Cold Spring Harbor: How Can We Improve Clinical Trial Design?
As the pipeline fills with potential therapies for Alzheimer (AD) and other neurodegenerative diseases, the research community is paying increased attention to how those therapies will be evaluated in clinical trials. Biomarkers, which are sorely needed for both AD and Huntington disease (HD), emerged as a central theme at the fourth meeting on Neurodegenerative Diseases at the Cold Spring Harbor Laboratory from 30 November through 3 December. Not only can biomarkers sharpen the diagnosis, but, equally importantly, they can improve the efficiency of clinical trials by monitoring treatment response and, in the best case, provide surrogate endpoints of the disease.
Since these diseases progress slowly, monitoring treatment response simply by assessing symptoms is too slow and requires too many subjects to allow an efficient trial. Moreover, by the time symptoms appear, much of the damage has already been done. In HD, for example, approximately half of the striatum has degenerated before the onset of symptoms, said Elizabeth Aylward from the University of Washington, Seattle. Aylward used magnetic resonance imaging (MRI) to track striatal volume in people who carry the HD gene. She claimed that this measure fulfills all the requirements of a biomarker: it can be objectively and reliably measured, it changes in a predictable manner over time, it predicts a known endpoint (onset of symptoms), and is associated with a known mechanism of pathology (neurodegeneration).
Aylward plotted volumes of the caudate and putamen, the two areas of the striatum, against estimated years to onset of disease. Years to onset was calculated using a formula based on the number of CAG repeats and the age of onset of the affected parent. The HD mutation represents an expansion of the trinucleotide cytosine-adenine-guanine (CAG) in the huntingtin gene. Having more than 39 CAG repeats virtually guarantees that the person will get Huntington’s, while fewer than 27 means (s)he will not. Aylward’s studies showed that the changes in putamen and caudate volume were linear and predictable: over a period of 2 years, the caudate shrank by some 4.2 percent, and the putamen by 5.6 percent. In addition, Aylward showed that the rate of change becomes significant about 10-12 years before onset. Subjects with caudate volumes smaller than 4.6 cc were symptomatic, while those with volumes greater than 5.3 cc were not. Control subjects had a caudate volume of 9.8 cc. In Aylward’s hands, caudate volume predicted which patients would be diagnosed with HD within 2 years with 100 percent accuracy (Aylward et al., 2004).
In designing clinical trials, MRI volumetric measures could serve to enrich for people who would be expected to develop symptoms during the course of the trial. This would allow onset to be used as a primary outcome measure, said Aylward. As the MRI studies show, however, significant striatal degeneration has already occurred by the time symptoms appear, suggesting that drugs to prevent neuron loss must be given sooner. Even more valuable would be to use striatal volume as a surrogate endpoint that would herald a clinical benefit during a treatment trial in presymptomatic patients. “We don’t have a treatment, so we can’t demonstrate this yet,” said Aylward. For the meantime, she recommended that MRI striatal volume be used provisionally as a surrogate endpoint to screen candidate treatments for full clinical trials that would use clinical outcome measures. The advantages, she said, are that striatal volume can be used in subjects who are many years away from onset, that its longitudinal use generates no practice or placebo effects, and that it could help flag the efficacy of a treatment even in small sample sizes.
While monitoring changes in the striatum may be particularly useful during the presymptomatic stages of HD, Herminia Diana Rosas from Massachusetts General Hospital said that cortical changes during the early symptomatic stages may help explain why, despite more than 50 percent striatal loss at the time of diagnosis, clinical symptoms continue to progress as further loss occurs and spreads. Data from 33 early symptomatic HD patients demonstrated significant thinning across the sensorimotor cortex early in the disease, which appeared to extend to other areas of the cortex in a regionally specific and temporally defined manner (Rosas et al., 2005). Rosas has investigated whether cortical thinning might serve as a surrogate marker in a pilot study of a neuroprotective agent in patients with HD. Earlier studies in HD transgenic mice had shown that the drug slows progression. In the human trial, after six months, patients showed a statistically significant slowing of the rate of change of cortical thinning in several regions; data on clinical symptoms are too mature for publication at this point, Rosas said. While further validation of these results is needed, Rosas concluded that MRI measures of cortical degeneration may provide a surrogate endpoint, and thus may improve the efficiency of clinical trials. Moreover, these methods may serve to clarify the role of the cortex in the pathogenic process.
