Main

The main aim of first-in-human trials (FIHTs) is to explore the safety of multiple escalating doses of a drug in order to identify the highest dose associated with a tolerable toxicity. This is usually defined as the maximum-tolerated dose (MTD) (Eisenhauer et al, 2000; Le Tourneau et al, 2009) and is frequently selected for the subsequent drug development, on the basis of the assumption that a positive correlation exists between the drug dose and its effect. It is then indicated as the recommended phase II dose (RP2D) (Le Tourneau et al, 2009). Available data convincingly showed that FIHTs are the most important step in determining the dose of FDA-approved anticancer drugs, because for most molecules the RP2D is determined on the basis of the MTD, and the dose tested in registration trials (RTs) is within 20% on either side of the RP2D (Jardim et al, 2014). More uncertainty exists about dose selection for targeted agents, because in this case, the RP2D coincides less frequently with the MTD and predicts poorly the dose used in RTs (Jardim et al, 2014). Indeed, designing and interpreting FIHTs for targeted agents is difficult because of their limited acute toxicity (the first-cycle toxicity is usually the endpoint for MTD selection), and because of the scarce correlation between pharmacokinetic (PK) or pharmacodynamic (PD) parameters and drug efficacy in this setting (Parulekar and Eisenhauer, 2004; Jardim et al, 2014; Janne et al, 2016; Sweis et al, 2016). The challenge is even greater in FIHTs of monoclonal antibodies (mAbs) due to the low risk of acute toxicity associated with these molecules as a consequence of their lack of off-target effects (Sachs et al, 2016), and also because conventional FIHTs cannot capture the medium- and long-term toxicity of tested drugs. We recently conducted a comprehensive analysis of the design, implementation and outcome of FIHTs on mAbs published between 2000 and 2013 (Tosi et al, 2015). We found that, for most of the tested molecules, early-occurring adverse events were rare and dose escalation could be continued up to the highest planned dose level in all trials. Consequently, the MTD could be identified only in a minority of trials. Conversely, the RP2D was indicated in an important proportion of FIHTs, mainly in the absence or independently of the MTD and on the basis of PK or PD considerations. The PK data used to justify the RP2D choice mostly relied on comparisons between the drug concentrations found to be effective in preclinical studies and the clinical PK findings. PD data often focused on receptor occupancy assessment. However, the correlation between PK or PD parameters in preclinical models and in patient samples is far from being clearly established, which makes RP2D recommendations based on these observations at least doubtful. Despite these uncertainties in RP2D selection, mAb clinical development achieved several important successes for the treatment of malignancies and immunologic disorders (Nelson et al, 2010); however, comprehensive reviews are not available on the strategies of mAb clinical testing following FIHTs.

The aim of this study was to evaluate the strategies of mAb clinical development by analysing single-agent non-FIHTs (NFIHTs) of mAbs the FIHT of which was included in our previous analysis (Tosi et al, 2015), as well as the RTs of all FDA-approved mAbs. After retrieving from MEDLINE all publications on these NFIHTs and RTs, we examined the trial design and results, with a particular focus on the relationship between FIHT data and doses tested in these trials.

Materials and methods

Article search

In June 2016 we performed a MEDLINE search to identify articles on single-agent trials of mAbs the FIHT of which was included in our previous analysis (Tosi et al, 2015). Separately, we identified mAbs approved as single agents by FDA up to 31 June 2016, and for each molecule we performed a MEDLINE search (using all the known names of each drug) to identify the FIHT and RTs. We excluded trials reporting on immunoconjugates, radioimmunoconjugates and non-systemic routes of administration (topical administration or ex-vivo treatment), trials on Asian patients performed to confirm previous results obtained in Western patients, phase III trials where the evaluated mAb was used as standard treatment, as well as articles not written in English language. The phase I and phase II parts of phase I/II studies were analysed separately when possible.

