensures balance at any time, and control for time
  effects
• Stratified, blocked
• primigravidae ABBA BBAA ABAB AABB BBAA BABA BAAB
• multigravidae BABA BAAB BBAA ABAB BBAA ABAB AABB
• Ensures control of confounding due to to
  stratifying variable
• Check adequacy of randomisation and repeat if
  inadequate

15
Confounding in RCT

• Randomisation will ideally result in
• Equal distribution of known and unknown
  potential confounders
• No confounding possible
• Randomisation may fail sloppiness or randomly
• The stronger predictor, the smaller differences
  cause confounding
• How to judge if randomisation was successful?
• Assess for baseline equivalence
• If there are differences, then
• repeat the randomisation, or
• assess and adjust for possible confounding

16
Baseline comparison
 

• Do we have baseline equivalence?
• A. Yes, baseline equivalence - no problem
• B. Yes, reasonable baseline equivalence
  probably no problem
• C. No, baseline inequivalence confounding could
  be a problem
• D. No, baseline inequivalence there is
  confounding

 
17
(No Transcript)
18
Baseline comparison

• Does iron supplementation increase HIV viral
  load?
• Do we have baseline equivalence?

 
19
Baseline comparison
 

• Does iron supplementation increase HIV viral
  load?
• Can we conclude, that iron increase HIV viral
  load?

p0.01
20
Baseline comparison

• Does iron supplementation increase HIV viral load?

p0.01
p0.01
 
What is the problem here?
21
Non-compliance in RCT

• Study participants may not comply with treatment
• Some in the intervention group may not be exposed
• Some in the control group may be exposed
  (co-intervention)
• Always assess compliance (pill-count, biochemical
  assessment)

Randomisation
What to compare?
A
B A to B ?
A1 B1 A1 to B1 ?
A1 B1
22
Intention-to-treat analysis

• Always base analysis on intent-to-treat, i.e.
• comparison based on assigned, not actual
  treatment
• compare those assigned to active vs placebo
• then maybe repeat the analysis among compliant
  ptts only
• never put non-compliant ptt into the placebo
  group!
• Non-compliance does not occur randomly
• Only intent-to-treat analysis benefits from
  randomization

23
Allocation concealment

• Random sequence has to be concealed
• Treatment allocation should be kept unpredictable
• If allocation concealment is inadequate, then
• investigators are likely to decipher the code
  sooner or later
• investigators next recruitement could then depend
  on prognosis
• patient could be held back if allocated to
  perceived wrong treatment
• Note, the following are distinctly different
• randomisation
• allocation concealment
• blinding

24
Loss to follow-up in RCT

• Study participants may be lost to follow-up
• What is the problem?

25
Placebo and blinding

• The purpose is to reduce bias
• Blind not aware of allocation to active or
  placebo
• Open label trial neither participant nor
  investigator are blinded
• Single only participant or investigator is
  blinded
• Double both participant and investigator are
  blinded
• Triple both participant, investigator and
  statistician are blinded
• Double-blinding requires use of placebo
• Single-blinding (examinator) does not require
  placebo

26
Effect modification in RCT
27
Subgroup effects in RCT
The effects of multimicronutrients on birth size
a randomised, placebo-controlled, double-blind
trial

• 1669 pregnant (22-35 wks) Zimbabwean women were
  enrolled
• randomised to a daily prenatal supplement
• multimicronutrients (12 vitamins and minerals, 1
  RDA)
• placebo
• data on birth size were obtained from 1106 giving
  birth at hospital
• What is the main problem and how would you assess
  it?

28
Subgroup effects in RCT
The effects of multimicronutrients on birth size
a randomised, placebo-controlled, double-blind
effectiveness trial

• Results
• What to conclude?

29
Special designs

• cross-over
• factorial
• cluster

30
Special design cross-over

• All are given both treatments, but in sequence
• Randomisation determines the treatment order
• Advantages
• The effect is within- rather than
  between-subjects more power
• Disadvantages
• Cannot be used when condition is cured
• Only useful for short treatments, which can be
  assessed quickly
• Drop-out after first treatment a major problem
• Period effect systematic differences between
  periods
• Carry-over effect from first to second
  treatment (treatment-period interaction)
• Analysis requires three tests

31
Special design factorial

• Two interventions assessed simultaneously
• Different outcomes interactions not expected
• Aspirin reduces risk of cardiovascular death
• ?-Carotene reduces risk of lung cancer
• If the two interventions as assumed - are
  independent
• The effect of each intervention can be assessed
  independently

?-caroten
PHSRG, NEJM, 1989
32
Special design factorial

• Same outcome interaction may be expected
• ?-Tocopherol reduces risk of lung cancer
• ?-Carotene reduces risk of lung cancer
• If interaction
• assess the effect of one intervention for each
  level of the other
• If no interaction
• greater power or able to detect smaller
  differences

?-caroten
ATBC-CPSG, NEJM, 1994
33
Special design cluster randomised

• If unit of randomisation is a group, not
  individual
• Questions
• Where is it useful?
• What is the problem?

