Abstract
The primary purpose of this study was to assess the translatability of preclinical to early clinical tolerable and pharmacologically active dose ranges for central nervous system (CNS) active drugs. As a part of this, IBs were reviewed on reporting quality. Investigator’s Brochures (IBs) of studies performed at the Centre for Human Drug Research (CHDR) reporting statistically significant results of CNS activity related to the drug’s mechanism of action were included. The quality of IBs was assessed based on the presence of a rationale for the chosen animal model, completeness of pharmacokinetic (PK) results in reporting and internal validity information of the preclinical evidence. The IB-derisk tool was used to generate preclinical and early clinical data overviews data. For each compound, the overlap between pharmacologically active dose ranges and well-tolerated levels was calculated for three pharmacokinetic (PK) parameters: human equivalent dose (HED), maximum plasma concentration (Cmax) and area under the curve (AUC). Twenty-five IBs were included. In general, the quality of reporting in IBs was assessed as poor. About a third of studies did not explore the entire concentration-effect curve (pre)clinically. Single dose tolerability ranges were most accurately predicted by Cmax. Human equivalent dose and AUC were the best predictors of pharmacologically active ranges. Tolerable and pharmacologically active dose ranges in healthy volunteers can be reasonably well predicted from preclinical data with the IB-derisk tool. The translatability of preclinical studies can be improved by applying a higher reporting standard in IBs including comparable PK measurements across all preclinical and clinical studies.
Similar content being viewed by others
Introduction
Drug development programs for neurological and psychiatric diseases have a high failure rate in both phase II and phase III [1, 2]. Reasons include the lack of safety and efficacy in clinical stages of drug development [1,2,3,4]. Next to that, a relatively large proportion of dose reductions of novel central nervous system (CNS) drugs is needed after marketing approval due to safety concerns [5]. Lastly, the poor translatability of preclinical experiments to clinical studies is often cited as a cause for these high attrition rates [6, 7].
For clinical researchers, the primary source of preclinical data for novel investigational products is the Investigator’s Brochure (IB). The IB is an obligatory part of a research file for clinical studies with an investigational medicinal product (IMP), which contains all nonclinical data relevant to studies in human subjects (supplemented with subsequent clinical results) [8]. According to the International Conference on Harmonization (ICH) guideline for Good Clinical Practice (GCP), the purpose of the IB is “to provide the investigators and others involved in the trial with the information to facilitate their understanding of the rationale for, and their compliance with, many key features of the protocol, such as the dose, dose frequency/interval, methods of administration and safety monitoring procedures” [8]. Minimum requirements for the IB are described in this guideline [8]. Despite this guidance, the content of IBs is highly variable in practice [9].
The European Medicines Agency (EMA) “guideline on strategies to identify and mitigate risks for first-in-human (FIH) and early clinical trials with investigational medicinal products” provides guidance on the quality and choices of preclinical safety and efficacy studies that should be performed prior to an FIH study and on how a safe starting dose should be determined [10]. According to this guideline, the starting dose should be based on both preclinical safety studies and efficacy or pharmacodynamic (PD) experiments [10]. Safety can be quantified by the no observed adverse effect level (NOAEL) in the most sensitive relevant species, and efficacy or (pharmacological) activity by estimations of the minimal anticipated biological effect level (MABEL), the pharmacologically active dose (PAD) and/or anticipated therapeutic dose (ATD) in humans [10]. In practice, however [10], the starting dose is often primarily based on safety considerations, as a fraction of the NOAEL [11]. This focus on NOAEL instead of a joint account of pharmacologically active levels can have disastrous consequences as observed with TGN1412 and BIA-10-2474 [12, 13] When TGN1412, a CD28 superagonistic antibody in development for the treatment of chronic lymphatic leukaemia and rheumatioid arthritis, was first administered to humans, it caused a cytokine release requiring intensive care treatment in all healthy individuals administered with the drug [12]. The starting dose for this study was based on the NOAEL and a factor 500 lower than the NOAEL, but at this dose cytokine release was already observed preclinically, indicating that the starting dose should have been even lower [12]. In the clinical trial with BIA-10-2474, a fatty acid amide hydrolase (FAAH)-inhibitor in development for diseases in which elevated endocannabinoid tone might be beneficial such as pain, glaucoma and post-traumatic stress disorder, several subjects developed neurological damage and one subject died [13]. The preclinical data identified safety concerns in the form of serious irreversible adverse effects that were observed at varying dose levels across species [14]. There were no pharmacological measures applied in the clinical study to counter this risk and dose escalation was based solely on tolerability findings [14]. Doses were escalated to Cmax values ~12 times higher than levels of maximal FAAH-inhibition, leading to fatal consequences [14]. Not taking into account pharmacological active levels in early-phase clinical studies can also have less dramatic effects, as illustrated by the example of CEP-26401 [15]. The first clinical study of CEP-26401, a histamine-3 receptor antagonist developed to improve cognitive functioning in for example Alzheimer’s disease, schizophrenia or attention deficit hyperactivity disorder (ADHD), demonstrated improved cognitive functioning at the lowest dose tested [15]. Therefore, a second study was required to find the dose with the best balance between wanted (improved cognition) and unwanted (sleep inhibition) effects [15].
