Clinically meaningful changes in cognitive and functional outcomes in a population‐based study of cognitive aging

Abstract

INTRODUCTION

Research is limited regarding meaningful change thresholds for individual patients on clinical outcome assessments (COAs) frequently used in clinical trials for Alzheimer's disease and related dementias (ADRD), particularly in population‐based studies early in the disease course.

METHODS

There were 646 study participants in the population‐based Mayo Clinic Study of Aging (MCSA), 54–99 years old (47% females), who developed the clinical syndromes of mild cognitive impairment (MCI) with complete data to establish clinically meaningful within‐patient change thresholds in cognitive and functional COAs.

RESULTS

Using the diagnosis of incident MCI as the anchor, mean (95% confidence interval [CI]) annualized estimates of change were: Clinical Dementia Rating (CDR) scale Sum of Boxes (SB) 0.49 (0.43, 0.55), Mini‐Mental State Examination (MMSE) −1.01 (−1.12, −0.91), and Functional Activities Questionnaire (FAQ) score 1.04 (0.82, 1.26).

DISCUSSION

This study provides within‐patient estimates of clinically meaningful change early in disease progression in a community‐based sample.

Highlights

• We studied incident mild cognitive impairment (MCI) participants from a population‐based study.

• Within‐patient change thresholds were estimated for clinical outcome assessments (COAs) used in clinical trials for Alzheimer's disease and related dementia (ADRD).

• These estimates may be used to plan and evaluate clinical trials involving disease‐modifying therapies (DMTs).

1. INTRODUCTION

The number of individuals living with dementia is projected to triple by 2050, posing significant challenges for affected patients, their families, healthcare systems, and society at large. ¹ Alzheimer's disease (AD) is the most prevalent form of dementia. ² Recent advances in treatment have focused on the development of anti‐beta amyloid monoclonal antibodies (mAbs)—the first disease‐modifying therapies (DMTs) for AD—and have demonstrated the potential to decrease amyloid‐beta (Aβ) plaques and slow disease progression in patients diagnosed with early‐stage biomarker‐confirmed symptomatic AD. ³

To assess the efficacy of these and prospective treatments, clinical trials are often dependent on clinical outcome assessments (COAs) of both cognitive and functional domains. ⁴ Often, the application of thresholds is at the group level ⁵ rather than focusing on the potential within‐patient benefit. ⁶ Studies have begun proposing methods to assess within‐person change, such as reporting proportions of patients who observe meaningful deterioration in COAs. ⁷ Regulatory authorities have preferred anchor‐based approaches because they rely on a clinician's, patient's, partner's, or care provider's opinion on meaningful decline. Anchor‐based methods link COAs to an external criterion considered to be clinically meaningful, which serves as the “anchor.” Distribution‐based approaches have been proposed as a statistical approach aimed at describing the spread of data and, in combination, provide ranges of meaningful decline. ⁸ , ⁹

Estimates of meaningful change have recently been published, ⁷ , ¹⁰ , ¹¹ , ¹² , ¹³ although the methods and terminology have left readers unclear whether individual or group‐level thresholds are reported or needed, and the interpretation ⁸ , ¹⁴ of these COA thresholds is left for debate. ⁶ Most thresholds have focused on the symptomatic stages of disease, ⁷ , ¹⁰ and far fewer have focused on early stages, specifically, thresholds for the onset of mild cognitive impairment (MCI).

Lessons may be learned from studies that have investigated the natural history of cognitive symptoms in patients with MCI. In a study comparing rates of decline in AD, MCI, and normal cognition, Wilson et al. ¹⁵ showed that decline rates were steeper in patients with AD compared to MCI and MCI compared to normal cognition with abnormal amyloid. Cloutier et al. ¹⁶ showed that episodic memory may exhibit a plateau period of stability after an initial decline before accelerating again with advanced disease.

RESEARCH IN CONTEXT

1. Systematic review: Traditional literature review suggests that meaningful change thresholds for individual patients on clinical outcome assessments (COAs) commonly used in clinical trials for Alzheimer's disease and related dementias (ADRD) are not well defined or agreed upon. In particular, population‐based estimates for early‐stage disease are lacking.