Data collection and analysis

From articles on the NFIHTs of mAbs with an FIHT included in our previous review (Tosi et al, 2015), we extracted treated disease, trial phase, rationale for dose(s) selection, administration route, dose calculation unit, schedule, presence of loading dose, tested dose(s), number of included patients, and availability of PK or PD data. From dose escalation trials, we also extracted the starting dose (SD), the maximum planned dose, the maximum administered dose (MAD), the MTD, the RP2D and the rationale for RP2D selection. For these trials, we calculated the ratio between FIHT MAD and NFIHT MAD, the ratio between NFIHT RP2D and FIHT MAD and the ratio between NFIHT RP2D and FIHT RP2D. For phase II and III trials, we calculated the ratio between the tested dose and FIHT MAD or FIHT RP2D. For the analysis of the RTs concerning mAbs approved by the FDA, we extracted treated disease, administration route, dose calculation unit, schedule, presence of loading dose, tested dose(s), number of included patients, and the three most frequent grade 3/4 toxicities. From the relevant FIHT, we recorded MAD, MTD, RP2D and the three most frequent grade 3/4 toxicities. We calculated the ratios between RT dose and FIHT MTD and MAD, respectively. When more than one trial was available for a given mAb in a data set, we used the mean of the ratios from all the trials of this mAb to calculate summary statistics on the dose ratios for the entire data set. We used descriptive statistics to report whether the top-three grade 3/4 toxicities in the RTs of each mAb were detected in the corresponding FIHT, and their grade in the FIHT. Statistical analyses were performed with the R software (version 3.3.2).

Results

General results on NFIHTs

After reviewing the 139 articles retrieved with the MEDLINE search, we selected for analysis 144 NFIHTs of 42 mAbs (1–15 NFIHTs for each molecule). The study design and drug administration data of the selected NFIHTs are shown in Table 1. Specifically, 39 studies (27%) were phase I and 103 (72%) phase II or III trials. Most trials concerned patients with solid cancers or haematological malignancies (n=111, 77%), while the others focused mainly on immunologic disorders. In 131 trials (91%), the mAb was administered only by intravenous route and a loading dose was used in 20 (14%). For most mAbs, the same dose calculation method was used in NFIHTs and the corresponding FIHT. However, in 16 NFIHTs, a flat dose was administered instead of the dose tested in the FIHT and calculated according to weight (mg kg−1) or body surface (mg m−2).

Table 1 Characteristics of the 144 NFIHTs included in the study

Analysis of dose escalation NFIHTs

In 37 of the 39 of phase I trials, a dose escalation procedure was implemented (for 21 mAbs). We found that the highest planned dose corresponded to the FIHT MAD for nine of the 19 (47%) mAbs tested in the NFIHTs that used the FIHT dose calculation method. Indeed, the range of ratios between the highest NFIHT planned dose and FIHT MAD was quite wide (0.1 to 6); however, for 15 mAbs (71%) in 21 trials (65%) the highest planned dose level was lower or equal to the FIHT MAD, and for 15 mAb (71%) in 19 trials (59%) it was within 33% on either side of the FIHT MAD (Figure 1). Like in the FIHT, in all NFIHTs the mAb favourable safety profiles allowed dose escalation up to the highest planned dose level that, therefore, coincided with the MAD. An MTD was found for only seven of 21 mAbs (33%) tested in eighth (22%) dose escalation trials. An RP2D was indicated for 11 of the 21 mAbs (52%) tested in 15 of the 37 NFIHTs (40%), but it matched the FIHT RP2D for only three mAbs in four trials (Figure 1). The rationale for RP2D selection was described for only 11 mAbs (Table 2) and was based on considerations about safety (n=6), PK (n=4), and PD (n=1). The medians of the NFIHT RP2D/FIHT RP2D and NFIHT RP2D/FIHT MAD ratios were 2.2 (range: 1 to 6) and 0.65 (range: 0.3 to 1), respectively.