The aim of the current paper was to investigate the accuracy with which the IB preclinical package can predict tolerable and pharmacologically active dose ranges for CNS drugs in humans. We assessed the overlap between preclinical and clinical well-tolerated dose levels and pharmacologically active dose ranges. Furthermore, we checked whether both preclinical safety and in vivo pharmacology experiments were used when determining the starting dose for FIH studies as recommended by current EMA guidelines [10]. Also, we investigated the reporting quality in the IBs, and, for compounds where phase II or III clinical trials were performed in patients, we conducted an exploratory analysis of the translatability of preclinical pharmacological active ranges to therapeutic effective ranges.
Methods
Study and IB selection
We performed a structural review of all IBs in CNS drug development conducted between 2003 and 2019 at the Centre for Human Drug Research (CHDR). To adhere to confidentiality agreements agreed upon with sponsors and clients of the performed clinical trials, compounds were anonymised and no individual study results are described. In order to be suitable for analysis of the overlap of tolerable and pharmacologically active dose ranges, a dose range of tolerability and pharmacological activity had to be reported both preclinically and clinically. This meant that IBs were included if they included a statistically significant effect (as reported in the document) on any CNS activity related to the compound’s mechanism of action for at least two dose levels across the reported preclinical studies and in the associated Clinical Study Report (CSR). Preclinical in vitro studies were not taken into account as we aimed to investigate the predictivity of in vivo pharmacology experiments in animals. Studies testing combinations of drugs and studies performed for the purpose of method development with drugs already approved by the Food and Drug Administration (FDA) or the European Medicines Agency (EMA) were excluded. For approved drugs, the Summary of Product Characteristics (SmPC) is provided to the investigator instead of an IB and animal pharmacokinetic (PK) and PD data are not systematically reported. In case phase II and III therapeutic efficacy studies were performed for a compound, and an exploratory translatability assessment of therapeutic efficacy was performed.
IB-derisk tool
In this paper, the IB-derisk tool was applied to IBs of included compounds. The IB-derisk tool (https://www.ib-derisk.org) is a tool that can be used to integrate preclinical and clinical data reported in the IB and for the comparison of preclinical studies results with predicted or emerging human data while the clinical study is ongoing [16]. An illustration of this can be found in the publication by Cohen and colleagues, who applied the IB-derisk tool to the BIA-10-2474 study [14].
For detailed information on the IB-derisk tool, we refer to the original publication [16]. In short, all reported in vitro findings and in vivo efficacy and safety findings and their according exposure parameters, such as the human equivalent dose (HED), maximum plasma concentration (Cmax) and total exposure (AUC) are entered in a spreadsheet-like document. By doing so, missing PK parameters can be estimated by interpolation and extrapolation, taking dose duration (single or multiple doses), route and bioavailability, species and sex into account. All effects are then colour-coded. Desired pharmacological effects are indicated by green, mild manageable adverse effects in yellow, more severe adverse effects that could not be accepted in a clinical situation but without unacceptable health risks by orange, and severe irreversible adverse effects in red. The no observed adverse effect level (NOAEL) in the most sensitive species is indicated in purple, and in vitro experiments are in light blue. By applying this colour coding, sorting the data on dose (HED), concentration or AUC provides a quick overview of dose-, concentration- or exposure-response patterns of pharmacological and toxicological effects, and of deviations from predictable relationships [16].
Data collection
Data were collected from the source documents by the authors GSF and FMD. All in vivo behavioural pharmacology and safety preclinical study results reported in the included IBs were extracted, regardless of species or outcome. GSF and FMD also assessed the quality of reporting in IBs in a similar manner as described by Wieschowski et al [17]. As a part of this it was assessed whether a rationale for the chosen animal model was provided, the completeness of reporting of pharmacokinetic results of in vivo pharmacology experiments (including strain, sex, and route of administration) was assessed and internal validity information (randomization, blinding) of the preclinical evidence was assessed. GSF and FMD discussed their quality assessment in case there were any unclarities and the consensus answer was incorporated in Table 1.
In addition to preclinical data retrieved from the IBs, the aims and results of the actual early human studies were extracted from the protocols and clinical study reports (CSRs). If available, these results were complemented with other clinical data provided in section 4 of the IB (“Effects in Humans”). For all included studies, information on the mechanism of action of the IMP and whether the clinical study was an FIH study was collected. For FIH studies, the section in the protocol describing the rationale for starting dose selection was assessed on whether the starting dose was primarily based on preclinical safety or also on in vivo pharmacological effects. As part of this, it was assessed whether the lowest preclinical and clinical pharmacologically active doses were established.
Data analysis
Two separate analyses were performed to determine the predictability of preclinical findings in humans: one for tolerability and another for pharmacologic activity. The purpose of these analyses was to compare the tolerable and pharmacologically active ranges in laboratory animals to those found in humans. The ranges were calculated for all three exposure parameters: dose (HED), Cmax and AUC.
For tolerability, the ratio between the starting level of the clinical study and the NOAEL, and the ratio between the highest well-tolerated level in the clinical study and the NOAEL were calculated. The starting level for the clinical study was defined as ‘human lower range adverse events’ (HLRAE), and the highest well-tolerated level in the clinical study as ‘human upper range adverse events’ (HURAE).