2. Interpretation: This study provides population‐based estimates of within‐patient change thresholds on key COAs that coincide with the clinically meaningful transition to mild cognitive impairment (MCI). These estimates may be useful in planning or assessing results of current and future clinical trials involving disease‐modifying therapies (DMTs) for ADRD.

3. Future directions: The findings provide potentially useful meaningful change thresholds that can be applied in a variety of settings. Extending these findings to other COAs and to a wider range of disease stages will enhance the literature regarding population‐based meaningful change thresholds for ADRD research. Applying and assessing these thresholds in different populations will be critical to evaluate their sustaining usefulness.

To keep the terminology consistent with previous work, although Mayo Clinic Study of Aging (MCSA) participants are not clinical “patients,” we will use the “within‐patient” term to denote “within‐participant.” The present study was conducted to identify meaningful within‐patient change thresholds for COAs often considered for the clinical transition from cognitively unimpaired (CU) to MCI. For anchor analyses, the onset of incident MCI—that is, a clinically meaningful outcome, was used to identify thresholds among a cohort of community‐dwelling CU participants of the MCSA. Distribution‐based methods were used as supportive information, consistent with recommendations. ⁸ , ⁹

2. METHODS

2.1. Study population

The MCSA is a population‐based cohort study that examines the epidemiology, risk factors, and biomarkers of MCI and dementia in older adults living in Olmsted County. ¹⁷ Briefly, participants are community‐dwelling older adults, randomly selected using the Rochester Epidemiology Project (REP) ¹⁸ medical records linkage system resources. When the study was established in 2004, 70–89‐year‐olds were invited to participate. The study population was later expanded in 2012 to include individuals ages 50–69. This study included CU MCSA participants with complete data, including at least one or more follow‐up visits to establish clinically meaningful within‐patient change thresholds in cognitive and functional COAs (N = 646 developed incident MCI and N = 3,284 remained CU).

2.2. Clinical evaluation

At each MCSA study visit, a study coordinator conducted a risk factor assessment; participants were asked questions regarding their personal and family medical history, medications, and neuropsychiatric symptoms. The Clinical Dementia Rating (CDR) scale ¹⁹ and the Functional Activities Questionnaire (FAQ) ²⁰ were administered to an informant who knows the participant well. A physician discussed the participant's medical history and conducted a neurological examination, including the Short Test of Mental Status (STMS). In addition, a psychometrist administered nine neuropsychological tests to the participants to assess cognitive performance in four domains, that is, memory (Rey Auditory Verbal Learning Test [AVLT] Delayed Recall, ²¹ Wechsler Memory Scale‐Revised [WMS‐R] Logical Memory II, and Visual Reproduction II), language (Boston Naming Test, ²² Category Fluency ²³ ), attention/executive (Trail Making Test B, ²³ , ²⁴ Wechsler Adult Intelligence Scale Revised [WAIS‐R] Digit Symbol ²⁵ ), and visuospatial (WAIS‐R Picture Completion & Block Design ²⁵ ). Participants were noted as test naïve at the first visit for which they completed neuropsychological testing.

A consensus panel including the study coordinator, physician, and neuropsychologist reviewed the visit data and determined the participant's cognitive diagnosis (MCI, ²⁶ dementia, ²⁷ , ²⁸ , ²⁹ or CU), blind to any prior MCSA visit diagnosis. Apolipoprotein E (APOE) ε4 status was determined from a blood draw at the MCSA baseline assessment. Participants were determined to be APOE ε4 carriers or non‐carriers if they had at least one ε4 allele or none, respectively.

2.3. COAs

The CDR scale assesses impairment caused by cognitive decline, including six domains (memory, orientation, judgment and problem solving, community affairs, home and hobbies, and personal care). ¹⁹ Ratings in each domain range from 0 (no impairment) to 3 (severe impairment), and the CDR Sum of Boxes (SB) is the sum of the scores for each domain, with a score range from 0 to 18. ¹⁹ , ³⁰

The FAQ ²⁰ is used to assess instrumental activities of daily living (IADLs), which require more complex skills (e.g., writing checks, preparing a balanced meal, keeping track of current events, etc.). The ability to perform each of 10 items is rated by the informant (0 = normal to 3 = dependent; 8 = “Not applicable (e.g., never did)” was not scored); the total score ranges from 0 to 30, with higher scores indicating greater impairment. The FAQ questionnaire was considered complete if 70% of the items were answered.

