Dose ranging for trials through biomarkers of drug effects

Abstract

To test the hypothesis that excess amyloid deposition is an essential step in the pathophysiology of Alzheimer's disease appropriate biomarkers are essential in selecting agents that modify amyloid formation or clearance. Cerebrospinal fluid concentrations of relevant analytes and PET measures of total brain load have been developed. These are directly applied to testing whether drugs reduce various soluble forms of amyloid as well as whether they enhance elimination of material deposited in brain parenchyma. The ideal profile of a drug that can fully test the amyloid hypothesis can be understood in terms of effects on currently available and future biomarkers. Dose selection for clinical trials should require a quantitative threshold effect on the most relevant biomarker.

As highlighted in Science magazine (1) there has been an appropriate re-assessment of the relevance of negative trials that were seen by some as evidence against the amyloid hypothesis. The two critical aspects of an adequate test of that hypothesis – matching the right patient(s) with the right drug(s) – are just beginning to be addressed through the application of appropriate biomarkers. Recent advances promises that we will be able to diagnose individuals who will progress to symptomatic Alzheimer’s years before anything more than age related cognitive decline is apparent. The field had hoped that by now we would have drugs that powerfully blocked Aß formation and/or cleared “toxic” and deposited forms from the brain. But no drug with proof of such powerful biology has yet been introduced into the twelve to eighteen month studies needed to explore disease progression. It may be that we will actually have identified the biomarkers of early illness before we have a drug which safely and maximally inhibits Aß formation and/or dramatically clears “toxic” or deposited species from the brain.

The focus of this discussion is on the criteria for deciding whether we have identified a drug with the right properties to adequately test the hypothesis that reduction of Aß formation and/or the clearance of toxic or deposited forms from the brains of affected individuals will prevent or slow the progression of Alzheimer’s disease. The ideal properties of drugs capable of testing the two forms of the stated hypothesis – blocking formation before potentially toxic deposition occurs or clearing already formed undesirable species – would include one or more of the following:

• 1.Ability to completely block a rate limiting enzyme without toxicity such that no new Aß is formed and any that is measured in the CNS compartment represents previously formed and “stored” material.
• 2.Appropriate pharmacokinetic characteristics such that a sustained maximal desired effect could be achieved with once a day or less frequent dosing – the more one targets prevention the more desirable such a characteristic. Vaccination is the extreme example of such an approach.
• 3.Ability to fully bind and remove soluble, oligomeric, fibrillar or plaque Aß depending on the extent to which dynamic equilibria exist between these forms. In other words, binding and removing a soluble form could lead to clearance of the other forms under conditions in which a continued input of newly formed A6 was essential to maintaining the concentrations of the other more complex forms.
• 4.Ability to activate some to be identified process for clearing any form of A6 that is toxic to the brain. If it proved true that microglial activation had the potential of clearing plaques then one would look for a drug that produced maximal safe activation of microglia and the biomarkers of such activation in living human brain.

We are almost certainly years away from any drug which meets these ideal criteria. On the other hand, there is no definitive evidence so say that some of these ideal criteria are impossible. For instance, complete inhibition of BACE in the CNS might prove to be safe and well tolerated.

Potentially disease modifying drugs such as the gamma secretase inhibitors from Lilly and Merck that have been introduced into humans have been limited to doses that do not produce sustained decreases of soluble A6 in the cerebrospinal fluid ( 2., 3., 4.). Based on some preclinical evidence during periods of rapid A6 deposition in the brains of transgenic mice, it has been suggested that even a sustained 25% reduction of formation could dramatically slow deposition in human brain (5). This suggestion depends on extrapolations from a critical period of rapid deposition in genetically engineered mice, a pattern of change which may have no analog in humans. By the time humans are diagnosed with MCI or AD a considerable amyloid load has accumulated. There is no evidence yet from PET imaging that there is a period of such rapid change in humans as seen in the mouse model. Therefore, as noted above, the only way to be sure one has fully tested the hypothesis is to completely block A6 formation.

