With growing interest in testing treatments for Alzheimer’s disease at early stages, trialists need better ways to detect subtle cognitive change. Both computerized cognitive assessments and electroencephalography (EEG) can track subtle decline, but neither has been validated in diagnoses or as quantitative trial outcome measures. What’s more, because these technologies are new, the Food and Drug Administration has no clear guidelines for how such validation might be achieved. On November 19-20, the FDA’s Office of Device Evaluation hosted a workshop seeking input on how to evaluate these neurodiagnostics. Stakeholders from academia, federal agencies, and advocacy groups discussed the devices’ ability to detect early symptoms, but also the risks of misdiagnosis that stem from their use. As is often the case with regulatory deliberations, much was said but few concrete conclusions were reached right away. “The information collected by the FDA will help the agency develop an appropriate risk-based strategy and regulatory framework for these devices,” FDA media spokesperson Deborah Kotz wrote to Alzforum. Some devices have received clearance for clinical use but remain short of approval as qualified biomarkers for trials.

“Computerized diagnostic tests will become the norm. This meeting served to jump-start dialogue and gather public feedback about the way forward,” Murali Doraiswamy at Duke University, Durham, North Carolina, wrote to Alzforum. Doraiswamy advises companies such as Neuronetrix in Louisville, Kentucky, which markets an EEG device for measuring cognitive performance.

Regulation is critical, some speakers suggested, because pharmaceutical companies are beginning to incorporate new neurodiagnostic measures into dementia trials and many commercial devices are being developed or already sold. Dallas Hack, a retired colonel who oversaw military brain health research programs, noted that in the absence of specific regulations, many companies are marketing devices that do not meet basic safety and efficacy standards. “It could become a Wild West,” Hack said at the workshop.

Interest in these devices is growing because researchers urgently need more sensitive cognitive outcome measures for early trials. In people at the earliest stages of cognitive decline, existing instruments like the ADAS-Cog are inadequate because everyone aces the test. Some trialists have turned to the CDR-sb, which detects subtle impairment and can stage disease accurately. However, some experts say CDR-sb does not measure changes finely enough to be useful as an outcome measure in trials of preclinical or early dementia (see Nov 2015 conference news). Researchers fear trials may miss a treatment effect due to the bluntness of the instrument used to measure it. To better capture decline, some groups have developed cognitive composites or functional questionnaires, which look promising in initial studies (see Feb 2013 WebinarJun 2014 newsMar 2015 news). However, most of these tools remain in research use only, and have yet to receive either marketing clearance or validation as biomarkers from the FDA.

Workshop Mulls Path to Market for Neurodiagnostics
The FDA workshop mainly discussed how new neurodiagnostic devices might enter the market for clinical use. How stringent the validation is depends on how the FDA categorizes the device, and that hinges on its intended use, agency representatives explained. Class I devices are low-risk gadgets, such as a toothbrush, and require no special regulation. Class II devices are higher-risk, for example a transcranial magnetic stimulation system, and need regulations specific to each device within the class. Class III devices, exemplified by deep-brain stimulation, are highest risk and require the most stringent controls, including proof of safety and efficacy. Jay Gupta at the FDA noted that neurodiagnostic devices can serve distinct purposes: they can screen people for cognitive impairment, probe specific cognitive domains, measure whether a therapy is working, aid in diagnosis, or provide a stand-alone diagnosis. These uses might involve different levels of risk and put devices into different classes. “A number of these device types do not yet have regulations,” Gupta said. 

The FDA’s Timothy Marjenin said that many neurodiagnostic devices will probably fall into Class II, whose regulatory standard is lower than that of drugs. Manufacturers typically do not have to demonstrate a benefit to patients to put such devices on the market. They only have to show that the device works as advertised. If the invention is similar to something already on the market, manufacturers can get clearance for clinical use by filing an FDA 510(k) premarket notification, usually without submitting any clinical data. Such devices are considered “cleared” for clinical use, but not “approved.” Approval requires proof of efficacy. Only for Class III devices must manufacturers submit a premarket approval application to the FDA.