Figure 1
figure 1

Relationship between NFIHT MAD, NFIHT RP2D and FIHT MAD for dose-escalating NFIHTs. (A) Ratio between the NFIHT MAD and the FIHT MAD for each NFIHT with dose escalation. Each bar represents the ratio between the highest planned dose of each NFIHT and the relevant FIHT. The names of tested mAb are indicated on the left. (B) Ratio between the NFIHT RP2D and the FIHT MAD for NFIHTs with dose escalation. Filled circles represent the ratio between NFIHT RP2D and FIHT MAD. Each circle refers to a dose tested in one or more NFIHTs. The names of tested mAb are indicated on the left. For comparison, hollow circles represent the ratio between RP2D and MAD in the corresponding FIHT. FIHT=first-in-human trial; MAD=maximum administered dose; NFIHT=non-first-in-human trial; RP2D=recommended phase II dose.

Table 2 Rationale for RP2D selection in trials with dose escalation

Analysis of phase II and III NFIHTs

We then analysed the doses tested in the 103 phase II and III trials (on 37 mAbs) with regard to the FIHT results to assess FIHT data relevance for the subsequent mAb development. First, we evaluated how the tested dose(s) was selected (Table 3). A rationale was indicated for 26 mAbs (70%) in 57 of the 103 trials (55%) and was based on the FIHT RP2D (19 trials), PK data (7 trials), efficacy (7 trials), FIHT MAD (4 trials), PD (2 trials), FIHT MTD (1 trial) or other considerations (17 trials). We then examined the relationship between FIHT RP2D and doses tested in NFIHTs (Figure 2). FIHT RP2Ds (one or more for each mAb) were available for 12 of the 37 mAbs and were tested for 11 mAbs, alone (n=5 mAbs) or in association with other doses (n=6 mAbs). The FIHT MAD was tested for eight mAbs (73%) in 17 trials (45%) of mAbs with available FIHT RP2D. The ratio between the doses tested in NFIHTs and the corresponding FIHT RP2Ds ranged from 0.1 to 5, and in 84% of cases the tested dose/FIHT RP2D ratio was not within 33% on either side of the FIHT RP2D (Figure 2). The FIHT MAD of 17 mAbs (46%) was tested alone or with other doses in 36 trials (35%). Only doses different from the FIHT RP2D or MAD were tested for 16 mAbs (43%) in 37 trials (36%). Finally, we verified that the tested doses were included in the range established as safe in the FIHT and compared them with the FIHT MAD (Figure 2). Only in nine trials on two mAbs, the tested dose was higher than the FIHT MAD. The median tested dose/FIHT MAD ratio was 0.71 (range: 0.25 to 2.5) in trials with comparable dose calculation methods.

Table 3 Rationale for tested dose selection in NFIHT without dose escalation
Figure 2
figure 2

Ratio between the phase II/III NFIHT doses and the FIHT MAD. (A) Ratio between the dose tested in NFIHT and the FIHT MAD for each mAb without an FIHT RP2D. Circles represents the ratio between the tested dose of each mAb and the relevant FIHT MAD. Each circle refers to a dose tested in one or more NFIHTs. The names of tested mAb are indicated on the left. (B) Ratio between the dose tested in NFIHT and the FIHT MAD for each mAb with an FIHT RP2D. Hollow circles represent the ratio between the tested dose of each mAb and the relevant FIHT MAD. Each circle refers to a dose tested in one or more NFIHTs. Filled circles represent the ratio between the FIHT RP2D and the FIHT MAD. The names of tested mAb are indicated on the left. FIHT=first-in-human trial; MAD=maximum administered dose; NFIHT=non-first-in-human trial; RP2D=recommended phase II dose.

