Comparison of post-authorisation measures from regulatory authorities with additional evidence requirements from the HTA body in Germany – are additional data requirements by the Federal Joint Committee justified?

At the time of marketing authorisation, the available information relating to a medicine may not yet be sufficient to fully assess the benefit/risk profile to the desired degree of certainty or may lack aspects of interests to the Committee for Medicinal Products for Human Use (CHMP) that are requested to be provided after marketing authorisation is granted. Therefore, marketing authorisation agencies such as the U.S. Food and Drug Administration (FDA) or the European Medicines Agency (EMA) may require the generation of additional data, e.g. in the form of clinical studies.

The FDA refers to these additional requirements as post-marketing requirements and commitments. These gather additional information about a product's safety, efficacy or optimal use [1]. The EMA can require post-authorisation measures (PAMs) if the Agency’s committees consider the generation of additional data necessary from a “public health perspective to complement the available data with additional data about the safety and, in certain cases, the efficacy of authorised medicinal products”. Thus, like their FDA equivalent, PAMs are mainly required in order to answer open questions regarding the risk/benefit profile or safety concerns of a product [2].

The European Public Assessment Report (EPAR) provides details of the measures required for each product and the agreed time frame for completion. PAMs are a well-established component of the centralised procedure for marketing authorisation in Europe and manufacturers are able to draw upon a wealth of guidance. Generally, each PAM is discussed in detail by the EMA and the manufacturer to define the clinical question and the trial design, including details of the patient population, intervention, comparison and outcomes of proposed studies [2]. PAMs are especially imposed on conditional marketing authorisations or when marketing authorisations have been granted under exceptional circumstances.

Since 2011, following the introduction of the Act for Restructuring the Pharmaceutical Market in Statutory Health Insurance (Arzneimittel – Markt – Neu – Ordnungs – Gesetz, AMNOG), a product has to undergo an early benefit assessment after marketing authorisation and launch in Germany. The assessment aims to determine the additional benefit of a product against a pre-defined appropriate comparator therapy in terms of morbidity, mortality, quality of life and safety. The extent of additional benefit granted will in turn influence the reimbursement price. Scientific information is presented by the manufacturer in the form of a benefit dossier according to a formalistic template and guidelines from the Federal Joint Committee (Gemeinsamer Bundesausschuss, G-BA). In general, the benefit assessment is conducted as a two-step approach encompassing i) scientific assessment by the Institute for Quality and Efficiency in Health Care (IQWiG) and ii) appraisal by the G-BA; for orphan drugs, both, assessment and appraisal, are performed by the G-BA. A decision is made on the extent of the additional benefit (major, considerable, minor, no benefit or benefit less than comparator) and the level of evidence (proof, indication, hint) [3‐5]. Details of the early benefit assessment procedure have been described previously [6, 7].

Based on German social law, the G-BA has the option to issue conditional appraisals [3, 5]. This is usually done if no reliable conclusions regarding the extent of additional benefit can be drawn from the submitted data [8]. After a pre-specified time frame, the G-BA will review the additionally submitted scientific data and subsequently re-assess a product’s additional benefit in a new procedure. The G-BA’s decision rationale (Tragende Gründe) provides further information on the requested data and the time frame for submission of the updated dossier containing the additional data [8]. In contrast to PAMs, experience regarding conditional appraisals and additional data requests by the G-BA is currently limited to a 5-year time frame [8].

In this study, we compared the requests for additional data by the EMA (PAMs) and the G-BA (conditional appraisals) to illustrate differences in the categories and types of requested data and to identify differences between PAMs and conditional appraisals in the still rather new and developing early benefit assessment process in Germany.

A step-wise approach was used for this analysis, comprising identification of all early benefit assessments of new medicines or new indications during the first 3.5 years after the introduction of the procedure in Germany, identification of the corresponding EPARs/EPAR variations and subsequent classification of data requests in conditional appraisals/PAMs. Parameters analysed included data categories, data types and clarity of data requests.

