Researchers conducting Alzheimer’s prevention studies face a dilemma: how to measure effects on cognition in normal populations? With no validated measure available for this purpose, scientists have pioneered new instruments by combining tests that best detect decline in observational studies of preclinical AD (see Feb 28 Webinar). Preliminary evidence for one such cognitive composite, the Alzheimer Disease Cooperative Study–Preclinical Alzheimer Cognitive Composite (ADCS–PACC), looks promising as reported in the June 2 JAMA Neurology. Researchers led by Michael Donohue at the University of California, San Diego, and Reisa Sperling at Brigham and Women’s Hospital, Boston, report that this composite picks up early signs of cognitive problems in people with brain amyloid or other risk factors for future Alzheimer’s dementia. 

The authors reached this conclusion based on retrospective analyses of data from three large observational studies. Although this approach does not validate the instrument, it suggests that the PACC will detect decline in prospective trials, Sperling told Alzforum. 

Other researchers agree. “The approach by Donohue et al. is a major first step in meeting the challenge of designing effective and informative secondary prevention trials,” Richard Kryscio at the University of Kentucky, Lexington, wrote in an accompanying editorial. Steven Ferris at New York University Langone Medical Center also believes the composite is promising, but cautioned that it needs to be validated. “There is good reason to think the PACC will be able to pick up a treatment effect; however, the ultimate validation of this composite will come from a prevention trial,” he said. Ferris chairs the instrument committee of the ADCS, but was not involved in the development of the PACC.

That validation is in the works. The ADCS–PACC will serve as the primary outcome measure for the Anti-Amyloid Treatment in Asymptomatic Alzheimer’s (A4) study, which enrolls cognitively normal older adults who have evidence of brain amyloid. Researchers are currently screening participants for this trial, and expect to soon begin treating the first volunteers with the anti-amyloid antibody solanezumab, said Sperling, the A4 principal investigator (see Jan 2013 news story). The trial will read out in about 2019. Other prevention studies conducted by the Alzheimer’s Prevention Initiative (API) and the Dominantly Inherited Alzheimer Network (DIAN) will use different cognitive composites, but test the same basic cognitive functions as the PACC, and they include some overlapping measures to facilitate comparisons, Sperling noted.

The ADCS–PACC combines four widely used paper-and-pencil cognitive tests. These include the list-learning task from the Free and Cued Selective Reminding Test, as well as a paragraph-recall test from the Wechsler Memory Scale, both of which measure episodic memory. The Digit Symbol Substitution Test from the Wechsler Adult Intelligence Scale tests executive function. The final component, the Mini-Mental State Examination, assesses global functioning and mental status. 

The inclusion of the MMSE might seem surprising, since most cognitively normal people perform at ceiling on this test, with scores of 29 or 30. However, Sperling noted that a drop of even one point, to 28, strongly predicts future cognitive decline. The MMSE questions that assess orientation to time and place best discern faltering cognition, but the authors chose to include the entire test because it is well-validated and provides a means to compare the A4 to other studies. Previously, the authors reported that the PACC had good test-retest reliability (see Nov 2013 conference story). 

To validate the test in a population similar to A4 participants, Donohue and colleagues turned to three observational studies: the Australian Imaging, Biomarkers, and Lifestyle (AIBL) Flagship Study of Ageing; the Alzheimer’s Disease Neuroimaging Initiative (ADNI); and the ADCS Prevention Instrument study. These included many of the same cognitive measures as those in the PACC, allowing the authors to get an idea of how their composite might perform over time.

Over a three-year span, scores on the composite dropped for cognitively normal participants in AIBL and ADNI who had accumulated amyloid in their brains, while staying stable for participants without evidence of brain amyloid. The results suggest that amyloid-positive participants in the A4 trial will experience a similar rate of decline in PACC scores, Sperling said. She added that the A4 trial is powered to detect a 30 percent slowing of cognitive decline. The ADCS Prevention Instrument study did not measure brain amyloid, but in that study, people who carried the ApoE4 risk allele did worse on the composite over time compared with the steady performance of those without an E4 allele, again suggesting the composite can pick up subtle changes in cognitive performance in at-risk individuals.

In his editorial, Kryscio noted some shortcomings of this retrospective validation. Because not all elements of the PACC were present in all the natural-history studies, the authors had to substitute similar tests in some cases. For example, AIBL and ADNI used a different word-list recall test, and the ADCS Prevention Instrument study used a different paragraph-recall test, as well as a modified MMSE. Sperling found it encouraging, however, that all the studies showed similar rates of decline, even when the elements of the composite were not identical. “It’s the combination of these four measures that seems to have the most power. The convergence of the data across the multiple natural history studies gave me confidence that this is a good composite,” she said.

