Testing new drugs in naked apes and getting the dose right in their young

NOAEL and MABEL

Historically, drugs such as digitalis, cocaine, opioids, marijuana and morphine were probably first used in man, or were used in man after only a few animal exposures to rule out major immediate harm [1]. Currently, however, formal animal toxicity testing and pharmacology usually precede first in man studies. (The discovery by Roger Altounyan that cromoglicate is effective against asthma represents an exception amongst modern drugs in this regard. An asthmatic, he tested extracts of Ammi visnaga, a plant used in traditional medicine in Egypt, on himself without benefit of animal pharmacology or toxicology (http://en.wikipedia.org/wiki/Cromoglicic_acid accessed 10th August 2010)).

Picking a suitable starting dose for a first in man study is an art based on science, as are protocols for dose escalation in early single ascending dose (SAD) studies. The sciences are those of evolution and genetics, which underpin the current approach of testing a new drug in one or more of our animal cousins, usually including non-human primates when the drug is a biologic, before we administer it to man. The art is to balance conservative starting doses and rules for dose escalation, with realistic expectations as to the dose at which a desired pharmacological effect may be expected and adverse effects deemed unlikely. We have written before encouraging submission to the Journal of studies that feature modern mechanistic approaches to early studies in man, rather than scientifically unsound and practically unhelpful claims as to tolerability [1]. Some drugs, e.g. cancer chemotherapy are subsequently studied up to maximum tolerated doses, where adverse effects are very likely, but judged acceptable. Therapeutic doses (explored in phase II and III studies) of many drug are very likely to involve a compromise between efficacy and some undesired effects (e.g. dry mouth during treatment with a tricyclic antidepressant).

Toxicity testing in animals is carried out on new drugs to identify potential hazards before administering them to humans [2]. A wide range of tests is used in different species, with long-term as well as acute administration of the drug, monitoring for physiological and biochemical changes, and post-mortem examinations to detect gross or histological abnormalities. Non-mammalian species, notably the transparent zebra fish, are promising intermediaries between studies in vitro and mammalian toxicity testing, which is performed traditionally at plasma or tissue concentrations well above the expected therapeutic range, and establishes which tissues or organs are likely ‘targets’ of toxic effects. Recovery studies are performed to assess whether toxic effects are reversible, and particular attention is paid to irreversible changes, such as carcinogenesis or neurodegeneration.

The basic premise is that the toxic effects of a drug, i.e. adverse effects that occur at supra-therapeutic doses, as well as adverse effects that occur in the therapeutic range (collateral effects) are likely to be similar in humans and other animals. This is inherently reasonable in view of the cellular and molecular similarities between higher organisms resulting from their common ancestry. There is, nevertheless, wide interspecies variation (as expected from descent with modification), especially in metabolizing enzymes. Consequently, a toxic metabolite formed in one species may not be formed in another. Pronethalol, the first β-adrenoceptor antagonist synthesized by James Black at ICI, was not developed, because it caused carcinogenicity in mice. It subsequently emerged that carcinogenicity occurred only in the ICI strain of mice, but by then other β-adrenoceptor blockers were already in development. Penicillin is another famous exception to the rule that toxicity in one mammalian species predicts problems in other mammals, in that it is extremely toxic to guinea-pigs in doses that are therapeutic and non-toxic in other mammals, including man. Even within breeds of domestic animals important differences can occur. For example, collie dogs lack the multidrug resistance gene (mdr1), a P-glycoprotein that contributes to the blood–brain barrier. This has important consequences for veterinary medicine, because ivermectin, an important antihelminthic drug, is consequently severely neurotoxic in the many canine breeds with collie ancestry [3, 4].