The Mini‐Mental State Examination (MMSE) ³¹ , ³² is a widely used screening tool, which assesses a person's cognitive state. The MMSE total score ranges from 0 to 30, with lower scores indicating greater impairment. In the MCSA, participants undergo the STMS test, which has been shown to be better at distinguishing between MCI and normal cognition. ³³ The MMSE score can be derived from the STMS score. ³³ , ³⁴

Raw test scores for the nine neuropsychological tests were transformed into age‐adjusted scores, averaged, and scaled to allow comparisons across domains. ¹⁷ A z‐score for each of the four cognitive domains and a global z‐score were then calculated by averaging and scaling the four‐domain z‐scores. Population‐based weights were applied using the 2013 Olmsted County 5‐year age and sex distribution. Scores at the individual, domain, and global levels can be interpreted as the number of standard deviations (SDs) above or below the population mean.

2.4. Statistical analysis

Descriptive statistics (mean [SD], count, and percentage) were used to summarize participants’ characteristics. An anchor‐based approach served as the primary method for estimating thresholds of meaningful within‐patient change on the target measures, complemented by supportive distribution‐based analyses. Analyses utilized data collected at all MCSA study visits up to and including the first visit where MCI was diagnosed. Participants who progressed from CU to dementia without an intermediary MCI visit (n = 68) were assumed to have passed through the MCI state and were included as incident MCI. For non‐progressors, the last two study visits were used to determine a change in outcome measures. Since MCSA visits occur approximately every 15 months, all estimates of change from prior visits were standardized to 1 year, “annualized” hereafter.

Using the conversion to diagnosed incident MCI as our anchor, analyses aimed to identify thresholds of meaningful score change, particularly for the CDR‐SB, ¹⁹ the MMSE derived from the STMS, ³³ , ³⁴ the FAQ ²⁰ score, and global and cognitive domain z‐scores (memory, attention, visuospatial, and language). Anchor‐based methods are the preference for regulatory authorities ⁷ , ³⁵ and are the focus of threshold identification in this work. Supportive distribution‐based estimates included calculations for one‐half SD at the prior visit, effect size (ES)—defined as the mean difference in score divided by the SD of the prior visit score and the standardized response mean (SRM)—defined as the mean difference in score divided by the SD of change from the previous visit score. ¹³ , ³⁶ General guidelines for interpreting ES and SRM could be applied (a magnitude of 0.2 for small, 0.5 for medium, and 0.8 for large). ¹⁰ , ¹¹ The one‐half SD method provides a lower bound perceived to be significant enough to warrant change in disease management. ¹⁰ , ¹¹ These distributional methods were done to support the anchor‐based approach outlined in prior published work to guide and to apply triangulation methods to derive the minimal clinically important difference (MCID). ⁹ As a sensitivity analysis, thresholds of meaningful change before incident MCI were also calculated in the subset of participants with positive amyloid‐positron emission tomography (PET). Amyloid PET imaging was performed with 11C‐Pittsburgh Compound B (PIB), and amyloid positivity (A+) was defined for the current study as a Centiloid value ≥25. ³⁷ , ³⁸ In supplemental analysis, we examined whether each of the participants reached the estimated MCIDs, as previous reports did, ⁷ having in mind that MCI is a clinical syndrome, and cognitive and functional trajectories can be very heterogeneous. ³⁹ Lastly, all observations for the COAs occurring in the past 3 years of follow‐up for this study were plotted to describe the shape of decline observed in those who progressed and, for completeness, those who have yet to progress to MCI. Analyses were considered statistically significant at a p‐value <0.05 and were performed using the SAS statistical software version 9.4 (SAS Institute, Cary, North Carolina).

3. RESULTS

In Table 1, participants who progressed to MCI (progressors, hereafter), data are summarized at the time of the incident MCI assessment visit. For non‐progressors, data are summarized at the time of their last study visit. The overall sample had a mean (SD) age of 77.3 (9.8) years; 49.5% were female, and 27.0% carried at least one APOE ε4 allele. The mean (SD) time between prior and final study visits was 17.3 (5.2) months. Compared to those who have not progressed to MCI, progressors were older [82.1 (7.2) vs. 76.4 (10.0) years], had lower average years of education [13.8 (2.8) vs. 14.9 (2.6)], and were more likely to have at least 1 APOE ε4 allele (36.1% vs. 25.2%). The mean (SD) in progressors versus non‐progressors for CDR‐SB was [0.9 (1.2) vs. 0.1 (0.3)], MMSE was [25.7 (1.8) vs. 28.5 (1.2)], FAQ total score was [2.2 (4.1) vs. 0.2 (1.0)], and global z‐score was [−1.7 (1.0) vs. 0.2 (1.0)].

