Clinical trials are studies that compare the effectiveness of two or more treatments. They are important for determining if a new treatment is better than no treatment, an old treatment, or a placebo. Key features of clinical trials include randomization of patients, use of controls, appropriate sample size, blinded assessment, and intention-to-treat analysis. Proper design and conduct of clinical trials can limit bias, but biased interpretation of results remains a risk.

1. Clinical Trials Methodology
Professor Neil Poulter
BMS Lecture
18th January 2018

2. Studies of association
1. Clinical observation case
series
2. Vital statistics trends
ecological
3. Animal-experimental
4. X-sectional
5. Case-control observational
6. Cohort
7. Trials

3. Hierarchy of data
Clinical observation
Cross-sectional study
Case-control study
Cohort study
RCT - vs active
- vs placebo

4. CLINICAL TRIALS
Studies comparing and assessing the
effectiveness of >2 treatments (drugs/methods
of care/intervention)
Uses
1. Is Rx better than no Rx?
2. Is new Rx better than old Rx?
3. What/how common are S/E?
4. ‘Proof’ of causation?

5. BASIC PLAN
Study sample
Randomize
Treatment A Treatment B
(?old Rx/placebo
= control)
BLIND
Assess results of Rx A + B
COMPARE

6. CLINICAL TRIALS
Design: critical features
Objectives
Patients selection
Controls
Study size
Unbiased data collection
Specific design
Ethics
Analysis

7. OBJECTIVES
specific
incl. power
1 + 2
define
- entry criteria
- disease etc
- measures of response
- end points

8. PATIENT SELECTION – IS IT
REPRESENTATIVE?
Target population
Study population generalise?
Study sample

9. PERCENTAGE SCREENED WHO WERE
INCLUDED IN TRIALS OF
HYPERTENSION OF THE ELDERLY
MRC 3.5%
SHEP 1.0%
STOP-HT 3.1%
C & W 8.2%

10. PATIENT SELECTION
Is it biased?
- randomisation
- after entry
- investigator blind

11. BIAS
A systematic error in design,
conduct or analysis of a study which
produces a mistaken estimate of an
exposure on the risk of disease

12. PATIENT SELECTION
Advantages of randomisation
Validates statistics
Excludes biased allocation
Equally distributes prognostic factors?
–known
–unknown
- need stratification?
- restricted/block randomisation
nb: how?

13. CAPPP Trial: baseline data
Characterisation ACE
(n = 5492)
Conventional
(n = 5493)
SBP mmHg 161.8 159.6
DBP mmHg 99.8 98.1
Previous MI / IHD 104 136
Previous CHF 19 10
DM 309 263
Lancet 1999

14. Drug treatments
ANBP2
Values are % of patients
At randomisation At end of study
ACE
inhibitor
Diuretic ACE
inhibitor
Diuretic
n=3044 n=3039 n=3044 n=3039
Therapy including
allocated treatment
83 83 58 62
Monotherapy on
allocated treatment
82 83 38 38
No drugs 16 15 4 3

15. CONTROLS – WHY BOTHER?
Placebo effect
Regression to the mean
Acclimatisation
Seasonal effect
Basis of study qu?

16. REGRESSION TO THE MEAN
Any intervention at bad times will be
followed by improvement!
Mean
Distribution
2nd screen
1st screen
Mean of 1st and

17. CONTROLS
- are the basis of the comparison
Without them there is no reason to
assume that the observed effect is
due to the intervention.

18. eg 1 TOMHS
Rx
Total chol. TG’s
Acebutolol - 11.9 - 10.1
Amlodipine - 6.7 - 26.3
Chlorthalidone + 3.2 - 16.2
Doxazosin - 12.9 - 37.7
Placebo - 5.7 - 26.3

19. e.g. 2 TOMHS - S/E Prevalence
Acebutolol 16.0%
Amlodipine 18.7%
Chlorthalidone 16.8%
Doxazosin 21.3%
Placebo 19.1%

20. SAMPLE SIZE: TYPES OF ERROR
Truth
Study
Yes No
Yes -  or Type 1
No ß or Type II
Power = 1 - ß (%)
-

21. SAMPLE SIZE
Need to know:
Objective?
Difference to be detected (minimum
important)
Significance of obs. difference =  (p)
Confidence by which a-ve results is
genuine = ß (1-power)
Variation of data (SD)

22. A CALCULATION FOR SAMPLE SIZE
N = 3.62 .  2 X 2
m
 = SD : rough estimate
= max-min
4
m = min. value important
3.6 = variable power
p-value
[
[

23. Bias in RCT’s
Due to:-
Sponsorship
Study patient selection/allocation
Prejudice of patient
Prejudice of observer
Faulty method
Faulty analyses
Faulty interpretation

24. BIASED ERRORS
distortion
n - no good
- worse? – credibility

25. Sponsorship and RCT’s
1. Critical requirement
2. Data ownership
3. Design issues
4. Independent analyses
5. Independent reporting

26. Risk of venous thrombosis with
third generation contraceptives
3.0
2.0
1.0
No Yes
Funding
Vandenbroucke JP

27. ‘BLINDNESS’ – DOUBLE/SINGLE
Pros
Doctors knowledge bias - gone
Patient’s knowledge bias - gone
Withdrawal bias- gone
Cons
Impossible?
Ethics?
‘Cost’ – complex
Titration

28. Trial Conduct

29. ABCD: ‘Additional follow-up’
“During the remaining year
of the study, a private
detective identified 6
additional documented non-
fatal MI’s!”
Schrier et al, NEJM 2000

30. ABCD: The Detective Impact
Before After
End point CCB ACE CCB ACE
MI 25 5 27 9
CHF 6 5 8 10

31. BIAS IN ANALYSIS
Must consider withdrawn subjects
- after entry to study
- due to treatment failure?
- due to S/E
Intention to treat
- BUT - what is the real effect?
- use of on treatment analyses?

