Treatment allocation by minimisation
BMJ 2005; 330 doi: https://doi.org/10.1136/bmj.330.7495.843 (Published 07 April 2005) Cite this as: BMJ 2005;330:843
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
In their recent article Altman and Bland in discussing randomisation
claim that ‘The only widely acceptable alternative approach is
minimisation’. However it simply is not the case that minimisation is
widely accepted. There are two constituencies that reject it. The first is
of those who are interested in randomisation based inference. It remains
an open question as to whether this can be validly applied to minimised
trials 1,2. For this reason regulatory agencies remain wary of it, as was
pointed out by Buyse and McEntegart3 in the article cited by Altman and
Bland and subsequently confirmed by former and current regulators in the
pages of Applied Clinical Trials4. The second constituency is of those who
are interested in optimal design of trials. Minimisation is not, in fact,
the best algorithm at improving efficiency. Atkinson’s biased coin
approach is superior5, although the advantage of either to randomised
designs is slight6. Those who are interested in wide acceptability would
be advised to avoid minimisation.
1. Senn SJ. Consensus and controversy in pharmaceutical statistics
(with discussion). The Statistician 2000;49:135-176.
2. Senn SJ. Added Values: Controversies concerning randomization and
additivity in clinical trials. Statistics in Medicine 2004;23(24):3729-
3753.
3. Buyse M, McEntegart D. Achieving balance in clinical trials.
Applied Clinical Trials 2004;13(5):36-40.
4. Day S, Groulin J-M, Lewis JA. Achieving balance in clinical
trials. Applied Clinical Trials 2005;14(1):24-26.
5. Atkinson AC. Optimum Biased Coin Designs for Sequential Clinical-
Trials with Prognostic Factors. Biometrika 1982;69(1):61-67.
6. Senn SJ. Statistical Issues in Drug Development. Chichester: John
Wiley, 1997.
Competing interests:
The author consults regularly for the pharmaceutical industry
Competing interests: No competing interests
Sir,
Altman and Bland have provided a useful summary of treatment
allocations by minimisation. In such a short note they perhaps
understandably did not mention the current controversy that surrounds the
use of minimisation in the pharmaceutical industry. The controversy
arises from the fact that in 2003, the European Committee for Proprietary
Medicinal Products (CPMP) issued a guideline1 claiming that dynamic
methods such as minimization “remain highly controversial” and are
“strongly discouraged”. Professor Buyse and I provided a commentary2 in
response to this guideline and this led to a set of three further letters3
-5 including a final response from the regulators where they explained
their position in more detail. All four pieces are written with a general
audience in mind and as three of them are available on the web, interested
readers can readily inform themselves of the debate.
To summarise, the CPMP’s position5 is that essentially “the use of
minimization may result in more harm than good”. This is because in their
opinion, the technique brings little statistical benefit in moderate sized
trials and that the logistical complexities need considering with this in
mind. They have seen “situations where programming algorithms have been
incorrect, the choice of factors to include in the minimization algorithm
has been poorly thought out or where telephone systems or Web-based
systems have proved unreliable”. The guideline1 itself states that if the
technique is used, then a marketing application is unlikely to be
successful without adequate supporting/sensitivity analyses which I take
to mean analysis by a re-randomisation test rather than the conventionally
used asymptotic tests.
The counter argument2,4 is that the technique does bring benefits in
the form of increased credibility as noted by Altman and Bland. This is
especially the case for as long as regulators and journals request
sensitivity analyses to account for chance imbalances on prognostic
factors; this is something that the CPMP specifically mention in their
guideline. Provided that practitioners use a validated randomisation
algorithm and a reliable telephone/web system, any concerns about
implementation of the technique can be overcome. It should be for
practitioners to weigh up the costs and benefits associated with any
randomisation technique and decide which best suits their situation. The
current position is unsatisfactory in that the pharmaceutical industry is
discouraged from using a technique that is used without question in trials
that are not sponsored by the industry. If the technique is used, after
the appropriate justification has been made to CPMP, then again the
requirement for a re-randomisation test is an unreasonable burden that to
my knowledge has ever been requested by any medical journal.
Finally Altman and Bland mention that minimisation is the only widely
acceptable alternative technique to stratified randomisation. In my own
review6 I refer to other valid alternatives such as the urn and dynamic
hierarchical techniques. Rather the point should be that minimisation is
the most widely used alternative technique.
1. Committee for Proprietary Medicinal Products. Points to consider
on adjustment for baseline covariates. CPMP/EWP/2863/99, 2003.
(http://www.emea.eu.int/pdfs/human/ewp/286399en.pdf). (accessed 11 April
2005)
2. Buyse M, McEntegart D. Achieving balance in clinical trials: an
unbalanced view from the European regulators. Applied Clin Trials
2004;13(5):36-40
http://www.actmagazine.com/appliedclinicaltrials/article/articleDetail.j...
(accessed 11 April 2005)
3. Senn S. Unbalanced claims for balance. Applied Clin Trials
2004;13(6):14-16
http://www.actmagazine.com/appliedclinicaltrials/article/articleDetail.j...(accessed
11 April 2005)
4. Buyse M, McEntegart D. More NonSENNse about balance in clinical
trials. Applied Clin Trials 2004;13(7) :14-15
5. Day S, Grouin J_M, Lewis JA. Achieving balance in clinical
trials. Applied Clin Trials 2005;14(1) 24-25
http://www.actmagazine.com/appliedclinicaltrials/article/articleDetail.j...(accessed
11 April 2005)
6. McEntegart DJ. The pursuit of balance using stratified and
dynamic randomization techniques: an overview. Drug Information Journal
2003;37 (3):293-308.
http://www.clinphone.com/files/Stratfied%20Dynamic%20Randomization%20Tec...(accessed
11 April 2005)
Competing interests:
Provider of phone/web randomisation and medication management system. But no vested interest in which randomisation technique is used
Competing interests: No competing interests
Minimisation may be viewed parallel to stratified randomisation though not equivalent
The method of treatment allocation by minimisation may be viewed as
parallel to stratified randomisation but not equivalent. Minimisation
principle may be an alternative to stratified randomisation in certain
situations. Stratified randomisation may be good in larger study with
smaller number of stratifying variables whereas minimisation which ensures
minimising overall imbalance may be suitable for smaller studies.
Stratified randomisation is a conventional approach offering reliable
scope for the application of inferential statistics. Minimisation
allocation lacks proper randomisation which questions the validity of
results preferably obtained through parametric test procedures. It may be
safely argued that minimisation based study might prove to be valid when
tests of association between variables are studied.
If some element of randomisation is introduced in the process of
minimisation, it may affect the desirable benefits intended from
minimisation. Debate on acceptability of minimisation principles may
continue long and it may face greater resistance from explanatory
trialists. However, acceptability of a tool lies in the situation where it
is applicable. Both scissors and blade used for cutting are needed but
needed in appropriate situation.
Competing interests:
None declared
Competing interests: No competing interests