TABLE 1.

Participants’ characteristics, by progression to MCI.

		Progression to incident MCI
Characteristics	Total (N = 3930)	Progressor a (N = 646)	Non‐progressor b (N = 3284)	p‐value
Age, in years	77.3 (9.8)	82.1 (7.2)	76.4 (10.0)	<0.001
Female sex, n (%)	1945 (49.5%)	306 (47.4%)	1639 (49.9%)	0.24
Years of education	14.7 (2.6)	13.8 (2.8)	14.9 (2.6)	<0.001
APOE e4 carrier, n (%)	1018 (27.0%)	231 (36.1%)	787 (25.2%)	<0.001
Abnormal amyloid	259 (27.7%)	133 (76.4%)	126 (16.6%)	<0.001
Time between visits, months	17.3 (5.2)	17.2 (4.8)	17.3 (5.3)	0.51
Test naïve at prior visit	746 (19.0%)	155 (24.0%)	591 (18.0%)	<0.001
CDR‐SB	0.2 (0.6)	0.9 (1.2)	0.1 (0.3)	<0.001
MMSE	28.0 (1.7)	25.7 (1.8)	28.5 (1.2)	<0.001
FAQ score	0.5 (2.0)	2.2 (4.1)	0.2 (1.0)	<0.001
Global z‐score	−0.1 (1.2)	−1.7 (1.0)	0.2 (1.0)	<0.001
Memory z‐score	0.0 (1.2)	−1.5 (0.9)	0.3 (1.0)	<0.001
Attention z‐score	−0.4 (1.3)	−1.7 (1.4)	−0.1 (1.1)	<0.001
Visuospatial z‐score	0.0 (1.1)	−0.9 (1.0)	0.2 (1.0)	<0.001
Language z‐score	−0.2 (1.2)	−1.3 (1.2)	0.1 (1.0)	<0.001

Note: Mean (SD) unless otherwise specified. p‐values calculated using t‐tests for continuous variables and chi‐squared tests for categorical variables. Data were summarized at the visit of observed conversion for progressors and the last study visit for those who have not progressed.

Abbreviations: APOE, apolipoprotein E; CDR‐SB, Clinical Dementia Rating‐Sum of Boxes; FAQ, Functional Activities Questionnaire; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination (derived from the Short Test of Mental Status); SD, standard deviation.

^a Data missing in progressors: Abnormal amyloid (472), CSR‐SB (5), MMSE (3), FAQ (19).

^b Data missing in non‐progressors: Abnormal amyloid (2523), CDR‐SB (27), MMSE (8), FAQ (116).

The study used the incident MCI visit of progressors (N = 646) as an anchor. For comparison, data of non‐progressors (N = 3,284) were analyzed at their last study visit, measuring change from their prior visit. Observed annualized mean changes, ES, and SRM for progressors versus non‐progressors were 0.49 (ES: 1.14; SRM: 0.60) versus 0.012 (0.048, 0.045) in CDR‐SB, −1.01 (−0.67; 0.74) versus −0.067 (−0.061; −0.080) in MMSE, 1.04 (0.59; 0.38) versus 0.032 (0.039; 0.041) in FAQ score, and −0.32 (−0.38; −0.74) versus −0.021 (−0.022; −0.069) in overall global cognitive z‐score (Table 2). The CDR‐SB, MMSE, and FAQ scores all had ½ SD at index values slightly less than the anchor‐based estimates. For global and domain‐specific z‐scores, these values are all slightly larger than the anchor‐based estimates (Table 2).

TABLE 2.

Anchor and distributional estimates of annualized CMC.

