This study presents the first comprehensive overview of the additional risk minimisation measures introduced at authorisation for ATMPs in the EU, including the risks addressed. We included all ATMPs authorised in the EU until 2023, of which the majority were authorised with additional measures to minimise risks. All of these ATMPs had educational material for HCPs, and most also required educational material for patients/caregivers. A controlled access or controlled distribution programme was introduced for about two-thirds of ATMPs. The additional measures commonly addressed safety concerns related to ‘Injury, poisoning and procedural complications’, often for risks regarding ‘Accidental exposure to product’ and ‘Transplantation complications’.

Conditional MA or MA under exceptional circumstances was granted in half of the ATMPs. This proportion is about four times higher than the 11.4% observed for all centrally authorised medicinal products (July 2012 to December 2021) [19, 20]. These expedited regulatory pathways facilitate timely access to medicinal products, with a positive benefit–risk balance based on the limited available data, despite the lack of comprehensive clinical data at time of MA. In these cases, the medical need outweighs the uncertainty about comprehensive evidence regarding efficacy and safety [8, 19, 21]. Given the uncertainties, this underscores the importance of risk management post-authorisation [10]. It should be noted that seven ATMPs were withdrawn from the market for commercial reasons.

In risk management, aRMMs manage the risks and strengthen the positive benefit–risk profile of a medicinal product. We showed that almost all (92.0%) ATMPs were authorised with aRMMs in the EU, which is nearly four times higher compared with all centrally authorised medicinal products [20, 22]. Zolgensma® (Novartis Europharm Limited, Ireland, 2020), indicated for spinal muscular atrophy, and Ebvallo® (Pierre Fabre Medicament, France, 2022), indicated for Epstein-Barr virus positive post-transplant lymphoproliferative disease, had no aRMMs in place at time of MA. It was, however, noticed that aRMMs were introduced post-authorisation for Zolgensma® (Novartis Europharm Limited, Ireland) to mitigate risks of hepatotoxicity, thrombocytopenia, and thrombotic microangiopathy [23]. The higher prevalence of aRMMs in ATMPs compared with non-ATMPs in general might be attributed to their unique nature, which might include exertion of effects over a long period and administration-related risks, like engraftment failure/rejection, medication errors, and transmission of the product to HCPs/close contacts [14, 21]. Although our cohort does not demonstrate a clear relationship between aRMMs and orphan drug status, the inherent uncertainties surrounding the safety of ATMPs, particularly given their frequent use in rare diseases (76.0%) and novelty, may lead to the precautionary introduction of aRMMs. Additionally, EMA guidance highlights that aRMMs should be considered to minimise risks of ATMPs [14], whereas the GVP Module XVI (Rev 2 and Rev 3) states that aRMMs should only be introduced when deemed necessary [12, 24]. Nevertheless, the need for aRMMs and their specific type will be assessed on a case-by-case basis, considering product-specific characteristics. GVP Module XVI (Rev 3) also specifies points to consider for requiring and selecting aRMMs, including the seriousness of the risks, and highlights the role of past and formative evidence in this decision-making process [24].

The most frequently utilised aRMMs were educational materials, as previously also observed for all centrally authorised medicinal products [20, 22]. Furthermore, controlled access or controlled distribution programmes were deemed necessary for 16 out of 25 ATMPs. This is high considering only 20 out of the 526 centrally authorised medicinal products (July 2012 to December 2021) had such programmes in place, also including 12 ATMPs [20]. Although deemed necessary, these programmes might impose significant burdens on healthcare systems. Our study found that qualified or certified treatment centres and/or HCPs were required for all but one ATMP, probably due to ATMPs’ specialised administration processes. Goula et al. emphasised that the success of ATMPs depends on the ability of the centres and HCPs to perform such treatments [25]. Further programme requirements varied among ATMPs, such as the six CAR T cells requiring on-site access to tocilizumab to minimise the risk of cytokine release syndrome and neurotoxicity. Furthermore, traceability measures were mentioned for four ATMPs, to ensure tracking of patients’ cells and the manufactured drug product between the hospital and manufacturing site and/or facilitate the management of potential safety issues. Lastly, patient information provision or patient consent before treatment was mentioned for three ATMPs.

Whilst GVP Module XVI (Rev 2) provided definitions of both controlled access and controlled distribution programmes, they were used interchangeably within our ATMP cohort [12]. A draft revision of the EMA ATMP guidance and the recently updated GVP Module XVI (Rev 3) implies that the terms ‘controlled access programme’ or ‘risk minimisation control tools’ might best describe the necessity of qualified/certified treatment centres and/or HCPs [24, 26]. The use of unambiguous definitions is in the interest of the wider community, including regulators, MA applicants/holders, and healthcare professionals.