Toxic effects can range from trivial to so severe as to preclude further development. Intermediate levels of toxicity are more acceptable in drugs intended for serious illnesses that are currently poorly served, and decisions on whether or not to continue development are often difficult. If development does proceed, safety monitoring can be concentrated on the system ‘flagged’ as a potential target of toxicity by the animal studies. The value of toxicity testing is illustrated by experience with triparanol, a cholesterol-lowering drug marketed in the USA in 1959. Three years later, a team from the Food and Drug Administration, acting on a tip-off, paid the manufacturer a surprise visit that revealed falsification of toxicology data demonstrating cataracts in rats and dogs. The drug was withdrawn, but some patients who had been taking it for a year or more also developed cataracts (quoted in [2]). Safety of a drug (as opposed to toxicity) can be established only during therapeutic use in humans. The level (i.e. dose) of a new drug that produces no observable adverse effect (‘NOAEL’) is established in animals. The art of translating this into man includes scaling the first dose to produce an estimated exposure of the tissues to the drug that is substantially less than the NOAEL, an approach that when applied to traditional low molecular weight drugs served well in investigating many new molecular entities over many years but proved misleading in the case of a novel biological agent (TGN 1412), whose first administration led to a cytokine storm in all the healthy volunteers exposed, despite a dose that was 1/500 of that previously administered without harm to non-human primates (Cynomolgus macaque monkeys). The volunteers all survived the acute episode with intensive care for multiple organ failure, but the long term consequences are as yet unknown.

TGN 1412 is a humanized mouse monoclonal antibody designed as a super-agonist at the CD28 receptor on T lymphocytes. This is now known to be a co-stimulatory receptor in normal physiology, leading to activation of the T-cell, causing increased production of cytokines and an increased rate of cellular proliferation. It had been hoped that agonism at CD28 receptors would be of therapeutic benefit in patients with chronic B-cell leukaemias and in rheumatoid arthritis, and the drug had been given orphan status by the European Medicines Agency. Toxicity was related to the main pharmacological effect of this novel agonist and was highly species specific for reasons that are still incompletely understood. Further investigation [5] showed that TGN-1412 released cytokines from human lymphocytes in vitro and caused intense lymphocyte proliferation only when it was immobilized, either by drying onto a plate, by binding to endothelial cells or after capture by anti-Fc antibody. The concentrations used in these in vitro assays suggested that the doses administered to the Northwick Park volunteers were near to the maximal immunostimulatory dose. TGN-1412 did not elicit responses from human lymphocytes when presented in the aqueous phase (even when cross-linked), and did not produce responses from lymphocytes from Cynomolgus macaque monkeys (which express CD28 identical in structure to humans) even when immobilized. TGN 1412 did not produce severe adverse reactions when administered in vivo to Cynomolgus monkeys. This led to advice from an expert advisory group set up by the MHRA and chaired by Sir Gordon Duff [5] as to non-standard circumstances in which special caution should be exercised, for example when a drug is first in class, particularly if its anticipated physiological effects are potentially profound; when species specificity is plausible and when the drug is an agonist at cell surface receptors, especially when there is a potential for cross-linking of receptors. There was support for a starting dose in first-in-man trials of biological medicines equal to the dose needed to reach the lower threshold of a dose–response curve in man, estimated from human receptor occupancy and cellular dose–response studies, combined with information from animal models and experience with similar agents. This would give a ‘Minimum Anticipated Biological Effect Level’ (MABEL) at or below which the starting dose should be set, and this approach has been used with success subsequently. We shall visit other practical recommendations unrelated to dose selection in a future issue.

How different are children?

In a commentary in this present issue, Cella and colleagues identify enabling children to be given the right dose as an ‘unmet medical need’[6]. The Journal has previously highlighted regulatory efforts to improve dose optimization of drugs prescribed for children [7] and the training that is needed to be a paediatric clinical pharmacologist [8], so this present contribution is the more shocking. It identifies revisiting dosing recommendations for drugs that are already marketed but are used for unlicensed indications (i.e. ‘off-label’) in children, and early development of new drugs for which there are no previous data in children as two priorities. In this same issue of the Journal we publish a systematic review, including 33 studies in general paediatric populations, which shows that adverse drug reactions in children constitute a significant public health problem [9]. Adverse reactions were especially common with vaccines, antibiotics and psychotropic medications, and it seems likely that incorrect dosing will have contributed to the harm done, perhaps especially by the latter two groups of medication. Why is optimizing paediatric dosing such a difficult problem?