If, on the other hand, a device represents the first of its kind, manufacturers must file a de novo application and will typically need clinical data, even for Class II devices. The FDA recently designated a new device category called computerized cognitive assessment aids based on a de novo application for the Cognivue cognitive assessment system. Made by Cerebral Assessment Systems, a medical services company in Pittsford, New York, Cognivue evaluates cognitive health based on a person’s ability to click on specific regions of a computer screen in response to subtle visual cues. Computerized cognitive assessment aids are defined as Class II devices that interpret a user’s level of cognitive function based on a battery of test scores, Kotz wrote to Alzforum.

Will all neurodiagnostic devices fall into Class II? Workshop participants agreed that devices for certain purposes might require tougher regulations even if they are not invasive. While devices that simply provide a cognitive score or aid in diagnosis might be low-risk, those that claim to make a stand-alone diagnosis could be designated Class III because they could have serious consequences for patients and families if they get it wrong, Doraiswamy predicted. Complicating this classification is the eagerness of many diagnosed or suspected AD patients to slow or stop their disease. “Families impacted by Alzheimer's are willing to take on a higher level of risk than what was put forward in the technical discussions at the meeting … to me this underscores the importance of making sure that the patient voice is fully heard in these debates,” wrote Stacy Haller of the BrightFocus Foundation, which advocates for and funds AD research (see full comment below).

FDA personnel stayed mum about how the workshop discussions might affect their future decisions.

Could EEG Signals Be Useful Biomarkers?
Also notably lacking was discussion of EEG devices which, after all, have already received Class II device clearance. Alzheimer’s researchers are becoming more interested in these systems, as they measure synaptic function. Event-related potentials (ERPs), which are EEG signals generated in response to a stimulus, can detect declines in cognition (see Nov 2011 conference newsNov 2011 conference news; Nov 2012 conference news). In small studies, changes in the ERP signal correlated with progression from MCI to AD, hinting that it could serve as an outcome measure (see, e.g., Lai et al., 2010; Chapman et al., 2011; Papaliagkas et al., 2011). 

Despite thousands of scientific papers on ERP, the technology has only recently made it to primary care clinics. Traditionally, ERP measurements required complicated hardware, shielding from even weak magnetic fields, and lengthy data analysis. Neuronetrix developed a handheld, automated device and claims it works in primary care settings without specialized knowledge of EEG. The device measures brain activity as people discriminate “oddball” auditory tones from standard tones and white noise, a common ERP paradigm. Clinicians upload test data to an online database for comparison with population norms, simplifying analysis. The Cognision system was recently cleared for clinical use by the FDA and is on the market (see Feb 2015 news). The FDA cleared it via a Class II protocol because ERP systems have been used in research clinics for years. Other companies may also seek FDA clearance. For example, ElMindA, in Herzliya, Israel, is developing an ERP system that can map brain networks.

Getting devices on the market is but a first step. Trialists really want validated outcome measures for people with very early or preclinical AD. Here the path to FDA approval becomes hazy. For biomarkers, the FDA spells out clear requirements: researchers must perform a clinical trial that shows a reproducible difference between patients and controls whose diagnoses have been confirmed by standard clinical methods. However, the FDA does not define how to validate a device, said K.C. Fadem, Neuronetrix CEO.

In the absence of specific guidance, Neuronetrix attempted to meet the existing biomarker standard for Cognision by testing 103 people with probable AD and 101 controls at seven clinical study sites in the United States. The observational trial found the ERP brainwave signal had lower amplitude and was more delayed in people with AD than in controls, suggesting the measure could be a biomarker of AD, Fadem claims. Because the trial confirmed data from previous studies, he believes it met the standard of reproducibility. The data were published September 30 in Alzheimer’s & Dementia online (Cecchi et al., 2015).

