31 March 2010. Last week, scientists from around the world gathered in the Swiss-French border city of Geneva, nestled around the southern tip of handsome Lac Leman. Right from its shores, this old town affords vistas of the Alps’ highest mountain, the Mont Blanc. From 24-27 March 2010, scientists met for the 11th International Geneva/Springfield Symposium on Advances in Alzheimer Therapy to discuss the state of affairs in Alzheimerology with some 90 talks and 60 posters. Your roving Alzforum reporter was able to attend only the first of three days of presentations; hence, impressions of the conference cover merely its N-terminal, or Phase 1, as the case may be. Still, here’s the gist of some of Thursday’s talks for all those who were unable to travel or were in parallel sessions. As always, we invite Geneva-goers to make conference coverage a community project and contribute their impressions.
Why So Few Drugs in AD?
This plaintive title question consumed a full session, as speakers representing stakeholders in AD drug development tried to answer it from their respective perches. A mix of deliberation and some finger pointing provided, if no definitive answer, then themes with which many agreed. Serge Gauthier of McGill University in Montreal started things off with a review of existing AD drugs. He noted that physicians needed better guidelines to decide when to stop prescribing them after their temporary stabilization of symptoms had likely run its course. Gauthier argued for randomized drug trials to start recruiting more homogeneous patient groups as evidenced by biomarkers. He also argued for use of progression milestones as outcome measures, and for more tests of drug combinations.
It’s the Methods, Stupid!
Robert Becker is a co-founding organizer of the original Springfield meetings, which took place in Springfield, Illinois, between 1988 and 1994 at the time Southern Illinois University had an NIA-funded Alzheimer’s Research Center. (The meeting has since moved to various European cities, Montreal, and Hong Kong. Stockholm will host the next one, in May 2012.) Becker has worked in drug development as an academic researcher. He recently founded Aristea Translational Medicine Corporation, which offers error-management consulting to drug developers. In Geneva, Becker argued its message, namely that methodological and execution errors in Alzheimer’s drug development have undermined otherwise respectable drugs. Citing phenserine, flurizan, and metrifonate—a failed drug Becker co-developed—Becker insisted that the drugs themselves were not to blame for the failures but rather errors ranging from trial design and planning to variability introduced by poorly trained raters at far-flung sites. “The methods have really not evaluated the drugs,” Becker said. He, too, noted that biomarkers are the future in AD drug development as a way to reduce human measurement error.
It’s the Drugs, Stupid!
After Becker’s talk, a different wind blew in from the health authorities. Cristina Sampaio is a clinical pharmacologist at the University of Lisbon and serves on the Committee for Medicinal Products for Human Use (CHMP) of the European Medicines Agency. (Formerly known as EMEA, EMA is essentially a decentralized “FDA” of the European Union.) Sampaio noted that while rater and other errors happen in clinical trials, the recent string of Phase 3 failures in AD had multiple, and bigger, reasons. In her view, these came down to the drug companies’ unwillingness to jettison bad drugs early enough. “The regulatory view is that there is no definitive knowledge why so many drugs have failed. I think we have seen well-minded but wrong decisions on the part of companies in taking poor drugs forward,” Sampaio said.
First off, Sampaio noted, scant new drug approvals in AD couldn’t be the result of CHMP’s refusing applications: in the 15 years of its existence, the group has refused but one application for AD, that of metrifonate. Rather, the clinical landscape has changed in that trials are now more difficult to conduct, and fewer approval applications ever reach the authorities. Drug companies are asked to test their candidates on top of existing therapy, and a paradigm shift in the science has moved the focus beyond symptomatic trials toward disease modification, she said. This raises the bar, but has not stopped drug approvals in other chronic progressive diseases, for example, rheumatoid arthritis. There, symptomatic treatments (e.g., NSAIDs) exist side-by-side with disease-modifying ones (e.g., methotrexate or injected biologics).
If it isn’t CHMP’s fault, then whose is it? Target discovery is not to blame, Sampaio said. Plenty of targets exist with reasonable validation and have drugs directed against them in clinical trials. Things mostly go wrong at that late stage. Why? A key shortcoming of current regulatory requirements for trials is the authority’s insistence on clinical endpoints coupled with the lack of formally validated biomarkers. “We recognize that, and we are rushing to validate disease-related biomarkers,” Sampaio said. Trials further suffer from slow recruitment and dropouts that erode their statistical power. That power tends to be weak to begin with, and this is due to variability between patients and between centers, to subjective rating scales that depend on rater training, and to the fact that patients on placebo decline fairly slowly these days. Yes, yes, yes, she recognizes all this, Sampaio said, but added, “I doubt any of the recent failures with disease-modifying trials can be attributed to these types of problems.”
