Evaluating subjective cognitive decline: a systematic review of tools available for evaluating cognitive complaints in Portuguese

ABSTRACT Background: Subjective cognitive decline (SCD) is a perception that is not objectively measured in screening tests. Although many tools are available for evaluating SCD, no single gold standard is available for classifying individuals as presenting SCD, in the Portuguese-speaking population. The aim of this study was to systematically review the literature for tools used to evaluate SCD in the Portuguese-speaking population. Methods: Four databases (Web of Science, SciELO, LILACS and MEDLINE) were primarily utilized in this study (Phase 1). Subsequently, we conducted a manual search of the literature (Phase 2). We then retrieved tools for critical evaluation (Phase 3). Studies that matched the inclusion criteria were analyzed. We summarized the features of each tool in terms of the number of questions, scoring system, benefits and deficiencies, translation and validity. Results: A total of 30 studies utilizing four questionnaires and seven different single questions were found. The tools retrieved were the Memory Assessment Questionnaire (MAC-Q; 12/30 studies), single-question methods (7/30 studies), Subjective Memory Complaint Scale (SMC scale; 5/30 studies), Prospective and Retrospective Memory Questionnaire (PRMQ; 3/30 studies) and Memory Complaint Scale (MCS; 3/30 studies). Only two were formally translated and validated for the Portuguese speaking population (PRMQ and MCS). Conclusions: In summary, SCD is still underinvestigated in Portuguese-speaking countries. The MAC-Q was the most commonly used tool in Portuguese, despite its lack of formal translation and validation for the Portuguese-speaking population. Further studies are required in order to develop and validate a screening tool that includes questions for detecting SCD-plus features and affective symptoms, so as to improve its predictive value.


INTRODUCTION
Subjective cognitive decline (SCD) is a manifestation of self-observed concern about one's thinking processes, most frequently memory. This complaint is helpful in accessing non-objective features of cognitive decline in clinical practice. Functionally, SCD is defined as a cognitive complaint without objective cognitive impairment, and it represents an early stage of Alzheimer's disease (AD) 1 . However, this subjective perception is greatly influenced by affective symptoms and mood states such as depression and anxiety in older adults 2 .
SCD has been correlated with increased risk of cognitive decline, even though the majority of individuals presenting SCD do not exhibit progressive decline 3 . It has been estimated that SCD manifests approximately 15 years before the subsequent mild cognitive impairment (MCI) stage 4 . Thus, it represents a therapeutic window for improved clinical interventions among AD patients. However, evaluating SCD in clinical practice and research protocols may be challenging. Many tools for classifying individuals as SCD-positive have been described, ranging from a direct single question to complex questionnaires. Nonetheless, no single gold standard tool is available 3 .
An international working group was created to standardize the terminology and criteria for SCD, namely the Subjective Cognitive Decline-Initiative (SCD-I) 1 . This group established that the research priority was to develop and validate an SCD scale 3 . A great variety of tools has been described in studies that evaluated SCD and, ultimately, this has hampered the interpretation of studies and decreased the consistency of trials. The working group also suggested a number of features that should be addressed when selecting a tool to evaluate SCD. The results from different studies that each used a single question to classify individuals as presenting SCD may not be interpreted together, given that different questions have different meanings. Consequently, different assessment procedures may have variable results 5 , which are influenced by culture and education 6,7 . Therefore, a single question or brief assessment tool may not encompass the full meaning of a cognitive complaint 8 .
Although structured questionnaires may be more accurate, they are time-consuming. This is an important variable in applying a questionnaire, considering that general practitioners' time is usually too limited to be able to perform a full assessment. Moreover, questionnaires with multiple questions have higher consistency but answering them is complicated for individuals with low levels of education, which is often the situation in countries like Brazil 9 . Measuring memory loss may be challenging in this context. Thus, more efficient and reliable tools need to be developed to evaluate SCD in both clinical and research settings 3 .
One important recommendation from the SCD-I is that tools evaluating SCD should address the appropriate demographic context, undergo psychometric validation studies and have content that is appropriate for the target population 10 .
Therefore, the present study was conducted to provide a systematic review of tools that have been used to evaluate SCD in Portuguese-speaking population. We critically reviewed and discussed the pros and cons of each questionnaire and put forward a suggestion for a high-accuracy screening tool for use within research settings.

METHODS
This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement 11 and was registered in the international prospective register of systematic reviews, under identification number 140092.

Study selection
The data for this study were collected in three phases, as follows.

