4.2 Available Evidence
Personalized medicine is a data-driven approach, and a critical step in the devel-
opment of personalized medicine will be the collection and integration of vast
amounts of data from multiple sources (clinical and laboratory data, genetics,
genomics, biological data, lifestyle and environment information, etc.), which
needs to be kept and analyzed (for instance, within biobanks). This large amount
of data will need to be converted into evidence that informs decision-making
processes. Here we are interested in the type of evidence that feeds into economic
evaluation.
Currently, there exists a limited amount of clinical data on health outcomes and
costs data for companion diagnostic tests and their associated treatments that can
populate economic models.^38 The lack of data will likely be of greater significance
as the use of diagnostics to target patients and guide therapies quickens. Towse and
Garrison ( 2013 ) argue that the current regulatory and reimbursement systems are
not inspiring the generation of sufficient evidence about the value of biomarker
patient stratification (i.e., identifying and treating only those predicted to benefit
from treatments). This is a problem (also) because economic evaluations normally
require a large quantity of different types of data (epidemiological, demographic,
clinical, economic, etc.) and from different sources (random controlled trials—
RCTs, clinical studies, patients’registries, administrative databases, etc.). The
unavailability of the required data (i.e., data gaps) impedes on the usefulness of
economic evaluations in the area of personalized medicine products and may lead
to inaccurate results of economic evaluations and hence to suboptimal conclusions
regarding health care allocation based on economic evaluations.
To overcome the problems that analysts face with respect to data, they may
resort to creating assumptions about the future costs and benefits of companion
diagnostics and generate simple instead of complex economic models. For instance,
if there is little information about the adherence to treatment recommendations
related to companion diagnostics, analysts may simply assume perfect adherence in
the patient population^39 and hence introduce bias. Alternatively, analysts may need
to combine the evidence from different sources, and where a data gap will be
noticed, and no randomized studies are possible or available (for whichever rea-
son), it may further stimulate real-world data collections. For instance, clinicians
and patients will have to make decision about the best treatment choice in a
particular situation, without relying on a large body of evidence. As argued by
Annemans et al. ( 2013 ), the documentation about the rationale behind the particular
treatment decision can help in the treatment of similar patients and build towards a
stronger evidence base, consequently improving the validity and generalizability of
future economic evaluations. Ultimately, at this stage of technology development,
(^38) As reviewed by Payne and Shabaruddin ( 2010 ), pp. 643–646.
(^39) A problem highlighted by various authors, such as Vegter et al. ( 2008 ), pp. 569–587; Phillips
et al. ( 2009 ), pp. 5166–5174; and Wong ( 2014 ), pp. 284–285.
Economic Evaluations of Personalized Health Technologies: An Overview of... 125