Because a definitive genetic test is available for HD, and all HD is caused by huntingtin expansions, biomarkers are not needed for diagnosis. With AD the situation is more complex, according to Chester Mathis from the University of Pittsburgh, Pennsylvania. Although many experienced clinicians at Alzheimer’s Disease Centers (ADCs) report being able to diagnose the disease with up to 95 percent accuracy, neurologists and psychiatrists outside of specialized centers have a higher error rate. Moreover, at ADCs only 20 percent of patients do not have confirmed AD, said Mathis, while the proportion of people without AD likely is much higher at other practices. The specificity of diagnosing non-AD dementia cases hovers around 55 percent. There are no clear measures to predict whether, or when, someone with mild cognitive impairment (MCI) will progress to AD (10-15 percent convert to AD per year, while 40 percent prove to have some other cause). In light of all this uncertainty, an objective marker that identifies AD would help identify appropriate subjects for clinical trials, and track their response.
Mathis, William Klunk, and colleagues have developed a positron emission tomography (PET) imaging tracer known as Pittsburgh Compound B (PIB, N-methyl-[(11)C]2-(4'-methylaminophenyl)-6-hydroxybenzothiazole). This chemical binds selectively to Aβ plaques in vivo, allowing identification and quantification of the pattern of amyloid deposition in the brain. Control brain does not retain PIB, but AD brain does so selectively in the temporal, parietal, anterior cingulate, posterior cingulate/precuneus and frontal cortices, and striatum (Klunk et al., 2004). Colleagues at Washington University in St. Louis, Missouri, showed that one in 10 of nondemented control subjects, especially older people, had PIB binding that resembled the pattern seen in AD brains (Mintun et al., 2006). These individuals will be monitored over time to see whether they go on to develop dementia.
Mathis also reported on studies in people with MCI (Lopresti et al., 2005). About 60 percent of them had amyloid deposits suggestive of AD; the rest did not. The Pittsburgh group suspects that these are the 60 percent who will go on to develop AD; however, only continued follow-up can confirm whether these patients have AD. One big question in terms of clinical trials is whether PIB binding patterns change during treatment. The Pittsburgh group has initial data suggesting as much. Postmortem cortex from a subject in the AN-1792 anti-Aβ trial had shown plaque clearance (Masliah et al., 2005) and the same tissue bound as little PIB as does cortex from control. For upcoming trials, Mathis said that PIB should be able to reliably detect regional differences of 10 percent or greater, and that correlations will be made with other measures, including cognitive, volumetric, and biochemical markers such as CSF Aβ and tau.
Currently, about 20 sites worldwide are studying PIB binding in normal aging, familial AD, and Down syndrome, and some anti-amyloid clinical trials are evaluating it as a surrogate marker of efficacy. Christopher Rowe and colleagues from Austin Hospital in Melbourne, Australia, used PIB to examine Aβ deposition in healthy people, as well as in those with AD, dementia with Lewy bodies (DLB), frontotemporal dementia (FTD), and MCI. Concurrently, they performed [18-F] Fluorodeoxyglucose (FDG)-PET, MRI, neuropsychological evaluations, and ApoE4 testing. Rowe’s data indicate that PIB binding is strongly elevated in AD, less so in DLB, and not at all in FTD. Among MCI patients, 64 percent had elevated PIB binding similar to AD subjects. The amount of PIB binding correlated with cognitive status in people with MCI but not in AD or DLB patients. This might be because the cognitive status of AD and DLB patients declines with time, but the amyloid load has reached a plateau before symptom onset, so PIB binding remains more or less stable during the clinical phase. PIB binding was higher in people with the ApoE4 allele.