Overall, no distinct patterns were observed between the type of ATMP and associated aRMMs in our cohort. However, we did observe that almost all ex vivo and in vivo GTMPs had educational materials for patients. This might be related to the complex nature of GTMPs, which might require patients to understand specific aspects of treatment, including long-term follow-up and potential risks. Additionally, no controlled access or controlled distribution programmes were introduced for sCTMPs, whilst these were frequently required for other types of ATMPs, especially ex vivo GTMPs and TEPs. This could be explained by the relatively less complex administration processes of sCTMPs, which might not require a controlled access or controlled distribution programme. To illustrate, Ebvallo® (Pierre Fabre Medicament, France) should be intravenously administered, where the preparation steps included in the product information suffice [27]. In contrast, Strimvelis® (Fondazione Telethon ETS, Italy), an ex vivo GTMP, requires harvesting CD34+ stem cells for autologous use [28], and Spherox® (CO.DON Gmbh, Germany), a TEP, requires intra-articular implantation [29]. Additionally, some in vivo GTMPs also require more complex administration routes, like Upstaza® (PTC Therapeutics International Limited, Ireland), which should be administered by bilateral intraputaminal infusion in a surgical session [30]. Given the small number of authorised ATMPs in our cohort, definitive conclusions regarding potential differences in aRMMs between different types of ATMPs cannot be drawn.

Considering all safety concerns included in the EU-RMP, safety concerns related to ‘Injury, poisoning and procedural complications’ were most common among ATMPs. In contrast, Vermeer et al. found that safety concerns related to ‘Injury, poisoning and procedural complications’ were not often included in the EU-RMPs at initial approval for centrally authorised small molecules for chronic use and biologicals (2006–2009). This difference might reflect the more complex administration processes required for several ATMPs compared with non-ATMPs. Their research highlighted that safety concerns related to ‘Investigations’ and ‘Infections and infestations’ were common for small molecules for chronic use and biologicals, which were also frequently observed within our ATMP cohort. In our study, safety concerns related to context of use, i.e., ‘Long term effects’ and ‘Use in special populations’, were often included in the EU-RMPs of ATMPs. Whilst Vermeer et al. also observed that safety concerns related to ‘Use in special populations’ were common, they found that safety concerns related to ‘Long term effects’ were rarely identified for small molecules for chronic use and biologicals [17]. Unlike small molecules for chronic use or biologics, which are administered repeatedly and have defined half-lives, the risks of many ATMPs could persist lifelong. Long-term safety might be especially important for ATMPs targeting the paediatric population, although we did not identify any specific safety concerns for ATMPs with a paediatric indication.

Differences in safety concerns (regardless of aRMMs) were observed between type of ATMPs. First, safety concerns within ‘Investigations’ were commonly identified for ex vivo GTMPs, in vivo GTMPs, and sCTMPs, but not for TEPs. This mainly reflects the risks of immunogenicity for GTMPs and sCTMPs and generation of replication competent retro/lentivirus for GTMPs, which are less of concern for TEPs as they do not use viral vectors and are minimally immunogenic [31]. Moreover, safety concerns related to ‘Infections and infestations’ were not often included for in vivo GTMPs, whereas they were more commonly observed for other types of ATMPs. This could be due to the fact that in vivo GTMPs are typically administered without requiring extensive cell culture, transplantation, or direct handling of cells. Whilst safety concerns within ‘Neoplasms benign, malignant and unspecified (including cysts and polyps)’ were identified for almost all ex vivo GTMPs, these were identified in less than half of the in vivo GTMPs and none of the TEPs. The higher rate in ex vivo GTMPs, including CAR T cells, could be attributed to the genetic modifications involved and/or insertional mutagenesis. In contrast, in vivo GTMPs involve direct gene delivery without ex vivo manipulation of cells, potentially leading to a lower potential risk of neoplasms, despite the possible risk of insertional mutagenesis [32]. TEPs do not generally involve genetic modifications, possibly explaining the absence of reported neoplasm-related safety concerns. However, the small sample size limits definitive conclusions.

Important identified risks (54.7%) and important potential risks (41.0%) were most likely to be addressed with aRMMs, whilst only 7.3% of safety concerns classified as missing information had aRMMs. Previous research identified that educational material was introduced at MA for 53.1% and 8.1% of the important identified and important potential risks for 17 centrally authorised biologicals (2006–2009) (unpublished data, Vermeer et al. [17]). The higher proportion of important potential risks requiring aRMMs for ATMPs might be explained by a different nature of the risks, requiring a more precautionary approach. For example, aRMMs for ATMPs might be introduced based on theoretical risks identified through state-of-the-art knowledge before a causal relationship is established, like genetic effects caused by transmission to third persons. Missing information seldomly necessitated aRMMs, as they denote data absence rather than risks needing minimisation, which is consistent with the findings of Vermeer et al. (4.5%) (unpublished data, Vermeer et al. [17]). Hence, post-authorisation safety studies are crucial to further characterise (long term) safety of ATMPs in real-world settings. Post-authorisation safety studies are an important component in risk management of ATMPs, especially considering the uncertainties associated with ATMPs at time of MA [8, 21, 33]. Long-term follow-up (≥ 10 years) is especially important for GTMPs to assess potential delayed effects in real-world settings, such as for neoplasms. As previous research indicated that the submission timelines for post-authorisation safety studies of ATMPs extended up to 15 years, the results related to long-term safety of ATMPs and relevance of these studies have yet to be assessed [33].

