Similar studies to the present one have been conducted in the past, albeit focussing on different aspects. Both Yu et al. and Maeda et al. reviewed PK interactions broadly, including CYP mediated DDIs and focussed on the effect size of DDIs for drugs approved by the FDA and the Japanese authority PMDA, respectively [30, 31]. No new strong inhibitors of transporters were identified among drugs approved by the FDA in 2020 [30]. Agarwal and colleagues [32] identified an increase in information about transporters in FDA-approved package inserts before and after the introduction of the first guidelines mentioning transporters. The predominant transporter in their analysis was PgP, but there were a few texts about OATPs and BCRP. Recommendations were provided on how to present this information in FDA package inserts. This analysis was followed up by Fan et al. who investigated the transporter-related post-marketing requirements by the FDA from 1999 to 2015 and identified that 65% led to updates in labelling once fulfilled [33].

The present analysis brings in new perspectives on the EU product information. The present analysis focussed on the message conveyed by the SmPC text, not on the underlying data and their interpretation. A prerequisite was that the transporter was mentioned in the SmPC. The absence of hit could be due to the DDI risk being excluded or considered to lack clinical relevance but could also be due to a spelling error in the SmPC. The DDI could also be described without specification of its mechanism, for example when mentioning a specific substance or substance class such as HMG-CoA reductase inhibitors [4, 34].

The dataset showed a large degree of similarity between OATP1B1 and 1B3 texts throughout the analysis. This was anticipated given that the transporters belong to the same family. Substances are often substrates of both. Inhibition of both transporters is commonly seen, albeit sometimes at different concentrations, which may impact the clinical relevance of the interaction, for example in Lynparza (Olaparib) [35].

The BCRP text dataset shared some similarities with the OATPs; however, the warnings tended to be of higher clinical relevance for both perpetrator and victim interactions compared to OATP texts. The higher number of signals for perpetrator interactions could be due to higher in vitro concentrations being tested for orally administered drugs for BCRP. The higher concentrations are needed to cover for the risk of DDI in the gastrointestinal tract, which is not applicable to OATP1B1 or 1B3. BCRP substrates and inhibitors also overlap with PgP [36, 37]. Consequently, it is not always possible to select an in vivo study design that would enable the isolation of a single transporter effect. Further overlaps are also known with enzymes, where the enzyme effect may be larger and the main responsible for the interaction, such as for riociguat (Adempas), where DDIs with multipathway inhibitors of CYP, PgP and BCRP are warned for [17].

Interactions between statins and OATP inhibitors are well-known due to the established link to myopathy and in extreme cases rhabdomyolysis [38, 39]. Rosuvastatin, one of the probe substrates recommended for in vivo interaction studies, is a substrate of all three studied transporters [4]. Thus, many perpetrator warning texts refer to the same substance group as potential victims. It must be noted that the statins themselves were not part of this dataset as they were approved prior to 2012. The widespread use of statins [40] could also have led to an increased interest for including information about transporter DDIs.

Some major areas of possible SmPC improvement were identified. The analysis revealed that the majority of texts about OATPs, and a large number of texts about BCRP indicated the absence of an interaction (Table 1, Fig. 1), which is not in line with the SmPC guideline [7]. The location of these texts describing the lack of a clinically relevant DDI was astonishing, as many were described in Sect. 4.5. While not factually incorrect, this information may distract the prescriber from more relevant information. It is thus recommended to omit such text. The reader should be able to rely on that the absence of a DDI risk description indicates that data on all transporters covered in the current DDI guideline was available and a DDI risk was excluded. However, there are exceptions where results of an in vivo study denoting the lack of DDI could still be presented in Sect. 4.5, for example in therapeutic areas known for the propensity of DDIs (antivirals) [7]. According to the guideline, information on the involvement of transporters in the absorption or elimination of a drug could be presented in Sect. 5.2. However, there is no recommendation to inform on the lack of inhibition (perpetrator data) in Sect. 5.2. Despite this, 12% (for OATP1B3) of all approved products during the period included negative transporter data in 5.2 and could thus benefit from a SmPC revision to remove irrelevant information. It should be noted that the policy regarding the presentation of interaction data differs between the EMA and FDA. The EMA considers that information of clinical relevance should be described, whereas the FDA encourages the presentation of data denoting the lack of interaction in the clinical pharmacology section [41,42,43].

In addition, many texts were unclear with respect to whether an action was anticipated from the prescriber. For products of the category “unknown” which were classified as such due to the lack of explicit recommendation, additional information may be considered. Actionable recommendations could be added to the SmPC to support the healthcare professionals with information on how to use the medicine safely and effectively.

