Commentary on the EMA Guideline on strategies to identify and mitigate risks for first‐in‐human and early clinical trials with investigational medicinal products

Background

The European Medicines Agency (EMA) published in July 2017 a guideline for first‐in‐human (FIH) drug studies 1. The purpose of this document is to assist investigators, pharmaceutical companies, ethics committees and other regulators and stakeholders with the design and performance of early clinical studies of new compounds in humans. The focus of the guideline is on risk mitigation and promotion of safety. The guideline is a revision of an earlier version dated 2007 and extends the existing EU guidance to address FIH and early phase clinical trials (CTs) with integrated protocols 2.

The first edition of the EMA guideline on FIH studies followed the devastating events that occurred during the FIH study of TGN1412 in March 2006. The first administered dose of this CD28 superagonistic antibody caused a cytokine release syndrome in all healthy volunteers. The causes of these unexpected severe effects were carefully analysed by different authorities and experts in the field 3, 4, 5. This resulted in guidelines that put an increased emphasis on the relevance of animal models; a revision of strategies to determine the starting dose (including the concept of MABEL, the minimally active biological effect level in humans); and an adaptation of safety measures for FIH studies (e.g. ‘sentinel’ cohort and intensive care access). Further research of TGN1412 revealed interspecies differences between the biological function of CD28 and led to the development of a bioassay that was predictive for humans. As a result, in 2014, the compound could be reintroduced at much lower doses in human development 6.

Ten years after its first publication the EMA guideline on FIH studies has now been revised. This revision followed a tragic event that happened in January 2016 during an FIH programme with BIA 10‐2474, a fatty acid amide hydrolase (FAAH) inhibitor that enhances endogenous endocannabinoid concentrations being developed by BIAL 7. Although BIA 10‐2474 had never been administered to humans, FAAH inhibitors had been examined in numerous clinical trials. Other compounds of the class had generally failed to show therapeutic effects in a number of indications, but no concerns had been raised for safety. The study – which included several parts under the same ‘umbrella’ protocol – had an apparently uncomplicated single ascending dose (SAD) part, on which several lower multiple ascending doses (MAD) cohorts had followed before the events unrolled. It was therefore quite unexpected that severe neurological symptoms occurred in the third cohort on the fifth day of administration, which caused the death of one of the volunteers and neurological toxicity in four others. The root cause analysis of these tragic outcomes has not yet been completed, largely because of the ongoing legal proceedings that impose limitations on sharing the data and the results. The French authorities have issued a report, which outlines a number of factors that may have been involved 8. Despite strong public appeal from both clinical researchers 9, 10 and regulators 11, the pharmacokinetic (PK), pharmacodynamic (PD) and clinical data obtained in the SAD and MAD studies have not been published, and many questions regarding the preclinical characteristics of the compound remain unanswered.

Irrespective of the ongoing debate about transparency and access to the BIAL data, EMA amended its guidance dated 2007 considering the conclusions highlighted in the TSSC Report 8 and also – as stated in the concept paper published in 2016 – to bring it in line with current practices for FIH CTs and early clinical drug development.

This commentary aims to put the new guideline 1, which will become effective in February 2018, into perspective by guiding the reader of the regulatory document through its chapters and expanding on how the revised guideline might impact practical aspects of FIH trials. The section headings in this commentary provide a direct hyperlink to the corresponding section of the guideline under discussion in a copy of the guideline that is available as Supporting Information to this article online (see Supp Mat).

1. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=4

The introduction of the new guideline starts with the assertion that the purpose of FIH trials is to evaluate an investigational medicinal product (IMP) in humans for the first time, to study the human pharmacology, tolerability and safety of the IMP and to compare how effects seen in nonclinical studies translate into humans. The revised guideline further acknowledges the integration of PK, PD and safety information, not only from preclinical experiments but also from human data emerging during the trial itself. The emphasis on human pharmacology seems to be a change from the first edition of the guideline in 2007 12, where the term ‘pharmacology’ was mainly used in the context of ‘nonclinical safety’. The emphasis of both versions remains on safety, but the revision seems to recognise the importance of the compound's pharmacological characteristics more clearly. This is in line with current R&D and lead selection practices in pharmaceutical industry of developing highly targeted IMPs.

