Randomization
Janet Holbrook, PhD, MPH
Johns Hopkins University
Copyright © 2013 Johns Hopkins University and Janet Holbrook. All Rights Reserved.
Section A
Rationale for Randomization
Copyright © 2013 Johns Hopkins University and Janet Holbrook. All Rights Reserved.
Experimental Design
n Randomization (Rz)
- Key feature of experiment
n Other features
- Standardized treatment
- Prospective plan
- Adverse event reporting
- Regulatory requirements
3
The Lottery Wheel
Military draft lottery, London (“The Lottery Wheel,” by W. H. Pyne, 1805)
4
Fair Allocation in Health Care
“Let us take out of the hospitals . . . 200 or
500 poor people, that have fevers,
pleurisies. Let us divide them into halves,
let us cast lots, that one half of them may
fall to my share, and the other to yours; I
will cure them without bloodletting and
sensible evacuation; but you do, as ye
know. . . . We shall see how many funerals
both of us shall have.”
— Van Helmont, 1662
Source: Silverman and Chalmers. (2001). British Medical Journal, 323, 1467–1468.
Image sources: James Lind Library. Retrieved June 21, 2006, from [Link]
17th_18th_Century/van_helmont/van_helmont_tp.html and [Link] 5
Definitions
n Random (lay)
- Having no specific pattern; lacking causal relationship;
haphazard (e.g., random acts of nature)
n Random (scientific)
- Process in which there is associated with every legitimate
outcome a probability (e.g., coin-toss, dice throw, draw
poker)
6
Introduction of Theory
n Ronald A. Fisher (1890–1962)
- Question was how to allocate plots,
not whether controls were needed
- A measurement error issue
“One way of making sure that a valid estimate of
error will be obtained is to arrange the plots
deliberately at random . . .” (1926)
“The purpose of randomization . . . is to guarantee
the validity of the test of significance, this test being
based on an estimate of error made possible by
replication” (1951)
Other methods of allocation would either over-
estimate or underestimate the random error
- Some safeguard against non-normality
Image source: Wikipedia. Retrieved June 21, 2006, from [Link] 7
Application to Clinical Trials
n Sir Austin Bradford Hill (1897–1991)
- Epidemiologist concerned with con-
founding
“In many respects the reactions
of humans beings are under
any circumstances extremely
variable … and that is where
the trouble begins”
“… and having used a random
allocation, the sternest critic is
unable to say … that quite probably the groups were
differentially biased through our predilections or
stupidities” (1952)
Image source: James Lind Library. Retrieved June 21, 2006,
from [Link] 8
First Properly Randomized and Documented Trial
n First properly randomized and DOCUMENTED trial
- Trial of streptomycin for TB in 1946
Fair distribution of risk/benefits
Fair allocation of a limited resource
Image source: CDC Public Health Image Library. 9
Rationale for Rz
n Avoids selection bias
- Prognostic factors related to treatment assignment
Confounding by indication
- Can influence outcomes as strongly as or more strongly
than treatments
n Tends to produce comparable treatment groups
- Known and unknown confounders
n Assures statistical tests (e.g., t-test, chi-square) will have valid
significance levels (e.g., p = 0.03)
n Defined time-point for trial entry
10
Rationale for Unequal Randomization
n Example: 1:1 with 95% power, 2:1 with 92%, 4:1 with 82%
- Typically less powerful
n Important to acquire as much experience with new treatment
as possible including adverse events and toxicity
n Incentive for recruitment, e.g., 2:1 patients have a 67% chance
of getting new treatment
n Cost issues
n Variance in treatment effect may be different, to optimize
power more patients in group with larger variance
11
Types of Selection Bias
n Epidemiologic definitions
- Study group association between intervention and
outcome different than association in target population
n Trial definition
- Treatment associated with prognosis
- Link broken with randomization
12