Instead, she thinks that bad drugs are kept alive too long before they fail. “People often do not have the courage to kill drugs in Phase 2, because they have already spent too much on them. There is an emotional problem with not going to Phase 3.”
Ideally, intensive screening of potential drugs should weed out the losers sooner, well before Phase 2. Drug-related biomarkers can serve to assess target engagement, and in this way help kill some drugs and help find the right dose for others. Some drug sponsors have not taken this process to heart quite enough, Sampaio said. Nor have they invested sufficient effort into pharmacology and pharmacodynamics before tackling Phase 3, agreed Lon Schneider of the University of Southern California, Los Angeles (see below). This has become a topic of broad consensus among leading researchers. Ron DeMattos of Eli Lilly and Company in Indianapolis, Indiana, as well as researchers from other pharma companies speaking in other sessions in Geneva, detailed their efforts to buttress Phase 3 trials with ample biomarker data to characterize their drug’s behavior from animal research through Phase 2.
From her regulatory perspective, Sampaio described seeing companies take “enormous” risks going to Phase 3 based on flimsy Phase 2 data. In some cases, biomarker-driven proof-of-concept work is hardly done at all, she said; in other cases, when biomarkers flag problems such as CSF data suggesting inadequate brain penetration of tarenflurbil (aka Flurizan, see ARF ICAD story), that data is ignored. She further chided drug sponsors for hiring groups of “magicians” to reinterpret failed Phase 2 data and make them look more palatable for Phase 3. Such data often end up containing more hope than substance, she said. Instead, companies should set strict internal rules for the go/no go decision. “You need those rules. You don’t have to show them to anybody; write them in secret ink and put them in your pockets. But do abide by them,” Sampaio told the drug researchers in the audience, only partly tongue-in-cheek.
When drug sponsors come to CHMP with weak, reanalyzed Phase 2 data and request permission to conduct Phase 3, they should not interpret the committee’s green light as a hint that the agency believes the drug is good, Sampaio emphasized. “If we suspect a drug will fail for lack of efficacy, we will not stop Phase 3 from happening. We only block Phase 3 if we see a safety issue. So our ‘yes’ to conduct Phase 3 does not indicate that we will approve,” Sampaio said. “I really believe the failures happened because the drugs truly do not work and were not killed early enough.”
One force that drives this self-defeating practice is the speed delusion, Sampaio added. Company scientists are under so much pressure to be “first to market” that they skip the essential step of proving that their product crosses the blood-brain barrier, for example, because such a study would slow them down. The way out, she insisted, lies in more extensive use of drug-related biomarkers in preclinical and early clinical stages.
It’s Both, Stupid!
Sampaio’s frank assessment drew endorsement from Lon Schneider, an academic clinical trials expert whose overall take on why drugs appear to have been stalling at Phase 2a could be summed up as: It’s the methods and the drugs.
In listing 28 wannabe AD drugs that flamed out in Phase 2 or 3, Schneider noted that a number of those were actually acetylcholinesterase inhibitors, indicating that even if certain members of a drug class fail (as early immunotherapies and γ-secretase inhibitors did), that does not mean the class can’t be successful later on. One problem in evaluating drugs is that scientists have a poor grasp of where in the long AD process a drug should work best, Schneider said. For example, a drug that helps early on might be ineffective or even toxic later in the process. This issue has come up in trials of estrogen and NSAIDs, as well as anti-amyloids. In Geneva, John Breitner of the University of Washington, Seattle, addressed this issue for the ADAPT trial of the NSAIDs naproxen and celecoxib. For prior coverage of this example, see ARF ICAD story.
AD clinical trials have stayed unchanged in the past 20 years except for minor tweaks around the edges, Schneider said. Some 110 six- to 12-month-long trials and 34 18-month trials in AD by and large have employed the same design and methods because this design enabled approval of the AChE inhibitors. Likewise, patients and their rates of decline on placebo have not changed significantly, either, though it is true that AChE treatment and good medical care can obscure an effective drug if its effect size is very small. All drugs tested so far with statistically significant results have had small effect sizes ranging from about 1.6 to 4.5 points on the ADAS-cog scale for six- or 12-month trials, respectively. Often itself a target of criticism, this scale performs predictably but is relatively imprecise, Schneider said.