The need for incorporating emerging data is also associated with a tendency to include several early human drug studies – called ‘study parts’ in the guideline – under the same integrated ‘master’ protocol, or to run them in parallel rather than consecutively. The obvious advantage from a drug development perspective is efficiency: Doses and study schedules can be adapted to concurrent findings from other study parts, and precious time and resources can be saved. However, the process is only efficient and safe if the right information is shared effectively among sponsors, contract research organisations (CROs) and investigators. Decisions that involve deviations from the predefined design and dosing schedule will also require efficient contacts with ethics committees, participants and sometimes regulators. To optimise data sharing and integration, it is encouraged that a totality of evidence approach – including analysis of emerging PK, PD and safety data – is applied.

2. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=5

There is an interesting shift in the scope of the revised guideline, compared to the previous version. Ten years ago, the guideline was basically limited to nonclinical issues for consideration of doses and designs of FIH studies with a SAD design. The current revision covers a much broader scope and is essentially meant to provide guidance for the design and dose selection of all early human drug studies, with different dosing regimens in different populations. The new guideline is restricted to studies that provide ‘initial knowledge’ of a new compound, but the recommendations apply to a number of important clinical pharmacological questions that cover major parts of early drug development. These questions deal with the ways in which the body handles the drug and how the drug affects the body and typically also include the interactions with foods and other drugs, and the influence of demographic variables and (concomitant) disease. These issues are not limited to healthy young males typically used in most early human studies but are also relevant to other populations including patients. In general, studies in special populations will require an integration of all the relevant preclinical and emerging human data, taking clinical pharmacological principles and disease characteristics into account.

The expansion of the scope of the guideline also comes from the need to be in line with progress in CT science. The traditional stepwise drug development paradigm that classifies studies in phases is slowly changing to an approach related to specific objectives 13. Modern tools – such as clinical biomarkers of efficacy and safety – and practices nowadays allow more extensive data retrieval and interpretation from smaller and better designed studies.

Appropriate use of such tools has the potential to enable a faster progress into therapeutic trials. However, this faster progress should not come at the cost of an increased risk to study participants. A thorough understanding of the value and the limitations of such tools/methods/technologies and appropriate use of the data derived from these should ultimately serve to promote a safe achievement of the trial objectives and a better use of resources.

Clearly the higher volume and complexity of data to analyse must also be matched with an increased capacity of scientific analysis and scrutiny in support of decision making during the trial.

3. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=5

The first edition of the guideline only referred to the EU Directive 2001/83/EC and its Annex and amendments 14, which apply to the request for authorisation of a clinical trial to competent authorities. Many parts of this directive have since been implemented in national laws or described in specific guidelines or new directives. The new FIH guideline refers to a long list of ‘good clinical practice’ documents for different drug classes, relating to production and quality and preclinical experiments, which should be considered when designing an early human clinical study. Other listed documents are related to specific cases, such as the evaluation of QTc‐prolongation, cancer drugs and vaccines.

4. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=6

This section of the guideline acknowledges the risks that are inherent to administration of new drugs to humans. These risks are linked to uncertainties related to the drug, to the characteristics of study participants and to study procedures. The guideline states that potential risks and appropriate mitigation strategies should be identified during the design of each study. Important aspects to consider are the quality of the IMP (https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=7), the available non‐clinical data (https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=7), the selection of the doses to be administered (https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=11), and precautionary measures during the conduct of early drug studies (https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=14). While more detailed systematic approaches can be found elsewhere 5, 15, in most cases, risk evaluation should include a detailed analysis of the mechanisms of action of the drug, its PK properties (including relation to levels of biological activity), pharmacological selectivity (and secondary activities), the predictability, measurability and manageability of potential adverse effects and vulnerabilities and disease characteristics of the study population. The analysis should rely on preclinical experiments, previous experience with similar compounds and action mechanisms, and will generally require the input from different experts to ultimately make sure that the trial objectives can be safely achieved. It is not only of utmost importance that the best possible predictions are made, but also that the right data is gathered and analysed to confirm, amend or refute validity of these predictions. The complexity of this integrated approach to early drug development may increase the risk of human studies, which is an important reason why a high quality in scientific scrutiny, supportive documentation, collaboration and experimental setup is crucial – especially for first‐in‐class compounds where uncertainties are higher.

