both analysts and policy makers should be realistic about their expectations relating
to the quality and quantity of data that is available for assessment.
Performing an economic evaluation of companion diagnostics (or the combina-
tion of the diagnostic test and the following treatment path(s)) involves understand-
ing the entire care pathway, which can be very complex. For instance, if two or
more diagnostic tests need to be performed sequentially, then the economic eval-
uation of a personalized medicine strategy will involve a series of tests requiring a
thorough understanding of which test will be performed after the initial test; the
treatment decisions to be made after the tests are performed; the impact of these
tests and treatments on health outcomes and costs; the quality of the tests in terms of
accuracy, sensitivity, specificity; and how the tests correlate (particularly their
correlation in terms of the number of false positives and negatives). If all the
information is available, it will make up a very complex decision tree, where
complexity is additionally emphasized by the fact that diagnostic tests normally
do not provide clear-cut answers to treatment decisions. Indeed, with complex
testing scenarios, there is ample room for mistakes and misinterpretation along
the entire testing sequence.^40 Phillips et al. ( 2009 ) suggest that the greater stan-
dardization of test procedures and processes and better communication between the
laboratory and the clinician may reduce the error margin.
The complexity of personalized medicine and the developments of increasingly
tailored interventions will also require an adaptation of the clinical trial designs in
the direction of efficient clinical trials based on a more thorough understanding of
the genetic basis of disease.^41 There are specific issues related to the design of RCT
that will need to take an account of the shift from the “one size fits all” model to a
much more individualized approach. For instance, small patient numbers will lead
to longer and more complex trials yielding more uncertain evidence. Freidlin
et al. ( 2010 ) discuss in detail the appropriate randomized phase II trial designs
involving biomarkers and suggest that, in most settings, randomized biomarker-
stratified designs (i.e., designs that use the biomarker to guide analysis but not
treatment assignment) should be used to obtain a rigorous assessment of biomarker
clinical utility. Current RCTs are mainly limited to homogeneous populations and
not well-suited for evaluating personalized medicine products that require a sys-
tematic study of the variation in subgroup clinical responses.
The limited amount of data and the data surrounded by uncertainty that is used to
populate decision models will reflect on the certainty with which the results of
economic evaluations may be interpreted, as well as on their usefulness for policy
makers. In the ideal conditions, we would be able to assess the cost-effectiveness of
diagnostic technologies using patient data drawn from randomized controlled trials,
together with the data from follow-up studies monitoring patients from the initial
test to the final outcome of the treatment. According to Towse et al. ( 2013 ),
(^40) For further details, see Allison ( 2008 ), pp. 9–517.
(^41) A problem recognized by many authors, e.g., Hamburg and Collins ( 2010 ), pp. 301–304;
Freidlin et al. ( 2010 ); Simon ( 2010 ), pp. 33–47; Simon and Roychowdhury ( 2013 ), pp. 358–369.
126 A. Bobinac and M. Vehovec