Roos Jutten

Figure 1. of converging evidence for the Cognitive-Functional Composite as a measure for clinical progression in MCI and mild dementia, as provided in this thesis.

was addressed in chapter 4, with chapter 4.1 revealing differences in change on the CFC measures amongst SCD, MCI and dementia, and chapter 4.2 showing that individual cognitive tests vary in their sensitivity to change at different AD clinical stages as outlined in the NIA-AA research framework. Table 1 shows how the NIA-AA clinical staging and clinical syndrome staging schemes could be aligned, and which individual CFC cognitive tests were identified as sensitive in chapter 4.2. Although it should be noted the study in chapter 4.2 differed regarding study-sample and follow-up timeframe, the findings regarding the sensitivity of the CFC cognitive subtests were in agreement with the results obtained in chapter 4.1. Both studies suggest that the cognitive component of the CFC in its current form is 1) limited sensitive to changes in SCD (Stage 2); 2) partly sensitive to change in MCI (Stage 3), and; 3) evidently sensitive to change in dementia (Stage 4). Moreover, the finding that the majority of CFC cognitive tests were identified as sensitive in a large group of AD biomarker positive individuals at clinical Stages 3 and 4 (chapter 4.2), supports the use of the CFC in our target population of individuals with MCI and mild dementia due to AD. However, these results also imply that the CFC could potentially be further refined to improve its sensitivity to change in Stage 3, as some of the individual CFC cognitive tests did not seem to capture decline after one year in this stage.

Table 1. Sensitive CFC measures identified in chapter 4.2.
Clinical status NIA-AA staging scheme Measures included in the CFC
SCD Stage 2 Category Fluency, Word Recall
MCI Stage 3 Category Fluency, Digit Span, Word Recognition, Word Recall
Dementia Stage 4 Category Fluency, Digit Span, Word Recognition, Word Recall, COWAT, Symbol Substitution

Combining cognitive tests with a functional measure: the bridge to clinical meaningfulness?
A recurring finding in this thesis was the added clinical value of the Amsterdam IADL Questionnaire as functional component of the CFC. For example, we found that associations between the CFC score and other measures of disease severity were mostly driven by the Amsterdam IADL score (chapter 3.3). Additionally, we showed that Amsterdam IADL Questionnaire is able to detect subtle impairment in individuals with SCD (chapter 2.1), and to capture decline over one year in MCI, whereas the cognitive component of the CFC failed to do so (chapter 4.1). These findings seem counterintuitive, as the assumed clinical trajectory of AD entails that cognitive impairment induces and thereby precedes functional impairment [7]. However, our findings do not argue against this conceptual understanding of cognitive impairment preceding functional change, but rather imply that existing paper-and-pencil cognitive tests may not provide the right tools to capture subtle cognitive decline. This is in line with previous studies that pointed towards the limited sensitivity of existing cognitive tests in early clinical stages of AD [8, 9]. A functional measure on the other hand, may be capable of capturing meaningful decline as reflected by increasing difficulties in complex activities of daily living. Previous studies on the Amsterdam IADL Questionnaire already demonstrated its added diagnostic and prognostic value to neuropsychological tests in dementia [5, 6]. The current thesis provides further evidence on the importance of the inclusion of a sensitive IADL measure to capture clinically cognitive decline in predementia stages of AD. More specifically, our results imply that the Amsterdam IADL Questionnaire could provide the missing link between cognitive test scores and clinical meaningfulness.

Worldwide, researchers and regulatory agencies have expressed the need for novel, sensitive tools that capture clinically meaningful decline in predementia stages of AD [10, 11]. This is of particular relevance in the context of clinical trials, since evidence of efficacy on a clinical meaningful measure is required to achieve regulatory approval of novel therapeutic interventions [12]. In an attempt to develop novel measures and methods that yield components of both functional and cognitive skills, several other endeavors have been undertaken. Examples include the Clinical Dementia Rating scale (CDR-SB) [13] and the recently designed Alzheimer’s Disease Composite Score (ADCOMS) [14]. However, the clinical meaningfulness of those measures is yet uncertain, especially for the ADCOMS procedure which has been largely statistically-driven [15]. Furthermore, we showed that both the CDR-SB and ADCOMS are prone to ceiling effects in MCI and mild dementia as compared to the CFC (chapter 3.3). We also showed that the CFC could provide a more refined and meaningful measure of clinical progression (chapter 4.1), and thereby offers advantages over the use of the CDR-SB as a measure of efficacy to effectively evaluate novel treatments targeting early symptomatic stages of AD.

Methodological considerations
There are several methodological considerations that should be considered when interpreting the findings of this thesis.