Phase 1 (search in literature databases)
We performed a search in MEDLINE, Web of Science, LILACS and SciELO for all data up to February 2020. Peer-reviewed journals and grey literature were investigated for original studies, in order to obtain an overview of the available literature.
The search strategy included "subjective cognitive decline" and all its possible synonyms found in the literature: ("self-reported" or "subjective") and ("memory" or "cognitive") and ("decline" or "impairment" or "complaint" or "concern") and "Portuguese". Boolean operators were used depending on the structure of each search engine. For a broader search of Lusophone people, there were no language restrictions. No meta-analysis was performed because of the qualitative nature of this investigation.
Two authors (ELC and VNL) independently assessed potentially eligible studies for their suitability for inclusion in the review. Any disagreements were resolved by discussion or by a third reviewer (WVB). During the screening of titles and abstracts, papers were defined as relevant if they mentioned aspects of SCD, such as "subjective memory decline, " "selfreported memory complaints, " "subjective cognitive impairment" or "memory complaints. " The abstracts and titles were analyzed in accordance with the inclusion criteria, and all articles that met these criteria were included for full reading. From this, the articles that identified questionnaires used in samples from Portuguese-speaking populations were included in this review.
In order to recognize all questionnaires available for Portuguese speakers, the inclusion criteria were rigorously defined. Articles were required to show original data, include a group of individuals presenting SCD or similar, and needed to clearly describe the inclusion criteria for participants in that study. The exclusion criteria comprised unclear definition of SCD and tools that were used to evaluate conditions other than neurodegeneration.

Phase 2 (manual search of the literature)
A manual investigation of the literature was performed to search for studies that might have been missed in the primary strategy. This phase was conducted to search for references within other studies and in Google Scholar. This produced a large number of findings (over three million), so we added the words "questionnaire and Portuguese" to improve study selection (Figure 1). Only the first 100 references were evaluated. According to previous evidence, Google Scholar is a useful tool and can help in retrieval of even the most obscure information, but its use is marred by inadequate citation information 12 .

Phase 3 (critical review of tools)
Data extraction was conducted by two authors (ELC and VNL), from papers that met the inclusion criteria. A summary of the studies was created, which included the country in which the study was conducted, total sample size, the method to classify individuals as SCD and the main outcome of the study. The scoring for SCD-positive features was determined as described by Jessen et al. 3 : the tools were evaluated for subjective decline only in memory; the onset of SCD within the past five years; onset of SCD in individuals aged 60 years or over; associated concerns, persistence of SCD over time; seeking of medical help; and confirmation of cognitive decline by an observer. The instrument validation process was also evaluated through its factors and psychometric properties when described ( Figure 1).

RESULTS
A total of 1,444 studies were retrieved ( Figure 1). Phase 1 of this study led to the retrieval of 21 studies that matched the inclusion criteria, while phase 2 included 9 studies from the grey literature and Google Scholar. In phase 3, seven studies were found to have used a single-question method to classify individuals as belonging to an SCD group, while 24 used questionnaires. During this last phase, the tools were evaluated for authors, year of publication, translation process and validity. To facilitate the interpretation of the results, the tools were evaluated together, but their properties were analyzed separately.
Studies that included other conditions not related to neurodegenerative diseases, such as multiple sclerosis, stroke and epilepsy, were excluded from this review. Some studies were excluded because they conducted evaluations on samples of older adults presenting with cognitive impairment, which was not the aim of this study. Eleven studies analyzed SCD using a group comparison (presence vs. absence of SCD), while the majority analyzed it using regression analysis (association between SCD and other variables).
The largest number of the studies was published in 2010 (5/30), while the first was published in 1998 (Brazil). There were 5 studies (16.7%) from Portugal and 25 studies from Brazil, among which the majority were from the São Paulo area (19/25) or from Porto Alegre (6/25 studies). The total number of participants recruited in all the studies was 7,035 (mean 234.5±274.23), with a mean age of 52.87 years (SD±7.79) and female predominance (4,355 individuals; 61.9%). Only one study analyzed biomarkers in individuals with SCD 13 , while another study analyzed SCD and individuals who converted from nondemented to dementia 14 .

