Since the late 1950s, investigators [1,2,3] have documented the connection between genetic background and response to medications as well as the importance of maximizing the benefits and minimizing the harm of treatments. As more pharmacogenetics (PGt) evidence accumulated over the decades, a “genotype-tailored” approach to pharmacological therapy emerged, whose advantages extend not only to the patients but also have significant implications for drug developers and National Health Systems (NHSs). However, in certain regions, implementation of PGt still needs to be consolidated.

In general, two different types of PGt testing are crucial to provide patients with the most effective treatments: somatic testing, used for tailored anti-cancer therapies, and germline testing, useful in predicting individual drug responses based on inherited genetic variations.

Somatic PGt testing is commonly employed in oncology to predict the efficacy of cancer treatments or drug resistance; examples are  EGFR mutations and response to tyrosine kinase inhibitors [3]; BRCA1/2 variants and PARP inhibitors [4]; KRAS/NRAS/BRAF mutations and cetuximab resistance [5]. In Italy, molecular testing in oncology is standardized by the Italian Association of Medical Oncology (AIOM) in collaboration with The Italian Society of Pharmacology (SIF), which promote guidelines for cancer therapy in different clinical scenarios [6].

Germline PGt testing can be used in two different frameworks: i) the evaluation of potential benefits of a treatment in presence of a specific genotype, as exemplified by the use of lexacaftor/ ivacaftor/ tezacaftor, prescribed for cystic fibrosis (CF) therapy only in patients with at least one copy of the F508del mutation (NM_000492.4:c.1521_1523del) in the CFTR gene [7]); ii) the investigation of functional genetic variants influencing the pharmacokinetics (PK) and/or the pharmacodynamics (PD) of a drug, thereby affecting the drug’s efficacy, appropriate dosage, and adverse effects.

Pharmacokinetics (PK) genetic variants influence drug absorption, liver metabolism, distribution, and excretion (ADME genes); a key role of CYP2D6 -- a highly polymorphic gene involved in the metabolism of up to 25% of the approved drugs – has been clearly demonstreated in PK regulation [8,9,10,11,12,13]. One of the most remarkable examples is siponimod, a drug prescribed for secondary progressive multiple sclerosis (SPMS) whose dosage must be adapted after genetic testing to maximize efficacy. The drug is metabolized in the liver by the polymorphic pharmacogene CYP2C9; accordingly, both the Food and Drug Administration (FDA) and European Medicinal Agency (EMA) require CYP2C9 genotyping before treatment with siponimod [14, 15]. In Italy, an official AIFA note refers to genetic testing for prescribing siponimod: “[…] Before starting treatment, it is necessary to determine the CYP2C9 genotype of patients with the aim of establishing their CYP2C9 metabolizer status […]. In patients homozygous for the allele CYP2C9*3 (NG_008385.2:g.48139A > C), siponimod should not be used” [16]. However, the Italian version of the leaflet of the drug Mayzent (commercial name of Siponimod) makes no mention of these indications, although the test is required by the ‘Summary of Product Characteristics’ also released by AIFA. [17].

Pharmacodynamics (PD) genetic variants, on the other hand, may influence the interaction between the active drug and effector molecules. An illuminating case is represented by the potent broad-spectrum aminoglycoside antibiotics, often used for the treatment of suspected infections in neonatal intensive care units (NICU): aminoglycoside-induced hearing loss (AIHL) is associated with at least three variants in the MT-RNR1 gene (m.1095T > C, m.1494 C > T, and m.1555 A > G) [18].

In this manuscript, we focus on the germline PGt tests, which are currently applied to a limited number of drugs within in the Italian healthcare system.

Factors driving the introduction of pharmacogenetics in healthcare systems

Many primary factors drive the implementation of pharmacogenetic testing within NHSs.

The cost reduction of genotyping and sequencing technologies has led to the widespread adoption of genetic testing across various domains, both clinical and direct-to-consumer.