	Progression to incident MCI
Measure	Progressor (N = 646) a	Non‐progressor (N = 3284) b
CDR‐SB
Mean change (95% CI)	0.49 (0.43, 0.55)	0.012 (0.004, 0.020)
ES	1.14	0.048
SRM	0.60	0.045
½ $S{D}_{(index)}c$	0.22	0.10
MMSE
Mean change (95% CI)	−1.01 (−1.12, −0.91)	−0.067 (−0.097, −0.036)
ES	−0.67	−0.061
SRM	−0.74	−0.080
½ $S{D}_{(index)}c$	0.75	0.57
FAQ score
Mean change (95% CI)	1.04 (0.82, 1.26)	0.032 (0.006, 0.058)
ES	0.59	0.039
SRM	0.38	0.041
½ $S{D}_{(index)}c$	0.88	0.39
Global z‐score
Mean change (95% CI)	−0.32 (−0.35, −0.29)	−0.021 (−0.031, −0.011)
ES	−0.38	−0.022
SRM	−0.74	−0.069
½ $S{D}_{(index)}c$	0.42	0.49
Memory z‐score
Mean change (95% CI)	−0.30 (−0.33, −0.26)	0.012 (−0.002, 0.026)
ES	−0.36	0.012
SRM	−0.64	0.029
½ $S{D}_{(index)}c$	0.42	0.50
Attention z‐score
Mean change (95% CI)	−0.37 (−0.43, −0.32)	−0.059 (−0.073, −0.046)
ES	−0.34	−0.059
SRM	−0.54	−0.151
½ $S{D}_{(index)}c$	0.55	0.50
Visuospatial z‐score
Mean change (95% CI)	−0.12 (−0.17, −0.08)	−0.003 (−0.017, 0.011)
ES	−0.12	−0.003
SRM	−0.22	−0.007
½ $S{D}_{(index)}c$	0.50	0.49
Language z‐score
Mean change (95% CI)	−0.23 (−0.27, −0.19)	−0.017 (−0.031, −0.003)
ES	−0.22	−0.018
SRM	−0.45	−0.041
½ $S{D}_{(index)}c$	0.52	0.48

Note: Data summarized at the visit of observed conversion for progressors and the last study visit for those who have not progressed.

Abbreviations: CDR‐SB, Clinical Dementia Rating‐sum of boxes; CI, confidence interval; CMC, Clinically Meaningful Change; ES, effect size; FAQ, Functional Activities Questionnaire; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination (derived from the Short Test of Mental Status); SD, standard deviation; SRM, standardized response mean.

^a Data missing in progressors: CSR‐SB (11), MMSE (5), FAQ (32).

^b Data missing in non‐progressors: MMSE (9), FAQ (198).

^c Index refers to the score from the prior visit.

One hundred seventy‐four of the 646 progressors had an available amyloid‐PET prior to progression to MCI (N = 133 A+ and N = 41 A−). Thus, a subset analysis of participants with positive brain amyloid‐PET who progressed to MCI (N = 133 of 646 progressors) was also performed following the same anchor‐based approach. This subset observed annualized mean changes in CDR‐SB of 0.41 (ES: 1.08; SRM: 0.62), in MMSE of −1.09 (−0.75; −0.92), and in FAQ score of 0.81 (0.68; 0.41). Overall cognitive z‐score changed by an annualized mean of −0.39 (−0.52; −0.95). Annualized mean changes in the four domains were −0.39 (−0.48; −0.78) for memory, −0.42 (−0.39; −0.57) for attention, −0.079 (−0.085; −0.16) for visuospatial, and −0.35 (−0.38; −0.80) for language (not shown in tables).

Table 3 summarizes the frequency and percentage of participants who reached the identified MCIDs for each COA, using anchor‐based annualized change thresholds, as estimated from the previously defined group of progressors (those who developed incident MCI) for each measure applied to the overall study population's last two visits. Applying the proposed change thresholds to the CDR‐SB (6.6%) and FAQ scores (7.4%) identified the smallest percentages of participants who reached the MCIDs. The percentage of participants who reached the identified MCIDs were 14.6% for MMSE, 24.3% for memory z‐score, 19.2% for attention z‐score, 17.8% for global cognitive z‐score, 39.0% for visuospatial z‐score, and 30.8% for language z‐score. Comparatively, the overall percentage of those who progressed to the clinical syndrome of MCI was 16.4%. We have included the frequency (%) of identified participants who reached the estimated MCIDs within each COA who also developed incident MCI in a separate column.

