Interventional Studies

These are either RCTs (randomized controlled trials) or uncontrolled trials. The major features of RCTs differentiating it from other
types of studies are

1. Experimental intervention – one group remains control (or standard) and the other(s) receives some intervention
2. Randomization occurs – i.e. after a subject enters the study he/she is randomly allocated to either control arm or intervention arm
(or other arms) through a process of randomization.
a. This may be simple randomization e.g. randomizing each entrant separately using a computer generated list; or
b. Block randomization – where randomization occurs in a set of specified numbers, say blocks of 6 etc., to ensure equal number
distribution between two arms. Larger the block size, lesser the ability to predict allocation.
c. In stratified randomization baseline characteristics which may have an implication on final outcome are stratified so that equal
distribution across groups is enforced. The gain in precision through stratification is minimal for large studies. Stratification does
not eliminate the need for adjustment for differences in the baseline composition of the study groups if found.
d. Cluster randomization – where unit of randomization and analysis is various centers or catchment-zones and not individual
patients – so after a cluster is allocated to an arm, all individuals in the cluster receive same intervention (or non-intervention).
(See minimisation below). Within a cluster, patients are more likely to respond in a similar manner, and thus can no longer be
assumed to act as independent units. This lack of independence in turn leads to a loss of statistical power in comparison with a
‘patient’ randomized trial. As a result, in cluster RCTs, effective sample size (ESS) is used in place of actual sample size when
undertaking power calculations. To calculate ESS, statisticians use a measure called intracluster correlation, which represents the
degree to which the various individuals in a cluster resemble each other in the outcome measure.
e. Minimisation: this is useful in two situations.
i. You want to do a stratified randomisation as you are worried Characters  of  sound  randomisation  
about unequal distribution some very important covariates.
But your trial is too small with too few participants for stratified •  Reproducible  order  of  assignment  (not  possible  for  coin-­‐flips  though  
randomisation. this  method  adequately  generates  a  random  sequence)  
ii. You are conducting a cluster randomisation and units of •  Documented  methods  for  assignments  
randomisation being clusters you end up with small numbers. •  Assignments  remain  concealed  (see  below)  
You want to minimise gross differences in confounder •  Future  assignments  not  predictable  from  past  assignments  
(covariant) distribution. •  Ability  to  detect  departures  from  established  procedures  
In the above situations, we can use minimisation to achieve a balance between treatment groups. In minimisation schemes, the
next allocation depends on characteristics of those already allocated. Allocation of each participant aims to ensure a balance of

© SPMM Course
 prognostic factors between groups. The disadvantage is that this method is inferior to proper randomisation as it allocation is
somewhat exposed and ‘controlled’ manually; in fact only the first allocation is truly random. Using computerized programs can
reduce bias while minimizing.
f. Quasi-randomisation refers to ‘randomizing’ using even/odd numbers of the date of birth, day of the week patient was seen,
etc. These are not reproducible methods, and the sequences cannot ensure equal distribution of variables. These must be
avoided.
g. Uses of randomisation
i. It permits use of probability theory in making inferences
ii. Eliminates effects of bias
iii. Facilitates blinding
iv. Distributes baseline characters in an unpredictable fashion
3. Randomisation improves generalizability, but it does not ensure the results will always be precise –precision depends very much
on the sample size. Having a large sample size does not mean randomization is unnecessary.

Special types of RCTs:

¬ Crossover trial – this refers to an interchange of study and control groups after a washout period so that all subjects in a study
receive both placebo and treatment but only in a different
order. This is an economic design as number of required Factorial design Naltrexone
subjects is lesser. But this cannot be used for diseases that
YES NO
are cured by the use of medications. Parallel RCT is a term
used to denote the common form of RCT, which is the Cognitive YES Both Naltrexone & CCT only (placebo)
opposite idea of crossover trials – i.e. both control and coping skills CCT
intervention happens in a parallel, not serial fashion as in therapy (CCT)
crossover trials. NO Naltrexone only Placebo only
¬ N of 1 trial – here a single subject is administered placebo (supported) (supported)
and active intervention (or two different interventions) in
tandem under double-blind controlled conditions. This is the best way of optimizing a treatment schedule for an individual patient –
but results cannot be generalized to other patients.
¬ Factorial trials - Most RCTs evaluate a single therapeutic factor of interest, often a novel intervention that is compared with one or
more alternatives, or a placebo. In a trial using a 2x2 factorial design, participants are allocated to one of the four possible

© SPMM Course
 combinations (2 treatments, 2 levels). A 2x2x2 trial indicates 3 treatments at 2 levels. For example in a 2x2 factorial RCT participants
would be allocated to drug A alone, drug B alone, both drug A &B, placebo. In this way, it is possible to test both the independent
effect of each drug and the combined effect. Consider the design (shown in the table) from Heinala P, et al. (2001) Targeted use of
naltrexone without prior detoxification in the treatment of alcohol dependence: a factorial double-blind, placebo-controlled
trial. J Clin Psychopharmacol 21: 287–292. For the same question if a multi-arm trial is conducted, the cost may be more but
power reduced. Factorial trials are difficult to operationalise.