As knowledge of the IMP increases (and uncertainty is reduced proportionally) along the course of drug development, the general principles of this section might be read as going beyond the scope of the guideline and actually apply to other later stages of drug development. Another, but not less important, general consideration is that for every individual trial a risk management system that is fit to deal with the uncertainties and potential risks needs to be in place and undergo constant update and improvement as data come in.

5. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=7

While the quality section has changed relatively little since the previous version, it stresses that quality of the IMP should never be a source of hazard to humans. One of the objectives of drug development is optimisation of the IMP for its use in clinical practice. This usually leads to changes of the formulation (solvents, additives, release matrices etc.) during the programme, but often also the production process or the drug itself (salts or other conjugates). For a new drug, it is inevitable that the rigour of ‘good manufacturing practices’ will improve during the development process. However, the guideline stipulates the need for an adequate level of quality characterisation at every stage. It is important that the quality considerations address study‐specific risk factors. For an FIH study, dose predictions from preclinical studies should consider the possibility that the drug formulation which was used in the animals could differ across experiments, and also to the drug that will be used in humans. This may influence the kinetics of the compound. This should be carefully considered during the prediction of dosing schedules for FIH studies and SAD or MAD protocols, and the selection of the appropriate dose and formulation. The properties of the study drug can also affect the stability, solubility, absorption to containers or infusion material and so on, which may all influence the extent and variability of exposure and bioavailability. Dilution to very small doses (e.g. to increase the versatility of dose selection) further increases the risks of error. Similarly, if changes to pharmaceutical quality are implemented during early human development of the IMP (or if novel formulations are explored in the course of an FIH), this should be kept in mind as a factor of increased uncertainty.

6. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=7

The information from preclinical experiments that should be obtained to support the design of early human drug studies is largely described in other guidelines (e.g. ICH M3(R2) 16). This current guideline, as do most of the recent nonclinical regulatory guidelines, discourages the use of nonrelevant animal species – often performed just to fulfil formal regulatory requirements – and specifically encourages the use of in silico tools and (humanised) in vitro methods in replacement of in vivo animal studies whenever possible, potentially foreseeing a future in drug development not entailing the use of animals. For the present, in accordance with the other two 3Rs principles of reduction and refinement, the guideline encourages scientific rigour and high‐quality animal study design/conduct in order to ensure reliability of results, avoid duplication of studies and minimise distress for the animals involved.

The Investigator's Brochure (IB) of an FIH dossier contains a wide array of preclinical studies, so it requires considerable expertise to understand the outcomes of each type of experiment by itself. Every preclinical result can affect a different aspect of the FIH design, and a team of experts is typically involved in decisions on doses, measurement schedules, safety measures, pharmacological biomarkers, specific in‐ and ex‐clusion criteria and other aspects of the design. This process is increasingly aided by complex translational models such as physiology‐based PK/PD modelling. Such integrated assessments are an important reason why early drug studies in humans are generally safe. However, the process can be difficult to follow for most parties outside the sponsor team. Moreover, integrated PK/PD‐type analyses can in cases be time‐consuming and complicated, and not easily applied to adaptive early human trials. Section 6 suggests that a tabulated summary of all relevant nonclinical data should be provided, as an appendix to the IB. This is undoubtedly useful, but a table does not automatically lead to a better understanding of the information. A few approaches to integrate data from different sources have been suggested, including Napier ‘knowledge plots’ 17 and more recently a relatively simple spreadsheet tool to summarise and integrate all results from the IB and emerging human studies 18.

6.1. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=8

The demonstration of the relevance of animal models chosen for preclinical testing is a paramount consideration to enable the correct integration of preclinical data for human risk assessment.

Predictions of pharmacologically active doses in humans (including starting dose and maximum effect dose) require an assessment of the binding and activity characteristics of the molecule in vitro and, in most cases, in vivo. Potential differences in PD activity between animals and humans should be identified and then factored, together with differences in target distribution/expression levels and PK aspects, to assess the relevance of the animal model for the clinical study design.

This is particularly the case for drugs that act on key nodes of complex regulatory systems, which are increasingly being developed. TGN1412 was an example of such a target, and of the fact that its pharmacological and physiological properties may differ considerably among species. Relevance of animal models is also important in relation to complex human diseases, such as neurodegenerative disorders, stroke or psychiatric conditions. The complexity of the human brain, the long duration and cascadic dysregulations and interactions of degenerative and regenerative processes and the poor understanding of psychopathology are all aspects that thwart the development of predictive animal models for central nervous system disorders.