TABLE 3.

Frequency and percent of participants who reached the identified MCID thresholds.

Measure	Total participants N	Estimated annualized a threshold	Threshold met b n (%)	With incident MCI c n (%)
CDR‐SB	3866	0.49‐point	256 (6.6%)	202 (78.9%)
MMSE	3930	−1.01‐point	574 (14.6%)	286 (49.8%)
FAQ score	3700	1.04‐point	275 (7.4%)	171 (62.2%)
Global z‐score	3930	−0.32‐point	698 (17.8%)	285 (40.8%)
Memory z‐score	3930	−0.30‐point	956 (24.3%)	300 (31.4%)
Attention z‐score	3930	−0.37‐point	756 (19.2%)	272 (36.0%)
Visuospatial z‐score	3930	−0.12‐point	1531 (39.0%)	321 (21.0%)
Language z‐score	3930	−0.23‐point	1209 (30.8%)	313 (25.9%)

Abbreviations: CDR‐SB, Clinical Dementia Rating‐Sum of Boxes; COA, clinical outcome assessment; FAQ, Functional Activities Questionnaire; MCI, mild cognitive impairment; MCID, minimal clinically important difference; MMSE, Mini‐Mental State Examination (derived from the Short Test of Mental Status).

^a Anchor‐based annualized change thresholds, as estimated from those who developed incident MCI.

^b Frequency and percentage of participants who reached the MCID thresholds for each specific COAs, using anchor‐based annualized change thresholds, as estimated from those who developed incident MCI.

^c Frequency and percentage of those who reached the MCID threshold for each specific COA who progressed to incident MCI.

To visualize the natural history of the COAs (CDR‐SB, MMSE, FAQ total score, global z‐score, memory z‐score, attention z‐score, visuospatial z‐score, and language z‐score), we looked at the scores up to 3 years before incident MCI diagnosis for progressors and the last 3 years of follow‐up for non‐progressors. Figure 1 shows the natural history of these scores by progression using cubic splines to allow non‐linear trends to emerge. Visually assessed, we observed flat, linear trends for non‐progressors over the last 3 years of follow‐up for each outcome. Progressors showed lower cognitive z‐scores and steeper rates of decline over the 3 years prior to MCI, with the CDR‐SB, MMSE, and FAQ exhibiting acceleration prior to progression.

FIGURE 1.

Performance on COAs as a function of time to incident MCI (for progressors) or in the past 36 months of assessments (for non‐progressors). Cubic splines were used to allow non‐linear trends to emerge. COA, clinical outcome assessment; MCI, mild cognitive impairment.

4. DISCUSSION

The study aimed to quantify clinically meaningful within‐patient change thresholds for the CDR‐SB, MMSE, FAQ score, and global and domain‐specific cognitive z‐scores in CU individuals at baseline who progressed to MCI. An anchor‐based approach was used, utilizing the diagnosis of MCI as the anchor, a clinically meaningful event. Distribution‐based methods were used to support the anchor‐based thresholds. Using both anchor and distributional methods in combination reveals a range of threshold values to consider. ⁸ , ⁹ Annualized change thresholds estimated from the anchor‐based analyses in those who developed incident MCI were estimated to be 0.49 for CDR‐SB, −1.01 for MMSE, and 1.04 for the FAQ score. Distributional methods were supportive of these estimates. In particular, the ES for CDR‐SB was very large (1.14), whereas the ES for MMSE and FAQ were in the medium‐to‐large range (0.67 and 0.59, respectively). This study adds to the current and growing literature by providing within‐patient estimates of clinically meaningful change early in disease progression.

As the work on DMT for ADRD evolves, a central question with any new treatment will always be whether an improvement achieved in COAs represents a beneficial change for patients and/or caregivers, that is, a clinically meaningful change in cognition or daily function. However, there is no established method to assess clinical meaningfulness. The United States Food and Drug Administration (FDA) suggests assessing the individual‐level benefit to answer this question more suitably. ⁶ , ⁷ , ³⁵ As more DMTs for ADRD are becoming available, findings on group‐level or individual‐level differences must be presented clearly, allowing for easy comparisons between studies and interpretation of findings into patient populations. ⁴⁰ , ⁴¹ Between‐group differences and statistical significance depend on within‐individual differences but also other details, such as trial study design, like the per‐arm sample size and group‐level estimated differences, and cannot be directly translated into individual‐level treatment effects. ⁶ Thus, including individual‐level analyses will help interpret findings, inform all stakeholders, including patients and their families, and add clear information to treatment decision‐making discussions.