Comparing FDG-PET with PIB-PET for detection of AD, Rowe found that PIB was more accurate, specific, and sensitive than FDG-PET, particularly among older subjects. FDG-PET, however, may be useful for unusually early detection and tracking of AD, said Eric Reiman from the Banner Alzheimer’s Institute in Phoenix, Arizona. He presented a series of studies showing decline in cerebral glucose metabolism, as assessed by FDG-PET, in cognitively normal ApoE4 carriers (Reiman et al., 2005). Reiman suggested that FDG-PET could measure a presymptomatic quantitative endophenotype to evaluate putative modifiers of AD risk. In this manner, FDG-PET could assess the effectiveness of promising prevention therapies in a select group of higher-risk subjects over a shorter period of time, as compared to studying thousands of healthy volunteers and waiting many years to determine how many of the volunteers develop symptoms, and when. While PET and MRI measures were the hottest topics at the meeting, scientists also proffered a number of other measures as possible biomarkers. Peter Snyder of the University of Connecticut, Storrs, discussed using measures of voice acoustics in patients with Parkinson disease (PD). Paul Maruff from the University of Melbourne presented data on an assessment of cognitive decline in AD patients that measures the time it takes a person to complete a simple reaction time task. While a single measure would not be meaningful, said Maruff, multiple assessments over time show 100 percent specificity and sensitivity for AD. In another novel biomarker approach, Holly Soares, from Pfizer Global Research and Development, in Groton, Connecticut, presented data from a clinical trial of atorvastatin for AD, in which investigators used a multiplex panel that included 78 cardiovascular and inflammatory endpoints, as well as Aβ peptide levels, to evaluate the effect of the treatment. The study not only collected additional data regarding possible biomarkers for AD, but also provided a window to ask other questions about the pathogenesis of AD and the mechanism of action of statin drugs.
PET, MRI, biochemical markers, and clinical and neuropsychological assessments are all being evaluated as part of the Alzheimer’s Disease Neuroimaging Initiative (ADNI), as was discussed by Clifford Jack, from the Mayo Clinic and Foundation in Rochester, Minnesota, and Leslie Shaw, from the University of Pennsylvania in Philadelphia. The biggest impact of biomarkers will come in conjunction with innovative clinical trial design, said Michael Krams of Wyeth Research. In a provocative talk, Krams presented a new way of designing clinical trials for disease-modifying AD drugs. The gist of Krams’s presentation was the idea that trial design must adapt to new information, including biomarker data, that emerges during the course of a trial. Consequently, the trial should adjust the sampling and allocation of subjects to different arms as it progresses. This kind of design would facilitate the implementation of seamless phase II/III trials that would require fewer subjects and take less time. In Phase I/II trials, Krams suggested scrapping traditional dose escalation studies in favor of continuous reassessments of subjects given increasing doses over time. Computer analysis of outcome data captured at different time points would be incorporated into the decision making process to determine whether a study should be continued as is, stopped, or tweaked. John Trojanowski, of the University of Pennsylvania, said this kind of innovation is desperately needed in the evaluation of treatments for AD. “We need a revolution in clinical trials for AD,” he said. “This is a great start.”—Lisa J. Bain.
Lisa Bain is a freelance science writer based in Philadelphia.
- Aylward EH, Sparks BF, Field KM, Yallapragada V, Shpritz BD, Rosenblatt A, Brandt J, Gourley LM, Liang K, Zhou H, Margolis RL, Ross CA. Onset and rate of striatal atrophy in preclinical Huntington disease. Neurology. 2004 Jul 13;63(1):66-72. PubMed.
- Rosas HD, Hevelone ND, Zaleta AK, Greve DN, Salat DH, Fischl B. Regional cortical thinning in preclinical Huntington disease and its relationship to cognition. Neurology. 2005 Sep 13;65(5):745-7. PubMed.
- Klunk WE, Engler H, Nordberg A, Wang Y, Blomqvist G, Holt DP, Bergström M, Savitcheva I, Huang GF, Estrada S, Ausén B, Debnath ML, Barletta J, Price JC, Sandell J, Lopresti BJ, Wall A, Koivisto P, Antoni G, Mathis CA, Långström B. Imaging brain amyloid in Alzheimer's disease with Pittsburgh Compound-B. Ann Neurol. 2004 Mar;55(3):306-19. PubMed.