The uncertainty concept is further detailed in the nonclinical section, which lists points to consider when selecting the test systems that will support entry into FIH CTs. Depending on the type of IMP, interspecies differences between animals and humans can be larger or smaller. The degree of relevance to humans of the nonclinical test systems can be considered a base factor increasing more or less the level of uncertainty that ultimately guides the CT design and monitoring set‐up.

6.2. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=9

A good understanding of the nature of the target and of its related downstream effects is essential for the understanding of the efficacy and safety of the compound. Furthermore, it can improve the interpretation of nonclinical safety profile of the drug and ultimately a more solid translation of this profile to human risk assessment. This does not mean that everything must be known about that target to progress to CTs, but it rather encourages identifying the limitations in knowledge. Knowledge gaps can be acceptable, but only if they do not translate into higher risk for subjects via appropriate mitigation in study design.

These considerations pertain to the primary as well as secondary targets. The insufficient characterisation of potential drug‐related targets is one of the hypotheses put forward in relation to BIA10‐2474. It is known that many FAAH inhibitors also inhibit other enzymes 19. Recently, independent studies with cultured human neurons provided indications for neurotoxicity of BIA10‐2574 that could be attributable to off‐target inhibition of other enzymes 20. This suggests that the events in Rennes could be related to poor specificity of BIA10‐2474 8 in combination with the high doses tested. Secondary pharmacological effects are inferred from in vitro experiments that screen for nonspecific binding of the molecule to a wide array of pharmacological targets. Nonspecific binding should be carefully considered in the estimation of the safety window of the compound and examined in more details if the secondary targets seem relevant and the selectivity margin is small.

6.3. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=9

This section of the guideline describes the importance of molecular pharmacodynamic properties for risk estimation. Understanding of PD should not be limited to the target engagement, but form part of the integrated human risk assessment that precedes the start of an FIH trial. This takes into account that a prolonged effect (by slow turn‐over of the target, long half‐life of the compound and strong binding) may lead to slower or poor reversibility of effects in an FIH study or to an unwanted long‐lasting target inhibition during multiple dosing schemes. Other factors that may increase the risk of unexpected adverse effects are a steep concentration‐effect curve, multiple off‐target effects or the novelty of the target.

In support of multiple dosing in humans, the new guideline requests specific attention to potential differences in PD effects between a single and a multiple dose regimen. This should not necessarily be read as a requirement for repeated‐doses in animal PD studies, but rather as a plea to appropriately assess how PD‐driven effects could change in magnitude/duration as a result of multiple dosing. In the case of BIA10‐2474, FAAH inhibition persists well over 24 h, despite complete clearance of the drug from plasma 8. Similar dissociations can be caused by long‐lasting secondary pharmacological or downstream effects, by an active metabolite with a long half‐life, or accumulation of drugs in target tissues (lipophilics). Such aspects should deserve specific scrutiny during the study design, since this may signify an increase of effect over time without accumulation of plasma drug concentrations.

For all the above‐mentioned reasons, the inclusion of a clinical PD biomarker, where possible, is encouraged in order to verify assumptions made from animal studies and data modelling. PD markers should be considered for the primary pharmacological effect, but also for secondary effects that might become clinically relevant in the planned dose range.

6.4. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=10, 6.5 https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=10 and 6.6 https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=10

The new version of the guideline does not introduce any specific outline of the nonclinical safety studies needed to support FIH studies, since the relevant information is provided by the well‐established overarching nonclinical ICH guidelines (ICH M3 16, S6 21 and S9 22). Traditionally, significant toxicological effects seen in animals are used to define a range of exposure where these are unlikely to occur in humans, usually based on the no‐observed‐(adverse)‐effect level (NOAEL). This is used to determine a maximum exposure that is not to be exceeded in the study subjects. To screen for the development of similar toxicities in humans, appropriate provisions for monitoring are included in early clinical studies, provided that these are available and feasible. The new guideline stresses that consideration of toxic mechanisms is essential to reduce the risk in human studies. Some toxicities are irrelevant for humans because they involve species‐specific immunological (e.g. reaction in animals to injection of a human antibody) or metabolic mechanisms.