In this analysis, we compared post-authorisation constraints from the EMA and the G-BA during the first 3.5 years after the introduction of early benefit assessment in Germany.

A considerable number of data requests by the G-BA went beyond those issued by the EMA (80 %), which may appear surprising considering a court ruling from the Federal Social Court stipulating that the G-BA may not issue a judgment on the quality, efficacy or safety of a drug that deviates from that of the licensing authorities [10]. However, according to the G-BA, the therapeutic benefit required for a drug to be reimbursable by the German sick funds is not identical to the therapeutic efficacy and the favourable benefit-risk ratio proven by the marketing authorisation [11]. In line with this, the Federal Social Court acknowledged that there may be discrepancies between the evidence gathered from pivotal trials and the requirements concerning the use of a drug in clinical practice [12].

In particular, the G-BA considers itself not bound by decisions made by the licensing authorities regarding appropriate comparator and patient-relevant endpoints [13], two areas of special relevance in the determination of additional benefit. Indeed, the majority of additional requests for both efficacy and safety data specifically concern these two topics, which have previously been identified as aspects in which G-BA requirements substantially deviate from the EMA’s viewpoint [7, 14‐16]. A recent evaluation of parallel scientific advice meetings demonstrated that the level of disagreement between the EMA and health technology assessment (HTA) bodies regarding the comparator and endpoints was 30 % and 12 %, respectively [17], indicating that the different remits and perspectives of regulators and HTA bodies frequently cause discrepancies regarding these points.

Our results demonstrate that the reasons for issuing conditional appraisals are not transparent as the quality of the submitted data for oncology assessments was overall comparable between conditional vs. unconditional appraisals, except for minor deviations in the potential for bias. In line with this, an analysis of G-BA appraisals concerning products approved under special regulatory circumstances showed that clear reasons for issuing conditional or unconditional appraisals were not always reported in the documentation of the assessment procedure [18].

Data requests by the G-BA are issued without any methodological guidance to assist the manufacturer in generating acceptable data [8]. In contrast, the majority of PAMs requesting additional studies stipulated that the study protocol should be approved by the CHMP prior to study start [9]. This ensures alignment between the manufacturer and the EMA and enables the manufacturer to adequately address PAM requests.

Other European HTA bodies also provide guidance regarding their preferred methodology. The French authority, the Haute Autorité de Santé, can request additional data, which may require new studies, during its assessment procedure. In order to make sure the agency’s requirements are met, a guidance document regarding general methodological considerations has been published, and the study protocol has to be submitted for evaluation prior to the start of the study [19].

The re-assessment of benefit after the expiry of the conditional appraisal for vemurafenib illustrates the consequences of the absence of such guidance by the G-BA. In its initial assessment, the G-BA had suggested a historic comparison of dacarbazine vs. vemurafenib to allow an evaluation of the differences in overall survival. Although this was provided by the manufacturer in a revised dossier, the G-BA concluded that it was obsolete for methodological reasons and the initially assigned additional benefit rating was not changed [8].

Our analysis reveals obvious differences in data handling between G-BA and EMA, with the EMA showing considerably more flexibility in terms of data acceptance, which may explain why almost half of the products with a conditional appraisal received no PAM. Of note, the EMA frequently had similar concerns to the G-BA, but took additional data into account to answer open questions. This is partly due to the consideration of studies involving different comparators and studies in which the therapy under assessment was not administered at the approved dose; most importantly, however, the EMA is not limited to RCTs in its evaluations. Whereas the G-BA mainly follows the methodology laid down by the IQWiG, which requests RCT data wherever feasible [20], the EMA also uses sources such as non-clinical studies or even expert advice to put data from pivotal trials into context.