The preclinical pharmacologically active range was determined by the ‘animal lowest’ and ‘animal highest’ dose-, concentration- or exposure-level, at which any effect associated with the proposed mechanism of action of the drug was reported in any animal species. These values were called ALRa ‘animal lower range active’ and AURa ‘animal upper range active’. Preclinical safety findings were often observed at higher levels than the highest level tested in efficacy experiments. If the safety issues were judged to be related to a compound’s pharmacological effects, this was used to determine the AURa. The human active ranges HLRa and HURa were defined as the lowest and highest level, respectively, in which pharmacological effects associated with the proposed mechanism of action of the drug were reported. An illustration of the calculations of the overlap between the preclinical active dose range with the corresponding clinical dose ranges for each exposure parameter (HED, Cmax and AUC) can be found in Fig. 1. Calculations were performed according to the formula below:
-
If (minimum value of either AURa or HURa–maximum value of either ALRa or HLRa)/(HURa-HLRa) < 0, then 0% overlap.
-
If (minimum value of either AURa or HURa–maximum value of either ALRa or HLRa)/(HURa-HLRa) > 1, then 0% overlap.
-
Otherwise: ((minimum value of either AURa or HURa–maximum value of either ALRa or HLRa)/(HURa−HLRa))*100%.
Results
General characteristics of studies
The search in CHDR’s database from 2003 up to and including 2019 returned 164 finished clinical CNS drug studies in healthy volunteers with corresponding IBs. Of these, 25 studies met the inclusion criteria (Fig. 2). The most common reasons for exclusion were studies being non-interventional (method development), or studies with a registered compound or patient studies (n = 106), studies focusing on drug–drug interaction (n = 13) and no clinical efficacy (pharmacological activity) results at two different dose levels (n = 8) (Fig. 2). The four studies that were excluded because there were no preclinical behavioural in vivo pharmacology experiments in animals performed concerned three studies in which only cell and cytokine responses were measured preclinically and a study in which only in vitro experiments were performed preclinically to assess the pharmacodynamics of the novel compound. The two studies that were excluded because there were no preclinical in vivo pharmacology results at two dose levels concerned one study in which several doses were tested in preclinical behavioural experiments, but only one dose had a statistically significant effect and one study in which only one dose was tested in a preclinical in vivo pharmacology experiment. These studies were not included in the analysis as the outcome could then be only 0% or 100% overlap. The two studies that were excluded because there was no efficacy (pharmacological activity) measurement in the clinical study concerned PK studies.
Of the eight studies that were excluded because there was no clinical efficacy (pharmacological activity) result at two different dose levels, there were four studies that measured a statistically significant effect in the clinical study, but only at one dose level. These studies were not included in the overlap calculations as the overlap could then only be 0% or 100%. There were four studies in which the results of the efficacy or pharmacological activity measurement in the clinical study were negative. This concerned a study measuring the effect of a single dose esketamine on driving performance compared to the effect of a placebo and positive control. A study measuring the effect of different doses of a novel compound (a selective muscarinic M 1-acetylcholine receptor agonist) on cognitive performance as measured by a battery of neurocognitive and neurophysiological tests in healthy elderly with below-average cognitive functioning. A study into the effect of different doses of a novel compound (a dual enkephalinase inhibitor) on neurocognitive and neurophysiological tasks and on a nociceptive test battery, and a study measuring CNS effects of different doses of a novel compound (guanylate cyclase stimulator) in healthy volunteers measured by a battery of neurocognitive and neurophysiological tests. In all four of these studies, the preclinical pharmacological activity experiments did demonstrate statistically significant effects.
Table 1 shows that 12 (48%) of the included studies concerned a First-in-Human (FIH) study. In five of these studies (42%), the starting dose was based on both safety and in vivo pharmacology findings. In seven (58%) studies the starting dose was only based on safety findings.
In 8 of the 25 IBs (32%) a rationale for the chosen animal model was provided. The motivations included availability (commonly used models), similar phenotype (symptoms), response to effective drugs (pharmacology), histology and biomarkers. There was no explanation as to the relevance of model choice when compared to other available options. In none of the IBs pharmacokinetic reporting was complete for all in vivo pharmacology experiments (Table 1). In most cases, PK values of behavioural or disease models in mice were not reported (Table 1). The strain and sex were frequently missing. None of the preclinical experiments in the included IBs fulfilled the criteria of internal validity as most animal experiments included a placebo arm, but blinding or randomization was not reported in any of the experiments (Table 1).
Tolerability assessment
Predictions of tolerability based on human equivalent dose
Table 2 shows that in most studies (N = 22, 88%), the starting dose (HLRAE) for the clinical study was at least a (rounded) factor 10 below the HED of the NOAEL (Ratio HLR/NOAEL). This is as expected since FDA guidelines propose 10 as the default safety factor [18]. Three studies that did not apply this method were the IB2, IB4 and IB21 studies. The studies of IB2 and IB4 investigated novel cannabinoid receptor agonists. In these studies, a starting dose of, respectively, factor 1.8 and 5.3 below the HED of the NOAEL was deemed safe by the investigators, as the NOAEL was based on transient, species-specific, monitorable effects on blood pressure, and considering the safety profile of other well-known cannabinoid agonists. The study of IB21 was no FIH and the starting dose was based on the results of a previous clinical study.