Ferris agreed that the combination of measures of episodic memory, executive function, and temporal and spatial orientation seems to be the key to detecting the earliest signs of impairment in AD. However, he noted that although the individual PACC tests track decline, they are not sensitive enough to distinguish between people with and without brain amyloid at baseline. “This is a reasonable composite test battery, but I don’t think this is the optimal composite measure for preclinical Alzheimer’s,” he told Alzforum.

Sperling agrees. The A4 trial includes two secondary, exploratory measures that may turn out to be more sensitive. The first comprises a set of self-reports, which will ask participants to assess how they think they are doing cognitively. Studies have shown that subjective memory complaints correlate with brain amyloid and predict future cognitive decline (see Jul 2013 conference story). 

The other measure consists of a computerized test battery administered on an iPad. It includes items from the CogState test battery, as well as a face-name pairing memory task and a pattern separation task. These last two measures have been shown by functional MRI studies to engage brain regions that are affected early in Alzheimer’s disease, Ferris noted. For example, face-name pairing depends on the perforant path of the hippocampus and entorhinal cortex. These areas accumulate tau pathology and show synaptic defects early in AD. These tasks may be more sensitive to the presence of Alzheimer’s pathology than the tests in the PACC are, Ferris suggested. “I think the computer battery will probably outperform the primary composite,” he speculated. “At the end of the study, researchers will have a lot more information to help design the cognitive measures for the next group of prevention studies.”—Madolyn Bowman Rogers.


  1. Donohue and colleagues have assembled a composite that shows excellent promise as a clinical trial endpoint. The composite is straightforward, easy to interpret, and strengthened by the theoretical and empirical bases underlying its construction. For obvious reasons, it is critical that the composite is optimally sensitive to detect treatment effects in Alzheimer's prevention trials. A treatment that impacts fluid or neuroimaging biomarkers alone would represent a remarkable step forward, but the ultimate Alzheimer's biomarker is cognitive function. Without evidence of cognitive stabilization or amelioration of cognitive deficits, a prevention trial cannot be considered truly successful. Thus, the importance of selecting a cognitive endpoint cannot be overstated.

    It was somewhat surprising to see inclusion of the full 30-point MMSE, which of course has a well-known ceiling effect. In addition, the rationale behind using the total-recall score rather than the much more sensitive free-recall score from the Free and Cued Selective Reminding Test was not provided. In our experience at the Knight Alzheimer's Disease Research Center at Washington University in St. Louis, we have seen considerable ceiling effects on the total recall score from the FCSRT in cognitively normal populations, including those with neuroimaging and CSF evidence of Stages I and II preclinical AD (Sperling et al., 2011). Since two of the measures will likely have pronounced ceiling effects in preclinical populations, it seems that there will be much leverage put on Logical Memory and the Digit-Symbol Substitution test. However, both of these measures are tried and true measures of cognitive functioning and have scores of studies establishing their validity in Alzheimer's disease.

    In the Dominantly Inherited Alzheimer Network Trials Unit (DIAN-TU), we are still in process of determining our cognitive endpoint and have thus far considered several candidates. We have evaluated these candidates in the DIAN observational study cohort, which began data collection in 2009. Many participants from this study will roll over into the DIAN-TU, making the observational cohort an ideal population in which to explore different cognitive endpoints. To date, we have considered endpoints that are similar to ADCS-PACC, using an unweighted z-score composite that includes a word-list recall task, Logical Memory (narrative recall), the five orientation items from the MMSE, and an associative-memory task. We have also focused on theoretically based composites that are specific to domains of episodic memory, executive function, attention, and language. Our biostatistics core has developed a method for optimizing cognitive composites using a stepwise procedure that first selects a single cognitive measure that performs best alone then uses an exhaustive search to find additional measures that add increasing power. This procedure then assigns weights to each measure included in the composite, thus optimizing the weights and increasing statistical power. In the near future, we will nominate our candidate endpoints for consideration. Careful and detailed efforts such as that by Donohue et al. will undoubtedly inform our ultimate decision and we congratulate their team on tackling a complex problem in what is essentially uncharted territory.


    . Toward defining the preclinical stages of Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement. 2011 May;7(3):280-92. Epub 2011 Apr 21 PubMed.

    View all comments by Jason Hassenstab

Make a Comment

To make a comment you must login or register.


Webinar Citations

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

Therapeutics Citations

  1. Solanezumab

News Citations

  1. Solanezumab Selected for Alzheimer’s A4 Prevention Trial
  2. Are New Cognitive Tests Ready For Preclinical Trials?
  3. Are Subtle Memory Concerns a Sign of Future Dementia?

External Citations

  1. Anti-Amyloid Treatment in Asymptomatic Alzheimer’s (A4) study

Further Reading

Primary Papers

  1. . The preclinical Alzheimer cognitive composite: measuring amyloid-related decline. JAMA Neurol. 2014 Aug;71(8):961-70. PubMed.
  2. . Secondary prevention trials in Alzheimer disease: the challenge of identifying a meaningful end point. JAMA Neurol. 2014 Aug;71(8):947-9. PubMed.