- Mintun MA, Larossa GN, Sheline YI, Dence CS, Lee SY, Mach RH, Klunk WE, Mathis CA, Dekosky ST, Morris JC. [11C]PIB in a nondemented population: potential antecedent marker of Alzheimer disease. Neurology. 2006 Aug 8;67(3):446-52. PubMed.
- Lopresti BJ, Klunk WE, Mathis CA, Hoge JA, Ziolko SK, Lu X, Meltzer CC, Schimmel K, Tsopelas ND, Dekosky ST, Price JC. Simplified quantification of Pittsburgh Compound B amyloid imaging PET studies: a comparative analysis. J Nucl Med. 2005 Dec;46(12):1959-72. PubMed.
- Masliah E, Hansen L, Adame A, Crews L, Bard F, Lee C, Seubert P, Games D, Kirby L, Schenk D. Abeta vaccination effects on plaque pathology in the absence of encephalitis in Alzheimer disease. Neurology. 2005 Jan 11;64(1):129-31. PubMed.
- Reiman EM, Chen K, Alexander GE, Caselli RJ, Bandy D, Osborne D, Saunders AM, Hardy J. Correlations between apolipoprotein E epsilon4 gene dose and brain-imaging measurements of regional hypometabolism. Proc Natl Acad Sci U S A. 2005 Jun 7;102(23):8299-302. PubMed.
No Available Further Reading
Cold Spring Harbor: A Grab Bag from the Drug Discovery Folks
Drug discovery for neurodegenerative diseases continues on all fronts, from studies of fundamental biology to the development of new therapeutic compounds. The Cold Spring Harbor conference, Neurodegenerative Diseases: Biology and Therapeutics, held 30 November through 3 December, 2006, offered something for everyone involved in this process. A few highlights from the meeting are presented below.
The National Institute of Health (NIH), foundations, academia, and pharmaceutical companies have all entered the fray by developing tools and new strategies to facilitate drug discovery. Christopher Austin from the NIH opened the meeting with a presentation about the Molecular Libraries Screening Center Network (MLSCN), a major component of the Molecular Libraries and Imaging Initiative of the NIH Roadmap for Medical Research. Ten NIH-funded screening centers around the country make up the MLSCN. They perform high-throughput screening (HTS) and develop chemical probes for investigators in academia who have identified novel therapeutic targets but lack the tools needed to identify hits and develop them further. Each of the screening centers offers different technologies. Investigators submit their assay to the MLSCN, which organizes a peer review to determine whether the assay is ready for HTS and which center would provide the best match. The centers have access to the 100,000-compound Small Molecular Repository, a collection of compounds the NIH has assembled for HTS through a contract with the company BioFocus DPI in South San Francisco, California. Scientists at the MLSCN center optimize the assay and then screen the compound collection to find a candidate probe. Analogs synthesized during the optimization may be added to the compound repository; then the MLSCN center returns data to the investigator. MLSCN also deposits the data in a publicly accessible database called PubChem, which is linked to other NIH databases. Austin said that although data are freely shared on PubChem, investigators are unlikely to get scooped. Not only do they usually have the data first, but they also are likely to be ahead on synthesis of the compounds. PubChem provides a rich source of pharmacologic data that can be applied to other research questions.
The resources available to pharmaceutical companies are even more elaborate, according to John Houston of Bristol-Myers Squibb Pharmaceutical Research Institute in Wallingford, Connecticut. Houston described an HTS program that can produce 500,000 to one million data points in 2 to 3 weeks, using many different assay technologies and screening systems. This integrated program allows a candidate compound to flow seamlessly through the drug discovery process, from selection of the initial chemical entity to evaluation of lead compounds, and it yields a compound profile with highly annotated data. The neuroscience group at Bristol-Myers Squibb focuses on schizophrenia, anxiety/depression, and Alzheimer’s (AD), Houston said, but compound information gathered during this process is available to scientists working in other therapeutic areas within the company, as well.
