Patient self-report for evaluating mild cognitive impairment and prodromal Alzheimer's disease

Patient-reported outcome (PRO) measures are used to evaluate disease and treatments in many therapeutic areas, capturing relevant aspects of the disorder not obtainable through clinician or informant report, including those for which patients may have a greater level of awareness than those around them. Using PRO measures in mild cognitive impairment (MCI) and prodromal Alzheimer's disease (AD) presents challenges given the presence of cognitive impairment and loss of insight. This overview presents issues relevant to the value of patient report with emphasis on the role of insight. Complex activities of daily living functioning and executive functioning emerge as areas of particular promise for obtaining patient self-report. The full promise of patient self-report has yet to be realized in MCI and prodromal AD, however, in part because of lack of PRO measures developed specifically for mild disease, limited use of best practices in new measure development, and limited attention to psychometric evaluation. Resolving different diagnostic definitions and improving clinical understanding of MCI and prodromal AD will also be critical to the development and use of PRO measures.


Introduction
Patient-reported outcome (PRO) measures are used to evaluate the impact of disease and treatment in many therapeutic areas. Among the advantages of patient report is the potential to capture aspects of the disease and treatment experience uniquely accessible to patients and, relatedly, to improve the measurement of therapeutic intervention eff ects [1]. Th e clinician's specialized framework of knowledge makes the clinician the most accurate reporter for some aspects of the disease experience. For which is the patient the more accurate reporter?
Th e most recent recommendations for core clinical criteria for the diagnosis of mild cognitive impairment (MCI) due to Alzheimer's disease (AD) [2] note that despite 'preservation of independence in functional abilities' some impairment in complex functional tasks may be evident, such as higher error rate, taking longer, and/or being less effi cient. Th e companion statement on research criteria for preclinical stages of AD [3] raises the possibility that biomarkers in combination with 'subjective assessment of subtle change will prove to be useful. ' Subtle but potentially important features of the disease experience may be inaccessible to those other than the patient, raising the interesting possibility that the patient may have the most comprehensive and accurate knowledge of performance [4].
Although impairment in social or occupational functioning is part of AD diagnostic criteria [5], the place of functioning in diagnostic defi nitions of MCI is still evolving [2,[6][7][8]. Initial defi nitions of MCI were based on cognitive impairment and intact activities of daily living [9], but empirical data support the presence of functional defi cits encompassing skills and activities beyond instrumental activities of daily living (ADLs), many of them subtle [10][11][12][13][14][15]. Functioning therefore emerges as an area of potential value for patient self-report. Two other areas with substantial prior research on patient selfreport in AD and MCI are neuropsychiatric symptoms and health-related quality of life.
Th ere are of course several important obstacles to use of patient self-report in cognitive impairment. Diseaserelated disruptions to memory and cognition may interfere with the ability to complete a questionnaire accurately, as might loss of insight with progressive disease [16], leading to reliance on informant and clinician report [15]. However, accuracy of informants, especially family caregivers, can also be suboptimal for multiple reasons, including the distortions introduced by caregiver depression and lack of caregiver awareness of some symptoms (for example, [17]).
Th e focus of this overview is on the value of patient report for evaluating disease course and treatments in MCI and in prodromal, or 'early' AD [18]. Th e emphasis is on early disease, corresponding to newer terminology referencing prodromal AD, as well as to the less specifi c 'mild cognitive impairment' referenced by Petersen and colleagues [9].

Methods and fi ndings
Domains important for patient report in cognition were identifi ed based on literature reviews completed for the Cognition Initiative, now the Cognition Working Group of the Critical Path Institute, between August 2009 and January 2011. Initial searches were limited to the period from January 2004 to June 2009 with subsequent updates through March 2011. Functioning, variously defi ned, emerged as an important area for self-report in early disease. Th ere has been recent PRO measure development and empirical studies in the areas of complex ADL functioning and neuropsychological aspects of func tioning (for example, executive functioning); additional work in self-reported neuropsychiatric symptoms and healthrelated quality of life was also identifi ed. Each of these areas is considered briefl y below, followed by a discussion of the role of insight in patient self-report. Details of the search and literature review are available below. A summary of selected measures is presented in Table 1.