The case of a positive in vitro signal without additional in vivo or in silico data was of interest, as it is challenging to define an adequate level of warning and risk mitigation. In the OATP1B1 perpetrator analysis, 10 DDIs were classified as clinically relevant, and 11 as of “unknown” clinical relevance when based only on in vitro data. For OATP1B3, eight DDIs were classified as clinically relevant, and four as “unknown”, while for BCRP 26 DDIs were classified as clinically relevant, and 13 as of “unknown”. There was thus a high frequency of observed in vitro interactions without clearly defined regulatory or clinical action to follow (i.e. request for an in vivo study or for a warning text). The SmPC texts addressing this scenario varied from lack of warning to different levels of caution, all based on an IC50 value compared with a reference therapeutic concentration. Even if the in vitro experiments only provide a “yes”/ “no” answer to the question whether a clinically relevant interaction can be excluded, sometimes the applicant and the assessor also consider how close the IC50 is to the relevant cutoff. For victim interactions, it could be possible to identify the clinical relevance if information on the therapeutic window is available.

Defining the level of clinical relevance was the most challenging and subjective part of this project. Available DDI-checking tools were consulted to support the classification of clinical relevance or action [44, 45]. As there was no consensus on how to classify the DDIs, the analysis was data driven. The consequence categories (Table 2) differ from the scenarios selected as potential starting points for standard texts (Table 3), as the latter were based on the experience gathered when assessing new drug applications.

Some texts were classified as of “unknown” or “lacking” clinical relevance despite a positive signal. These could be cases such as Mulpleo (lusutrombopab) [46] where the clinical DDI study demonstrated an impact on PK, but the therapeutic window was sufficiently broad to conclude on the lack of clinical relevance. In other cases, there were in vivo data but the relevance of the DDI was uncertain, such as for Orgovyx (relugolix) [47].

All contraindications identified in this analysis were found in hepatitis C products [13, 48,49,50]. The clinically more serious DDIs, i.e. contraindications, were generally unambiguous. DDIs slightly less severe, where dose adjustments or monitoring may suffice, had a high degree of variability in the provided [not always] actionable recommendation. This highlights the larger room for interpretation of SmPC information and that standard texts would be helpful.

Whenever a DDI is potentially clinically relevant, it is vital that this message is clearly conveyed to the prescriber with a description of any required action to take. This is important as warning texts can be taken out of context and misinterpreted. An effective text would include information regarding what to monitor, and the consequences of a higher or lower exposure, along with relevant examples of substances. Keeping the SmPC updated with relevant DDI examples is challenging over time. At the EMA level, there is no listing of clinically relevant substrates and inhibitors of the different transporters. The clinical relevance may also differ depending on the therapeutic context, which could be the reason why examples given for the same mechanistic DDI may vary. The FDA however has published a list that may be used as a starting point to find relevant examples of concomitant drugs at risk [51].

The next step of the project was to identify SmPC texts that were perceived as easy to understand and actionable, as a basis for developing standard texts for different DDI scenarios (Tables 3, S6 and S7). There are ongoing projects aiming at the introduction of standard sentences for the improvement of package leaflet [52]. Other initiatives analysed the overall content in product information to understand where harmonisation would have the greatest impact [9].

Standard texts would be particularly helpful in the perpetrator scenario. While there are currently no standard texts to describe DDIs with transporters, the scenarios identified here (Tables 3, S6 and S7) and the corresponding example SmPC texts could form the basis for developing such standard texts. These could be used not only for OATPs and BCRP but could also be adapted to DDIs with other transporters and enzymes.

Harmonised texts for DDIs would result in streamlined, unambiguous texts where the SmPC reader would not be left to make their own interpretation. Besides achieving consistency for similar scenarios, the DDI texts could be harmonised across indications. Further advantages could be that the availability of multiple choices of standardised texts for each DDI scenario could simplify the writing of SmPCs for applicants and reduce assessment time for regulators. By using reference data vocabularies on standard sentences as structured data, spelling errors would also be avoided. Standardised texts are compatible with EMAs electronic product information initiative [53]. Other projects also aim at developing standard texts, for example for the reproduction and lactation sections of the product information [54]. It must be noted that not all scenarios may be covered by standard sentences, and it will remain essential that the documents are created and assessed by experts, keeping the human in the loop throughout the process. Ultimately, generating standard texts for DDIs with the possibility to technically structure information for better filtering and searchability would be most valuable for healthcare professionals in facilitating interoperability with e-health systems using standardised information. This would ensure potential DDIs are properly warned for and handled adequately, maintaining efficacy and safety for each patient regardless of concomitant medications.