The known methodological limitations of trials became a disabling problem because the effect sizes of the candidate drugs tested to date have been très petit. Heterogeneity among patients, ratings, sites, and countries, combined with small effect size, create a situation where a trial is virtually set up to fail if it enrolls fewer than 100 people per group, or if it does not treat for a long time, such as 18 months or longer. Worse, because the trials are underpowered to detect these small effect sizes, play-of-chance rules, meaning that an ineffective drug can come out looking effective.
This same old trial design might still prove serviceable once a drug with a truly strong effect comes along. But for the most part, its limitations make late-stage trials supremely vulnerable to failure, Schneider said, and call for change. Some ideas, besides developing plain better drugs: test them in groups of patients recruited to match the target, for example, anti-amyloid drugs in patients who are known to have brain amyloid; focus on change in individual patients, not groups of patients; use multivariate outcomes that are in line with the expected action of a drug; and bulk up pharmacology.
Hang in There, It Takes Time!
Eric Siemers of Eli Lilly and Company sidestepped the session’s title question. He gave a pharmaceutical perspective on why it takes so long to get to the treatments patients are hoping for. To counter the perception that AD drug development is especially slow, Siemers first noted that while President Nixon’s 1971 declaration of war on cancer has led to more than $100 billion spent by the National Cancer Institute alone, nearly 40 years later therapy has improved, but cancer is far from cured. In Parkinson disease, pharmaceutical drug development to this day revolves around improving dopamine-based symptomatic drugs whose development grew out of a scientific discovery in the 1960s, even though in the interim a handful of PD genes have been identified starting in 1998. No drugs related to protein deposition of α-synuclein, the first PD gene, are in the clinic yet. “It takes a while for industry to shift to other targets,” Siemers said.
Glacial as these timelines seem to someone who has a neurodegenerative disease, by such standards, AD drug discovery would not seem quite such a laggard. Approved in the 1990s to about 2001, current AD drugs followed the scientific discovery of acetylcholine loss by some 20 years. The autosomal-dominant gene discoveries of the 1990s defined a new set of targets against which a “flotilla” of compounds are wending their way through mid- to late-stage clinical trials, Siemers said. One Phase 3 drug, Lilly’s γ-secretase inhibitor semagacestat, first entered the clinic in the year 2000, nine years after the first APP mutation and five years after the first presenilin mutations were reported. In contrast, ApoE has only recently inspired renewed interest in pharma firms. “ApoE is not a highly druggable target,” Siemers said.
It can take a company several years between deciding an initial idea might be druggable and really starting the development process in earnest. “It is difficult for us to shift from dopamine to α-synuclein, or from acetylcholinesterase to Aβ compounds. This is where collaboration and intense discussion can be key,” Siemers told the audience. Moreover, in taking on a new target, companies may use an earlier, inferior compound to explore the concept and keep a better compound just behind the front line. This was in evidence at the Springfield symposium poster session, for example, where scientists from Janssen Pharmaceuticals and Cellzome presented proof of concept for a new γ-secretase modulator in cell culture and in the Tg2576 mouse model, but said they are truly interested only in a less advanced sister compound on which they are not presenting any data. While common in industry, this strategy is expensive, Siemers said.
Constraints outside of the scientific program can delay a drug’s march to market, as well. For example, the number of clinical investigators available to run multicenter trials has been declining for the past decade, Siemers said. Part of the reason might be that companies tend to demand more in the way of procedures and record-keeping from investigators these days, while their payment for a given trial has stayed the same, Siemers said. To enlist enough sites and patients, Phase 3 trials nowadays take place in many different countries all around the world. This provokes some anxiety, Siemers said. While capable people live everywhere in the world, site experience and rater training vary, increasing demand for monitoring. Just like a single national leader arrives at a decision faster than does the United Nations (which incidentally is located across the street from the Geneva International Conference Center, where the conference was held), multinational trials increase the complexity of regulatory oversight, slowing the process down.
Siemers summed up his talk with a wish for more sites, more investigators, more patient involvement, and a simpler regulatory process. On the Holy Grail of finding a Phase 2 strategy that decreases the risk of Phase 3 failure without requiring long treatment periods, he said: “I don’t know what that is. The person who finds it will be able to give plenary lectures for the next five years.”—Gabrielle Strobel.