Analysis of the correlation of doses and toxicities in RTs and the corresponding FIHT

We retrieved 27 FDA-approved mAbs with a FIHT and 60 RTs on these molecules (Supplementary Table S1). The mAb indication was cancer (solid tumours for eight mAbs, haematological cancers for three mAbs), immune system diseases (13 mAbs) and other diseases (four mAbs). The FIHT MTD was available for only one molecule, whereas the FIHT RP2D was indicated for seven mAbs (26%; five cancer trials and two other trials). We then evaluated the relevance of the FIHT results for the 17 mAbs with the same dose calculation method in FIHT and RTs. The RP2D was tested in RTs of five mAbs (but only in two with the same schedule), and the MAD in RTs of four mAbs (Figure 3). The median RT dose/FIHT MAD ratio was 0.78 (range: 0.1 to 2.5). When considering the nine mAbs for which an RP2D was not available, at least one RT dose was lower than 75% of the MAD for six of them (specifically, lower than 50% for four mAbs and lower than 25% for one). We determined whether the top-three grade 3/4 toxicities in the RTs of each mAb were reported in the corresponding FIHT, and their grade in the FIHT. For only seven mAbs (25%) at least two of the top-three RT grade 3/4 toxicities were reported as grade 3/4 in FIHT. Conversely, for 16 (57%) none of the top-three grade 3/4 toxicities described in the RTs was reported as grade 3/4 in FIHT. In addition, for seven (25%) of mAbs none of the top-three grade 3/4 toxicities was reported in FIHT.

Figure 3
figure 3

Ratio between RT doses and FIHT MAD. (A) Ratio between the dose tested in RT and the FIHT MAD for each mAb without an FIHT RP2D. Circles represent the ratio between the tested dose of each mAb and the relevant FIHT MAD. Each circle refers to a dose tested in one or more RTs. The names of tested mAb are indicated on the left. (B) Ratio between the dose tested in RT and the FIHT MAD for each mAb with an FIHT RP2D. Hollow circles represent the ratio between the tested dose of each mAb and the relevant FIHT MAD. Each circle refers to a dose tested in one or more RTs. Filled circles represent the ratio between the FIHT RP2D and the FIHT MAD. The names of tested mAb are indicated on the left. FIHT=first-in-human trial; MAD=maximum administered dose; NFIHT=non-first-in-human trial; RP2D=recommended phase II dose; RT=registration trial.

Discussion

In our previous analysis concerning the FIHTs of mAbs published between 2000 and 2013, we showed that, for most of the tested molecules, acute toxicity events were rarely observed and did not allow the identification of an MTD. This frequently led to doubtful or questionable recommendations about the RP2D that was determined on the basis of surrogate endpoints (Tosi et al, 2015). Here, we analysed the NFIHTs of the same mAbs to evaluate how these trials used the FIHT-derived results, particularly the FIHT RP2D and MAD. We also investigated whether the FIHT results were relevant in the case of FDA-approved mAbs, relative to the tested doses and toxicities observed in RTs.

In the examined dose escalation NFIHTs, the dose level scheme was conservative relative to the FIHT, severe toxicities were infrequent, the MTD was rarely determined and the RP2D was indicated in a minority of trials. In addition, the rationale for RP2D selection was infrequently available and not always stringent, when present. These results indicate that even when a dose escalation trial for an mAb is performed in post-FIHT settings, a significant uncertainty persists over the RP2D indication. Moreover, we found that the FIHT RP2D had a limited influence on dose selection in phase II–III NFIHTs. This is in striking contrast with what generally occurs in anticancer drug development, where the final FDA-approved dose is within 20% on either side of the FIHT RP2D in 73% of cases (Jardim et al, 2014). In addition, the FIHT MAD was frequently tested in trials of mAb with available RP2D, suggesting a lack of confidence in the RP2D selection criteria. On the other hand, the FIHT MAD constituted a widely accepted upper limit for dose selection in phase II–III NFIHTs. Frequently, we could not retrieve a convincing justification for dose selection in NFIHTs of mAbs. In a significant percentage of trials, the dose tested in NFIHTs without dose escalation did not correspond to the RP2D or MAD and no rationale for dose selection was available, which did not allow evaluating whether the assumptions underlying the choice were appropriate. When the FIHT RP2D or MAD was not used to inform decisions about the dose to be tested in NFIHTs, preclinical data on the drug effective concentration and clinical PK data were frequently the parameters of choice, notably the serum concentrations attained in clinical trials. However, due to their size, mAb penetration in tissues occurs mainly by convective transport and is characterised by slow diffusion rates in tumour tissue (Tabrizi et al, 2010; Dostalek et al, 2013; Tibbitts et al, 2016). This poor tumour uptake is mostly explained by the scarce tumour vascularisation and the increased interstitial fluid pressure, secondary to vessel abnormalities, fibrosis and interstitial matrix contraction (Heldin et al, 2004). Consequently, it appears hazardous to simply infer the mAb tumour tissue concentration from their serum level.

