Journal of Orthodontics, Vol.

30, 2003, 337341

FEATURES
SECTION
How to do a randomized controlled
trial
K. D. OBrien, J. L. Wright and N. A. Mandall
University Dental Hospital of Manchester, UK

In this section of the series on orthodontic research we Objectives

intend to give a brief outline of how to carry out a
randomized controlled trial (RCT). The RCT is one of The first stage in planning a trial is the generation of a
the most simple and powerful research tools in which question. In this example, our question is concerned with
people are allocated at random to receive one of several whether nickel-titanium coils springs are more effective
clinical interventions. We hope that this will not only than power chain in space closure. The null hypothesis is
provide information to anyone who is aiming to carry out There is no difference in the rate of space closure with Ni-
research of this type, but also help in the interpretation of Ti springs when compared to power modules.
RCTs that have been published.
Before we consider the mechanics of how to carry out an Outcomes
RCT we should consider the advantages of this type of
research over other study methods. These are: It is important to know before you start how you are
going to measure the effect of each intervention. In our
The RCT is prospective. As a result, the subjects and study, this will be the rate of closure of extraction spaces.
the data are under some control by the investigator. It is also necessary to determine what size of difference
The treatment or intervention is randomly allocated. will indicate a clinical difference between interventions.
Therefore, the perceptions of the investigator on the
value of a particular treatment should not influence
treatment allocation that could bias the results. The study population and site of the
The study is planned before the data is collected. This study
is the important distinction between the RCT and the
We now need to consider our study population. This is an
retrospective investigation, and this results in a mini-
important step because it is important that this popu-
mization of bias that is inherent in the retrospective
lation is relevant to both the question that we hope to
study.
answer and to the provision of orthodontic treatment. It
We hope that this simple guide will stimulate efforts to is fairly obvious that the study population for this
carry out randomized trials of some of our treatment example is easily defined and may be confined to children
methods. As with most of the How to do papers in this under 16 who are wearing the same type of appliance with
series, the ideas and concepts are a simple interpretation extraction spaces that require closure. This is not as
of the work of other authors. simple as it sounds. We have to make it clear that we are
We have decided to use the CONSORT guidelines only going to include patients who have had first
(http://www.consort-statement.org/) as a framework premolars extracted. It is tempting to include all patients
for this chapter.1 These are a set of guidelines that have who have had an extraction, but this complicates the
been formulated to aid the reporting of RCTs. They also study by introducing the additional variable of tooth
help with study design. We will use these guidelines to extracted and this would need to be included in the
plan a theoretical investigation into different methods of analysis. As a result we would need to recruit a larger
orthodontic space closure, based on the study by Dixon sample. Furthermore, we should ensure that all patients
et al.,1 as an example of the type of study that can be are being treated with the same appliance type and
carried out.2 prescription. Finally, it is best to produce a list of
inclusion and exclusion criteria for the study. For our
study the inclusion criteria are:
Address for correspondence: Professor Kevin OBrien, Orthodontic Department, University Dental Hospital of Manchester, Higher Cambridge Street,
Manchester M15 6FH. Email: [email protected]
 338 K. D. OBrien et al. Feature Section JO December 2003

Children under 16 years old at the start of treatment. to investigate. This is a difficult situation, but in ortho-
Appliances will be MBT brackets. dontics it can be approached by considering the level of
Teeth extracted will be first premolars. the evidence that the perceptions of any operator are
Space closure will start one month after the placement based upon. If this is based on evidence from retro-
of 019  025 stainless steel archwires. spective studies or more commonly expert opinion, this
may also be considered to be unethical and, perhaps,
There are also exclusion criteria:
equipoise is the best place to be. Furthermore, if the
The patient has had Phase I treatment with a functional operator has a preference this may influence the way that
appliance. they enter patients into the study and could lead to bias.
The patient has a craniofacial syndrome. In our hypothetical study the interventions may be clearly
Teeth other than first premolars have been extracted. stated as:
While we need to consider the study population, we must Nickel-Titanium coil springs; or
not forget to pay some attention to the operators in the Power chain.
study. It is important that the findings of the study have
An RCT may have a treatment compared with a no
generality and the results are relevant to current ortho-
treatment or control group. Ethically, it may not always
dontic practice in the setting of care where most treatment
be possible to randomize to a control group and not
is provided. This, however, is not always possible, espe-
provide treatment to some patients. Therefore, most
cially when the RCT is investigating a new method of
RCTs in orthodontics will compare two or more treat-
treatment. As a result, most orthodontic RCTs have been
ments or interventions.
carried out in Dental Schools. This has the advantage of
being able to keep close control of the operators and
patients in the study. However, the trade-off for this Patient registration
control is the potential lack of generality. In our planned
Once ethical committee approval has been obtained, the
study, we would like to make the study results applicable
next stage is patient recruitment. This may be considered
to current practice and the operators will be selected from
in several stages.
local specialist orthodontic practitioners.