Search methods
Th e initial literature search strategy targeted publications on AD and MCI (specifi cally ' AD, moderate to severe' and 'MCI or very early AD'), crossing this literature with specifi c domain terms (functioning, functional status, executive functioning, HRQL, aff ect/mood/behavior). Th e search was limited to English language publications from 2004 to 2009 in MedLine and Embase. To ensure that relevant measures used in clinical trials for currently marketed AD drugs were included, separate searches were conducted for MCI and AD in each domain of interest, limited to 1999 to 2009, with the main focus on ' Alzheimer's disease' OR 'mild cognitive impairment' OR 'cognitive impairment no dementia. ' Since treatment effi cacy was not the focus of this review, but rather measures used to assess effi cacy and eff ectiveness from the perspective of patients and caregivers, this part of the search was limited to review articles. Searches were conducted in PubMed initially, followed by Medline, Cochrane Library of Systematic Reviews, PsychINFO, and Embase.
Full articles were retrieved if information on measure development, psychometric evaluation, and/or use were mentioned in the abstract. Information from retrieved articles was abstracted into tables addressing each of these elements. All relevant titles and abstracts were screened (level 1). Full papers were obtained for any studies considered potentially eligible or where un cer tainty existed as to whether a paper should be included in the review. Full papers were formally assessed for relevance (level 2). Level 1 and 2 reviews for the literature review conformed to pre-determined inclusion and exclusion criteria, including focus on early AD/MCI patients, and caregiver-and patient-reported outcomes were included. Electronic data extraction forms were completed by reviewers trained in the critical assessment of evidence. A third reviewer independently examined any inconsistencies in extracted data elements between extractors and missing data fi elds. Any discrepancies in extracted data were resolved by consensus and any disagreements were resolved by consulting with a third investigator, as necessary. Th e consensus version of the extracted data was subsequently exported to the evidence tables. Th e extracted data elements from each accepted study included study design and measures, instruments, and domains and items of interest.
Consensus on the specifi c functional defi cits that charac terize MCI or prodromal AD has not been reached, especially since early defi nitions of MCI required the absence of functional defi cits. Th e presence of MCI, as well as subtlety of functional defi cits relative to AD, is now recognized [34].
Many AD functioning measures exist given the centrality of functioning to the expression of disease, but most are informant reported, including in: the Physical Self-Maintenance Scale [35][36][37]; the Blessed Dementia Scale [36][37][38]; the Dependence Scale in Alzheimer's Disease [39]; the Disability Assessment for Dementia Scale [40,41]; the Interview for Deterioration in Daily Living Dementia [42,43]; and the Progressive Deterioration Scale [44].
Like most measures of functioning used in AD, the Alzheimer's Disease Cooperative Study Activities of Daily Living (ADCS-ADL) was developed as an interviewbased informant-reported measure of level of independence in specifi c tasks [45]. Subsequently, a version for use with MCI, the ADCS ADL-MCI, was developed with both informant-and patient-completed versions; item content includes complex and instru mental ADLs, such as handling fi nances, shopping, travel, and remembering appointments [46]. To meet the need for a brief in-home rated ADL measure, the Activities of Daily Living Prevention Instrument was developed by the Alzheimer's Disease Cooperative Study Prevention Instrument Project, and is based in part on items from the ACDS ADL-MCI, the Functional Activities Questionnaire [47], and the Disability Assessment for Dementia Scale [46,[48][49][50][51][52]. Th ere are both patient-and informant-rated versions; item content overlaps substantially with the ADCS ADL-MCI. Th e ADCS Prevention Instrument Project also developed the Mail-In Cognitive Function Screening Instrument, with patient-and informant-completed versions. Although intended as a screening tool, item content includes a range of everyday functioning, including social activities and work performance [41,42,[51][52][53][54].
Th e Patient-Reported Outcomes in Cognitive Impairment (PROCOG) [14] measures the impact of MCI and early AD-associated cognitive impairment on multiple domains, including specifi c everyday functioning skills and social functioning. Similarly, the Perceived Defi cits Questionnaire addresses a range of symptoms and functional impacts of memory loss based on patient selfreport and has proven useful for signal detection in a treatment trial for MCI, although it was originally developed for use in multiple sclerosis [4]. Th e Perceived Defi cits Questionnaire is an example of a measure of 'subjective memory complaints' , most of which include cognition symptom report along with functioning (for example, Questionnaire d'auto-évaluation de la mémoire (QAM)/Self-Evaluation Complaint Questionnaire [55]; Self-Rating Scale of Memory Functions [56]).
A summary of some relevant measures is provided in Table 1. As noted by others, few published reports on functioning measures include psychometric performance [32], although for the measures with patient-reported versions, available test-retest reliability data and concurrent or predictive validity data generally indicate good psychometric performance, providing some evidence of accurate measurement. Of note is that despite content overlap in existing measures, some domains are relatively under-represented, such as social functioning or functioning related to language skills -both areas for which patient report may be particularly well-suited.
Th e domain of functional status in cognitive disorders is one with a long history of scale development and use, and AD research is currently well-served by existing informant-reported scales for assessing moderate to severe disease. However, most item content fails to capture subtle defi cits, and few patient-reported measures have been developed to date.
Some performance-based assessments address areas that could be promising for adaptation as self-rated measures, including fi nancial capacity [57,58], facial emotion processing [59], and route navigation [60]. Linking functioning to specifi c cognitive skills through these and other areas may expand clinical characterization of prodromal AD [61].
Because of limited use of qualitative data collection from patients in the measure development process, a step key to best practice in measure development [1], further refi nement of 'functioning' measures may be warranted, including through identifying and measuring aspects of functioning most relevant to early disease, and establishing consensus on the defi nition of everyday functioning and complex ADL functioning.

