Skip to Content

Drug Development

A Question: Monoclonal Antibodies in the Clinic

A reader sends along this query, and since I’ve never worked around monoclonal antibodies, I thought I’d ask the crowd: how much of a read on safety do you get with a mAb in Phase I? How much Phase I work would one feel necessary to feel safe going on to Phase II, from a tox/safety standpoint? Any thoughts are welcome. I suspect the answer is greatly going to depend on what said antibody is being raised to target.

16 comments on “A Question: Monoclonal Antibodies in the Clinic”

  1. David Borhani says:

    Assuming you get drug exposure (i.e. you can measure it), I don’t think you get any more or less of a read on safety with mAbs compared to small molecules. Timing may be a bit different, in that most mAbs have long half-lives. But, if your mAb is contaminated, for example, you’ll quickly see some sort of inflammatory/anaphylactic response (cf. TGN1412). Long-term safety: same ball park as small molecules, IMO.

  2. steve says:

    TGN1412 wasn’t a contamination problem it was a rare side effect that wasn’t picked up in preclinical. Antibodies are a lot cleaner than small molecules and it’s possible to do extensive preclinical studies on cross-reactivity. The MOA, PK and PD are usually well-understood and surrogate efficacy assays are usually easily developed. Therefore you can get a much better idea in Ph1 about efficacy for an antibody than you can with a small molecule.

  3. David Borhani says:

    I really shouldn’t have written “contaminated” — quick & sloppy wording, sorry. That would be very rare.
    Rather, a mAb (or any drug) that somehow causes an acute or delayed inflammatory response, for example via:
    – allergenicity, either to the mAb or one of the formulation components;
    – off-target binding, with immediate downstream results; or
    – on-target binding, with immediate downstream results (cf. TGN1412)
    would show a clear safety signal in Phase 1.
    Note that some mAbs (e.g., Rituxan) are only infused after having prepped the patient with antihistamines, to try to stave off the milder types of allergic responses.
    See: Classification and Descriptions of Allergic Reactions to Drugs.
    Long-term (e.g., mechanism-based) safety would need longer studies, as for any drug.

  4. annonie says:

    My answer to Derek’s question is his final statement….it depends.

  5. thorazine says:

    @1, 2 – You’re both wrong about TGN1412 – the problem wasn’t contamination, and it wasn’t “rare side effects” – it was extremely potent on-target toxicity not exhibited in animal models.
    Let Wikipedia take it away: “The trial resulted in hospitalization of all six volunteers administered the drug, at least four of whom suffered multiple organ dysfunction.”

  6. steve says:

    By rare side effect, I mean it was rare among monoclonals, not rare among those who received the drug. It was not seen in primate studies and only manifested in the clinical trial. TGN1412 is pretty much a unique example in eliciting that type of cytokine storm.; it did so because of the particular way it interacted with the T cell receptor (its target) when clustered. My point is that it’s usually much easier to determine off-target effects with an antibody than with a small molecule prior to clinical trials because of the higher specificity, the ability to screen for the antigen and the now extensive clinical experience with a wide variety of antibodies. For the same reasons it’s also easier to get a clinical efficacy signal in early trials.

  7. SomeGuy says:

    A lot of the tox is discovered pre-clinical. Monkey and Rat give a pretty good idea of what you’re going to see safety-wise in humans. Most truly toxic mAbs never even see Phase I if due diligence is done.

  8. johnnyboy says:

    None of the commenters have answered the actual question, which was whether Phase 1 gives a better read on safety in biologics than in small molecules. I don’t know myself, and I doubt anyone does, but an informed opinion would be interesting.

  9. Ken says:

    In a lot of ways I think you get more useful safety data with a long-acting biologic like a mAb. Especially for single dose studies since the exposure is so much longer. If you are modulating a target for weeks instead of hours, you can see a lot more things manifest. Safety signals like changes in bone biomarkers are a good example. You also have arguably better specificity for your target than a small molecule, meaning the safety signals you do see (absent immunogenicity) are target related. Of course there are operational details that make ph1 studies with biological challenging. Long time periods between safety/escalation reviews, long washout periods, expensive API, etc.

  10. P.K. Peedee says:

    Outside the immune system, exposure typically drives both toxicity and efficacy, regardless of the modality.
    Phase I trials are a fairly reliable place to look for large molecule immunogenicity, whereas idiosyncratic immune reactions to small molecule drugs (e.g., Stevens-Johnson syndrome) are typically so rare that they don’t show up until phIII or later.
    The consequences of anti-drug antibodies are typically limited to reduced exposure rather than frank toxicity. Toxic anti-drug antibody reactions (e.g., pure red cell aplasia and anti-erythropoietin antibodies in patients treated with EPO) have been reported for recombinant proteins but I’m unaware of any similar ADA-driven toxicity from monoclonals…yet.