32. Intention-to-treat analysis?
Withdrawals from randomised treatment?
What else are they taking?
%
IPPPSH 26
MRC 40
EWPHE 26
Australian 34
SHEP 37
STOP 20
MRC elderly 54

33. CCU vs. MOUNTAIN CLIMBING AFTER AMI
200
patients
100 CCU 100 MC
18
dead
82
home
1
dead
90
lost to follow-up
Survival rate = 82%
9 survive
Survival rate = 90%

34. Primary Result
ANBP2
ACE Diuretic
Events Rate Events Rate HR (95% CI) p
All Cardiovascular Events
or Death (Any Cause)
695 56.1 736 59.8 0.89
(0.79,1.00)
0.05
First Cardiovascular Event
or Death (Any Cause)
490 41.9 529 45.7 0.89
(0.79,1.01)
0.06
Death (Any Cause) 195 15.7 210 17.1 0.90
(0.75,1.09)
0.27

35. Hazard Ratio (95% CI) p ACE better Diuretic better
0.2 1.0 5.0
All CV Events or Any Death 1.00 (0.83,1.21) 0.98
First CV Event or Any Death 1.00 (0.83,1.20) 0.98
Any Death 1.01 (0.76,1.35) 0.94
Primary Result - Females
ANBP2All events

36. Bias in Interpretation

37. The Lancet – Saturday 13 August 1988
RANDOMISED TRIAL OF INTRAVENOUS
STREPTOKINASE, ORAL ASPIRIN, BOTH, OR
NEITHER AMONG 17 187 CASES OF
SUSPECTED ACUTE MYOCARDIAL
INFARCTIONS: ISIS-2
ISIS-2 (SECOND INTERNATIONAL STUDY OF
INFARCT SURVIVAL) COLLABORATIVE GROUP*
ISIS-2 Collaborative Group. Lancet 1988;2:349–60

38. Astrological star sign and vascular death
ISIS-2, Days 0–35
Gemini/Libra
0.5 1.0 1.5
Steptokinase better Placebo better
Odds Ratio (95%
CI)
Sever PS, et al. Lancet. 2003;361:1149-1158.ISIS-2 Collaborative Group. Lancet 1988;2:349–60
25% SD4
odds reduction
Other birth
signs
All patients
Gemini/Libra
0.5 1.0 1.5
Aspirin better Placebo better
Odds Ratio (95%
CI)
23% SD4
odds reduction
Other birth signs
All patients
Area of squares is proportional to the
amount of statistical information

39. HANG-GLIDING ACCIDENTS
BMJ 1987; Feb 18; 400-4-2
Results:
73% occurred between 11am - 3pm
Recommendation:
Avoid flying between 11am - 3pm

40. CORONARY DRUG PROJECT
Clofibrate treatment group
Adherence n Deaths/1000/5 years
< 80% 357 22.7 p<0.01
80%+ 708 15.6
NEJM 1980;303:1038-41

41. CORONARY DRUG PROJECT
Placebo treatment group
Adherence n Deaths/1000/5 years
< 80% 882 26.0 p<0.01
80%+ 1813 15.1
NEJM 1980;303:1038-41

42. Association between adherence to
placebo and mortality
Simpson SH. BMJ 2006;333

43. In the MRC trial, 850 people
had to be treated for 1 year
to prevent 1 stroke

44. MRC TRIAL: PROBLEMS OF
INTERPRETATION
1. ‘Healthy volunteer’ effect
2. Baseline BP level
- mean diastolic: 98 or 91mmHg?
3. Active treatment of placebo group
- random?
- 18% > 110mmHg
4. Drugs: Composition and dosage

45. In the MRC trial, which dealt with
mainly middle-class people (who have
less than the usual levels of other
cardiovascular risk factors) whose
average DBP was 91 mmHg and in
which treatment was given to the most
endangered control subjects, 850
people had to be treated for 1 year to
make a difference of 1 stroke between
the treated and control groups.

46. X-OVER STUDY
RA
RA
RB
RB
wash-outNB: • Carry over
• Order effect
• ‘n’ = smaller
• BP effects (short term)
• Difficult/no baseline
• Within subject . . variation
.

47. PARALLEL STUDY
RA
When?
Big numbers
Curative
Long R effect
Surgery
Carry over effect
Long duration
RB

48. ETHICS
No unnecessary risk
Benefit?
Placebo?
Committee approval
Informed consent
Entry criteria
Withdrawal criteria
Stop the study criteria

49. CLINICAL TRIALS
Learning Objectives
To be able to define and appreciate the uses of clinical trials.
To appreciate those features of a clinical trial which differentiate it from
other non-experimental study designs
To appreciate the effects, benefits, and logistics of randomisation in clinical
trials.
To appreciate the importance of sample size on the statistical power and
interpretation of clinical trials.
To understand the relevance of patient selection to the generalisibility of
results and the importance of a control group.
To appreciate those features of a clinical trial which should be considered
when evaluating a published report of a clinical trial.
To appreciate the role and importance of blinding (or masking) of
interventions and/or endpoint evaluations.
To appreciate the differences and uses of parallel and cross-over designs.
To appreciate the reference for intention to treat analysis over ‘on-
treatment’ analysis.
To appreciate the importance of the unbiased accurate and scientific
interpretation of results, including evaluation of false positive and false
negative results.