POC-CT methodology is well suited for studies with the following features:
- Interventions already approved by the FDA.
- A clinical question where there is equipoise regarding clinically
relevant alternative interventions.
- Interventions that are part of routine practice, well tolerated, and have
well-recognized toxicities which mitigates the need for adverse event
monitoring beyond that in routine clinical care.
- Subject identification, inclusion and exclusion criteria, and endpoints
that are accurately obtained from the EMR.
- Outcomes are objective and require little or no adjudication.
- Study protocol requiring minimal deviations from usual care.
- No systematic laboratory or clinical follow-up required for either safety
or comparative effectiveness.
This trial is designed to be on the pragmatic extreme of the clinical trial spectrum
with the subject consent process being the sole perturbation of the clinical care
‘ecosystem.’ The absence of study specific interventions,
procedures, and monitoring together with passive data capture attempts to maximize
the relevance of the findings to current practice at the VA Boston Healthcare
System. Adaptive randomization is designed to assign subjects preferentially to the
treatment arm that, in real time, appears superior, with an
‘efficacy’ stopping rule that has acceptable Type I error. If the
study terminates without reaching its ‘efficacy’ boundary, it will
reliably rule out a substantial difference, in which case cost, convenience, and
other factors will dictate which treatment arms continue to be supported. Such
direct translation of study results into clinical practice defines a
‘learning healthcare system.’
The clinical question posed in this protocol, comparison of insulin administration
methods, was chosen because it is amenable to a maximally pragmatic study as defined
by the PRECIS criteria and because:
- Broad participation by healthcare providers is expected. The clinical
question is compelling and in practice there is apparent equipoise
between the two regimens in that roughly half of patients are currently
treated by each technique.
- The inclusion/exclusion criteria will allow enrollment of nearly all the
VA Boston patients who require the intervention.
- The study interventions are currently utilized at VA Boston, have known
toxicities that are monitored as part of usual care, and thus require no
specific study related monitoring.
- All study data elements are objective, resident in the EMR and do not
require study specific interactions or visits for capture.
- Adaptive randomization methodology leads to real-time incorporation of
study results into practice, if one treatment proves superior.
The ability to implement this study is made possible by the VA’s EMR
environment. CPRS is in use at all the VA’s 1500-plus points of care and was
designed to incorporate clinical data as part of efforts to improve clinical care.
As a result, it features several packages that allow end users to automatically
generate reports, ‘listen’ for certain values associated with
patient data objects, consider these values with programmatic logic, and introduce
information and workflows directly into the EMR. To capitalize on this level of
flexibility, most VA healthcare systems employ Clinical Application Coordinators,
who use these tools to create and report measures of the quality of care, to
implement guidelines, and to create clinical reminders based on the priorities of
each hospital. This infrastructure will allow for the relatively easy roll-out of
this and other POC-CT studies system-wide as well as systematic implementation of
The ability to use existing functionalities, as opposed to developing custom software
is important for a number of reasons. First, development of new software
functionality is constrained by time for development, testing, and approval, and
development resources. Second, by capitalizing on existing system functionality, we
increase the likelihood of a successful deployment to other VA hospitals or clinics,
each one of which employs CPRS. Finally, although this particular use of CPRS may be
novel, the POC-CT processes are presented through familiar interfaces and into a
culture of robust CPRS use, which we hope will facilitate adoption of this
The ability of institutions to implement POC-CTs is dependent on the ability to use
the EMR to: (1) identify events as they present in real time; (2) intervene in the
clinical care workflow; and (3) track longitudinal data. It is worth noting that
these functionalities are critical to the creation and implementation of many novel
approaches to learn from and improve healthcare based on real data and that few
systems offer such capabilities to end users. The need for such functionalities is
of particular relevance in light of the US Federal Government’s upcoming
investment of $19 billion to support the adoption of EMRs [13
]. Much of this funding
is contingent on the adoption of ‘certified’ EMR systems and the
‘meaningful use’ of such systems. Definitions that require flexible
integration with EMR data and workflows are essential to meeting the goals of such
enormous investments [14
The ethical and practical considerations of informed consent have been extensively
discussed and debated [15
] as have methods such as cluster randomization which might obviate or
preclude individual informed consent [20
]. Detailed analyses of these
considerations are outside the scope of this article. However, as POC-CTs or
similarly designed trials become an important component of clinical research, it
will be incumbent on investigators, ethicists, and IRBs to fully consider the
potential benefits and apparently minimal incremental risks of a POC-CT, and to take
responsibility for helping their healthcare systems to lower the barriers to
successful study design and implementation of improvements in care.
A study coordinator will obtain written informed consent for all subjects entered
into this trial. This requirement accounts for a significant proportion of the study
cost and introduces the single most tangible perturbation to the usual care
workflow. We recognize that replacement of such full written informed consent by an
alternative (such as simple ‘notification’ by the healthcare
provider and verbal consent by the subject with subsequent randomization through a
fully automated computerized process) would result in an even more efficient design,
with a closer match to clinical care. The IRB could consider such a variation on the
usual research informed consent, on a study-by-study basis, especially when the
POC-CT results in care materially identical to usual clinical practice. Parallel
requirements would be a waiver of HIPAA authorization to obtain study data from the
EMR and acknowledgement that treating clinicians who authorize automated
randomization are not ‘engaged’ in research.
A POC-CT will likely require significantly less study-specific infrastructure and
cost than traditional RCTs (after the up-front investment in coordinating center and
informatics, already made by the VA). These advantages together with an economy of
scale once an investment in the methodology has been made could lead to low
incremental cost per study as well as allowing study designs of sufficient duration
to capture clinically relevant (as opposed to surrogate) endpoints.