Foundations that focus on a particular disease are also developing new strategies for drug discovery. Doug MacDonald works at CHDI, Inc. in Los Angeles, California, a non-profit biotech organization dedicated solely to finding treatments for Huntington disease (HD.) He described a three-pronged approach to drug discovery. It includes internally driven programs, external collaborations with other biotech organizations, and compound development agreements with pharmaceutical companies who have an advanced compound that looks promising for HD. For the internal “prong,” CHDI functions as a virtual drug discovery engine. It uses contract research organizations (CROs), for assay development, screening, medicinal chemistry, structural biology, analytical chemistry, bioinformatics, pharmacokinetics, formulations, and toxicology to accomplish the tasks that CHDI has identified as essential for a given project. For example, one project in the CHDI portfolio focuses on transglutaminase 2; this protein that been validated as a possible HD therapeutic target in several reports, but there are currently no good validating ligands for it. CHDI established collaborations with scientists and CROs to identify potent and selective inhibitors that would cross the blood brain barrier. Working with the biotechnology company Evotec AG in Hamburg, Germany, they screened nearly 300,000 compounds and are now pursuing 21 chemical “clusters” that were identified based on structure-activity relationships. Next, they plan to conduct cell-based assays and then tests in HD mouse models.
Secondly, CHDI has established collaborations with biotechnology companies to apply specific technologies and expertise to HD drug discovery. As an example, MacDonald described one such collaboration with CombinatoRx, Inc. in Cambridge, Massachusetts, where combinations of existing drugs are screened in high throughput HD assays to find compounds that act synergistically against HD targets. And thirdly, CHDI has contracted with PsychoGenics, Inc. in Tarrytown, New York, to test candidate drugs from large pharmaceutical and other biotech companies companies in proof-of-mechanism studies using HD mouse models.
Meanwhile, academic institutions and medical centers are making their own inroads into drug discovery. Linda Van Eldik is at the Center of Drug Discovery and Chemical Biology at Northwestern University in Chicago, Illinois. Van Eldik talked about a “crisis in drug discovery,” where the number of new chemical entities is flat or declining despite increased research funding. She called the situation especially grave for neurologic diseases. She and other speakers agreed that planning of the early stages of drug discovery must improve to decrease risk, lower cost, save time, and reduce late-stage failures.
Van Eldik described a de novo lead discovery approach to find compounds that target glia as mediators of neuroinflammation in AD (Wing et al., 2006). Her team started with a particular chemical skeleton called an inactive aminopyridazine fragment. That’s because compounds built on this structure have proven safe and effective as CNS drugs. The scientists then diversified the fragment chemically, designing compounds with properties such as low molecular weight, moderate lipophilicity, as well as good solubility, safety, and bioavailability. Next, they used cell-based primary screens to identify compounds that inhibit glial activation, followed by secondary screens of compounds in animal models. In less than 2 years, the scientists identified a lead compound, refined it with medicinal chemistry to identify a candidate for clinical development, and worked out a chemical process to produce it (Hu et al., 2006). In mouse AD models, the compound, Minozac, suppressed upregulation of proinflammatory cytokines, decreased astrocyte and microglial activation, prevented loss of synaptic proteins, and attenuated behavioral deficits. It is currently in clinical development for AD and related disorders at a biotechnology company that has licensed Minozac and similar compounds, VanEldik said.
Minozac is one of a dozen compounds discussed at the meeting as potential therapies for AD or other neurodegenerative diseases. Many of the other AD treatments are focusing on amyloid. Paul Aisen of Georgetown University identified three categories of anti-amyloid therapies: those that affect secretase cleavage, those that target the amyloid peptide itself, and those that have anti-inflammatory, antioxidant, or neuroprotective properties. Among those that act on the secretase cleavage site or directly on the amyloid peptide, there are both immunological and non-immunological approaches.