Moreover, there is a mounting body of evidence demonstrating the clinical utility of validated germline pharmacogenetic testing [19]. An interesting analysis carried out across 15 US institutions provided an overview of current efforts to implement a preventive (pre-emptive) PGt testing strategy in clinical practice aiming for the “Right Drug, Right Dose, Right Time” approach [20, 21]. The genes pre-emptively tested varied among sites, but generally included CYP2C19, CYP2C9, VKORC1, and CYP2D6. These genes were consistently analyzed prior the prescription of selective serotonin reuptake inhibitors (SSRIs), voriconazole, clopidogrel, opioids, and warfarin.

The clinical utility of a pre-emptive strategy was further demonstrated by the Pre-emptive Pharmacogenomic Testing for Preventing Adverse Drug Reactions (PREPARE) trial, a recent multi-center, controlled, cluster-randomized study [19] in which a 12-gene PGt panel was used to accurately genotype the selected pharmacogenes in 18 hospitals, 9 community health centers and 28 community pharmacies in seven European countries. Italy was one of the partner countries. Participants were genotyped for 50 germline variants, and those with an “actionable” variant were treated according to the Dutch Pharmacogenetics Implementation Working Group (DPWG) recommendations, whilst patients in the control group received standard treatment. The results indicated a 30% reduced risk of clinically relevant adverse reaction (OR = 0.70 [95% CI 0.54–0.91]; p = 0.0075), demonstrating that genotype-guided treatment significantly reduced the incidence of clinically relevant adverse drug reaction (ADR) and is feasible in different organizations and health system settings. This kind of approach is based on multiple genotyping which may be useful also for guiding future treatments.

The rapid spread of genetic biobanks is also driving the growth of pharmacogenetics; indeed increasing amount of data require the adoption of recommendations by the NHSs [22]. Population-scale studies revealed that over 95% of the general population carries at least one actionable genotype or diplotype [23, 24]. On the other side, half of all prescriptions in the United States is potentially affected by actionable germline PGt variants [25]. With the advent of population-scale research initiatives, it is also becoming feasible to estimate the prevalence of the most relevant pharmacogenetic variants in the European [26] and Italian populations [24, 27]. Therefore, it is this becoming possible to assess which part of the population may be targeted by preventive pharmacogenetic initiatives.

Furthermore, with a view to a cost efficiency, it would be interesting to consider taking advantage of routine comprehensive genetic tests performed using NGS analysis for other purposes (such as exome sequencing or targeted panels) to extract pharmacogenetic information for benefit the individual. This approach would provide a clear cost saving for the analysis, bearing in mind that in most cases these subjects require chronic drug therapies. Recent studies have investigated the feasibility of identifying incidental findings, namely pathogenic or likely pathogenic variants in pharmacogenes, as a secondary finding of NGS analyses initially performed for other diagnostic purposes [28,29,30,31]. However, currently, there are no specific recommendations from the American College of Medical Genetics (ACMG) regarding this topic, and the interpretation of such variants remains challenging. This complexity arises from the high probability of detecting incidental finding and the fact that, unlike disease-related genes, the influence of drug-related genes is often modulated by environmental factors. According to the ACMG recommendations for incidental findings from WES/WGS analyses reporting [32], Malignant Hyperthermia Susceptibility (OMIM #145,600) is the sole condition associated with genes included in the Clinical Pharmacogenetics Implementation Consortium (CPIC) guidelines, for which the use of volatile anesthetics and succinylcholine should consider RYR1 and CACNA1S genotypes. Consequently, it is advisable to report the presence of pathogenic or likely pathogenic variants in these two genes as a secondary outcome [33].

Finally, a phenomenon that cannot be ignored is the market of “direct-to-consumer genetic testing” (DTC-GT). In this regard, the company 23andMe, which has more than 14 million customers worldwide [34], provides an FDA-approved pharmacogenetic report based on the genotype of CYP2C19, DPYD, and SLCO1B1 genes.

Pharmacogenetic guidelines

Best practices and algorithms to assist the choice of therapy according to genotype are codified by guidelines formulated by international consortia and scientific societies, edited by groups of experts, and usually published in scientific journals.

According to PharmGKB [35] – the primary pharmacogenomic knowledge-based resource created to aggregate all clinically relevant pharmacogenetic information -- there are currently 260 clinical guidelines for 194 active substances and combinations, published by different networks and scientific societies, including CPIC [36], DPWG [37], and other scientific societies, including the Canadian Pharmacogenomics Network for Drug Safety (CPNDS) [38] and the French National Network of Pharmacogenetics (RNPGx) [39], which has around 30 members throughout France and other French-speaking countries (Belgium and, more recently, Switzerland and Canada). Guidelines are also established in other countries, including Germany and Spain.