Patient requires treatment and is eligible

Sample size
It is important to ensure that patients entered into a trial
To be able to state with sufficient probability that any are representative of the population. This is achieved by
difference found between groups is likely to be due to the operator considering that all patients who s/he sees
the intervention, rather than to the particular samples with the entry criteria are eligible for the study. The
you have, it is necessary to have a large enough sample. clinician should not be selective.
Tatiana Macfarlane has written an excellent introduction
to this topic in the June 2003 edition of the Journal.3
Agreement to randomize
The clinician should be in equipoise for a patient who is
The intervention
eligible for the trial and s/he should be willing to accept
This is the treatment of interest in the study and it is vital the randomization.
for the success of the study that the interventions are
clearly stated. It is essential that the existing literature
Patient consent
does not already strongly suggest that one intervention is
better or more effective than another. Furthermore, the The patient should be informed of the theoretical risks
operators in the study should not have a preference for and benefits of the interventions under test, both verbally
any of the interventions that are being tested. This is and in writing. This allows the patient to give fully
termed equipoise. Importantly, if there is no equipoise it informed written consent before being registered and
cannot be ethical to randomize people to different inter- randomized. Ideally, patients should be given the trial
ventions (or to intervention v. control) because we information and then given at least 24 hours to consider
already know the answer to the question we are trying whether they would like to participate.
 JO December 2003 Feature Section How to do a randomized controlled trial 339

Formal entry Allocation

Details of the patient are then entered onto a log sheet The next stage is the method by which the operator finds
of the trial or, more commonly, onto a computer data- out which treatment the patient has been assigned to. It is
base. The information that is collected is frequently the essential that the operator does not know what the assign-
patients name, their hospital number, date of birth and ment will be in advance and there are several methods of
institution (if the trial is multi-centred) and date of entry. concealing this. One popular method is to transfer the
The reason for this step is that the trial organizers need randomization list to a series of sealed envelopes each
know about every patient entered. This enables them to containing the allocation on a card. The clinician then
obtain information on trial drop-outs and patients who opens the next envelope in the series when the patient
are not entered, and guards against the deviant investi- formally enters the trial. This method is particularly rele-
gators who do not give the randomized treatment. vant when the clinician registers his/her own patients.
The method of registration depends upon the setting of However, care needs to be taken to ensure that the clin-
the trial. In a multi-centre trial this is usually carried out ician does not reseal the envelope having discovered that
by the clinician making contact with a central registration the allocation was not what he/she was hoping!4
office by telephone. In a single centre study, this should The best method of allocation is to make use of a
be carried out by a person who is not a participating central registration office. In this method the treatment
clinician. However, if the trial only has one investigator, assignment is read from a prepared list and given to the
then patient registration may need to be left to the investi- investigator while still on the phone, following the
gating clinician, although this is not ideal. In this case, registration of the patient. While this method is more
it is important that care is taken that no bias is intro- expensive and requires more preparation than using
duced, for example, through the investigating clinician envelopes it does provide an almost foolproof method of
having access to the randomization, which may influence allocation.
whether they approach particular patients.4

Blinding for orthodontic studies?

Random assignment One important concept of medical studies is blinding.
The assignment of an intervention must only be made This is important because we should consider that if a
after the patient has agreed to take part in the trial. It is patient or operator knows the identity of the treatment
important that the clinician does not know in advance the results of the study could be distorted. The effect of
what the allocation will be. this is minimized by concealing the identity of the treat-
ments and by the use of placebos. Blinding may occur in
many ways, for example, blinding the patient, the
Randomization operator, the investigator who measures the outcomes
and the statisticians. However, when we consider the
This stage is central to the mechanics of the trial, because
nature of orthodontic treatment it is impossible to blind
by allocating participants randomly, patient character-
treatment allocation to both the operator and patient. As
istics are likely to be similar across the groups at the start.
a result, the only type of blinding that we can practice is
By keeping the groups balanced at baseline, the outcomes
blinding of the person who records and analyses the data.
can be attributed to the intervention with minimal effects
This is important because if, for example, the evaluator
from other factors that may influence the treatments.
knows that a group of patients have had a new treatment
The method of randomization should be decided before
then they may record outcome data in a favourable man-
the trial starts. There are many methods of random-
ner. Blinding can be done by concealing the identity of the
ization and we will not go into detail but interested
patient and the treatment allocation using numbers, or by
readers should refer to two excellent texts that cover this
having the data recorded by a different person from the
in detail.5,6 In brief, the object of randomization is
one who is going to analyse the data. If handled carefully,
to allocate one or more interventions (or control), in a
in our space closure RCT the patient, the evaluator, the
manner that ensures that the samples that you are going
data recorder and the analyst can all be unaware of the
to compare, are similar in every respect apart from the
treatment method allocated.
intervention. In most trials, a randomization list has been
prepared in advance using random numbers.
 340 K. D. OBrien et al. Feature Section JO December 2003