Ron Black of Wyeth Research, in Collegeville, Pennsylvania, reported results from a phase 1 trial of 30 patients with mild to moderate AD who received a single dose of bapineuzumab, a humanized monoclonal antibody to Aβ. These data were originally presented at the Geneva-Springfield Symposium in April, 2006. Bapineuzumab is a second-generation immunotherapeutic from Wyeth and Elan Pharmaceuticals in South San Francisco, California. At the highest dose of 5 mg/kg, three of 10 patients developed MRI abnormalities, which were not seen among the 12 patients who got lower doses. Four months after the injection, cognitive performance as assessed with the Mini-Mental State Exam (MMSE) improved with lower doses, reaching statistical significance at 1.5 mg/kg. Treatment also caused a transient increase in plasma Aβ. A phase 2 trial started in 2005, but Black did not present data on it.
Beka Solomon from Tel-Aviv University in Tel-Aviv, Israel, developed a different immunological approach using antibodies against the β-secretase cleavage site on the amyloid precursor protein (APP, Arbel et al., 2005). In transgenic mice, both the Tg2576 line and another also expressing the APP Swedish mutation, animals treated with the antibody performed better on an object recognition test in a dose-dependent manner, and also showed less microglial activation in the hippocampus, dentate gyrus, and parietal cortex. Treatment with this antibody led to fewer brain microhemorrhages than were seen in mice treated with some other anti-Aβ therapies that involve a redistribution of amyloid from the parenchyma to blood vessels (Morgan, 2005). Solomon said that her proposed immunotherapy does not involve such a redistribution, and thus is less likely to cause microbleeds.
Among non-immunologic anti-Aβ therapies, Paul Aisen of Georgetown University in Washington, D.C., and Lara Fallon of Neurochem in Laval, Quebec, Canada, discussed tramiprosate (AlzhemedTM). This small molecule therapeutic is believed to prevent amyloid deposition by binding to soluble Aβ. Phase 1 and 2 clinical studies suggested that tramiprosate safely reduces CSF Aβ42 levels in mild to moderate AD patients, and in an open-label, uncontrolled extension of the phase 2 trial, people with mild AD so far have shown no significant cognitive decline at 20 months, said Aisen. Results for the North American phase 3 trial are expected in spring 2007.
On the basic science front, Scott Small of Columbia University in New York City studies the basic cell biology of AD to understand key mechanistic pathways that may be vulnerable to disease pathogenesis. Small initially demonstrated with brain imaging and gene expression profiling that late-onset AD patients have defects in the retromer trafficking complex, which is important for protein sorting (Small et al., 2005). Then Small went on to show in cell culture and in a knockout mouse model that such defects lead to accelerated Aβ production, thus implicating retromer dysfunction as a therapeutic target. More recently, Small and colleagues have found a polymorphism in a retromer-related molecule that significantly increases the risk for AD (in press).”—Lisa J. Bain.
Lisa J. Bain is a freelance science writer in Philadelphia.
- Wing LK, Behanna HA, Van Eldik LJ, Watterson DM, Ralay Ranaivo H. De novo and molecular target-independent discovery of orally bioavailable lead compounds for neurological disorders. Curr Alzheimer Res. 2006 Jul;3(3):205-14. PubMed.
- Hu W, Ralay Ranaivo H, Roy SM, Behanna HA, Wing LK, Munoz L, Guo L, Van Eldik LJ, Watterson DM. Development of a novel therapeutic suppressor of brain proinflammatory cytokine up-regulation that attenuates synaptic dysfunction and behavioral deficits. Bioorg Med Chem Lett. 2007 Jan 15;17(2):414-8. PubMed.
- Arbel M, Yacoby I, Solomon B. Inhibition of amyloid precursor protein processing by beta-secretase through site-directed antibodies. Proc Natl Acad Sci U S A. 2005 May 24;102(21):7718-23. PubMed.
- Small SA, Kent K, Pierce A, Leung C, Kang MS, Okada H, Honig L, Vonsattel JP, Kim TW. Model-guided microarray implicates the retromer complex in Alzheimer's disease. Ann Neurol. 2005 Dec;58(6):909-19. PubMed.
No Available Further Reading