For HED, the human upper tolerability range HURAE surpassed the NOAEL in 9 out of 25 studies (36%): IB2, IB3, IB4, IB6, IB8, IB9, IB11, IB14 and IB20 (Table 2, Fig. 3). In all except two of these studies, the compound was well tolerated up to the highest administered dose range (HURAE), probably explaining why human doses could be escalated to levels beyond the NOAEL. Studies with IB2 and IB4 both involved experimental cannabinoid receptor agonists. The preclinical NOAEL for these compounds was based on cardiovascular side effects, which can be intensely monitored in humans and therefore be used to guide dose escalation. Although such effects also occurred in humans, they were not considered dose-limiting. For both cannabinoid receptor agonists, dose escalation was only halted when undesirable (reversible) psychiatric events occurred, which were in line with the action mechanism.
Predictions of tolerability based on C max
As shown in Table 2 and Fig. 3, the Cmax value of the HURAE surpassed the Cmax value of the NOAEL in 8 of the 25 studies (32%; IB2, IB4, IB5, IB6, IB8, IB10, IB15, and IB21). The findings with the cannabinoid receptor agonists of IB2 and IB4 were mentioned before. In the clinical studies of IB5, IB6, IB8, IB10, IB15, and IB21 the compound was well tolerated up to the highest HURAE, allowing the dose levels to be escalated to levels beyond the NOAEL.
Predictions of tolerability based on AUC
For AUC, Table 2 and Fig. 3 show that the HURAE surpassed the AUC associated with the NOAEL in 12 out of 23 studies (52%; IB2, IB4, IB5, IB6, IB8, IB15, IB16, IB18, IB20, IB21, IB23, IB25). In all these studies (except IB2 and IB4 as explained above), the compound was tolerated well enough to be escalated to levels above the NOAEL.
Other observations on tolerability
In none of the clinical studies serious adverse events or irreversible adverse events, related to the investigational compound, were reported. Four human studies (IB7, IB12, IB13 and IB17) showed unacceptable adverse events at levels below values associated with the NOAEL for all exposure parameters (Table 2, Fig. 3). In the study of IB7 with a GABA receptor modulator, volunteers experienced ataxia, imbalance, tiredness and drowsiness. Symptoms of comparable nature, such as somnolence and ataxia, were also observed preclinically, albeit at higher dose levels. In the clinical study of IB12 with a histamine receptor agonist, participants reported pseudo-hallucinations and experienced hypotension. Preclinically, decreases in blood pressure were also observed, but at much higher dose levels. Pseudo-hallucinations could obviously not be observed preclinically, but behavioural changes were observed in monkeys at much higher dose levels than those given in the clinical study. In the study of IB13 with a histamine receptor antagonist, subjects experienced moderate nausea and insomnia. Preclinically, increased wakefulness was also observed, but this was considered a desired effect that was observed at similar dose levels. Emesis only occurred in dogs at much higher dose levels than in the clinical study. For IB17 with a trace amine-associated receptor (TAAR) partial agonist, cardiovascular AEs of tachycardia, palpitations and orthostatic hypotension were observed. Increased heart rate was also observed preclinically, but at much higher dose levels.
Taken together, Cmax values seem to be the most accurate predictor of tolerability limits for CNS active compounds, with only 32% of clinical studies reporting well-tolerated doses above the NOAEL value, compared to 36% for HED and 52% for AUC. The percentage (16%) of clinical studies reporting unacceptable side effects at values below the NOAEL was similar for all three exposure parameters.
Pharmacological activity assessment
Preclinical and clinical pharmacologically active ranges
Eight (32%) IBs (IB2, IB5, IB6, IB8, IB18, IB22, IB23, IB24, IB25) reported statistically significant pharmacological effects at the lowest tested preclinical dose (Fig. 4). Hence, 32% of the preclinical studies did not cover the full concentration-effect range.
Nine (36%) clinical studies (IB1, IB13, IB14, IB15, IB16, IB19, IB20, IB21, IB23) showed a pharmacological effect at all administered doses (Fig. 4). For one study (IB23 with a GCase modulator) both the preclinical and clinical lowest tested dose was effective, meaning that no no-effect level was defined.
Predictions of pharmacological activity based on HED
On average, HED was the best predictor of pharmacologically active ranges, with 84% overlap between preclinical and clinical pharmacologically active ranges (Table 3, Fig. 4). For one compound (IB20) there was no overlap between the preclinical and clinical ranges for HED or any other exposure parameter. This involved a muscarinic receptor partial agonist. Memory testing showed improvement in healthy volunteers, at lower levels than in the preclinical experiments. Memory functioning in the clinical study was tested using the NeuroCart, which consists of a battery of drug-sensitive neurophysiological and cognitive tests [19]. Possibly the NeuroCart is more sensitive to drug effects than preclinical models available to test memory functioning, explaining why drug effects in humans were observed at lower exposure levels than in animals. Further dose escalation in the clinical study was prevented by adverse events that could be expected with muscarinergic agonists, such as increased blood pressure and hypersalivation. Blood pressure increases were also described in the IB in rats, at a similar HED but with higher Cmax and AUC values.