Few studies have presented change thresholds for the CDR‐SB, MMSE, and FAQ score in a cohort of initially CU individuals. Our anchor‐based estimates of annualized within‐patient change thresholds for clinically meaningful change of 0.49 in CDR‐SB, −1.01 in MMSE, and 1.04 in FAQ scores in a cohort of CU participants are quite similar to Andrews, et al. ¹⁰ Their estimates (0.54 in CDR‐SB, −0.89 in MMSE, and 1.42 in FAQ score) were found in a cohort of cognitively normal participants included in the Uniform Data Set (UDS) of the National Alzheimer's Coordinating Center (NACC). Using similar methods (anchor and distribution‐based), Lansdall, et al. ⁷ presented thresholds for both “minimal” (1 for CDR‐SB and 2–3 for MMSE) and “moderate” (2.5 for CDR‐SB and 6–7 for MMSE) deterioration in a cohort with amnestic MCI. As participants are evaluated in the MCSA not based on symptomatology but on scheduled follow‐up visits approximately every 15 months, we expect that MCI will be detected earlier in the MCSA than in patients visiting a clinic. Therefore, our estimated thresholds might be smaller than those detected in clinical patients. However, as ADRD clinical trials move earlier in the disease process, such thresholds in population‐based studies are very insightful. Others who have presented within‐patient change thresholds have done so in cohorts with more advanced disease. ¹¹ , ¹² , ¹³ , ⁴¹

Table 1 presents the characteristics of progressors and non‐progressors, with the reminder that this is an observational, natural history study. Participants who progressed or did not to MCI present many differences in known risk factors for cognitive impairment. Of note, progressors were on average older, with fewer years of education, and more likely to be APOE ε4 carriers, known as important risk factors for cognitive impairment. ⁴² , ⁴³ Progressors were more likely to be test‐naïve in the visit prior to the incident MCI diagnosis visit. Prior research ⁴⁴ suggests that CU individuals who progress to incident cognitive impairment present practice effects only in memory (which declines significantly thereafter), while stable CU individuals present practice effects in all cognitive domains (sustained in memory and visuospatial reasoning). Practice effects can be important confounders in studies of longitudinal changes in cognitive performance, including clinical trials, with suggested approaches that could diminish these effects. ⁴⁵

In sensitivity analysis, we observed that MCIDs were greater for those with an elevated amyloid burden in the brain. Previous studies have observed that CU individuals with elevated amyloid burden show faster cognitive decline and higher progression risk to MCI or dementia. ⁴⁶ , ⁴⁷ , ⁴⁸

Complementary analysis for within‐individual change could be aided by reporting the proportion of participants who had the within‐individual clinically meaningful change thresholds for progression from baseline. ⁶ We have applied such analysis to help design or apply results from DMTs in a community‐based setting. A similar approach was taken in the Phase 3 donanemab trial, ⁴⁹ examining which participants achieved prespecified meaningful within‐patient change thresholds of clinically important progression during the trial period in both the treatment and placebo groups. In the present study, when comparing the percentage of participants who reached MCIDs, the CDR‐SB and FAQ total score stand out as under‐representing the actual percentage who progressed to the clinical syndrome of MCI (16.4%). This could be attributed to the fact that these scales have functional components not expected to be exhibited in the early stages of disease. Conversely, the visuospatial and language domain thresholds represent values that identify far more participants who reached MCIDs compared to those who progressed to incident MCI. Incident MCI was found in just 21.0% and 25.9% of those having met the proposed thresholds for visuospatial and language z‐scores, respectively (Table 3). The ranges of percentages reaching the MCIDs in COAs illustrate the heterogeneity of the MCI clinical syndrome.