Among the 291 variant-drug combinations categorized in PharmGKB [40] as “Level 1A” (accessed on 15 Jan 2024) --indicating inclusion in at least one major clinical guideline -- half are associated with genetic variants that elevate the risk of drug toxicity. The remaining 50% concerned variants affecting efficacy (17%), dosage (5%), drug metabolism (24%) or combinations of these mechanisms. In addition, 106 combinations are of interest in paediatric care settings.

CPIC guidelines were created in 2009 to provide dosing guidance related to PGx information present in the medical record, to assist healthcare providers in making informed decisions based on genetic testing results [41, 42]. In a scenario where preventive and clinical genotyping are becoming increasingly common and accessible, CPIC guidelines represent a key resource.

Currently, CPIC [43] has published level A or B guidelines for of 110 gene/drug pairs, mainly regarding cytochromes CYP2D6 (16% of gene/drug pairs), CYP2C19 (13%) and CYP2C9 (10%); the most frequently found gene after the 3 cytochromes is MT-RNR1 (10%) [Suppl. Table “CPIC (A and B)”]. A total of 85 drugs are considered in these guidelines.

According to PharmGKB [Suppl. Table “PharmGKB guidelines (w. rec)”], and excluding the drugs with a “no recommendation” statement (e.g., the case of aspirin, for which CPIC states “CYP2C9: no recommendation; G6PD: no recommendation”), there are 107 active substances for which at least one guideline provides advice for alternate drug or change in dosage: 80 active substances are covered in the CPIC guidelines; 54 in DPWG guidelines; 23 in other guidelines.

Furthermore, guidelines from the CPIC cover three-quarters of drugs; among these, 46 active ingredients are not currently considered by guidelines from other scientific societies or consortia guidelines. Moreover, there are 27 drugs for which no CPIC guidelines exist.

Interestingly, for 5 of the 20 most widely consumed active ingredients in Italy [44], CPIC claims that there is sufficient evidence to recommend at least one prescriptive action (CPIC level A and B gene/drug pairs), i.e.: pantoprazole, lansoprazole, omeprazole (CYP2C19), with “Actionable PGx” indications in the FDA labels and “Moderate/Optional” CPIC strength of evidence; atorvastatin (SLCO1B1), with “Informative PGx” in the FDA label, with “Moderate” CPIC strength of evidence; rosuvastatin (ABCG2 and SLCO1B1), with “Actionable PGx” in the FDA label, with “Strong type” CPIC strength of evidence for SLCO1B1 and “Strong/Moderate” type for ABCG2.

Several studies have explored the appropriateness of patients genotyping to improve patient outcomes and prevent the discontinuation of treatment due to either ineffectiveness and/or the occurrence of side effects. These issues indirectly increase the costs for the Italian NHS. For instance, a recent study developed a cost-effectiveness model for the introduction of multiple pharmacogenetic tests (CYP2C19, CYP2C9, CYP4F2, VKORC1) in a hypothetical cohort of patients with acute coronary syndrome and/or atrial fibrillation, supported by electronic and informatics tools. This approach has suggested a real improvement in patient quality of life of, along with a reduction in clinical events and costs for the NHS [45]. Of note, this approach considered drug efficacy, which is often excluded from assessment of the potential impact of PGt testing. Future research could benefit from integrating these endpoints for a better characterization of pre-emptive PGt tests.

The DPWG was founded in 2005 by the Royal Dutch Pharmacist’s Association (KNMP).

The DPWG’s recommendations are available on the KNMP website (in Dutch) and are updated periodically [46,47,48]. According to PharmGKB reports, the DPWG guidelines include recommendations for genetic testing for 54 active substances. Among these, 18 active substances are exclusive to the DPWG guidelines and not mentioned in other guidelines [Suppl. Table “ST PharmGKB guidelines (w. rec)”]. These recommendations have been instrumental in selecting actionable drug–gene interactions for the PREPARE study [19]. The implementation of this panel, known as the “PGx-Passport”, which include the most prevalent genes among the DPWG annotation (CYP2B6, CYP2C9, CYP2C19, CYP2D6, CYP3A5, DPYD, F5, HLA-B, SLCO1B1, TPMT, UGT1A1 and VKORC1), has been demonstrated to be adaptable to various European health-care systems.