Predictions of pharmacological activity based on C max
The preclinical and clinical pharmacologically active range showed overlapping Cmax values in 64% of studies. There was no overlap for IB2, IB3, IB4, IB11, IB19 and IB20 (Table 3, Fig. 4).
In the clinical studies of IB2 and IB4, some participants experienced typical mental effects of cannabinoid receptor agonists, at Cmax values well below preclinically active levels. The psychiatric effects were considered dose-limiting in view of the anticipated therapeutic indications (analgesia and sedation).
In the clinical studies of IB3 and IB11, both with an orexin antagonist, preclinical pharmacological effects were observed at higher Cmax levels than in the clinical study. It is possible that more sensitive PD measurements for orexin antagonists are available in humans (using the NeuroCart) than in animals, to detect (subjective) reduced alertness, attention and vigilance.
In the clinical study of IB19 with a subtype selective purine antagonist, decreased interleukine-1β release was observed in humans at a somewhat lower Cmax value than preclinically, suggesting that humans are more sensitive to the effects of purine antagonists than animals.
Pharmacological activity was measured at a lower Cmax value in humans than in animals in the clinical study of IB20 with a muscarinic receptor partial agonist, again suggesting that memory tests are more sensitive to this class of compounds in humans than in animals.
Predictions of pharmacological activity based on AUC
Pharmacologically active AUC ranges in animals and humans overlapped in 78% of studies. There was no overlap between animals and humans for AUCs of compounds in IB3, IB4, IB15 and IB20 (Table 3, Fig. 4). In the clinical study of IB3 with an orexin antagonist, the preclinical effects were observed at higher AUC levels than in the clinical study. It is possible that—partly subjective—PD measurements are more sensitive to orexin antagonists in humans than in animals [19]. However, the overlap between the preclinical and clinical pharmacologically active AUC range in IB11 with an orexin antagonist as well was 97%. In general, the PK profile of this orexin antagonist was much more comparable between animals and humans than for the orexin antagonist of IB3. The same was found for the two cannabinoid receptor agonists. The AUC overlap of pharmacologically active ranges for the compound of IB2 was 100%. For IB4 there was no overlap because, in the clinical study, dosing was stopped for unacceptable (albeit pharmacological) mental effects at lower AUC levels causing detectable pharmacological effects in animals.
In the clinical study of IB15 with a subtype selective nicotinic receptor agonist, the preclinical pharmacologically active range was lower than the clinical active range. However, the lowest dose in humans already showed pharmacological activity, meaning that there could be an overlap, but this was not assessed. For the muscarinic receptor partial agonist of IB20, the NeuroCart could demonstrate pharmacological activity in humans at lower AUCs, than where effects occurred in animals.
Overall, the preclinical data predicted the pharmacologically active range in humans to a high degree, as indicated by an overlap of ≥ 80% in 18 out of 25 (72%) for HED, 15 out of 25 (60%) for Cmax and 19 out of 23 (83%) for AUC. A particularly poor preclinical prediction of the clinical active range (as indicated by ≤ 20% overlap) was shown in 2 out of 25 studies (8%) for HED, 7 of 25 (28%) for Cmax and 4 out of 23 studies (17%) for AUC.
Therapeutic efficacy assessment
Therapeutic efficacy studies were reported for only six of the 25 included compounds in our paper. For three out of the six compounds, there was a large overlap between the therapeutic effective ranges in patients and the preclinical and human pharmacologically active ranges. For the three other compounds, the therapeutic effective ranges were lower than the preclinical pharmacologically active range, but they corresponded closely with the clinical pharmacologically active ranges in healthy volunteers.
Discussion
The results show that, in general, tolerable dose ranges for clinical studies with novel CNS active compounds can be reasonably well predicted from preclinical data. Overall, Cmax corresponding to the preclinical NOAEL was the best predictor of the tolerable range in humans, although the observed adverse effects in animals (or any other dose-limiting effect) did not occur in 32% of the healthy volunteer studies. HED and AUC predictions based on this ‘default’ safety level were even more conservative (with poor predictability in 36% and 52%), particularly when the effects could be readily monitored in healthy subjects and used for dose escalation (e.g., using intensive cardiovascular monitoring, or repeated NeuroCart measurements for CNS-effects).
In 4 out of 25 studies (24%), the highest tolerated (administered) doses in humans (HURAE) were much lower than expected based on the NOAEL for all three exposure parameters. This concerned the clinical study of IB7 (GABA modulator), IB12 (histamine agonist), IB13 (histamine antagonist) and IB17 (TAAR partial agonist). In three cases (IB7, IB13 and IB17) the dose-limiting AEs reported by volunteers, including ataxia, hypotension, drowsiness, insomnia and nausea, were observed preclinically as well, but only at higher dose levels. Thus, a considerable proportion of CNS active agents (28%) seem to have more prominent effects in humans than animals. For some compounds, dose escalation was limited by psychiatric side effects, which are difficult to observe preclinically. This was the case for the histamine receptor agonist of IB12 at doses well below NOAEL. Mental effects also limited the dosing of the two cannabinoid receptor agonists (IB2 and IB4), but here the highest tolerated dose in humans (HURAE), was higher than the values associated with the NOAEL that was based on cardiovascular effects. These results emphasize that in FIH studies with CNS active compounds, researchers should pay special attention to the psychiatric effects of new compounds, as these cannot be reliably predicted from animal experiments.
