This is Part 1 of a two-part series. See also Part 2.
What a difference a year can make. At the 2011 Alzheimer’s Association International Conference in Paris, France, leading scientists in the CSF Alzheimer’s biomarker field met to tackle the vexing problem of measurement variability with the available commercial assays. Alas, back then a listener could be forgiven for thinking of the Tower of Babel, as groups presented their own stance and talked past each other as much as finding common ground.
That did not stand with Maria Carrillo of the Association, and 10 teleconferences, thousands of collective lab hours (and uncounted nudges) later, things have begun to change. When the same group—which goes by the name of Global Consortium for the Standardization of CSF Biomarkers (GCSB)—met again on 14 July 2012 in Vancouver, Canada, it was clear that the big guns have moved in on two fronts. For one, in-vitro diagnostics companies are now developing fully automated assays that do not let the user influence parameters affecting the measurements. For another, the official international bodies for certification and reference/metrology have formally agreed to take on the issue. These two developments bring the vision of stable assays grounded by single certified reference material within reach.
When it comes to CSF Aβ and tau measurements, pharma companies developing Alzheimer’s disease treatments and diagnosing clinicians share similar goals. “We need a robust assay that we can send to CROs and get the same result in the U.S. and Europe and Asia. We also need a reference standard and reference material,” said Holly Soares of Bristol-Myers Squibb in Wallingford, New Jersey. “We were in a much different place last year. Every group was doing their own thing and we did not have the diagnostics companies engaged.”
The new developments could make CSF Aβ and tau measurements as reproducible as a blood cholesterol or troponin test, said Henrik Zetterberg of the University of Gothenberg, Sweden. Specifically, the hope is that each kit manufacturer and clinical lab around the world can obtain certified reference materials at cost, to which they can calibrate their own local CSF Aβ/tau measurements.
The problem is serious. Over the past two decades, the handful of assays that are available to measure CSF Aβ and tau have enabled major scientific advances through AD research done in local research groups. They have helped transform the view of Alzheimer’s into a disease with a 15-year prodrome, and have prompted revised diagnostic criteria to now call for the use of CSF biomarkers. At the same time, these assays are research-use only (RUO). They underperform in the next step, that is, in multicenter and longitudinal studies and therapeutic trials. For its part, the Food and Drug Administration (FDA) indicated earlier this year that the current body of scientific literature on the value of CSF Aβ and tau to flag AD is insufficient for formal qualification of those markers. The big push throughout the therapy development field—articulated in dozens of presentations at the AAIC meeting in Vancouver—is to find a reliable way to pick out people within the heterogeneous clinical syndrome of MCI whose symptoms are due to Alzheimer’s, and enroll only those people into early-stage AD treatment studies.
CSF could do that well if there weren’t the problem of measurement variability. RUO assays tend to be somewhat uncontrolled, said Adam Simon, an assay developer formerly at Merck who now consults on the topic. There is variability among assays, across platforms, within centers, and between lots. In 2009, this brought into being the Alzheimer’s Association Quality Control Initiative (see ARF related news story) as an effort to survey how big the problem really is, and to see if monitoring and giving feedback to labs and kit manufacturers alone could quell the variability over time. In Vancouver, Niklas Mattsson, also of Gothenburg, told the group that three years of doing that have not mitigated the problem.
The initiative itself is a success. By now, 85 labs in more than 20 countries have signed up, and more are still joining. Three times a year, each lab analyzes two unique and one longitudinal QC sample shipped from Sweden with whatever commercial assays the lab uses locally for its AD research. But results from measurement rounds one through seven show that the ELISAs generate a stable and unacceptably high variance of about 25 percent for Aβ42, 20 percent for total tau, and 20 percent for phospho-tau. Researchers agree that these numbers must decrease to below 10 percent. All variability is not the same, Mattsson said; Aβ42’s Achilles' heel seems to be batch-to-batch changes; for tau it is differences between centers.
The study found that experience helps. Much like surgery error rates appear to be lower in hospitals that perform a given procedure frequently, the tightest CSF measurements came in from the four reference sites in the QC program, which tend to perform the assays as a clinical service for surrounding physicians or on a large scale for all ADNI samples.
Communication alone is insufficient to reduce variability, Mattsson said. The QC program gathers data on exactly how labs perform the assay with a checklist and analyzes those data. However, to date that kind of self-reporting has not pinpointed what specifically might be going wrong at individual sites. Nor has knowing one has been off by 25 percent in round one necessarily brought down a site’s variability by round seven.
In the long run, the QC program can only track and evaluate how global standardization progresses. Real change will require implementation of standard operating procedures, training courses, and certified reference materials, Mattsson said. Novel assays on fully automated platforms may also be needed, he said.