Cohorts. Different cohorts were used in the studies included in this thesis. Studies on the CFC were performed using data from the Catch-Cog cohort, which is an international, observational cohort enrolling individuals with SCD, MCI, AD dementia and DLB. The prospective character of this cohort enabled us to perform an independent validation of the CFC, which is a unique aspect of this study. However, limitations of this cohort include possible heterogeneity due to differences in recruitment strategies across study centers (clinically-based versus community-based), and the fact that not all participants had biomarkers available to confirm neurodegeneration. Furthermore, the relatively small sample-sizes of the SCD and DLB groups limited the interpretation of our findings in these groups. As such, our investigation of the CFC in SCD and DLB remained of rather explorative nature.

Other studies in this thesis were conducted using data from the Amsterdam Dementia Cohort (ADC), which is an observational memory-clinic cohort of the Alzheimer Center Amsterdam [16]. A major strength of this cohort is its extensive phenotyping. However, it should be noted that the Alzheimer Center Amsterdam is a tertiary memory-clinic specialized in young-onset dementia, resulting in a relatively young cohort that is less generalizable to older populations. In our last chapter, we combined data from the ADC with data from three North-American cohorts, including the community-based observational Harvard Aging Brain Study (HABS) cohort [17], the Alzheimer’s Disease Neuroimaging Initiative (ADNI) [18] and National Alzheimer’s’ Coordinating Center (NACC) [19] research cohorts. Combining these four well-defined cohorts provided a large and unique sample of AD biomarker positive individuals covering the entire clinical spectrum of AD. However, pooling data across these studies had some limitations due to differences across cohorts regarding 1) cognitive tests that had been assessed; 2) time-intervals between follow-up visits; and 3) assessment methods to determine amyloid positivity. Furthermore, a general limitation of these observational cohorts includes the loss to follow-up, which can be regarded as non-random in that more severely affected individuals are probably earlier lost to follow-up [20]. This could have induced a selection-bias that may have affected the internal validity of our results.

Study design. The longitudinal study design of the Catch-Cog study including assessments at baseline, 3, 6 and 12 months may have influenced our results regarding the CFC’s sensitivity to change over time. First, participants were only followed for one year, and it could be argued that this is rather short to observe evident decline in individuals with MCI or mild AD. However, both the A-IADL-Q and subtests of the CC were shown to be able to capture changes within the one-year timeframe [21, 22], so therefore we expected that one year would be sufficient to detect decline. Second, the time intervals between follow-up visits were relatively short, which may have led to practice effects on cognitive tests, particularly at 3 and 6-months follow-up visits. Practice effects reflect improvements in cognitive test performance that result from repeated exposure to the test [23]. They are a potential threat for longitudinal cognitive assessment, as they can underestimate true cognitive decline [24]. Previous studies on practice effects in the early stages of dementia are contradictory [25], but some of them have shown that practice effects on memory measures can be observed in people with MCI [26]. The results presented in this thesis are contradictory as well, as we found negligible practice effects in our test-retest study, while in our longitudinal study an improvement was observed in CFC cognitive test performance in SCD and MCI participants. Interestingly, this improvement was observed while alternate versions of word lists were used at each follow-up visit, implying that not the actual test material but rather the ‘familiarity of being tested’ may have induced practice effects [27]. More specifically, this could be attributed to reductions in anxiety, as levels of anxiety are often higher during first assessment and thereby negatively affect cognitive performance [28]. This further highlights that the concept of practice effects is rather complex, and that variables that impact these effects are not fully understood yet. As a result, it remains difficult to distinguish whether ‘no change’ in cognitive performance in our MCI cohort reflect absence of progression, or whether decline caused by progression was diminished due to factors such as practice effects.

Construct validation approach. A main challenge for the validation of the CFC was the absence of a gold standard for ‘clinical progression’. We aimed to obviate this with a construct validation approach, by including other clinical and biological measures as reference measures of disease severity. We included other cognitive and functional tests as reference measures of progression, however, a potential limitation of those tests was their expected limited sensitivity to change in our target population [29, 30]. Therefore, we also selected measures that would be less likely to suffer from range restrictions in scoring, such as informant-reports of cognitive decline, quality of life measures, and global cortical volume [31-33]. Altogether, the associations between the CFC and these ‘silver’ standards of clinical severity could provide converging evidence for the clinical relevance of the CFC.

Furthermore, the inclusion of the traditional tests also enabled us to perform a direct head-to-head comparison between the CFC and traditional tests, which is a unique aspect of our study. In fact, we did not only perform an independent validation of the CFC, but of the traditional measures as well. This led to further evidence for the quality limitations of those traditional measures when assessing clinically meaningful cognitive decline in MCI and early dementia.