In all studies included in this report, the observed adverse events were exaggerated pharmacological effects in line with the working mechanism of the compound and therefore predictable based on preclinical data. This illustrates the importance of monitoring pharmacological effects of compounds based on the mechanism of action. Monitoring based on translatable and thus predictable pharmacological mechanisms of actions can also include important off-target effects, which in the IB are presented as ex vivo or in vitro pharmacological binding studies.
The average overlap values of preclinical and clinical pharmacologically active dose ranges demonstrate that the prediction of clinical pharmacologically active dose ranges based on preclinical data of behavioural experiments is fairly reliable. With an average overlap of 84%, the HED was the best predictor for the pharmacologically active dose range. Possibly, this reflects a bias in reporting as the MABEL or PAD were most often based on the HED in the IBs included in this report. When looking at the PK parameter with the highest percentage of high preclinical and clinical active dose range overlap, AUC was the best predictor with 83% of the compounds having more than 80% overlap.
In cases where no overlap between preclinical and clinical pharmacological active dose ranges could be observed, humans were more sensitive to the effects of the compound. In the clinical studies with cannabinoid receptor agonists, it was not possible to dose up to the levels of desired pharmacological (analgesic, sedative) effects due to unacceptable psychiatric (but still pharmacological) effects observed at lower dose levels. Not only for psychomimetic effects but also for some other CNS effects, more sensitive methods are available in humans than in animals. Complex measures of memory or eye-hand coordination in the NeuroCart showed effects of orexin antagonists and cholinergic/muscarinergic agonists, at lower levels in humans than predicted from animal models.
Four studies were excluded from the quantitative analysis of overlapping exposure ranges because no statistically significant effects were observed on measurements of pharmacological activity/pharmacodynamics in the clinical study. In all four of these studies, statistically significant effects were observed in preclinical, behavioural experiments, but not in humans. One of the omitted studies in humans was not designed to measure pharmacodynamic effects of the compound, but to assess continuous driving performance (after tmax), which explains why no effect was observed. The other three studies concerned compounds that aim to modulate neuronal processes in the brain in the longer term instead of in the acute phase, which might explain why no statistically significant effect was observed in the single-dose clinical studies.
The EMA guideline on how to determine the starting dose for an FIH study was updated after the TGN1412 study. In this new guideline, published in 2007, it was recommended to base the starting dose not only on the NOAEL but also include the MABEL [8]. In our sample of FIH studies performed between 2003 and 2019, in 58% of the included studies, the starting dose for the clinical study was based on preclinical safety experiments (NOAEL) only. One of these studies was performed prior to 2007. This percentage is in line with other publications reporting that the NOAEL-based approach is still the most common method to determine the starting dose for an FIH study [11,12,13].
In line with previous research, our data show that important details of animal studies are poorly reported in IBs [17]. In none of the IBs blinding or randomization of the preclinical experiments was reported and most lacked important information, such as animal sex or route of administration. PK measurements were often missing for animal models of behaviour or disease. Although regulatory guidelines do not require the reporting of PK-analyses in each preclinical study, the translation to effective human dose ranges is not possible without exposure data. Since poor reporting of study design is often associated with an overestimation of efficacy outcomes, it means that ethics committees and other regulatory bodies could be allowing first-in-human trials to start on the basis of spurious results [17, 20].
For only six of the 25 compounds therapeutic efficacy study results were reported. For three of those, the therapeutic effective ranges in patients were lower than the preclinically pharmacologically active range, but in all cases, there was a good overlap between the pharmacological effect ranges in healthy volunteers and the therapeutic dose range. This relatively high translatability is contradictory to existing literature reporting high failure rates of translation of new investigational compounds that seem effective in preclinical experiments but fail in clinical therapeutic studies [21]. These findings may be biased to some extent as the decisions to advance these compounds to clinical trials in patients relied on consistent results from the preclinical and human phase I studies. Next to that, phase I studies often solely focus on tolerability, safety and pharmacokinetics instead of also including relatively basic human pharmacological characteristics of new compounds such as blood-brain barrier penetration, as done in included studies [22]. Another factor often cited as a cause for the high attrition rate in CNS drug development is the limited knowledge of receptor occupancy [23]. A possible solution to this problem is to perform more PET studies to study receptor occupancy [23]. The high attrition rate in CNS drug development can also be explained by poorly understood human diseases as psychiatric disorders are usually diagnosed based on a cluster of symptoms instead of a biological basis [24]. This leads to several problems, such as the animal model being a mismatch or simplification of the human disease. There are current initiatives to overcome these problems, such as the Research Domain Criteria (RDoC) initiative introduced in 2009, which aims to more precisely link treatment targets to dysfunctional mechanisms relevant to clinical manifestations [24]. By doing so, biomarkers aiming to characterize the pharmacological activity of novel compounds in early-phase clinical trials are being developed as recommended in several publications [19, 22, 25].