The study displayed the data as a function of the time of diagnosis rather than being anchored at study entry, presenting the natural history of COAs before conversion to MCI (progressors) and the comparative group (non‐progressors) in a community‐based study. In each COA, progressors show steeper declines compared to a non‐progressor group that is cognitively stable. In particular, the CDR‐SB, MMSE, and FAQ total score show an accelerated slope leading up to their diagnosis of MCI. The presentation of this data is complementary to both the primary goal of identifying clinically meaningful changes for the current study and the growing literature of natural history studies. ¹⁶ , ⁵⁰

The current study has several strengths. The MCSA is a population‐based study of community‐dwelling older adults, with comprehensive serial cognitive evaluations and state‐of‐the‐art neuroimaging. However, we need to consider the study's limitations. We acknowledge circularity between changes in cognitive scores and cognitive impairment diagnosis; however, our consensus diagnosis conference does not utilize prior scores when assigning current diagnosis. The study did not include tau‐PET, as it was implemented later (2015), with only 22 participants having a positive tau‐PET in the sensitivity analysis. In addition, 98.8% of all participants were White individuals, and 99.5% of participants were not Hispanic or Latino individuals; additional studies in more diverse populations are warranted. As we only used a binary anchor, only an overall threshold may be established, limiting the applicability to less and more severe disease.

Research is limited regarding meaningful change thresholds for individual patients on COAs frequently used in clinical trials for ADRD, particularly in population‐based studies early in the disease course. This study provides within‐patient estimates of clinically meaningful change early in disease progression in a community‐based sample.

AUTHOR CONTRIBUTIONS

Jeremiah A. Aakre, Ronald C. Petersen, and Maria Vassilaki conceived the study aim and contributed to the study design. Jeremiah A. Aakre and Maria Vassilaki drafted the manuscript. Jeremiah A. Aakre and Anna M. Castillo performed the statistical analysis. All authors contributed to data acquisition and interpretation, critically revised the manuscript for important intellectual content, and approved the final draft of this report.

CONFLICT OF INTEREST STATEMENT

Jeremiah A. Aakre has no disclosures. Anna M. Castillo has no disclosures. Jonathan Graff‐Radford receives support from the NIH and serves on the DSMB for StrokeNET and is an investigator in clinical trials sponsored by Esai, Cognition therapeutics, and the Alzheimer's Treatment and Research Institute at USC. He is an associate editor for JAMA neurology. Prashanthi Vemuri receives research funding from NIH. Mary M. Machulda receives research funding from NIH. Clifford R. Jack Jr. has no financial conflicts to disclose; he receives research support from NIH and the Alexander Family Alzheimer's Disease Research Professorship of the Mayo Clinic. David S. Knopman serves on a Data Safety Monitoring Board for the Dominantly Inherited Alzheimer Network Treatment Unit study sponsored by Washington University St Louis, and for the SMART‐HS clinical trial (Univ of Kentucky). He was an investigator in Alzheimer clinical trials sponsored by Biogen, Lilly Pharmaceuticals and the University of Southern California, both of which have ended, and is currently an investigator in a trial in frontotemporal degeneration with Alector. He has served as a consultant for Roche, AriBio, Linus Health, Biovie and Alzeca Biosciences but receives no personal compensation. He receives funding from the NIH. Ronald C. Petersen serves as a consultant for Roche, Inc., Genentech, Inc., Eli Lilly and Co., Novo Nordisk, Eisai, Inc. (no funding), receives royalties from Oxford University Press and UpToDate, contributes to Medscape education and receives NIH funding. Maria Vassilaki consulted for F. Hoffmann‐La Roche Ltd, unrelated to the current study; she currently receives research funding from NIH and has equity ownership in Johnson and Johnson, Merck and Amgen. Author disclosures are available in the supporting information.

DATA ACCESS STATEMENT

The Mayo Clinic Study of Aging makes data available to qualified researchers upon reasonable request.

CONSENT STATEMENT

Study approval was obtained from the Institutional Review Boards of the Mayo Clinic and Olmsted Medical Center in Rochester, Minnesota. Participants provided written informed consent before participation. In the case of participants with cognitive impairment sufficient to interfere with capacity, assent was obtained from a legally authorized representative.

Supporting information

Supporting Information

Aakre JA, Castillo AM, Graff‐Radford J, et al. Clinically meaningful changes in cognitive and functional outcomes in a population‐based study of cognitive aging. Alzheimer's Dement. 2025;11:e70160. 10.1002/trc2.70160