The FDA has not approved Cognision as a trial outcome marker. “I don’t think the FDA is ready to pick an electrophysiological measure as a primary endpoint in Alzheimer’s,” Fadem said. However, ERP measures have been used as primary endpoints in schizophrenia trials. Even without FDA approval, several companies have incorporated Cognision as a secondary endpoint, for example in the HIV-Associated Neurocognitive Disorders (HAND) trial. At the moment, these trials are using Cognision as a pharmacodynamic marker to demonstrate that the drug hit its target in the brain. “Pharma companies are very interested in biomarkers of synaptic function,” Fadem noted.

Meanwhile, Neuronetrix continues to gather data on the device’s performance. The researchers are currently analyzing data from their observation trial to develop an algorithm for discriminating AD patients from controls. They will report sensitivity and specificity of this method and then confirm its diagnostic accuracy in a new trial. They are also following up the existing cohort with ERP measurements at one- to two-year intervals to find out whether the signal changes over time. This will be important in deciding whether ERP would make a good trial outcome measure.

Researchers also want to know how well ERP signals reflect brain amyloid. The amyloid hypothesis predicts that the presence of amyloid damages synapses. To test this, Neuronetrix is starting the Amyloid Correlation with Clinical Electrophysiology and Psychometric Tests (ACCEPT) study to link ERP signals with psychometric tests and brain amyloid as seen by PET. The hope is that ERP correlates better with amyloid than psychometric measures do, Fadem said.

Overall, people in the field agree that neurodiagnostic devices are up and coming, as technology is evolving and beginning to turn out inexpensive, easy-to-use devices.—Madolyn Bowman Rogers


  1. These technologies hold great promise. They can help better inform families and their doctors, and if they are low-cost and non-invasive they could help researchers by accelerating the recruitment of patients and caregivers into clinical trials.

    To me, one of the interesting discussion points during the day was about risk—both institutional and individual. Families impacted by Alzheimer's are willing to take on a level of risk higher than what was put forward in the technical discussions at the meeting. This is a tough question and one that will not be easily resolved, but to me it underscores the importance of making sure that the patient voice is fully heard in these debates.

    This is a rapidly evolving field, a point well made by speakers such as Dallas Hack, who suggested a number of improvements on validation of these new technologies and standardizations between trial designs. It is particularly important that these tests be useful to clinicians and, for example, provide data that is meaningful at the individual level, not just to a larger population. We must invest the resources to fully validate these tests at a high level of rigor so they truly help people.


Make a Comment

To make a comment you must login or register.


News Citations

  1. Outcomes, Outcomes: Cognition is Crux of New Alzheimer’s Trials
  2. Test Battery Picks Up Cognitive Decline in Normal Populations
  3. Test Tracks Preclinical Functional Decline
  4. EEG: Coming in From the Margins of Alzheimer’s Research?
  5. EEG: Old Method to Lend New Help in AD Drug Development?
  6. CTAD: EEG Gains Luster as More Trials Incorporate Biomarkers
  7. FDA Has Cleared New Cognitive Screening Tests

Webinar Citations

  1. New Frontier: Developing Outcome Measures for Pre-dementia Trials

Paper Citations

  1. . The role of event-related potentials in cognitive decline in Alzheimer's disease. Clin Neurophysiol. 2010 Feb;121(2):194-9. PubMed.
  2. . Brain ERP components predict which individuals progress to Alzheimer's disease and which do not. Neurobiol Aging. 2011 Oct;32(10):1742-55. PubMed.
  3. . Cognitive event-related potentials: longitudinal changes in mild cognitive impairment. Clin Neurophysiol. 2011 Jul;122(7):1322-6. PubMed.
  4. . A clinical trial to validate event-related potential markers of Alzheimer's disease in outpatient settings. Alzheimers Dement. 2015 Dec;1(4):387-94.

External Citations

  1. workshop 
  2. computerized cognitive assessment aids
  3. Cognivue cognitive assessment system
  4. Cognision
  5. ERP system
  6. observational trial 
  7. HIV-Associated Neurocognitive Disorders (HAND) trial

Further Reading