  11. Cellbio says:

    In my opinion, the actual question needs to be answered with discussions like prior posters have initiated. Mabs are quite clean, so Ph1 studies, though from a regulatory perspective are about safety and tolerability, are more informative about PK/PD. A well chosen antibody that is characterized in detail (affinity, specificity, stability, homogeneity) and not carrying contaminants is highly unlikely to show issues in single dose studies or in small studies exploring repeat dose. Antigenicity seems to run, in my experience (small n) about the same as for recombinant protein therapeutics, so seeing issues here requires larger populations and multiple dosing. Never seen this to be a big issue but perhaps others have.
    Most issues are mechanism related. To achieve favorable dosing intervals with non-linear PK, circulating concentrations are rather high compared to kDs so biological systems are essentially shut down, at least systemically. In tissues this may be different, but ablating functions is more likely than modulating them. Consequences such as infections take time, larger populations and external events to show true risk. Really known post-marketing approval.
    Regarding TGN1412 and perhaps anti-CD40 ligand antibodies, my own work leads me to think the one preclinical challenge we have in predicting toxicology is how antibodies may interact with surfaces that cause formation of aggregates (either the antibody bound to Fc receptors or the antibody bound to clustered ligand). Modeling this in vitro has shown that antagonists can convert to partial agonists when clustered. Careful in vivo follow-up for a clinical candidate showed that there was a first dose wave of agonism and partial cytokine storm in cynos like for TGN1412. If circulating levels dropped prior to repeat dose, the phenomenon was observed again. Exploring CD40L mabs showed that the clinical candidates associated with adverse events (thrombosis in patients prone to thrombosis) had a pronounced ability to form large aggregates at the right molar ratios (dimer antibody binding to trimer antigen). This could change the dynamics of platelet function (they express the antigen) that, in the context of other drivers of thrombotic risk, would be very hard to measure accurately until large populations were exposed.
    The other potential driver of risk hard to assess in Ph1, especially in normal volunteers without higher levels of antigen, is the potential for mabs to alter clearance of circulating growth factor targets. In one program, ligand targeting appeared to have an inverse dose response and waning effect. Though unsure why, the thought was that perhaps antibody bound ligand levels override antagonism after the complex rises to a level where the off-rate of ligand is sufficient to provide stimulation.
    BTW, did not find that rats or monkeys were good for tox. One often has to use surrogate antibodies to test impact of similar mechanism since there is species specificity. Also have seen cross species binders display very different bioactivities. This can occur with identical kDs.
    So yep, it depends!

  12. steve says:

    I think I did answer the question but I can elaborate. First, let’s please dispose of the canard that TGN1412 has much to say about antibody development nowadays. Quite simply it was an aberration. There have probably been more than a hundred antibodies that have gone into the clinic and none have had that side effect, which was caused by the particular MOA of that antibody.
    Contrary to the previous post, animal models are in fact very good predictors of how most antibodies will act in humans. The reason that you can get better efficacy answers in early clinical testing with antibodies than with small molecules is that you’re dealing with a standard scaffold that varies primarily in the binding site. Therefore, there is a lot of accumulated understanding about PK and PD issues that you don’t have with a new small molecule drug.
    Also, as I said, the same antibody being used therapeutically can also be used very often diagnostically so you can get an early read on efficacy. Finally, most antibodies are to protein epitopes and there are large databases that can be searched to see if similar epitopes appear in non-target proteins. This can be confirmed before conducting any clinical trials through the use of immunocytochemistry on tissue blocks. Unlike small molecules, you can see beforehand if there are cross-reaction issues.

  13. cynical1 says:

    TNF antibodies have a history of leading to TB, MS and rare lymphomas. Alemtuzumab and interferons (not an antibody) lead to autoimmune thyroid disease. Rituxan, efalizumab and natalizumab can cause PML. Avastin is associated with necrotizing fasciitis. None of this stuff came out in the Phase I studies and those are some pretty nasty toxicities. Some of those above can’t be explained by anything to do with their mechanism either. I’m not sure you get a better read with antibodies but they’re usually being developed for more severe medical conditions (autoimmunity and cancer) so they don’t get yanked when they eventually show up. Sort of my take.

  14. Rob says:

    For those who are interested this topic is given special attention in a report of the Working Party on Statistical Issues in First-in-Man studies issued by the Royal Statistical Society:

  15. cellbio says:

    Agree with cynical1. Those safety concerns do not present themselves in Ph1 nor in animal safety assessment. Additionally, direct guidance from the FDA in one program was to not include any rodent, dog or lower order primate studies in our IND as they were not relevant to safety assessment of our lead. My experiences, drawn from only a handful of programs, is in agreement with steve in that animal models predict how mabs will work, in general, which is by a safe and tolerated mechanism. But, from a regulatory perspective and from true insights into risks to humans exposed for years, Ph1 and animal models are of limited value. Would love to hear other examples that highlight the opposite experiences. In contrast, several small molecules I worked with showed safety concerns in Ph1 that were not evident in tox studies, so seems more valuable as a weed out step.
    The company I was with came around to thinking that since no issues arose in early clinical trials with mabs and PK/PD was predictable, we had to be committed to efficacy end points and had to build budgets and rationale to support the full investment up through largish Ph2s. Lots of boxes to check and lots of review board meetings for approval as the ticket was in the several 10s of millions depending upon indication. Blockbuster mentality reigned supreme as a result. NPV had to be really positive.

  16. clinicaltrialist says:

    I’ve worked on dozens of antibodies, many through to approval. In general, antibodies don’t have off-target effects so most safety issues are predictable, with a few exceptions. Even for the exceptions, such as HER2 cardiac side effects, are not true exceptions in that once the biology is understood, you can predict the AE. Out of the all the antibodies I worked on, I can only think of 2 instances where we were surprised by the side effect profile.
    To cynical1’s point, It is hard to explain TNF and MS-like diseases, in that immunosuppressive agents usually improve MS. However, the the rest of the AEs he mentions are mechanism-related. TNF Ab are immunosuppressives, which is why you get TB and lymphomas. Ditto with the other immunosuppressive agents. Avastin inhibits angiogenesis, which is required for wound healing, which is in turn required to prevent necrotizing fasciitis. All are predictable.
    TGN1412 – almost anyone who know the target could have predicted it was going to be a high risk target.
    Also, antibodies generally have a very wide therapeutic index. so Phase I is really a formality rather than anything else. You get a bit of PK information but most antibodies have a half life of a few days so Phase I’s are usually a non-event.
    So the answer to the question is that Phase I’s are correlated with Phase III because Ab are very safe compared to small molecules. But you don’t really need to run Phase I to know that – you already know that at the point when you start with the hybridoma.

Comments are closed.