Monitoring progress Treatment intervention and stopping

rules
So now you have set up your trial, and you think that you
can just sit back and the trial will run, and all you have to It is important that only the treatment interventions of
do is to collect and analyse the data. Unfortunately, this is interest are carried out during the trial. In our study, the
not the case! It is essential that the progress of the trial is operator wishing to use inter-arch elastics may com-
closely monitored. Several areas should be evaluated as plicate the treatment effect of the different space closing
part of this process. The first of these is protocol com- mechanics. In this example, inter-arch elastics should not
pliance. You need to check that the study protocols are be used for the trial duration as this operator decision
being followed by the operator(s) in the study. The easiest could bias any results
way of doing this for an orthodontic study is to period- Stopping rules are defined at the start of the trial to
ically look at the records of the patients in the study and ensure that there is a safety valve. If, for example, it
check for protocol deviations that are recorded. You becomes obvious during a trial that one or more
should also check for adverse effects. While these are treatments is significantly worse or better than another,
unlikely for an orthodontic study you could find, for then the trial should be stopped.
example, that a new type of archwire is constantly frac-
turing, and patients are beginning to complain about this
and are withdrawing from the trial. Data analysis
Another error is to allow the patient records to pile up Methods of data analysis for RCTs do not markedly
so that there is no organized check on trial progress. It is differ from other orthodontic studies and these shall not
far better to record the data as the trial progresses. This be discussed in this paper. However, it is important to
allows you to identify any problems with your outcome consider the difficult question of how to handle data from
measures or even your method of data collection. patients who dropped out of the investigation.
Finally, a careful record of all study withdrawals or When this occurs we are left with several choices. The
drop-outs should be made, and as much baseline data as first is to report the number of patients who withdrew
possible recorded. This will ensure that a statistical check from the investigation and emphasize that the two
can be made to discover whether the drop-outs were interventions under investigation had certain success and
similar to those people who remained in the trial. failure rates. Or the data analysis should include the
results of the treatments on all the patients who entered
Interim data analysis the study, regardless of successful compliance or com-
pletion of the treatment. This is termed an intention to
An area of controversy is the analysis of the interim treat analysis (ITT analysis). This type of approach
results of the trial. It is very tempting, particularly in a results in a measure of the true effectiveness of the treat-
lengthy study, to run an interim analysis and have a look ment and should be attempted wherever possible. One
to see how things are doing. This is a common occurrence possible drawback of this approach with orthodontic
if the trial is attracting a degree of attention and you need treatment is that we may not have collected data on the
some data to present at a conference, etc. The problem patients who dropped out of the investigation, as they
with this is that the patients that are analysed first may may not have returned to the clinic. One solution to this
not be representative of the trial population and any is to statistically impute data to compensate for the lost
conclusions that are released are incorrect. data. Several statistical packages have the ability to be
However, the counterargument to this is that it is programmed to carry out this type of analysis.
necessary to run an interim analysis to check that the
treatments are not causing harm, which is important for
the ethics of the trial. While this may be essential in some Writing up
medical trials, this may not be necessary for orthodontics. The final stage in carrying out an RCT is to write up the
Importantly, if an interim analysis is done for this reason project. It goes without saying that the dissemination of
then the results should not be published. the results of a study, even if the results are negative is a
vital step. There is little point in carrying out a project and
keeping the results a secret! When an RCT is written
up, many journals are now requesting that RCTs are
 JO December 2003 Feature Section How to do a randomized controlled trial 341

presented using the CONSORT guidelines as a frame- References

work (http://www.consort-statement.org/).1 On first read-
1. Altman DG, Schulz KF, Moher D, Egger M, Davidoff F,
ing, these may seem to be rather complex and difficult to Elbourne D, Gotzsche PC, Lang T. The revised CONSORT
follow; however, we find that, when the guidelines are statement for reporting randomized trials: explanation and
followed, then writing a paper is made easier as the frame- elaboration. Ann Intern Med 2001; 134: 66394.
work provides headings for the prospective author to 2. Dixon V, Read MJ, OBrien KD, Worthington HV, Mandall
follow. NA. A randomized clinical trial to compare three methods of
orthodontic space closure. J Orthod 2002; 29: 316.
3. Macfarlane T. Sample size determination for research projects
Conclusions (Editorial). J Orthod 2003; 30(2): 99100.
In this review we have attempted to provide a brief outline 4. Schulz KF. Subverting randomization in controlled trials.
JAMA 1995; 274: 14568.
of how to carry out a randomized controlled trial. You
will see that carrying out this type of research is relatively 5. Jadad A. Randomized Controlled Trials: a users guide. London,
BMJ books, 1998.
difficult and time consuming, but prospective investi-
6. Friedman LM, Furberg CD, DeMets DL. Fundamentals of
gators should not be put off from carrying out this type of
Clinical Trials. New York, Springer-Verlag, 1998.
research. The effort is worth it, because a well prepared
and run RCT will provide highest levels of evidence base
for the treatment that we provide.