What about the assays, then? The GCSB Vancouver meeting features talks from representatives of the two companies whose RUO assays are in widespread use, followed by talks from in-vitro diagnostics companies. Pankaj Oberoi spoke for Meso Scale Diagnostics (MSD), a company that sells a chemiluminescence-based multiplex test mostly to industry and to a select few academic labs that can afford the $100K reader (most bemoan that they cannot). In June 2012, MSD released Aβ42 and tau kits that have been analytically validated in a multiyear collaboration with Soares and others at BMS.
In particular, Oberoi’s group worked to reduce the nagging problem of dilution linearity, where measurement values don’t drop as expected when the operator dilutes the CSF sample. This problem dogs all current assays and, incidentally, was difficult to overcome for cholesterol tests, too. It is blamed on the so-called matrix effect, whereby other proteins in CSF form a web of interactions that changes with each dilution step and, hence, sequesters more or less free Aβ. “We have removed the interference associated with the protein matrix,” Oberoi claimed. At the AAIC conference, Oberoi and colleagues presented posters describing the performance of these assays in vitro and in an initial clinical study of 50 control, 50 MCI, and 50 AD cases. A multisite clinical validation with the University of Gothenburg; Edith Cowan University in Perth, Australia; and other groups is underway.
Oberoi said that MSD recently became certified by the International Organization of Standards. He noted that his company trains local scientists and technicians, and is the only manufacturer to provide on-the-ground support, though some scientists privately quibbled that this support for what they perceive to be an expensive platform could be improved.
For Innogenetics, the manufacturers of the most widely used RUO CSF assays for Aβ and tau, Manu Vandijck said that the purchase last year of his company by the Japanese diagnostics company Fujirebio had not watered down Innogenetics’ commitment to AD or its freedom to operate. He acknowledged the matrix effect, and said Innogenetics researchers were developing new, automated chemiluminescent enzyme immuno assays on an integrated platform called LUMIPULSE®.
Next up were representatives of two large in-vitro diagnostics companies, each presenting similar results of their ongoing work to develop what is called a companion diagnostic. This concept received a widely noticed endorsement when the FDA issued a guidance in July 2011. The guidance encourages the twin development of a drug with a diagnostic that can identify the type of patient the drug will benefit. This is part of a trendy concept referred to as "stratified medicine." Ideally, the diagnostic should be ready for approval concurrently with the drug. Recently, several cancer drugs have been approved together with a molecular diagnostic test that identifies in which subgroups of patients they are likely to work.
Toward that end, Salvatore Salamone from the companion diagnostics company Saladax Biomedical gave an update on a joint project with Ortho-Clinical Diagnostics and BMS to develop automated Aβ and tau assays for a machine called the Vitros ECiQ analyzer. “These should be easy to use in every lab around the world,” Salamone said. Claiming precision of better than 2 percent, Salamone rattled off data points about detection limits and the stability of the assays’ reagents. He noted that adding certain (non-disclosed) dilutants rendered the assays’ linearity “near-perfect,” but acknowledged that “getting past these matrix elements was a tough nut to crack.” According to Salamone, both assays are now in design lock and their reagents in process development at an Ortho-Clinical Diagnostics production facility in the UK.
The next speaker was Tobias Bittner for Roche Diagnostics, who told much the same story. That company is part of the overall Roche Group, whose other division includes Roche Pharma, Genentech, and Chugai in Japan. Bittner develops immunoassays that run on a platform called Cobas E601. The 25,000 machines of this type that are placed in labs around the world each run automated, 18-minute-long tests at a capacity of 180 samples per hour, Bittner said.
The group develops individual sandwich ELISAs for Aβ42 and tau. They are meant to enable identification of prodromal AD patients for treatment with the antibody gantenerumab (see Alzforum interview with Luca Santarelli), and also to work as a stand-alone diagnostic. The Aβ assay—a work in progress—currently achieves a precision of 1.5 to 3 percent, and variation due to dilution linearity is below 4 percent, Bittner said, noting that the tau assay is not ready to show data.
At the GCSB meeting, Diane Stephenson of the Coalition Against Major Diseases, Phoenix, Arizona, said that in a February 2012 advice meeting with the FDA about the status of AD fluid markers, regulators recommended that CAMD help the field rise above the current scientific literature. Specifically, the FDA wants to see precision-based assays with data showing how the test relates to the level of cognitive impairment. The FDA also seeks precise definition of the assay performance characteristics, as well as analyses of individual patient level data. CAMD is a precompetitive initiative to help CSF biomarkers achieve qualification with the FDA. CAMD combines industry and academic experts with representatives of patient advocacy groups and government agencies who work collaboratively toward improving success of therapeutic trials in patients with mild cognitive impairment (see ARF related news story).
“It is gratifying to see the advancement of all the assays. We have struggled with these issues for a long time. As academics and pharma scientists, we would usually go back to our own silo and struggle alone. Without this group, this would not have happened,” said Zetterberg.
See upcoming Part 2 of this series for news on how the international certification institutions are weighing in on CSF tests for Alzheimer’s.—Gabrielle Strobel.
This is Part 1 of a two-part series. See also Part 2.