Privacy issues and prescription: the legal framework

In an era characterized by the rapid increase of databases containing large collections of human genomic data, the protection of personal information has become a critical issue. Genomic data represents a crucial resource for biomedical and clinical research, but protecting the privacy of personal data from illegal or unauthorized for-profit uses is increasingly challenging.

On May 25, 2018, the European Union (EU) regulation 2016/679 known as the General Data Protection Regulation (GDPR) came into force, abrogating all previous laws concerning data protection [49]. Under GDPR statute, the processing of special categories of personal data, including genetic data, should be allowed only when strictly necessary [“Article 10”]. Such processing must be preceded by the individual’s consent, which should be informed, complete and specific [“Article 13”]. For pharmacogenomic testing, the same general regulations apply. Hence, consent to process genetic data is always required and it would be desirable to specifically supplement the informed consents currently in use for diagnostic examinations.

In our opinion, with regard to the practice of testing, the prescription of the test could be requested by a specialist doctor or by a general practitioner, whilst the role of the geneticist should remain central in counselling, testing and interpretation of the result. This approach is particularly valuable when genetic testing is not contingent on an evaluation of family history or a risk calculation [50].

Current status of pharmacogenetics guidance by EMA and pharmacogenetic testing prescriptions in the context of the Italian NHS

Although there are examples of the implementation of pharmacogenetic guidelines in european NHSs, a certain discordance exists between these national initiatives and the regulations set by the EMA. Furthermore, there is a lack of uniformity in the management of pharmacogenetic testing and clinical practices across various European countries.

According to PharmGKB (accessed on 15 Jan 2024), EMA has approved 95 drugs for which “required” or “recommended” pharmacogenetic indications are available in the respective labels. At the time of access, there were also 135 “informative” or 44 “actionable” label annotations. As the active substance pegloticase is currently withdrawn from the European market, the actual number is 94 drugs with indications “required” (N = 91) or “recommended” (N = 3) [Suppl. Table “ST Table EMA”].

In detail, in 26 out of 94 cases, the genetic test is germline (27.7%), in 7 of them it predicts an enzymatic activity (7.5%), while for the remaining active ingredients the test is for somatic mutations (64.9%) (Table 1). Most of the drugs considered (64/94, 68%) are classified as antineoplastic agents according to the EMA classification [Suppl Table “ST Table EMA”].

Table 1 Category of genetic test in the EMA labels (for drugs approved in the European market), grouped by the Human pharmacotherapeutic group (EMA) (extended details in Suppl Table “ST Table EMA”)

The ‘germinal’ category includes those drugs for which germline genetic testing is intended. The ‘somatic’ category, on the other hand, includes those drugs for which genetic tests are performed on tissue, mainly in the contest of cancer therapy. In the category “other” we have included specific drugs for certain genetic mendelian disorders, such as inborn errors or enzymatic deficiencies, for which genetic or laboratory tests (e.g. biochemical assay or enzymatic activity) are required prior to intake, according to the EMA annotations reported on PharmGKB. In this case, the test is obviously performed prior to taking the drug, as the therapy is specific to the identified enzyme defect. The EMA annotations for this group of drugs generally refer to the diagnosis of the disease, which in this case is usually achieved by performing a biochemical assay. As a rule, it is only afterwards that genetic confirmation of the result is carried out. In the cases of 4 drugs for metabolic disease (betaine, carglumic acid, fosdenopterin, migalastat), on the other hand, diagnosis by genetic testing is known to be primary/obligatory or necessary to identify specific ‘amenable’ variants for therapeutic indications. Hence, these drugs are included in the “germinal” category. However, the analysis of these genes is intended for therapeutic purposes and not for dose modulation/prevention of adverse events, therefore it was not the object of this study

Considering the 20 non-antineoplastic drugs for which germline testing is required by EMA, in most of them the test is motivated by diagnostic purposes, and not to predict the efficacy or toxicity of the drug (e.g. drugs used for cystic fibrosis, such as elexacaftor and ivacaftor).

