Real changes can enhance information yield on novel psychopharmacologic agents

The excellent review of current efforts and issues in the field of psychopharmacology produced by Correll et al ¹ does not, unfortunately, provide much that would convince skeptical decision makers that the future of psychopharmacology will look that much different from the past. I write from the perspective of selection of compounds and mechanisms for clinical devel­opment as well as of implementation of clinical stud­ies across phases 1‐3, both from the industry and the US National Institute of Mental Health (NIMH) vantage points. As both a past decision maker and a current advisor, I will focus on what I believe has greatest promise for the future of psychopharmacology over the next five to ten years.

Three thematic areas are implicit in Correll et al's review: a) what have we learned that is most useful in terms of design and implementation of clinical trials which herald a better future?; b) what should we do to de‐risk both compound selection and dose setting for clinical trials that will improve productivity in terms of knowledge gained as well as advancing compounds?; c) what impact is likely to derive from emerging technologies provided by such US National Institutes of Health (NIH)‐funded efforts as the Brain Initiative ² , and from utilization of remote technologies to passively and actively monitor participants in studies?

In the review, what seems to be the basis for identifying promising compounds is that there is a positive phase 2 study. Given the history of positive phase 2 studies that do not lead to successful phase 3 development, most decision makers would not see that these are any more promising than those that have failed in the past. What would be more convincing is evidence of what we have learned that can make future phase 2 trials more informative and predictive. For instance, an analysis showing that use of adaptive designs resulted in more efficient and successful drug development programs, or even a post‐hoc analysis showing some common flaws in failed phase 3 programs that would allow focus on one or a limited number of variables that could be better managed.

One trial design element that is cited as having been shown not to work, based on meta‐analyses of trials dating back to 1994 and recently confirmed, the single‐blind lead‐in, provides an excellent example of how advances can be made when data are shared. The field might be able to align on eliminating other wasteful practices if there were some way to share relevant data from as many as possible well‐powered trials con­ducted over the last decade, whether or not they resulted in approval by the US Food and Drug Administration (FDA). Such an effort could include NIH‐funded studies as well. One current effort to generate support for this kind of data sharing is provided by a panel on this topic scheduled for an upcoming meeting of the International Society for CNS Clinical Trials and Methodology.

The point made in the review that “the strongest the rationale for the randomized controlled trial (RCT), the more de‐risked the trial will be” ¹ raises the question of what constitutes a strong rationale, given a history of rationales – such as the one for targeting amyloid in Alzheimer's disease – not so far delivering after cumulative investments in the billions of dollars.

Although questions remain, I believe that having solid information on the relationship between a dose of a potential new drug and the degree to which it interacts with its primary site(s) of action in the brain and can be linked to downstream changes in brain function will allow future clinical trials to be better interpreted. One would lower risk of failure by avoiding compounds without robust translational pharmacoki­netic‐phar­macodynamic (PK/PD) brain effect data. In­deed, a recent analysis of in­dustry success rates of compounds that had full target engagement packages across ther­apeutic ar­e­as reported that 12 of 14 yielded positive proof‐of‐concept studies, with eight advanc­ing to phase 3, versus none of 12 com­pounds for which evidence of target engagement was weak or missing ³ .

As a corollary, since animals do not pro­vide true models of syndromal clinical brain conditions (except perhaps drug dependence), the future is likely to use evidence of effects on some domains in an animal assay that might be translated into humans for either a broadly defined syndromal disorder or a domain of function, as a core part of build­ing the rationale for advancing a mecha­nism and/or compound. Such is the poten­tial ben­efit of building out the Research Domain Cri­teria approach ⁴ .

As an example, the so‐called Fast‐Fail ap­proach piloted by the NIMH ⁵ , which com­plements approaches being taken with in­dustry to generate rationales to pursue a do­main such as cognitive impairment in schiz­ophrenia, has shown promise. A specific kap­pa opiate receptor antagonist, for which brain receptor occupancy data were available, was investigated in terms of potential for the domain of anhedonia. The drug was shown to positively affect a reward task‐associated functional magnetic resonance imaging (fMRI) signal and to specifically improve severity of apathy in a group of individuals with DSM depression and anxiety spectrum disorders ⁶ . This finding was seen as de‐risking future studies, and led to large pharma investment in a phase 2 trial followed by a just initiated phase 3 program (NCT03559192 and NCT05518149).

This approach goes beyond examples of selecting subsets of a DSM diagnosed group, such as failure to respond to standard treatments, or restricting subjects to those below a certain age and fewer hospitalizations, as noted for the positive phase 2 trial of ulotaront in schizophrenia. For novel mechanisms, as part of a de‐risking strategy, one should first show whether any effect can be detected on some domain of function. Then, one should decide what syndromal disorder(s) might best benefit from the compound.

This domain approach might also help de‐risk compounds with three or more phar­macological mechanisms that might be affected in humans, which are problematic in terms of demonstrating target engagement across dose ranges. A functional brain measure that translates from animals to humans, or even one with some degree of “face validity” in humans, can be applied to any molecule, whatever its mix of known mechanisms, or even initially un­known mechanisms. For compounds such as ulotaront, a promising antipsychotic discovered with a phenotypic assay battery (Smart Cube) ⁷ , a functional brain measure can potentially be used to set doses in humans prior to identification of molecular mechanisms and development of specific target engagement tools. Assessment of brain function prior to clinical testing is like­ly to become more and more part of psychopharmacology.

The utility of emerging methods, such as differentiating pluripotent cells from individuals into a neuronal type in which compounds can be tested prior to be administered, to see if some functional effect detectable in vitro predicts activity in humans, remains to be demonstrated. Nonetheless, if early reports of predicting aspects of lithium response in cells from bipolar patients ⁸ generalize to drug response predictions, this approach may become an important addition to the future of psychopharmacology.

Similarly, by then we should have enough experience to know if remote measures that can be gathered passively on a device or those resulting from approaches such as eco­logical momentary assessments are more sensitive in terms of picking up systematic drug effects than traditional types of clinical measures. It seems likely that at least some of these assessments will reveal drug effects on one or another variable that we do not currently capture with existing methods.

In summary, beyond what is recommen­ded by Correll et al's review, I predict that the near future of psychopharmacology will include a greater emphasis on target engagement PK/PD studies that can be translated from animals to humans, a focus on functional domains as a core part of building the rationale for advancing a mechanism or a compound, and the development of means for all interested parties to have access to relevant data to decide on design elements that influence signal detection in a trial.