34
Special design cluster randomised

• If unit of randomisation is a group, not
  individual
• Where is it useful?
• If intervention targets groups
• Teaching methods, iodised table salt, etc
• Example classes or households as unit of
  randomisation
• If effect on one may impact others
• The effect of STI treatment on HIV incidence
• Example health-facility catchment area as unit
  of randomisation
• If more practically feasible
• Difficult to administer intervention to
  individuals spill-over
• Example administrative unit (ward) as unit of
  randomisation
• What is the problem?
• Individuals within groups are not independent
  design effect
• sample size is number of groups
• use group measure as unit of observation or
  control for design effect

35
CONSORT CONsolidated Standards Of Reporting Trials

• The quality of reporting of randomised,
  controlled trials is inadequate
• This leads to biased estimates of treatment
  effects
• The CONSORT Statement aims to improve the quality
  of reporting
• The Statement includes a checklist and flow
  diagram
• Many leading journals have adopted the CONSORT
  Statement

www.consort-statement. org
36
Flow chart
Assessed for eligibility (n ....)
Excluded (n ....) Not meeting inclu ... (n
....) Refused (n ....) Other reasons
(n ...)
Randomised (n ....)
Allocated to active (n ....)
Allocated to placebo (n ....)
Lost to follow-up (n ...)
Lost to follow-up (n ...)
www.consort-statement. org
37
Cochrane Collaboration

• Archie Cochrane, 1972
• "It is surely a great criticism of our profession
  that we have not organised a critical summary, by
  specialty or subspecialty, adapted periodically,
  of all relevant randomized controlled trials."
• Collaborative Review Groups
• Cochrane reviews are published electronically in
  The Cochrane Database of Systematic Reviews.
  Preparation and maintenance of Cochrane reviews
  is the responsibility of international
  collaborative review groups. The existing review
  groups cover all important areas of health care.
• Examples of topics reviewed
• Acupuncture for chronic asthmaAcupuncture for
  idiopathic headacheAcupuncture for induction of
  labour
• Acyclovir for treating varicella in otherwise
  healthy children and adolescentsAddition of
  anti-leukotriene agents to inhaled
  corticosteroids for chronic asthmaAddition of
  intravenous aminophylline to beta2-agonists in
  adults with acute asthmaAdenotonsillectomy for
  obstructive sleep apnoea in children
• Adhesives for fixed orthodontic bracketsAlarm
  interventions for nocturnal enuresis in children

www.cochrane.org
38
Cochrane review

• Background Acupuncture has traditionally been
  used to treat asthma in China and is used
  increasingly for this purpose internationally.
• Objectives The objective of this review was to
  assess the effects of acupuncture for the
  treatment of asthma or asthma-like symptoms.
• Search strategy We searched the Cochrane Airways
  Group trials register, the Cochrane Complementary
  Medicine Field trials register and reference
  lists of articles.
• Selection criteria Randomised and possibly
  randomised trials using acupuncture to treat
  asthma and asthma-like symptoms. Acupuncture
  could involve the insertion of needles or other
  forms of stimulation of acupuncture points.
• Data collection and analysis Trial quality was
  assessed by at least two reviewers independently.
  A reviewer experienced in acupuncture assessed
  the adequacy of the sham acupuncture. Study
  authors were contacted for missing information.
• Main results Seven trials involving 174 people
  were included. Trial quality varied and results
  were inconsistent. No statistically significant
  or clinically relevant effects were found for
  acupuncture compared to sham acupuncture. However
  the points used in the sham arm of some studies
  are used for the treatment of asthma according to
  traditional Chinese medicine. Only one study used
  individualised treatment strategies. Lung
  function could be compared statistically in only
  3 trials. Peak expiratory flow rate showed a
  statistically insignificant increase of 8.4
  litres/minute weighted mean difference (95
  confidence interval -29.4 to 46.2) when
  acupuncture was compared to sham acupuncture.
• Reviewers' conclusions There is not enough
  evidence to make recommendations about the value
  of acupuncture in asthma treatment. Further
  research needs to consider the complexities and
  different types of acupuncture.
• Citation Linde K, Jobst K, Panton J. Acupuncture
  for chronic asthma (Cochrane Review). In The
  Cochrane Library, Issue 3, 2003. Oxford Update
  Software.

www.cochrane.org
39
Summary

• The randomised, controlled trial (RCT) - an
  important tool
• need for evidence-based therapeutic and
  preventive interventions
• but not always practically or ethically feasible
• Special designs
• cross-over, factorial, cluster randomised
• Even data from RCT may be flawed
• inadequate randomisation and allocation
  concealment
• confounding possible baseline comparison
  important
• non-compliance intention-to-treat important
• inadequate blinding
• loss to follow-up
• subgroup effects
• Consort statement important guidelines for
  authors
• Cochrane an important ressource