It is interesting to note that the revised guideline specifies how to handle animal toxicities caused by the intended PD effects of the IMP. It states that these should not be dismissed in human risk assessment and their relevance is to be interpreted considering interspecies differences in PD and PK, in addition to the clinical dose range to be explored and the possibility to monitor progression into severe/irreversible consequences. Toxicities seen in animals in far excess of clinically relevant levels will usually not necessitate further mechanistic studies. However, this requires a reliable prediction of the pharmacologically active and therapeutically relevant dose range. Drug concentrations (C_max) can be used in this case to integrate the preclinical experiments in the IB 18, and this obviously requires pharmaco‐/toxicokinetic information for all species that are examined in the preclinical programme. In principle, unacceptable PD effects can be avoided if the concentrations and the pharmacological effects stay within the therapeutic range. During the early human studies, this can be monitored with an ongoing evaluation of PK and PD.

7. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=11

7.1. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=11

A multidisciplinary ‘totality of evidence’ approach is stressed throughout this section, which has been created to tie together the preceding considerations on the IMP/drug target/nonclinical knowledge to actual study design principles. It also encourages the use of in silico data modelling to support these. Given the differences in experimental set‐up and study objectives between healthy volunteer and patient trials, starting dose considerations have been separated based on the population of interest for the FIH study.

The guideline emphasises that observed ‘exposures’ in early human studies are more relevant than doses. The term ‘exposure’ is used because a given effect can be primarily related to maximum plasma concentrations (C_max for drugs with a reliable penetration and a direct effect at the target site), accumulated concentrations (AUC where the effect of the drug is more related to average concentrations or prolonged exposure), minimum concentrations (if drug activity should remain above a certain trough level, like for antibiotics or antiepileptics) or concentrations in cerebro‐spinal fluid or other relevant compartments. In some cases, the best measure of exposure can be a closely drug‐related PD effect (e.g. prothrombin time for vitamin K antagonists). Relevant measures of exposure can also differ for therapeutic and toxic effects, or for primary and secondary pharmacological effects. Exposure should therefore always be related to the observed effects in animals or the emerging effects from previous doses or other trials. The targeted exposure range for the early human studies (with a maximum) should be justified and predefined in the protocol. The guideline also states that if necessary, incoming information should be used to adapt the doses, and sometimes, this may require a substantial amendment. These new recommendations have important consequences for the performance of early human drug studies.

7.2. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=11 and 7.3. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=12

The FDA Guidance on Maximum Starting Dose from 2005 23 recommends that the maximum starting dose in humans should primarily be calculated on the basis of safety levels (NOAEL). However, safety predictions may be incorrect (and misleading) if the animal models are not sufficiently predictive, like for TGN1412. For other drugs, improved ‘safety’ is part of the design of a drug's pharmacological features (such as increased selectivity and indirect modulation). In either case, the NOAEL will be high if preclinical experiments show that the drug is (perceived to be) very safe, and so will be the calculated starting dose. This dose may even exceed the pharmacologically active dose (PAD) or the MABEL 24. The introduction of the MABEL in the first version of the EMA FIH guideline recognised the importance of the physiological impact of the drug, rather than its mere interaction with a pharmacological target. The MABEL and the PAD were suggested as supplementary approaches to the NOAEL, which in practice still forms the basis for many starting dose determinations because of its easy applicability. The new version emphasises the need for good predictions of exposures and effects in humans, also using appropriate humanised in vitro and in silico models. In most cases, the starting dose should be (just) below pharmacologically or biologically active levels. In addition, the anticipated therapeutic dose (ATD) should be estimated – a concept that was not used in the previous FIH guidelines of the EMA or FDA. The starting dose in patients can be adapted, for instance if the disease state influences the PK or PD of the drug or if a therapeutic effect is considered essential. The guideline also suggests a more cautious design in the first patient studies (e.g. to start with SAD), depending on the vulnerability of this group, interactions with the disease state or population‐specific uncertainties.

