Response to Schneider and Goldberg “Composite cognitive and functional measures for early stage Alzheimer's disease trials”

The work of Schneider and Goldberg focuses on the concept of “Composite” scores as measures for clinical trial outcomes in Alzheimer's disease (AD), with a special emphasis on pre‐symptomatic individuals, who have a biomarker of risk or a marker of pathology. ¹ These authors suggest that regulatory authorities are moving toward an imprimatur of cognitive composite as a viable outcome in these relatively asymptomatic individuals. They also suggest that true composites have been proposed for mild cognitive impairment (MCI) but not for early or mild AD. This interpretation of regulatory guidance is truly a “straw man” not embedded in the overwhelming opus of regulatory guidance. Specifically, regulatory agencies look to the work of content experts to define the most important outcomes. In this sense, Schneider and Goldberg have created a false or at least partial narrative around the limited U.S. Food and Drug Administration (FDA) and European Medicines Agency (EMA) guidance. ² , ³

Although the authors’ plea to better describe and improve psychometric properties of composites is of course wise, the absence of such information is not proof that existing cognitive composites are not useful. The literature review and supplemental materials focused only on the use of composites in clinical trials. This is unfortunate because most important is capturing meaningful aspects of the disease or “at risk” condition and not just the shortest path to approval. The review excluded theory‐driven composites that have been developed. For example, the Alzheimer Disease Centers (ADC) through the clinical task force have created a consensus on important cognitive domains, which include memory, language, attention, and executive function and although specific tests assessing those domains have changed, the composite should have stability. ⁴ Indeed there is collinearity among many of these cognitive domains, but the baseline performance in any domain as well as in the composite is a more potent variable in determining the trajectory than even disease condition. In a report comparing overall cognitive composite and domain‐specific subscores and their progression over time in non‐demented older adults, we observed that composite trajectory (Figure 1) may not perfectly reflect the trajectory of each domain (Figure 2), but it portends the direction of all. ⁵ Should a shift in only one domain really be defined as a meaningful outcome?

FIGURE 1.

Predicted cognitive composite score, by diabetes status and diagnosis group. Reprinted with permission from Ref. 5

FIGURE 2.

Predicted domain specific subscores, by diabetes status and diagnosis group. Reprinted with permission from Ref. 5

The discussion about “composites” should be driven by what is most important and most clinically meaningful to capture. We respectfully submit that a global impression of disease or risk stage or severity can be accomplished with a composite that captures cognition, behavior, and function. As early as 2011, the National Institute of Health (NIH) and the Alzheimer's Association acknowledged that dementia symptoms included “…cognitive or behavioral (neuropsychiatric) symptoms.” ⁶ Since that time, the concept of mild behavioral impairment has been introduced and mild behavioral and psychiatric symptoms have been identified, even among prodromal AD. ⁷ , ⁸ Subtle but measurable changes in disposition, emotional response, and initiation reflect behaviors that have been noted in asymptomatic individuals can contribute to sensitive composites. Furthermore, there is strong evidence that they are related to the biological and genetic mechanisms that underlie disease. ⁹ Subtle changes in function can be measured by assessing completion of high level financial (eg, banking, tax preparation) or operational (eg, driving, use of technology) tasks, even among those with intact cognition. Measures of the rate of adapting to new technologies may also be sensitive measures that could be added to cognitive and behavioral scores to create more comprehensive composites.

In recent years, we have seen clinical trials with strong negative effects on the core symptoms of AD, even in mild ¹⁰ , ¹¹ and prodromal ¹² cases. Current approaches to outcomes in these early disease stages have debated on the meaningfulness of changes on individual components, arguing that there is recovery of mild cognitive deterioration when the drug is stopped but a small and persistent deficit has been measured with continued treatment. Seldom are psychiatric symptoms even measured in the prodromal stage, although some agents have been noted to cause sleep and other psychiatric disturbances. The use of a composite will allow for a more comprehensive assessment of treatment interventions. Given the multiplicity of regional and global positive as well as negative impact on brain with current treatments, it would seem prudent to have a composite that captured all clinically relevant aspects of disease. Whether reversible, or persistent with treatment, these symptoms are the source of burden on patients and families and the constellation should be given weight. The goal of a composite or any outcome should not be to guarantee drug approval with the fewest subjects; rather it should be to identify the most effective and safest impact on disease and risk reduction.