Executive functioning
Executive functioning represents the cognitive skills required for the planning, initiation, sequencing, and monitoring of complex goal-directed behavior, such as household chores [5,62,63]. Executive functioning impairment is a criterion for dementia diagnosis [6].
Executive functioning skills underlie the everyday functioning skills discussed above, but are considered separately here because measures of executive functioning focus on a specifi cally defi ned set of cognitive skills rather than on the tasks those skills enable. Data from Farias and colleagues [64,65] support distinguishing between measurement of daily living skills and measurement of neuropsychological functioning, based on data showing a moderate correlation between measures of each in a sample with AD (see also [66]). More recently, data from the Advanced Cognitive Training for Independent and Vital Elderly (ACTIVE) Cognitive Intervention Trial also support this distinction, as well as the relationship between cognitive skills and everyday functioning [67].
Executive functioning measures that have been used in MCI include the Behavior Rating Inventory of Executive Function -Adult version [68] (BRIEF-A [69]) and the Frontal Systems Behavior Scale [70,71], with patient-and informant-reported versions for each. Th e BRIEF is a measure of everyday behavioral manifestations of executive control and is sensitive to subtle changes in MCI patients and those with cognitive complaints [68,72]. Similar to fi ndings from Farias and colleagues [65], BRIEF-A scores were only modestly correlated to neuropsychological measures of executive functioning, suggesting that self-and informant report provides unique information about executive functioning relative to performance-based measures. Th e Frontal Systems Behavior Scale is a rating scale of apathy, disinhibition, and executive function and has demonstrated sensitivity to impairment in an MCI sample [70].
Measures of executive functioning show promise for detection of subtle defi cits in MCI [70,71]. As noted above for everyday functioning measures, obtaining structured input from patients in accordance with best practice for measure development may usefully expand the set of relevant impacts to measure and/or aid with identifying diff erential importance of content from the patient perspective and by stage of disease.

Neuropsychiatric symptoms
Although neuropsychiatric symptoms are frequently an important part of the disease course of AD, their presence earlier in the disease is not as well-established. In recognition of the unique presentation and possible prognostic signifi cance of major depressive disorder within AD, the National Institute of Mental Health developed a modifi ed provisional set of criteria for depression in AD, distinct from the DSM-IV (Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition) criteria for major depressive disorder [73,74]. Work with these criteria has indicated that the prevalence of major depressive disorder is signifi cantly underestimated in this population relative to DSM-IV-based prevalence estimates [74,75].
Behavioral and psychological symptoms are evident among some MCI and mild AD patients [11], and at elevated rates relative to the normal aging population [76]. Increased apathy and executive dysfunction have been documented in MCI [70,77]. Th ere is preliminary evidence for higher rates of neuropsychiatric symptoms such as depression, anxiety, agitation, disinhibition, irritability, and sleep problems among those with executive dysfunction type MCI relative to both amnestic and non-amnestic MCI [76] and presence of depression (based on caregiver report) has been found to be predictive of progression from amnestic MCI to AD [75].
Few measures of neuropsychiatric symptoms have selfreport versions and few are validated for use in early disease. Further research is required to develop evidence for the validity of patient self-report for these symptoms.