7.4. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=13, 7.5. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=13, 7.6. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=14 and 7.7. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=14

These sections recapitulate the ‘totality of evidence’ considerations of Section 7.1. Dose escalation should be based on predicted and observed exposures, in relation to the effects in animals and study participants. Dose selection, escalation steps and maximum exposure should consider the dose/exposure‐response curves for relevant (desirable and adverse) effects, including the steepness and the plateau of the curve and the anticipated therapeutic dose. The reliability of these estimates may vary with the nature and novelty of the drug, the accuracy of biomarkers etc. The consequence of these recommendations is that the exposure and the effects of the drug should be monitored during the trial (in addition to safety) and that planned dosing schedules may have to be adapted to the findings. More cautious approaches are warranted if the level of knowledge/uncertainty before the trial starts is limited and if monitoring is impossible or unreliable.

The currently implemented FDA and EMA guidance documents mainly address safe starting doses and offer limited guidance on maximum allowable doses in clinical trials. The novel EMA guideline requires that a clear limit to exposure is provided in the protocol. If emerging results suggest that it may be necessary or useful to exceed the planned maximum exposure, a substantial amendment with justification is required. The maximum exposure level should be carefully justified in relation to the anticipated therapeutic dose, and the dose/exposure‐response and ‐toxicity curves of the compound. The expected variability of dose/expose‐effect relations in patient populations can be taken into account for the planned dose/exposure‐range. Irrespective of the characteristics of the study subjects, the use of doses above the pharmacodynamic range is discouraged and limited to the cases where a safe exploration of supra‐therapeutic doses is considered useful to inform safety aspects, such as potential for QT prolongation (ICH E14 25) or studies in individuals with limited drug clearance (EMA Guidelines on the evaluation of the PK of medicinal products in patients with decreased renal 26 and impaired hepatic 27 function). In essence, this limits the determination of the maximum tolerated dose (MTD) as a concrete objective for FIH studies. The size of this safety window should be rationally predetermined, based on the maximum pharmacological effect, the variability of responses and kinetics in different populations, the exposures at which unacceptable adverse effect are anticipated and their measurability and manageability.

The principles of ‘totality of evidence’ also apply to transition to the MAD study, which is specifically mentioned in the new version. Results from preclinical studies as well as emerging data from previous SAD studies should be used to plan the multiple dosing schedules. Analysis of both PK and PD is required to predict accumulation of a drug or a metabolite with a long terminal half‐life, prolongation of the effect of a compound with a long residence time, development of tolerance etc. Depending on the predictions, their uncertainties, as well as the possibilities for monitoring, the starting dose, the escalation steps and the design of the MAD study can be more or less cautious.

The choice of route of administration for dosing in humans should be based on the characteristics of the IMP, the intended therapeutic use and adequately supported by the nonclinical data. In the case of an intravenous administration, a slow infusion may be more appropriate than a bolus injection. This would allow for a timely discontinuation of the infusion to mitigate an acute adverse event.

8. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=14

8.1. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=14

The section on planning and conduct of early human studies has been the most extensively revised part of the original guideline. The main reason is the expanded scope of the guideline which now covers integrated protocols, given the more frequent inclusion of FIH administration in their context.

8.2. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=15

8.2.1. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=15 and 8.2.2. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=15

The underlying message of these new sections on protocol designs is that more complex trials are equally acceptable as more basic/traditional designs. However, a more articulated trial design often relies more on assumptions that are to be confirmed during the study itself. Given the single regulatory review under which such studies are approved, inevitably a more thorough description of potential risks in the protocol and a solid risk management framework are required. In short, a risk‐proportionate approach to CT planning and conduct is encouraged 28. Obviously, the documentation and supportive data to be submitted to the regulatory authorities should be of appropriate quality and detail to allow approval of such protocols.

The guideline advises on what overlap can be accepted between study parts. However, this is to be seen only as a general framework given IMP‐specific characteristics that influence the duration of dosing and washout period needed. The key aspect is that appropriate knowledge should be available for review before progressing to the following cohort/study part. The exact type/amount of supportive data for decision making should be outlined in the protocol and agreed by the regulatory authorities. Having tight deadlines for dose escalation/progression to other study parts should not mean that decision making is to go ahead with incomplete data sets.

If the complexity of the protocol makes it impossible to outline exact doses to be given in later parts of the study, the criteria for decision making should be outlined. These considerations can pertain to PK [exposure limits (cf. NOAEL), deviation from linearity etc.], PD (achievement of prespecified target effect, lack of effect progression/attainment of maximum effect etc.) and safety (related to (im)possibilities for monitoring, based on anticipated PD effects and animal observations – see https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=18 and https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=19).