Five main problems make the study of medicines in children more challenging than in adults [10]: such populations include premature as well as term neonates, infants, toddlers, older children, and adolescents and are therefore far from homogeneous, many diseases are uncommon in childhood, so recruitment to trials is relatively difficult and the potential market modest, veins and blood volume are both smaller than in adults (especially in very young children) making for practical difficulties in studying pharmacokinetics, furthermore, there are real (and perceived) ethical problems in conducting controlled trials in children and trial end points are hard to define in this population (see below).

Most marketed therapeutic drugs are developed for use in adults, and the licensed dosage regimen for the (adult) patient is based on compromises between efficacy, safety and convenience. Children have seldom been subject to dose escalation studies as are carried out in adults, so initial estimation of dose in paediatrics has usually been obtained by extrapolation (so called ‘bridging’) from the adult dose regimen using some scaling factor related to age, body weight or body surface area. It has been argued that, following such an exercise, pharmacokinetic–pharmacodynamic (PK–PD) or PK studies will be needed to determine the most appropriate doses for neonates, infants, children and adolescents [11]. Cella and colleagues argue for scaling for function rather than size or age alone, through the use of nonlinear relationships such as allometric scaling [6], writing that ‘it is clear that a shift in paradigm is required that focuses on the differences in (physiological) function between populations, rather than differences in size between adults and children. …’.

Certainly there are important age-related but non-linear changes in the renal excretion and metabolism of drugs, supporting the aphorism that children are not merely small adults. Examples include altered aminoglycoside excretion as kidney function matures in premature infants, and immature glucuronidation, which underlies, for example, the susceptibility of neonates to chloramphenicol toxicity (‘grey baby syndrome’). Furthermore, there is a potential for drugs to produce age-specific permanent adverse effects by acting on a target that is unique to a particular aspect of development as exemplified by the ability of antithyroid drugs to influence brain development irreversibly.

However, while acknowledging the above, Stephenson has argued [12] that children's responses to drugs have much in common with the responses in adults and indeed in other mammals, pointing out that most basic cellular and physiological processes and receptors are common to all mammals, irrespective of age or stage of development. He writes that: ‘Children have a loop of Henle and a distal collecting tubule just as adults do, and therefore it is not surprising that diuretics work on a child's kidney. The magnitude of the effect may be different but the basic response has more features in common than there are differences. Often, it is assumed that drug effects differ in children but in reality this perception often arises because the drugs have not been adequately studied in paediatric populations of different ages and with different diseases. There may also be difficulties in measuring small but significant effects because the outcome measures are more difficult to assess in children. For example, a study of an anti-asthmatic drug in adults might show a 10% benefit in peak expiratory flow rate. In children under 5 years old who cannot perform such objective measures, a 10% difference may be missed using cruder techniques such as symptom diaries. Finally, part of the reason for the perception that pharmacodynamics are different in children is because the pharmacokinetics may be different at different ages. As a result, the same dose kg⁻¹ does not result in the same circulating concentration because the absorption/metabolism/clearance is different.’ Two further papers in the current issue support relatively straightforward approaches. Zeng and colleagues studied the population pharmacokinetics of mycophenolic acid in children undergoing marrow and solid organ transplantation [13]. Body weight and concomitant ciclosporin treatment influenced the pharmacokinetics, and current dosing strategies may be suboptimal for children weighing <10 kg. Their model should be useful in dosing children requiring immunosuppression. Kumpulainen and her colleagues performed a highly novel study to characterize paediatric pharmacokinetics and central nervous system exposure of flurbiprofen in 64 children aged 3 months to 13 years undergoing spinal anaesthesia [14]. They found that flurbiprofen pharmacokinetics can be described using weight as the only covariate in children above 6 months, while more work is needed on younger infants and neonates. Cerebrospinal fluid concentrations of flurbiprofen were markedly higher than unbound plasma concentrations.

Much more remains to be done, and paediatric medicinal development is rightly a current focus of regulatory guidelines and policies [15]. We agree with Rose that Europe needs a solid regulatory framework for paediatric drug development, a strong academic infrastructure for clinical research, and a strong pharmaceutical industry, and we echo his conclusion that ‘an open and trustful dialogue between the key stakeholders for paediatric health care needs to be established and maintained for the good of our children’[16].

Competing interests

There are no competing interests to declare.