Regarding safety concerns addressed with aRMMs, ATMPs often had aRMMs addressing safety concerns related to ‘Injury, poisoning and procedural complications’, ‘General disorders and administration site conditions’, and ‘Immune system disorders’ in absolute numbers. A previous study investigating aRMMs for all centrally authorised medicinal products (1995–2009) highlighted similar trends [34]. Johnson and Priefer reported that risks identified in the US’s REMS were also commonly related to ‘Injury, poisoning and procedural complications’ [35]. Our findings suggest alignment with observations for medicinal products in general.

ATMPs with safety concerns related to ‘Injury, poisoning and procedural complications’,'Nervous system disorders', and ‘General disorders and administration site conditions’ often required aRMMs for these safety concerns. The safety concerns within ‘Injury, poisoning and procedural complications’ that often necessitated aRMMs seemed to be linked to administration processes, underscoring the importance of administration as recommended in the product information [9, 25]. In contrast, safety concerns related to ‘Poisoning and toxicity’, which covered safety concerns comprising carcinogenicity, were never accompanied by aRMMs. Whilst the potential risk of carcinogenicity cannot be prevented using additional measures, patient and HCP awareness might be crucial for early detection and monitoring. Among ATMPs with aRMMs in place for ‘Nervous system disorders’, safety concerns within ‘Nervous system disorders NEC’ were frequently covered. It should, however, be noticed that those safety concerns were related to neurological toxicity that might be specific to CAR T cells. Within ‘General disorders and administration site conditions’, ATMPs frequently required aRMMs for ‘Therapeutic and nontherapeutic responses’, involving concerns regarding lack of efficacy. Although many ATMPs presented safety concerns related to ‘Investigations’, these were infrequently addressed with aRMMs. Specifically, most ATMPs included safety concerns on ‘Therapeutic drug monitoring analyses’, with few requiring aRMMs. This indicates that aRMMs were occasionally required for risks related to immunogenicity. Less than half of the ATMPs with safety concerns related to ‘Neoplasms benign, malignant and unspecified (including cysts and polyps)’ had aRMMs introduced for those safety concerns, even though, as with carcinogenicity, aRMMs might be used to facilitate early detection and monitoring. Of note, no aRMMs were introduced at time of MA to mitigate the risk of secondary malignancies for CAR T cells, although aRMMs were introduced post-authorisation [36]. These aRMMs focus on the need for lifelong monitoring of patients for secondary malignancies. Safety concerns related to context of use were typically classified as missing information and rarely accompanied by aRMMs. We emphasise that the ATMPs in our cohort represent a heterogenous group of medicinal products, including in terms of safety concerns addressed by aRMMs. The small absolute numbers in the subgroups, combined with this heterogeneity, impede meaningful comparisons between type of ATMPs. This variability further complicates the establishment of a ‘standard’ set of aRMMs, as evidenced by the differences observed across our cohort. As stated before, it remains important to assess the need for aRMMs for specific safety concerns on a case-by-case basis for individual ATMPs at the time of MA application and post-authorisation.

This study is the first to provide an inventory description of the aRMMs implemented for ATMPs in the EU. This adds depth to our understanding of the regulatory approach and risk management for ATMPs and could support regulatory decision making for future ATMPs, next to the risk-based approach. Whilst we described our results for ATMPs as a group, we also explored potential differences across ATMP types regarding safety concerns and aRMMs and provided an overview of the aRMMs implemented at the product level. We also want to acknowledge the limitations of this study. Whilst the use of publicly available EPAR documents ensures transparency, these documents only provide summarised information and lack details on the regulatory discussions behind the necessity for aRMMs. Further research would be needed to explore the decision-making process. Another limitation is that safety concerns were manually categorised into the MedDRA® terms, context of use categories, or ‘Not classifiable’. To increase validity, independent dual data collection was performed. Besides, our analyses were restricted to aRMMs established at MA and did not include updated aRMMs. Whilst risk management is a continuous, iterative process, this approach was chosen as some ATMPs were no longer authorised and EU-RMPs were not always publicly available. Lastly, the small number of authorised ATMPs impedes definitive comparisons between type of ATMPs.