Health-related quality of life
Health-related quality of life (HRQL) is the subjective assessment of an individual's psychological, physical and social functioning or well-being [78,79] and is traditionally measured via self-report, although for AD, measures have both patient-and informant reported versions [80,81]. No MCI-specifi c HRQL scale exists; instead, existing AD measures have been used in MCI (for example, the Alzheimer's Disease Related Quality of Life instrument [82]) as have generic measures (defi ned as measures intended for use with any population or therapeutic area; examples are the World Health Organization Quality of Life questionnaire, short version [83] and Short-Form (SF)-12 [84]). A systematic review of clinical trials in AD found very low use of HRQL measures (in <5% for trials conducted through part of 2006) [85]. Data from a small sample suggest that reliability and validity of HRQL self-report in MCI and AD is correlated with insight level [86]. While HRQL measures have been central to the rise of PRO assessment over the past two decades, the value of disease-specifi c HRQL assessment to treatment evaluation in MCI and prodromal AD is limited by lack of consensus on domains to include and lack of clarity about how to weight domains for scoring. Th e HRQL impact of MCI, distinct from that of later disease, remains to be defi ned. Further work exploring the relationship of HRQL to functioning, neuropsychological disease eff ects, and neuropsychiatric symptoms would enhance the quality of HRQL measurement and improve its usefulness for research applications.

Insight and patient self-report
Insight into illness is a critical issue for patient self-report in MCI and AD given that insight into disease eff ects declines as the disease progresses [87][88][89][90][91]. Lack of insight is defi ned as lack of the ability to elaborate on the experience of a disease, label the symptoms of the disease as pathological, or have knowledge of the deeper eff ects that the symptoms or disease will have on one's environment [92]. Anosognosia is defi ned as unawareness of defi cits, specifi c cognitive dysfunction, and lack of insight [16,[93][94][95][96]. Th e terms 'lack of insight' and 'anosognosia' are used largely interchangeably in the cognitive impairment literature.
Th e relationship of insight to progression in MCI is less clear than it is for AD. For a review of insight in MCI see [95]. Th ere is currently no consensus on the best method to measure insight. Most methods rely on informant report as a 'gold standard' with patient/informant concordance taken as an indirect measure of patient insight. When the informant is the caregiver, accuracy of report bears critical examination. Caregiver burden, level of depression and anxiety, and caregiver health, including cognitive health, may infl uence accuracy of caregiver report (for example, [97,98]).
Within the AD literature, there has been examination of concordance along with caregiver factors in reporting [17,86,99,100]. Data on patient/informant concordance and informant accuracy are limited for the milder levels of cognitive impairment. In general, data support an inverse correlation between insight and severity of cognitive impairment and an inverse correlation between patient and caregiver report and severity of cognitive impairment [88,101,102]. Dementia patients likely underestimate their defi cits in comparison to caregiver informants [103], with concordance further reduced as disease progresses (for example, [104]). Some empirical reports conclude MCI patients have preserved insight. For example, Farias and colleagues [15] found that MCI patient self-report was concordant with reports of others, suggesting that MCI patients do not under-report actual defi cits in cognition and functioning. Other studies suggest lack of MCI patient insight (see [95] for a review). Confl icting fi ndings about insight and ability of patients to self-report may be due to diff erent defi nitions of insight, diff erent defi nitions of MCI, and/or diff erent methods of measuring insight. Most studies of insight focus on insight for memory functioning; few studies address insight for other cognitive skills, everyday functional abilities, behavior, or aff ect [95]. Th e current literature on insight in MCI is limited by lack of specifi city about domains aff ected, a critical point given evidence of diff erential insight by domain for MCI patients [91,[105][106][107]. Insight may be well-preserved in some domains across a range of disease severity, but may diminish more rapidly in others [95]. For example, Clement and colleagues [91] found that some but not all domains assessed corresponded to performance defi cits in global cognitive score and executive functioning for MCI patients, suggesting MCI patients may be aware of general cognitive defi cits but not specifi c memory defi cits. To date, the literature on MCI supports the conclusion that insight in MCI is not a single construct and that insight might be spared for some but impaired for other domains (see Roberts and colleagues [95] for a review).
Evidence suggests that MCI patients may have knowledge of defi cits in advance of when defi cits are clinically discernible [108][109][110]. Kalbe and colleagues [93] found that MCI patients overestimate cognitive defi cits relative to informants on a 13-domain complaint interview; mild AD patients underestimate their defi cits relative to an informant. Th e validity of the conclusion of 'overestimation' is worth challenging, however, as early cognitive loss may be apparent to the patient but no one else, in part because of the nature of the defi cits and in part because MCI patients may actively hide symptoms from others.
To optimize patient self-report, further research is warranted to determine the relationship of insight to level of disease severity, attending to potential diff erences in insight by domain rather than treating insight as a single global construct. It will be particularly interesting to identify those domains for which patients, especially MCI patients, may have the most accurate view of performance relative to other informants, including clinicians.
Some patient-reported insight scales are presented in Table 1.