While the trends identified in our study are worth investigating, the limited number, diversity, and non-randomness (only studies in our own research institute were included) of the included studies make our findings suggestive rather than confirmatory. We performed this overview solely with IBs of drugs for which at least two (pharmacodynamically) active doses were identified in phase I trials to allow a comparison between animal and human ranges. The lack of a dose range meant that we excluded six studies because preclinical efficacy was only established at one dose level or there was no preclinical in vivo efficacy data and ten studies because clinical pharmacodynamics was only established at one dose level or there was no clinical pharmacodynamic data (Fig. 2). As such, these data cannot be used to compare the HED, Cmax and AUC regarding their ability to predict the presence of an effect in humans. Next to that, we used linear inter- and extrapolation to determine missing pharmacokinetic parameters in animals. Although such a strategy is common practice, it might lead to prediction inaccuracies for drugs with a non-linear pharmacokinetic profile, and PK/PD-based analyses might have been more reliable (albeit unfeasible owing to the lack of data in many cases). Also, the analysis was limited to studies of unregistered compounds mostly in healthy volunteers. Despite these limitations, our sample is likely representative of IBs in practice. IBs are all investigators have at their disposal when they study the pharmacokinetics, pharmacodynamics and tolerability of a new CNS active compound.
In this report, we applied the IB-derisk tool on a selection of 25 IBs and compared the predictions of tolerable and pharmacologically active dose ranges based on preclinical data to the results of clinical studies. The results demonstrate that tolerable and pharmacologically active dose ranges in clinical studies can be reasonably well predicted from preclinical data. Tolerability was best predicted by Cmax and pharmacologically active ranges by HED or AUC. We noted that despite recommendations by the EMA to base the starting dose on both NOAEL and MABEL, the starting dose is often solely based on the NOAEL. In line with current literature [17], the internal validity of preclinical experiments was poor and preclinical in vivo CNS experiments are often performed without reporting PK results. The translation of preclinical to clinical studies would benefit from complete and comparable reporting of PK measurements of both toxicity and efficacy experiments. This report further demonstrates that an integrated presentation of the contents of the IB, such as provided by the IB-derisk tool, can improve the translatability of preclinical to clinical data [10, 14, 16].
References
Arrowsmith J. Phase II failures: 2008–2010. Nat Rev Drug Discov. 2011;10:328–329. https://doi.org/10.1038/nrd3439
Arrowsmith J. Trial watch: phase III and submission failures: 2007–2010. Nat Rev Drug Discov. 2011;10:87 https://doi.org/10.1038/nrd3375
Hay M, Thomas DW, Craighead JL, Economides C, Rosenthal J. Clinical development success rates for investigational drugs. Nat Biotechnol. 2014;32:40–51. https://doi.org/10.1038/nbt.2786
Hwang TJ, Carpenter D, Lauffenburger JC, Wang B, Franklin JM, Kesselheim AS. Failure of investigational drugs in late-stage clinical development and publication of Trial Results. JAMA Intern Med. 2016;176:1826–1833. https://doi.org/10.1001/jamainternmed.2016.6008
Cross J, Lee H, Westelinck A, Nelson J, Grudzinskas C, Peck C. Postmarketing drug dosage changes of 499 FDA-approved new molecular entities, 1980–1999. Pharmacoepidemiol Drug Saf. 2002;11:439–446. https://doi.org/10.1002/pds.744
van der Doef TF, Zaragoza Domingo S, Jacobs GE, Drevets WC, Marston HM, Nathan PJ, et al. New approaches in psychiatric drug development. Eur Neuropsychopharmacol J Eur Coll Neuropsychopharmacol 2018;28:983–993. https://doi.org/10.1016/j.euroneuro.2018.06.006
O’Donnell P, Rosen L, Alexander R, Murthy V, Davies CH, Ratti E. Strategies to address challenges in neuroscience drug discovery and development. Int J Neuropsychopharmacol. 2019;22:445–448. https://doi.org/10.1093/ijnp/pyz027
European Medicines Agency. Guideline for good clinical practice E6 (R2). EMA/CHMP/ICH/135/1995. London: European Medicines Agency; 2018
Langhof H, Chin WWL, Wieschowski S, Federico C, Kimmelman J, Strech D. Preclinical efficacy in therapeutic area guidelines from the U.S. Food and Drug Administration and the European Medicines Agency: a cross-sectional study. Br J Pharm. 2018;175:4229–4238. https://doi.org/10.1111/bph.14485
European Medicines Agency. Guideline on strategies to identify and mitigate risks for first-in-human and early clinical trials with investigational medicinal compounds, EMEA/CHMP/. London: European Medicines Agency; 2017
Baldrick P, Reichl A. Nonclinical & clinical interface—extrapolation of nonclinical data to support Phase I clinical studies. Regul Toxicol Pharm. 2021;121:104869 https://doi.org/10.1016/j.yrtph.2021.104869