Conclusion

By investigating how the measurement of clinical progression in early AD could be refined, this thesis provided several insights on the progression of clinical symptoms due to AD in general. First, our findings further support that cognitive decline due to AD starts years before the onset of dementia, and that decline in preclinical and prodromal stages of AD could be captured by certain, but not all, neuropsychological tests. Moreover, our results highlight that clinical endpoints to asses change should be tailored to a specific clinical stage, in order to effectively capture cognitive decline in that stage. Our findings further imply that longer time intervals and tests addressing specific cognitive domains are needed to capture change at earlier clinical stages, that AD-related cognitive decline exponentially increases over the years, and that broader cognitive functions become increasingly affected. Finally, this thesis demonstrates the added value of measuring IADL functioning when assessing clinical progression in early stages of AD, implying that the utility of an everyday functioning measure goes beyond just its use for diagnostic purposes in dementia. We conclude that combining measures of cognitive and function is recommended to capture clinically relevant cognitive decline due to AD, even in clinical stages that occur before the onset of dementia.

Clinical implications
Although it was our main focus to improve the measurement of progression in the context of research and clinical trials, the results presented in this thesis also imply the utility of the CFC for use in clinical practice. We showed that the CFC yields a concise measure that could be used to monitor disease progression after a diagnosis of MCI or dementia has been established. In many clinical settings such as memory clinics, it is not always feasible to perform a complete neuropsychological examination at each follow-up consult, due to time or financial constraints. Furthermore, patients often experience the examination as burdensome, which may cause them to abort the testing procedure [34].

However, a quantified measure of cognition and function may assist the clinician in determining whether there has been progression compared to a previous visit. The CFC provides an efficient measure for that purpose, as the cognitive tests can be administered within half an hour. In the meantime, the Amsterdam IADL Questionnaire can be independently completed by an informant. It should, however, be noted that the CFC is partly based on the ADAS-Cog, which is not commonly used in clinical practice. As we did not compare the CFC with a full neuropsychological test battery, we cannot say whether the CFC outperforms a complete neuropsychological examination. However, results in chapter 4.2 suggest that far not all neuropsychological tests are sensitive to decline in early clinical stages of AD. We think that the CFC is less time-consuming and thereby probably less burdensome as compared to a full neuropsychological test battery, while sufficiently comprehensive enough to capture clinically relevant change over time. Additionally, our results imply that the addition of the Amsterdam IADL Questionnaire could enhance the ecological validity of the neuropsychological tests scores that are generally used in clinical practice.

Future perspectives
As AD clinical trials are moving towards pre-dementia stages [35], the sensitivity to change of the CFC in the MCI stage warrants further investigation, because the cognitive component did not detect decline in this stage over a period of one year. One approach to improve the sensitivity of the selected cognitive tests could be the use of IRT, which is the same scoring technique that is applied to the Amsterdam IADL Questionnaire. When applying IRT scoring, individual items are weighted based on their difficulty level, and this information is incorporated when calculating a total score based on an individual’s responses [36]. For example, items that decline relatively early in the disease course require a higher level of the cognitive functioning to successfully endorse that item, and will thus be more heavily weighted for the total score. This may enhance the measurement precision of the cognitive component score, particularly at the early part of the clinical spectrum. Previous studies have indeed shown that IRT could improve the sensitivity to change of existing cognitive tests that were initially scored using classical test theory [37]. Furthermore, simulation studies have shown that IRT yields a better method to analyze repeatedly measured data as compared to sum-score based analyses, as sum-scores tend to overestimate within person variance and underestimate between person variance, which hampers the detection of change over time [38].

Whereas we found that the CFC score declined over time, we did not investigate whether this change was associated with biological changes related to progression. Therefore, it would be interesting to relate longitudinal change on the CFC to changes in biomarkers of AD neurodegeneration, such as hippocampal atrophy or tau pathology [32, 39, 40], to examine whether the CFC captures clinical progression due to AD specific processes. Next to that, it would be interesting to further validate the CFC as outcome measure for progression in DLB, which is the second common cause of neurodegenerative dementia [41]. Limited evidence is available on the clinical course of DLB so far, and proper methods to capture changes in clinical symptoms are still lacking [42]. The Catch-Cog study results suggest that the CFC could be of use to measures progression DLB, however, research including a larger sample-size would be needed to elucidate this.

Finally, the ultimate validation of the CFC would be to confirm its responsiveness to therapeutic intervention [43], which could, by definition, only be accomplished by implementing it in future clinical trials. A first step to achieve this is to convince regulatory agencies that the CFC is a suitable and feasible measure for use in clinical trials that aim to halt clinical progression in prodromal and mild AD, or the so-called Stage 3 and 4 patients [15]. In principle, both the FDA and European Medical Agency (EMA) encourage the use of composite endpoints to evaluate the efficacy of novel therapies and interventions aimed at those stages [12, 44]. However, they also stipulate that those composite measures should be 1) carefully designed, 2) validated in an independent prospective cohort-study, and 3) ‘bear some relevance to existing tools for which historical experience exists’. Most of these aspects have been addressed in the validation study of the CFC so far, which could advance the approval of the CFC as primary outcome measure of efficacy [45]. In this way, the CFC could contribute to the still ongoing quest for a disease-modifying therapy for AD.