Conclusion
Th e increasing interest in MCI due to AD [2], preclinical AD [111], and prodromal AD [18] presents an oppor tunity to advance outcomes measurement in cognitive disorders by addressing ceiling eff ects of existing measures and by expanding the range of measurement targets beyond neuropsychological assessments into the realm of patient-reported outcomes. Patient self-report also off ers a means of expressing, and perhaps quanti fying, the clinical signifi cance of specifi c clinical changes.
Progress in identifi cation of treatments for cognitive impairment depends on accurate measurement. Among the concepts for which patient self-report could be valuable, and for which measurement appears feasible based on available psychometric data, are aspects of everyday functioning and complex activities of daily living and some aspects of executive functioning. Few measures currently address these concepts. Further, domains included in existing measures vary and no measure is comprehensive; consensus on specifi c functioning domains relevant to early disease would improve measurement. Th e extent to which under-studied areas, such as social functioning and language skills, are useful to assess is uncertain given lack of data.
Subtle changes in mood and aff ect specifi c to MCI may be usefully captured by self-report but to date there are limited data on validity of patient self-report of neuropsychiatric symptoms in early disease. Measurement of the health-related quality of life impact of MCI has proceeded largely on the basis of measures developed for AD; relevance to the MCI experience remains to be established. Understanding the MCI experience in greater depth can improve conceptualization of HRQL. Currently, HRQL assessments in MCI and mild AD are based largely on existing AD measures with little psychometric performance data on suitability for measurement of milder levels of impairment.
Th e impact of fl uctuating or declining insight in mild cognitive impairment on patient report is unclear. At what point does loss of insight make patient self-report no longer reliable and valid? Current research suggests that this point may vary by domain, with patients demonstrating suffi cient insight to reliably and validly self-report about disease-related impairment in some areas well into mild to moderate AD.
Consideration of strategies for quantifying the impact of other variables on the accuracy of measurement should be part of measure validation. Cultural diff erences in symptom expression and interpretation are one example. Item response theory methods will likely be of value to identifying diff erential item functioning and quantifying cultural confounds [120][121][122]. In addition to the possibilities of new measure development, such as is being undertaken by the Cognition Working Group of the Critical Path Institute's PRO Consortium, existing AD measures could be tested in the MCI population and converted to self-report if feasible.
Th e increasing emphasis of research on symptoms, correlates, and impact of cognitive impairment at mild levels suggests that the time is right for development of new patient-reported measures for MCI. Although measure ment from the perspective of patients with MCI and prodromal AD is still at an early stage, the development of new measures and psychometric evaluation of existing AD measures for use in early disease should be pursued to increase the tools available and to expand our understanding of mild levels of cognitive impairment.

Competing interests
LF was a salaried employee of United BioSource Corporation (UBC) when some of the work included in this manuscript was completed. UBC held and still holds a contract with the Critical Path Institute for completion of work related to development of a new PRO measure and that work included literature reviews. Some content related to those literature reviews is included in this manuscript. During completion of this manuscript LF worked at MedImmune, LLC, owned by AstraZeneca, PLC, and served as an AstraZeneca industry sponsor representative to the Cognition Working Group of the Critical Path Institute. WRL holds stock in, and has a pension with, Pfi zer. MC was an employee of Merck until 2010 and still holds stock, and has been a consultant with Critical Path Institute and the Cognition Working Group.

Acknowledgments
Funding for this review was provided by the Cognition Working Group of Critical Path Institute's PRO Consortium. The pharmaceutical fi rms participating in the Cognition Working Group are: Abbott, AstraZeneca, Boehringer Ingelheim, Bristol-Myers Squibb, Eisai, Janssen, Merck Sharp & Dohme, Novartis, Pfi zer, and Roche. Critical Path Institute's PRO Consortium is supported by grant number U01FD003865 from the United States Food and Drug Administration and by Science Foundation Arizona under grant number SRG 0335-08. The authors gratefully acknowledge the comments and feedback from the Cognition Working Group members on earlier drafts of this manuscript. The authors also thank Leah Kleinman, Jill Bell, and Anne Brooks for their assistance and review. The views expressed in this article are those of Dr Frank and do not necessarily refl ect those of PCORI.