Kenter MJH, Cohen AF. Establishing risk of human experimentation with drugs: lessons from TGN1412. Lancet 2006;368:1387–1391. https://doi.org/10.1016/S0140-6736(06)69562-7
Rocha J-F, Santos A, Gama H, Moser P, Falcao A, Pressman, et al. Safety, tolerability, and pharmacokinetics of FAAH inhibitor BIA 10-2474: a double-blind, Randomized, Placebo-Controlled Study in Healthy Volunteers. Clin Pharm Ther. 2022;111:391–403. https://doi.org/10.1002/cpt.2290
Cohen AF, van Smeden J, Webb DJ. De-risking clinical trials: the BIAL Phase I Trial in Foresight. Clin Pharm Ther. 2022;111:362–365. https://doi.org/10.1002/cpt.2498
Baakman AC, Zuiker R, van Gerven JMA, Gross N, Yang R, Fetell M, et al. Central nervous system effects of the histamine-3 receptor antagonist CEP-26401, in comparison with modafinil and donepezil, after a single dose in a cross-over study in healthy volunteers. Br J Clin Pharm. 2019;85:970–985. https://doi.org/10.1111/bcp.13885
van Gerven J, Cohen A. Integrating data from the Investigational Medicinal Product Dossier/investigator’s brochure. A new tool for translational integration of preclinical effects. Br J Clin Pharm. 2018;84:1457–1466. https://doi.org/10.1111/bcp.13529
Wieschowski S, Chin WWL, Federico C, Sievers S, Kimmelman J, Strech D. Preclinical efficacy studies in investigator brochures: do they enable risk-benefit assessment? PLoS Biol. 2018;16:e2004879 https://doi.org/10.1371/journal.pbio.2004879
FDA. Guidance for industry: estimating the maximum safe starting dose in initial clinical trials for therapeutics in adult healthy volunteers. 2005. https://www.fda.gov/media/72309/download. Accessed 28 Jan 2021
Groeneveld GJ, Hay JL, Van Gerven JM. Measuring blood–brain barrier penetration using the NeuroCart, a CNS test battery. Drug Discov Today Technol 2016;20:27–34. https://doi.org/10.1016/j.ddtec.2016.07.004
Perel P, Roberts I, Sena E, Wheble P, Briscoe C, Sandercock S, et al. Comparison of treatment effects between animal experiments and clinical trials: systematic review. BMJ. 2007;334:197 https://doi.org/10.1136/bmj.39048.407928.BE
van der Worp HB, Howells DW, Sena ES, Porritt MJ, Rewell S, O’Collins V, et al. Can animal models of disease reliably inform human studies? PLoS Med. 2010;7:e1000245 https://doi.org/10.1371/journal.pmed.1000245
Cohen AF, Burggraaf J, van Gerven JM, Moerland M, Groeneveld GJ. The use of biomarkers in human pharmacology (Phase I) studies. Annu Rev Pharm Toxicol. 2015;55:55–74. https://doi.org/10.1146/annurev-pharmtox-011613-135918
Griebel G, Holsboer F. Neuropeptide receptor ligands as drugs for psychiatric diseases: the end of the beginning? Nat Rev Drug Discov. 2012;11:462–478. https://doi.org/10.1038/nrd3702
Sanislow CA, Ferrante M, Pacheco J, Rudorfer MV, Morris SE. Advancing translational research using NIMH research domain criteria and computational methods. Neuron. 2019;101:779–782. https://doi.org/10.1016/j.neuron.2019.02.024
Cai HL, Jiang P, Tan QY, Dang RL, Tang MM, Xue Y, et al. Therapeutic efficacy of atypical antipsychotic drugs by targeting multiple stress-related metabolic pathways. Transl Psychiatry. 2017;7:e1130 https://doi.org/10.1038/tp.2017.94
Centre for Human Drug Research. TRC Pharmacology. 2022. https://trc-p.nl
European Commission, Health and Consumers Directorate-General. The rules governing medicinal products in the European Union, vol. 10—guidance documents applying to clinical trials guidance on investigational medicinal products (IMPs) and ‘noninvestigational medicinal products’ (NIMPs) (rev. 1, March 2011). Bruxelles: European Commission, Health and Consumers Directorate-General
Mishra A, Sarangi SC, Reeta K. First-in-human dose: current status review for better future perspectives. Eur J Clin Pharm. 2020;76:1237–1243. https://doi.org/10.1007/s00228-020-02924-x
Author information
Authors and Affiliations
Contributions
GF, FD, FS, and JvG contributed to the study design. GF and FD contributed to the acquisition of data. GF, FD, DVG, WPCB, HS, EHMM, PJKvM, FES, JMAvG contributed to the analysis and interpretation of data, drafting and review of the manuscript.
Corresponding author
Ethics declarations
Competing interests
DVG reports personal fees from Danone Nutricia Research B.V. and MSD Animal Health outside of the submitted work. GSF reports personal fees from Merck KGaA, Curare Consulting B.V., Access to Medicine Foundation, 3D PharmXchange B.V. and Janssen Biologics B.V. outside of the submitted work. None of the other authors has any conflict of interest.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Ferreira, G.S., Dijkstra, F.M., Veening-Griffioen, D.H. et al. Translatability of preclinical to early clinical tolerable and pharmacologically active dose ranges for central nervous system active drugs. Transl Psychiatry 13, 74 (2023). https://doi.org/10.1038/s41398-023-02353-1
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41398-023-02353-1