Currently, considering only drug/gene pairs for which EMA provides annotations for preventing adverse events and/or dose modulation, only the following genes are included: HLA, CYP2C19, CYP2D6, CYP2C9 and DPYD [Suppl Table “ST Table EMA”].

To date, in Italy, the availability of molecular genetic tests for pharmacogenetic purposes has been paradoxically facilitated by unclear legislation generically referring to the EMA/AIFA recommendations. However, the publication of various decrees over the last 7 years has revealed certain contradictions and inconsistencies.

From a practical point of view, the list of medical prescriptions of any genetic test recognized by the Italian NHS is regulated through the Essential Levels of Care (LEA) framework. The LEAs consist of a comprehensive repository of treatments and services that is required to provide to all citizens, either free of charge or after payment of a participation fee (the so-called “ticket”).

Recently, the Italian government has updated this repository to better define the services provided and the corresponding cost. However, with regards to the implementation of pharmacogenetic testing, this novel decree [51, 52] has excluded most of the pharmacogenetic analyses already approved by EMA and some of the tests supplied so far.

The new fee schedule is based on a list of healthcare services approved in 2017, which included the analyses of CYP2D6, CYP2C19 and UGT1A1 genes, while excluded CYP2C9 and DPYD despite they are both required to avoid adverse events during specific pharmacological therapies (i.e., siponimod and capecitabine, respectively [14, 53]). Moreover, the prescription of these three tests would also be subjected to very restrictive criteria, relating to only 3 drugs. Therefore, this update does not consider either a large number of tests that are regularly prescribed and performed nationwide (e.g., DPYD, HLA typing, etc.) or the most recent international recommendations regarding the implementation of new pharmacogenetic tests.

A key point is that this update will significantly reduce the public offer of genetic testing for patients undergoing therapies that could benefit from specific genotyping, effectively limiting the EMA’s recommendations. However, the effective date of application of the new LEAs has been postponed to the next year (1 January 2025) [54]; this paper therefore aims to raise the issue of a revision of the regulation before the rule becomes operative.

Oncological patients deserve special mention, because most of the actionable PGt variants are associated to anticancer drugs in different guidelines. The impact of PGt testing on this subpopulation of patients is consistent with a high rate of prescription, due to the high risk of severe adverse events following therapy. A critical example is that of DPYD genotyping. About 30% of patients undergoing chemotherapy with fluoropymirimidines have ADRs due to a decrease of the DPYD activity, though additional gene variants have been associated with the ADRs. Therefore, DPYD genotyping is widely recommended as a pre-emptive test and its exclusion from updated LEA will likely result in a health-related and legal problem for oncologists as the test is recommended by an official AIFA note of May 2020 [55].

In order to resolve these conflicting and divergent recommendations at the national and international level, following the example proposed in this paper, other European countries could also assess the overlap between national regulations, EMA indications and the most recent scientific evidence, such as reported in the PREPARE study. This could make easier to align the different European regulations and keep the recommendations provided by EMA up-to-date.

Towards a full application of pharmacogenetic practices in the Italian NHS

The Pharmacogenetics Working Group of the Italian Society of Human Genetics (SIGU) aims to establish the groundwork for the standardization of pharmacogenetic practices within the Italian NHS. This initiative begins with the incorporation of international guidelines based on the pharmacogenetic panel outlined in the PREPARE study [19]. Additionally, the group aims to expand and clarify the existing EMA/AIFA germline tests recommendations in the framework of the new LEAs.

Of the 40 drugs in the PREPARE panel (clozapine, efavirenz, carbamazepine, sertraline and oxycodone were removed according to indications provided in Supplementary files of Swen et al.), each of them has a guideline from the DPWG. Additionally, 26 of these drugs (72%), also have a guideline from CPIC (Table 2).

Table 2 Full list of drugs included in the PREPARE study which are on the market in Italy with indication for PGt by at least CPIC

Currently, PGt testing “required” or “recommended” is indicated by EMA for predicting adverse events and/or dose modulation for six drugs (thus excluding all the therapeutic indications): abacavir, atazanavir, capecitabine, eliglustat, siponimod, tegafur / gimeracil / oteracil [Suppl Table “ST Table EMA”, genes and drugs highlighted in red]. However, by comparing the EMA gene-drug pairs with those included in the PREPARE panel (Table 2), only the DPYD genotyping for capecitabine administration is shared (Fig. 1).