The lack of association between the most frequent severe toxicities in FIHT and RTs suggests that the FIHT results are not useful to predict the actual mAb toxicity and that, consequently, an approach based on a toxicity-guided dose selection during the early clinical development of mAbs could be misguiding. Moreover, the absence of significant toxicity in FIHTs could complicate the choice of the doses to be tested in later trials. Selecting an unnecessarily high mAb dose can be unsafe because rare dose-dependent toxicities could appear later during the drug development process. Inappropriately low doses also can affect efficacy and tolerability because, in the presence of an abundant target mass, the mAb PK could be altered due to target-mediated drug disposition (Cartron et al, 2016; Meulendijks et al, 2016), especially when the mAb target is also expressed in healthy tissues (Azzopardi et al, 2011).

Other approaches for optimal mAb dose selection could be suggested, such as correlating the mAb serum concentration with PD marker variations, or implementing PK/PD models. The choice and accessibility to the measured PD markers are crucial in this setting. Quantitative data on serum (soluble) mAb targets, receptor occupancy on circulating tumour cells, serum markers that indirectly reflect the mAb effect (Mayer et al, 2015), or clinical parameters directly linked to disease activity (Azzopardi et al, 2015) represent useful PD endpoints for clinical trials. However, for mAbs that alter intracellular signalling, PD marker assessment in tumour cells is an elusive endpoint due to the limited availability of repeated biopsies. Integrative evaluations, including gene expression and phosphokinome profiling in tumour samples and liquid biopsies, could represent suitable tools for dose-finding clinical trials when preclinical studies have established clear correlations between a molecular signature and drug efficacy.

In addition, we previously showed that in mAb FIHTs the safety data relevant for dose selection are collected during a short observation window, which frequently corresponds to the first cycle of treatment (Tosi et al, 2015). Indeed, mAb PK could be far from the steady state throughout this time, because of the long drug half-life and dosing schedules that are frequently at least weekly (Tosi et al, 2015). In addition, the effect of target-mediated drug disposition (Azzopardi et al, 2011), and the rare administration of loading doses (Tosi et al, 2015) could contribute to delay reaching the maximal serum concentrations. Consequently, safety data or PK or PD evaluations obtained in this setting have limited value, suggesting that trial designs including a longer time frame for endpoint assessment at selected doses could be more appropriate.

Conclusions

We show that the results of FIHTs, particularly standard FIHT endpoints such as MAD, MTD and RP2D, are frequently not taken into account for the design of later clinical studies on mAbs. Moreover, while safety is the main endpoint of mAb FIHTs, other pharmacological aspects are often considered for dose choice in later clinical trials, although the relevance of these surrogate endpoints relative to the mAb clinical activity is questionable. New clinical development strategies are urgently needed for this class of molecules characterised by scarce toxicity, specific PK and high therapeutic potential. Particularly, these data strongly support shorter and more PD-focused phase I studies, as well as randomised phase II studies to compare different mAb doses.