Towards an Automated Assistant for Clinical Investigations
V. Nigam, LMU, Germany; T. Ban Kirigin, Univ. of Rijeka, Croatia; A. Scedrov, UPENN, USA; C. Talcott, SRI, USA; M. Kanovich, Univ. of London, UK; and R. Perovic, USA
CLINICAL INVESTIGATIONS POLICY
Regulations may be imposed by multiple governmental agencies as well as by institutional policies and protocols.
• Before drugs can be made available to the Due to the complexity of both regulations and activities there is great potential for violation due to human error,
general public, their effectiveness has to be misunderstanding, or even intent.
• Normally, at the final stages, clinical FDA Regulations Protocols
investigations (CI), involving human subjects, "Any adverse experience associated with the use of the drug Protocols are elaborated by specialists
are carried out. that is both serious and unexpected; Each notification shall be explaining how one should carry CIs, so that
made as soon as possible and in no event later than 15 calendar the most conclusive data is collected and the
days after the sponsor's initial receipt of the information." health of subjects is not compromised.
Safety of Subjects
One should avoid at all costs that
the health of subjects is compromised Pharmaceutical companies (Sponsor), clinical ●
Deviations from procedures and violations of regulations
during the tests. research organizations (CRO), health should be avoided as they may compromise both the
institutions (HI) and government agencies collected data and more importantly the health of subjects.
Conclusive Data Collection (FDA) collaborate in order to carry out CIs.
CIs are rigorously regulated by FDA inspectors.
CIs should be carried out in order to
Violations may also imply heavy penalties, both financial as
obtain the most conclusive results/data well as of bad PR.
without compromising the health of
HIs with record of deviations in the past may be punished
A better approach would be to have a by the market and not being hired for carrying out future CIs.
computer assistant for CIs.
Two possible applications (see Figure to
Plan Generation the left), Plan Generation and Execution
Monitoring. Deviations from procedures and violations of regulations
Subject Monitoring ●
Mathematical models do occur.
As illustrated above, regulations and ●
At each site it is common to have more than one on going
policies mention time explicitly. trial, typically 5-10 trials at the same time.
Novel models for collaborative systems ●
Trials take place along with normal hospital duties.
Team subject to explicit time constraints. ●
Easy to lose track of the global view of the process.
FDA / Sponsor / PI
Computational complexity for the plan ●
Therefore, there is great potential for violation due to
generation problem. human error, misunderstanding, or even intent.
Whenever a subject visits a HI, the staff receives
Ways to translate protocols into our
from the computer assistant a plan (Plan mathematical formalism.
Generation) with the actions that need to be
Investigate adequate human computer Next Steps
performed. interfaces. ●
Define a language for specifying protocols that can be
Data Management understood by protocol developers and at the same
The computer assistant can help FDA inspectors, ●
Formally certify that privacy policies are not time implemented in our model.
Sponsor monitors and the Principal Investigators of violated. ●
We expect this language to be based on existing
HI to monitor the execution of the CI ●
Facilitate the statistical analysis of data by standards for CIs, such as the model proposed by
building bridges to existing tools. Clinical Data Interchange Standards Consortium.