Fig. 1

Comparison between drugs included in gene-drug pairs listed by Swen and colleagues (limited to drugs marketed in Italy) and drugs with label indications for germinal PGt tests according to EMA. Drugs or genes included by PREPARE, EMA and/or Italian LEA are coloured in light blue. With regard to EMA recommendations, only PGt tests with a modulating dose/AE prevention indication are considered, as explained in the main text. The left panel (a) shows the overlapping drugs between PREPARE and EMA lists. The right panel (b) compares the genes included into the PREPARE study with the genes noted by EMA recommendations and the Italian LEAs. Drugs associated with each gene with a pharmacogenetic indication are also shown in boxes. Considering the PGt tests included in the PREPARE panel that will be suitable in Italy after the implementation of the new LEAs, paradoxically the only test reported as recommended by the EMA (DPYD) will be excluded in the 2024 LEAs version. With regard to irinotecan, it has been included in the LEA column as the analysis of UGT1A1 is scheduled, although the prescription of this test has not yet been fully clarified in the decree. Moreover, among the drugs reported by the EMA, only capecitabine overlaps with the PREPARE panel. It would therefore be useful to integrate the PREPARE list of drugs with that reported by EMA and to extend the genes noted in the new LEAs to those validated by Swen and colleagues

Furthermore, regarding DPYD, EMA recommends a preventive test for the drug combination tegafur / gimeracil / oteracil, which is not included in the PREPARE panel. It would therefore be useful to extend this panel to drugs approved by AIFA that are analogous to those collected in the PREPARE study.

Another relevant issue concerns siponimod, for which CYP2C9 genotyping is recommended by EMA. While the CYP2C9 gene is included in the PREPARE panel for therapies based on warfarin and phenytoin, there is no specific indication for siponimod.

Regarding CYP2D6 and CYP2C19, LEAs do not list drugs for which testing is appropriate, but rather provide a general indication for the pharmacogenetics of drug metabolism genes. Similarly, irinotecan is not directly mentioned in the LEAs, even though the analysis of known UGT1A1 mutations (which are of oncological pharmacogenetic interest) is specified in the 2024 version.

Concerning HLA genotyping, the PREPARE panel includes only the HLA-B testing for the drug flucloxacillin, while the EMA annotations also indicate the genotyping of other HLA genes (for abacavir). According to the new LEAs, however, it is in any case allowed to prescribe all these tests as different MHC genes are listed and specified in detail. However, no indications are mentioned for using the test for pharmacogenetic purposes.

Moreover, 22 drugs approved by AIFA and considered in the PREPARE trial contain pharmacogenetic indications in the respective drug labels [Suppl. Table “AIFA extended”]. In more details, the drug labels of only three drugs recommend genetic testing prior to administration (Table 2): the anticancers fluorouracil and capecitabine, for which genotypic and phenotypic testing of DPYD is suggested; and the antipsychotic pimozide, for which testing of the CYP2D6 gene is recommended to identify slow metabolizers. However, numerous efforts are underway to adapt the new decree to the local and national landscape by allowing the incorporation of certain adjustments to ensure an adequate implementation of pharmacogenetic testing.

It is of extreme interest to note that for several drugs (Table 2) -- some of which are widely used in Italy -- there is no indication of pharmacogenetic testing in AIFA-approved labels.

Striking examples are citalopram and escitalopram, for which the DPWG issued therapeutic dose recommendations based on CYP2C19 genotype. For CYP2C19 ultrarapid metabolizers, the recommendation is to avoid escitalopram, the most widely used antidepressant in Italy (OSMED Report2022).

To estimate the potential impact on the Italian NHS, it would be useful to consider data prevalence of pharmacogenetic variants in the Italian population and drugs prescription, but this has so far only been possible for certain subpopulations [24, 56].

In this regard, an Italian collaborative effort is ongoing to create a reference database for genomic data in the Italian Population (http://nigdb.cineca.it/), that can be searched for allele and genotype frequencies in the main macroareas [57,