8.2.3. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=16

The choice of study subjects is a complex ethical matter that goes beyond the remit of the EMA and its guidelines. However, this section outlines points to consider in this respect, duly complementing the considerations on dosing selection that are inevitably affected by this specific aspect of trial set‐up. In addition, the guideline encourages designing inclusion/exclusion criteria via the definition of a base set of vital sign/physiological parameters, as recently suggested in some publications 29. The guideline suggests that reference should be made to normal ranges, although deviations are acceptable. In this respect, it is important to consider that laboratory ranges in hospitals are selected for the identification of pathological conditions with a reasonable sensitivity. This is a different premise than the distribution of a parameter in the general population. The range of normality in a healthy (young) population is often larger than the reference ranges that are used for disease diagnostics 30. Similarly, apparently ‘abnormal’ ECG configurations frequently occur in completely healthy young well‐trained individuals 31.

8.2.4. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=16, 8.2.5. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=17, 8.2.6. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=17, 8.2.7. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=17 and 8.2.8. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=18

These sections address the design of the study (or study parts) and the transitions between consecutive doses (between individuals, dosing cohorts and study parts). All these aspects require ‘prespecification’ and ‘monitoring’. In essence, the framework of the protocol should be described unambiguously, including the criteria and the limits for anticipated adaptations (of doses or sometimes other design aspects), and the processes (methods, timing etc.) on which these decisions are based. If unexpected emerging information necessitates changes from the prespecified plan, this will require a substantial amendment with regulatory/ethics approval. Adaptation requires sufficiently rapid processing of emerging information. Since the most predictable safety issues are usually related to a compound's (primary or secondary) pharmacological properties, PD assessments are also important for the recognition of developing toxicity at an early stage. Limitations to monitor the properties of the drug that are relevant for safety will increase the uncertainty and hence require more cautious approaches. These principles apply to all the different sections relating to study design aspects.

As for the other sections in the guideline, the degree of uncertainty present before study start is mentioned here, as this is a key consideration that will influence the study design. This includes the number of dosing steps and size of study cohorts, what monitoring is to be routinely conducted, as well as its timing and duration. The monitoring plan, which should include also additional investigations as needed (such as radiological or PD assessments), should be adaptable based on emerging data. This will increase the understanding of treatment‐emergent adverse events (AEs) and their relevance for dose escalation and/or study continuation. The actual time course of drug concentrations and effects may affect the planned length of follow‐up (normally, until complete resolution of drug effects). The revamped section on sentinel dosing, with using sentinel advocated for any cohort and in any study part, might look very rigid at first read. However, it allows flexibility to initiate dosing in new cohorts also in the absence of sentinels. The important key to obtain this flexibility is having sufficient knowledge of the IMPs PD, PK and safety in order to be comfortable that specific key junctions – i.e. approaching target saturation levels, reaching the maximum clinical exposure levels defined in the protocol, and threshold of non‐linear PK – are not going to be encountered in the specific cohort where the sentinel approach is not to be applied. In general, if the ‘safety properties’ of the drug (including PK and PD) are more predictable, more subjects can be dosed concomitantly. This requires continuous monitoring of the ‘predictability’ of these safety features. The criteria for deviations from sentinel dosing should be part of the protocol (or amendments) and approved as part of the regulatory review. A critical appraisal of emerging data should be made before initiation of any study cohort or part. New data obtained should always be reviewed considering previous knowledge in order to readily halt or adapt dosing (or dose escalation). Review of all previous cohorts' data in a cumulative manner is to take place to inform decision making.

The type/amount of data to use for every decision‐making step should be defined in the protocol and be proportional to the level of uncertainty. The totality of safety data collected is required for review before progression, while sparse PK sampling and less comprehensive PD data can be accepted if appropriately justified.

8.2.9. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=18 and 8.2.10. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=19

In relation to stopping rules, a plea for clarity is made in the revised guideline. Stopping rules to dosing the individual subject, to dosing other subjects in the same cohort, to proceed with a subsequent cohort (or study part, or the study altogether) should be described and justified in the protocol. For all of these, the consequence to the stopping rule being met – usually a final end of dosing, a temporary halt with (or without) adaptation of the dose escalation/frequency of dosing – should be stated.

