4 | DISCUSSION
This study describes our experience from the implementation of a 16-gene PGx panel in routine clinical practice with a focus on clinical relevance. The 16-gene PGx panel test was able to detect variants that are clinically relevant according to the PHARMGKB classification in 100% of tested patients. More important, results of PGx testing led to an actual change of medication or specific recommendations to do so in a high proportion of tested patients. These adjustments of current medication and specific recommendations regarding potential future medication were supported by a PGx expert system and implemented through personalized clinical pharmacology consultations.
Overall, frequencies of PGx variants shown in Figure 2 are in agreement with previous studies in Caucasian populations [14-16]. The detection rate of 100% for at least actionable variants is also not an unexpected finding for a 16-gene PGx panel if one considers that in a previous study even a panel with only 5 genes had a reported detection rate of 99% [15]. Detection rates are typically based on the PHARMGKB classification of clinical relevance, which may be considered as the single best currently available PGx knowledgebase. SONOGEN XP further enhances PGx clinical decision support through additional reviews of other knowledgebases, thorough review of the original literature, collaborations with external experts, and an array of separate reports for different purposes. These range from concise reports written for patients, over specific therapeutic recommendations for prescribing physicians, to extensive summaries for experts of ten and more pages including references to original research publications. The very high detection rate of PGx panel tests for variants that are classified as “required”, “recommended” or “actionable” support the use of such multi-gene PGx panels with the automated interpretation from expert systems for preemptive testing with the ultimate goal to improve efficacy of pharmacotherapy, and to reduce adverse reactions and costs [15, 17].
Furthermore, the experience reported in our study looks beyond PGx panel tests with automated clinical decision support for PGx-based pharmacotherapy and their merely theoretical impact on pharmacotherapy. Whereas Table 2 lists a large number of PGx drugs for the identified PGx variants including some that are hardly ever used (e.g. pimozide or atazanavir), Table 3 provides a real-life insight into the prevalence of specific drugs plus relevant PGx variants that required a change of therapy in our patients. In our subpopulation of patients with a specific indication for PGx testing and a median number of 6 concomitant drugs we provided personalized clinical pharmacology consultations and issued personalized expert recommendations to adjust therapy with the PGx-triggering drug, current concomitant medication and potential future medication. We recommended or, if the clinical pharmacologist was directly involved in patient care, directly changed the PGx-triggering drug in 32.4%, and any other concomitant medication as a “bycatch” in 22.5% of patients based on PGx panel results. This high value supports the clinical relevance of PGx panels for actual clinical decision making and, to our knowledge, has not been investigated in this way before. Because additional costs of panel vs. single gene tests are moderate and likely to further decrease with advancing technology and widespread use, these findings further support the cost-efficiency of PGx panel testing and provide an alternative view at traditional cost-benefit calculations based on single drug-gene pairs.
However, a closer look also reveals that PGx-based management of pharmacotherapy in real-life clinical practice is a complex process, and that the standardized PHARMGKB classification can be highly heterogeneous within the same class. For example, PGx testing for clopidogrel and tamoxifen is merely classified as “actionable” according to PHARMGKB. But the lack of efficacy associated with the tested PGx variants is potentially lethal, and based on a review of the latest evidence, PGx expert guidelines, as well as our own clinical experience, we conclude that PGx testing indeed makes an important contribution to clinical decisions related to those frequently prescribed drugs and can even improve patient compliance [4-6, 18-20]. Furthermore, one must realize that most PGx variants do not have a high predictive value for efficacy or adverse reactions of a drug in individual patients. Rather, they act as one of several factors with complex and often poorly understood interactions, and their effect may be best described by a causative pie model [21]. Accordingly, our clinical experience from PGx-supported clinical decision making also taught us that PGx decision support algorithms are helpful, but that they do not comprehensively capture the complexity of (shared) clinical decision making. As shown in Table 1, we identified a considerable number of patients with comedication inhibiting CYP2C19 or CYP2D6, or renal impairment, and our therapeutic decisions considered all those factors and their interactions with PGx variants, as well as alternative therapeutic options. Indeed, the number of new drugs where the SmPC includes information on PGx variants is steadily increasing. For example, prescription of siponimod (Mayzent®) requires preemptive CYP2C9 PGx testing, and the prescribing information of bexpiprazole (Rexulti®) provides dosing recommendations that consider both, PGx variants as well as concomitant therapy with inhibitors of CYP2D6 or CYP3A4. And even for drugs that have been marketed for a long time, postmarketing studies may identify previously unknown relevant PGx variants [22]. Therefore, we expect a growing demand for PGx testing with integrated expert consulting in clinical pharmacology in the near future, also outside academic centres.
Some limitations of our study should also be addressed. Our study population was selected, partially through physicians that referred patients for specific drug-gene indications, and partially through “mere” screening indications. Characteristics of our patients are therefore transparently presented in Table 1, and one may consider that those may be different in other institutions that offer PGx services. Although our recommendations are a critical appraisal of clinical relevance, we were not able to conduct a larger study with longitudinal follow-up in order to evaluate outcomes of our PGx-based recommendations. These must be addressed in prospective large controlled studies for specific PGx-guided therapy [4, 20]. Nevertheless, we were able to perform a separate analysis for our PGx consultations in patients with clopidogrel therapy, and our results appear to be in line with those studies [18]. Another limitation concerns the 16-gene panel itself that we were able to use. Due to technical reasons this panel did not include relevant HLA variants associated with severe adverse reactions towards carbamazepine or abacavir [23, 24], but from a medical point of view this would certainly be desirable.
In conclusion, our study demonstrates the value of PGx panel testing in routine clinical practice and the valuable contribution of a PGx clinical decision support system. Additional costs of panel vs. single gene tests are moderate, and the efficiency of PGx panel testing challenges traditional cost-benefit calculations based on single drug-gene pairs. However, a closer look also reveals that truly personalized pharmacogenetic medication management will not achieve its full potential without individual patient consultations where additional factors and individual weighing of risks vs. benefits and pharmacotherapeutic as well non-pharmacotherapeutic care are considered. Limited availability of experts and specialized clinics may become a bottle neck for the implementation of PGx-guided pharmacotherapy, which is a challenge but also an opportunity and responsibility for clinical pharmacology and clinical pharmacy services to seek direct patient contact and involvement in PGx-guided medication management.