A similarly less restrictive approach explicitly accepting non-RCT data is used by the UK National Institute for Health and Clinical Excellence (NICE) for HTA [21]; in such cases, data adjustment by modelling and a critical appraisal of uncertainty are required. Modelling is generally used to extrapolate data that do not fit ideal requirements, e.g. surrogate endpoints, a shorter than desired study duration or a study population that is not representative of patients within the UK National Health Service [22]. NICE therefore tries to make the most of the available data and to obtain at least limited information, whereas the G-BA in most cases categorically rejects all data that do not conform to the IQWiG’s standards and does not take this information into account. As the vemurafenib example shows, the rigid data requirements of the G-BA, together with the lack of methodological guidance, also represent a significant obstacle to obtaining a superior benefit rating in a re-assessment, even if new data are submitted.

For medicines with conditional appraisals, as the initially presented data have not been considered to be fully conclusive by the G-BA, it can be expected that the benefit rating is conservatively low. This will influence initial reimbursement as under AMNOG, the price of a medicine is in part determined by its benefit rating [23]. In contrast, an improved benefit rating after re-assessment would result in new price negotiations and potentially in a higher price. Considering that the goal of AMNOG is to reduce the rapidly increasing drug costs to the statutory health insurance by 2.2 billion Euro per year [24], it could be hypothesised that this may not be the desired outcome in the eyes of the G-BA and the German sick funds. In fact, a potential political motivation regarding data evaluation, namely in the determination of the appropriate comparator by the G-BA, in order to contain costs has been suspected before [25].

In a worst case scenario, a drug can actually receive a worse benefit rating after re-assessment, which would very likely further lower its price. This occurred in the case of regorafenib [8], which was consequently taken off the German market after the re-assessment because the manufacturer was not willing to further reduce the price [26]. It has already been noted that AMNOG has decreased the traditionally high availability of innovative drugs on the German market through opt-outs following benefit assessments [27]; conditional appraisals and the resulting re-assessments may well contribute to this.

A structured collaboration between the German national licensing authorities and the G-BA, in the form of national joint scientific advice, has recently been established [28]. While this is a positive development, it has already been clarified that this dialogue “will not, cannot and is not intended to” harmonise study requirements [29]. This will continue to make it difficult for manufacturers to obtain data acceptable to both the EMA and the G-BA in one single study. The previously cited analysis of joint scientific advice meetings [17] showed that in cases where the EMA and HTA bodies did not agree on the comparator, the suggested solution was most frequently an indirect comparison. However, given the stringent methodological requirements in Germany, it will be challenging to obtain an additional benefit on this basis: of the first 23 indirect comparisons submitted after the introduction of AMNOG, only one was formally accepted by the IQWiG [25].

A genuine harmonisation of study requirements between regulators and HTA bodies has repeatedly been called for, not just in relation to the G-BA [30, 31] but for European HTA in general [17, 32]. More flexibility in HTA will be especially crucial as new methods for marketing authorisation emerge [33]. Alignment of data requests between the stakeholders, both around the time of authorisation and during a product’s life cycle, would markedly simplify the process of data generation to answer important questions concerning a given drug.

This analysis has some limitations. Firstly, when reviewing the Procedural Steps, only PAMs clearly identified as such were included. Therefore, studies originally conducted as PAMs but not appropriately labelled may have been excluded from the analysis. However, as only products with newly granted marketing authorisations were assessed, it can be assumed that the majority of PAMs have not yet been fulfilled, and therefore the number of PAMs not included is probably negligible. Secondly, the dataset is relatively small as only appraisals from the first 3.5 years of AMNOG were analysed and the number of available benefit assessments from this period is limited. Future studies on larger datasets will be required to validate the conclusions drawn from this analysis.

In conclusion, it would be helpful if the factual basis for conditional appraisals could be clarified. Moreover, clear instructions regarding the fulfilment of data requests, as typically provided for PAMs, would be beneficial in providing direction to the manufacturer and subsequently enabling the G-BA to make an informed decision on the additional benefit of a new medicinal product. Ideally, discussions between the licensing authorities, the G-BA and manufacturers early in a product’s life cycle should determine the extent and format of the required data and the strategies to be used for their generation. Undoubtedly, everybody, from licensing authorities over HTA bodies and payers to clinicians and patients would stand to benefit from this scenario.