There are two possibilities with regard to what one might see with current methods in terms of assessing blockade of Aß formation. Levels of soluble A6 post drug in the CSF could be eliminated. Alternatively, these could be reduced to an amount reflecting dissociation from oligomeric or other forms. In that event, one could use a PET ligand measure of total A6 load (assuming multiple forms are accessed by ligand) to show that even with a stable CSF value of soluble AB the total load was decreased. The recent progress from Bateman and colleagues (6, 7) in labeling APP (amyloid precursor protein) with 13C leucine may provide an alternate means of establishing full antagonism of AB formation even in the presence of some continued concentration of soluble AB in the CSF.

It is therefore possible with current methodologies to set some demonstrable threshold effect on Aß formation in the brain that must be achieved before going into any clinical study of disease modification. Industry is free to follow this strategy for setting doses for proof of concept studies. What has not been addressed is whether incorporating a biomarker of drug effect would allow one to explore fewer doses in a registration study.

One might argue that in such a serious and debilitating condition as Alzheimer’s one should study only the dose that gives the greatest possibility of affecting the condition, i.e, the dose with maximal efficacy as judged by the biomarker of drug effect within the bounds of acceptable safety. If a user friendly biomarker was available, each individual could be titrated to that dose and, in theory, a drug introduced with such an individualized dosing regimen. One raises this possibility as a point for discussion since if done properly, it would optimize the chance of therapeutic benefit per individual from the start assuming the underlying hypothesis is valid.

Are we at the same stage of markers to interpret the effects of an AB clearing compound as represented by antibody approaches? Evidence for a brain effect of antibodies ranges from MRI detectable peri-vascular changes seen as potentially toxic to post-mortem studies in a few individuals in whom “moth-eaten” remnants of what are presumed to have been plaques are seen (8). The latter is taken as evidence that an antibody can clear Aß containing plaques from the brain. Cerebrospinal studies of patients administered the Lilly antibody apparently reveal an increase in Aß, a particularly interesting pattern of changes (9, 10). Both replication and additional studies are needed to appreciate whether the findings that at the time points investigated, both soluble Aß 42 and a bound “fragment” are elevated in CSF can be explained by some antibody getting into the brain. The simple prediction would have been that bound AB would increase and soluble would fall.

Much remains to be done to understand the dynamics and relationship between each form of Aß and what degree of effect is desirable. The current assumption is that the amount of antibody that reaches the central compartment and interacts with one or another form of Aß is critical and directly related to the degree of efficacy. We do not, however, understand how to interpret the changes we observe in the CSF. One is therefore dependent on the Aß load across the brain that is detected with PET ligands to show that an antibody (or vaccine) is actually having the desired effect. Moreover, it remains to be established that one or another PET ligands bind to the form of

Aß that is most important to clear from the brain. From a practical viewpoint, however, PET is a much more costly as well as radiation exposure limited methodology for tracking drug induced effects compared to CSF (or any blood measure that may ultimately be developed to reflect the critical brain events). It is difficult to imagine titrating to a PET dependent drug effect in each individual. The hope remains, however, that as we learn more about Aß formation and turnover we will be able to interpret effects after antibodies in such a way as to be confident that we are using the optimal dose (from a biochemical effect viewpoint) in any clinical study.

As additional targets are identified to alter Aß (e.g. blocking aggregation into higher order forms) additional tools for measuring the drug effects in the brain will be needed.

The strategic position is that to test any mechanism, the field invest in whatever is required to establish that one has produced the greatest effect possible within the bounds of safety. Too often, trials take place with drugs that have little or no effect on the desired biochemical target in the brain. When these trials fail, the results can be misinterpreted as evidence against a mechanism when in fact they are simply uninformative. With a means of demonstrating that one is maximally testing a mechanism within an individual, for a mechanism that works, it will require a smaller population to establish therapeutic benefit and bring urgently needed drugs more quickly to patients.