A specific paragraph states that also the occurrence of moderately severe AEs should be considered as stopping criteria. This is particularly relevant if they are related to the same (primary or secondary pharmacological) mechanism.

When it comes to provisions on monitoring and communication of AEs, the revised guideline expands considerably on the level of detail to be provided in the protocol. The duration and the location of monitoring should be justified and can be adjusted to the level of potential risk to the subject in specific situations.

In addition, the level of training of study monitors, access to treatment allocation codes, and treatment protocols to initiate in case of specific AEs are all specifically mentioned.

In relation to the handling of AEs, regulators now request a description of the communication plan – between staff involved in the study, health care facilities that might come into play in case of emergency, the different study sites (if applicable), the sponsor and the regulatory authorities. This should mitigate the possibility that faults in communication – as the chronicles of events from the BIAL trial and the subsequent report from the Inspection Générale des Affaires Sociales, or General Inspectorate of Social Affairs (IGAS) have shown – should lead to dosing further subjects after a serious adverse event has occurred.

8.3. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=20 and 8.4. https://bpspubs.onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1111%2Fbcp.13550&attachmentId=2215672579#page=20

The plea for clarity in the protocol also extends to the need to clearly document the appropriateness of the composition of any data review group responsible for dose escalation (or any other relevant) decision. In addition, it is requested that a background rationale for proceeding to dose escalation should be on record at the study site and made available for potential future GCP inspection.

Finally, also in view of the fact that the EU does not have an FIH/Phase I unit accreditation scheme, it is required that the protocol includes appropriate information in relation to the adequacy of the facilities where the study is taking place. The incorporation of PD biomarkers in the study, and their rapid processing and analysis, is a major challenge for early clinical studies. Traditionally, such studies are under a high time pressure, which is perhaps understandable if the main outcomes are ‘pharmacokinetics, tolerability and safety’. However, the revised guideline puts more emphasis on the ongoing determination of dose/exposure‐effect relations of the drug, primarily because this is a safer approach for compounds carrying a higher uncertainty 5, 15. In addition, putting more weight on PD will deliver valuable information about target engagement, pharmacological activity and physiological effects of the drug and contributes to the validation of biomarkers that can be used in later clinical trials. These aspects are increasingly considered to be essential for a successful drug development programme 24, 32. The option of spending extra time and effort in early clinical trials proposed in the guideline may therefore be well spent, even in cases where following all provisions included in the guidelines might not be strictly required for safety purposes.

The leitmotif of the whole document is that a risk‐proportionate approach to the study logistics is seen as key to enable the safe achievement of the study plan. The different obligations of investigators, sponsors and regulators/ethics committees during the execution of an early human development study will require clear arrangements regarding the timing of communication, shared information, decision criteria, responsibilities and expectations. For many investigators and sponsors, the practical consequences are considerable. Considering their increasingly active roles, regulators and ethics committees will also need to adapt their processes. It should also be noted that a much closer involvement of the authorities in the actual decisions regarding dosing and safety might impact their legal positions as independent regulators and supervisors. For many sites, the practical and formal consequences of the FIH guideline for the different parties still largely need to be worked out in more detail and will be linked also to how the national competent authorities will interpret the provisions thereof.

The principles of the new guideline are sensible and should lead not only to safer but also to more informative early development programmes. These principles should be used conscientiously and judiciously by all the stakeholders involved. If the implementation is too rigid or literal, or if the system formed by the multiple stakeholders involved is pushed to the limits, this can easily reduce the feasibility of an FIH study without necessarily increasing the safety of participants. The novel provisions included are mostly based on pharmacological and biological principles of drug action, which should be ‘common sense’ to the clinical researcher. The same should also be kept in mind when the guideline is applied to actual studies.

Competing Interests

There are no competing interests to declare.

The views expressed in this article are the personal views of the authors and may not be understood or quoted as being made on behalf of or reflecting the position of the EMA or one of its committees or working parties.

Supporting information

Supp Mat Guideline on strategies to identify and mitigate risks for first‐in‐human and early clinical trials with investigational medicinal products

van Gerven, J. , and Bonelli, M. (2018) Commentary on the EMA Guideline on strategies to identify and mitigate risks for first‐in‐human and early clinical trials with investigational medicinal products. Br J Clin Pharmacol, 84: 1401–1409. doi: 10.1111/bcp.13550.