¬ Patient preference trials: When evaluating certain interventions some patients may have a strong preference for a particular
therapy. They are then allocated the therapy they prefer. Only those with equivocal or no preference are then randomised. Note that
the primary analysis should still be limited to randomised groups. But the researcher can obtain data from an observational cohort
(preference group) in a parallel fashion. This maximises recruitment in a trial. One can also analyse the effect of motivation on
intervention effectiveness. This makes the results more generalisable. But obtaining consent for such trials may be complex and
recruiting those with no preference may be hard. (BMJ 1998; 316: 360)
¬ Zelen’s modified RCTs: Generally informed consent is obtained before randomising a sample for RCT. But the problem with this
standard procedure is that a substantial number of patients may withdraw if they suspect they are not getting the treatment they hope
for. Their compliance may become poor if they have a hunch on what intervention they are getting. Zelen’s modification suggests
randomisation before obtaining consent. The consent is sought then from those randomised to experimental treatment. This
approach is controversial with respect to ethics. MRC does not accept this design as ethically valid. (BMJ 1998; 316: 606)

¬ Non-inferiority trials: Most RCTs are conducted to demonstrate the superiority of one intervention over the other. These can be
called as superiority trials. In some instances, especially if a well-established standard treatment is available, drug companies may
want only to demonstrate that the new intervention is ‘not inferior’ to standard intervention. This type of RCT is called as an
equivalence trial (same efficacy as the comparison) or non-inferiority (not significantly worse than the comparison by more than a
certain ‘allowed’ margin) trial. Note that the null hypothesis for such trials must be ‘there is a significant difference between two
interventions’ and not otherwise as seen in superiority trials. Also, a higher sample size is required for an equivalence or non-
inferiority trial compared to a superiority trial at the same level of power and significance. (In fact theoretically true equivalence can
only be demonstrated if the sample size is infinity as the effect size, by definition, is zero in such studies!) Also intention to treat analysis
may blur actual differences between the groups and, in fact, may support the claim of equivalence. Hence, it is best to report both
ITT and per-protocol analysis in these trials. (BMJ 1996: 313; 36–39).

© SPMM Course
Other types of interventional trials:

¬ Uncontrolled trials – here only one group is observed for effects of an intervention – there is no control group. The study
inference is made using a historical control i.e.the status of the treated group before intervention
¬ Before and after trials – here, the outcome is assessed before and after some intervention in one group of subjects. The
observed change may be in fact attributable to a factor other than the intervention itself, making this design inherently weak.
There is no randomization in this design.
¬ Multi Arm trial: This is a simple extension of RCT where more than one study arm is included. It is easy to design and allows
intervention-1 vs. intervention-2 comparison in addition to intervention-1/2 vs. placebo effect. But this requires a larger sample
size and power may be reduced as a consequence.
Unnatural  circumstances  in  a  RCT  compared  to  regular  
clinical  practice  
Pragmatic RCTs:
Recruitment:    from  specialist  centers  
o Efficacy refers to how well a drug can work under optimal circumstances. Patients  with  co-­‐morbid  medical  or  psychiatric  
o Effectiveness refers to how well it works under usual practice circumstances. disorders  are  excluded  
o Efficiency measures whether health care resources are being used to Patients  are  carefully  selected  to  generate  
maximize value for money homogeneous  diagnostic  groups    
Patients  are  allocated  the  treatment  at  random  (no  
The fundamental differences between efficacy trials and effectiveness trials relate to negotiation  during  prescription)  
objectives and motivation for the trial. An efficacy trial is undertaken to meet Patients  are  given  detailed  information  for  informed  
regulatory approval. An effectiveness trial is designed to convince formularies and consent  
payers of the actual usefulness of the drug in current practice. Such effectiveness trials Placebo  is  used  to  compare  active  treatment    
for which the hypothesis and study design are developed specifically to answer the Patients  are  followed  at  frequent  intervals    
questions faced by decision makers are also called pragmatic or practical clinical Patients  get  detailed  checklists  of  side-­‐effects    
Assessment  endpoint  is  typically  4–6  weeks    
trials (PCTs), e.g., CATIE was a pragmatic RCT. One need not model all details of
Often  symptoms  measures  are  the  outcomes  
intervention (such as when to escalate the dose, etc.) beforehand in pragmatic trials –
considered.  
these are attended to as per the need of the patients in trials.
Patient  and  clinician  are  blind  to  treatment.  

© SPMM Course
The characteristic features of pragmatic RCTs are that they
(1) Select clinically relevant alternative interventions to compare, Pragmatic RCTs
(2) Include a diverse population of study participants,
(3) Recruit participants from heterogeneous practice settings, and • Reflect the heterogeneity of patients in clinical practice
(4) Collect data on a broad range of health outcomes.
• Minimise exclusion criteria
(5) They do not tell how well a drug works (EFFICACY) – they inform
how EFFECTIVE is the intervention in the real world. • Focus on groups with a wide range of diagnoses

• Define patient groups by presentation rather than diagnosis

Problems with pragmatic designs include -
• May not employ placebos
1) Substantial drop-outs can occur
2) Blinding can fail • May not be blinded
3) Less rigorous methodology • Must carefully conceal allocation during randomisation
4) Higher rates of non-specific treatment effects.
• May evaluate real world, function based outcomes

© SPMM Course
 DISCLAIMER: This material is developed from various revision notes assembled while
preparing for MRCPsych exams. The content is periodically updated with excerpts
from various published sources including peer-reviewed journals, websites, patient
information leaflets and books. These sources are cited and acknowledged wherever
possible; due to the structure of this material, acknowledgments have not been
possible for every passage/fact that is common knowledge in psychiatry. We do not
check the accuracy of drug-related information using external sources; no part of
these notes should be used as prescribing information.

© SPMM Course