General evaluation of instruments
We found that 30 studies used some tools (questionnaire or question) to evaluate SCD in the Portuguese speaking population ( Table 1). The majority of studies (24, 80%) used a questionnaire to evaluate memory complaints. Seven authors (23.3%) classified individuals in the SCD group based on a  15 , Forgetfulness frequency scale 16 and Sunderland Everyday Memory Questionnaire 17 . The latter two were referenced as retrieved from books, not articles (Table 1). Several tools have been developed previously, such as MAC-Q, SMC-scale and Sunderland questionnaire. However, most of these have not been subjected to any associated in-depth psychometric analysis. A total of 18 studies analyzed here referred to other studies and used non-validated tools. Furthermore, only two instruments retrieved in this review had been analyzed for validity in their populations (PRMQ and MCS). Importantly, the PRMQ was subjected to both confirmatory 18 and exploratory factor analysis 19 . In each analysis, the number of questions was the same (16 items), but the latter found a higher Cronbach's alpha when the PRMQ was reduced to 10 questions.
Interestingly, application of the MAC-Q showed a decrease with time, and the last study that mentioned using this tool was carried out in 2017 20 , while other tools have been increasingly used in Portuguese-speaking populations. The first usage of the MAC-Q appeared in 2003 21 . However, a small number of studies have utilized this tool for evaluating SCD based on cross-cultural adaptation and validation within the population studied. It remains the tool that has most commonly been used.
One study used both a questionnaire and a single question to classify individuals with memory complaints, and the results were compared 21 . Two questionnaires could not be found online (Forgetfulness frequency scale 16 and Sunderland everyday memory questionnaire 22 ), despite an extensive search of the literature. Jacinto et al. 23 used a single question to classify individuals in the Subject Memory Complaints group (yes or no) and measured its correlation with the diagnosis established by cognitive neurologists using a short form of the IQ-CODE. This tool was used in their study by physicians to help in making diagnoses and not in classifying individuals as SCD-positive.

Evaluation of specific instruments
The instruments have been listed according to the frequency of their use in the studies included ( Table 2). The adapted version of the questionnaire on forgetfulness described by Lima-Silva 24 was used specifically in one study and could not be found in the search online. Similarly, the questionnaire used by Gil 22 could not be found and, thus, both tools were excluded from the critical review. • Memory Complaint Questionnaire (MAC-Q 19,20,21,24,25,26,27,28,29,30,31,32,33,34,35 ): This questionnaire uses a 5-item Likert scale from "much better now"=1 to "much worse now"=5, and the last question scores double. The scores can range from 7 to 35 points. The questions ask the subject to compare some aspects of his/her memory with a time when he/she was younger. A cutoff of 25 points or more indicates that the individual presents a memory complaint. The question about the telephone number (Q2) is relatively outdated due to the popularity of mobile phones. Despite being the questionnaire most used to evaluate SCD, it has not been translated and validated for use in Portuguese-speaking populations, unlike the PRMQ 38 27,28 in five different studies. It comprised 10 questions scoring from "No"=0 to "Yes"=1 to 3, depending on each question. The cutoff point for SCD was defined as 3 points. Its short and direct questions were easy for older adults to understand. Two questions (Q6 and Q10) evaluated naming abilities and attention, respectively, which may negatively interfere with the purpose of the questionnaire, i.e. evaluation of memory. However, no formal translation and validation were found in phase 2 of the present study. • Prospective and Retrospective Memory Questionnaire (PRMQ 18,29,38 ): The PRMQ is a well-tested and validated tool for use among Portuguese speakers. It contains 16 questions on a Likert scale from "Never"=1 to "Almost always"=5. The cutoff point is 16 or more. It is the only questionnaire that evaluates both retrospective and prospective memory. Two-factor analyses, namely exploratory 19 and confirmatory 18 , were performed to validate this instrument, which makes comparisons between studies difficult. However, because of the large high number of questions and the length of some questions, applying this questionnaire may be time-consuming. • Memory Complaint Scale (MCS 30,40,43,44 ): This scale comprises two sections with seven items each, scoring from 0 to 2 in each item. Its results are scaled as mild (3-6 points), moderate (7-10 points) or severe (11-14 points) memory complaint. It addresses both the patient and the companion, which increases its predictive value for dementia 31 . This tool was originally developed and validated for the Portuguese population, and it presented high internal consistency. However, except for P7, it relies excessively upon abstraction in questions. The validity analysis that was performed was limited to internal consistency, and there was no factorial analysis. Moreover, the validity analysis 30 had a small sample size (n=161) compared with other studies of this type, and the subjects were selected from a memory clinic. • Single-question methods 13,21,23,46,47,48,49 : To classify subjects as positive for SCD, questions like "How is your memory?", "Do you have memory problems?" or synonyms were used in seven studies. These studies presented different outcomes, including the cerebrospinal levels of amyloid, cognitive scores and affective symptoms. A few identical questions were put forward in different ways. In brief, concise questioning may be easier to apply in the clinical setting 21 , but is vulnerable to the level of education 32 . Even small differences in the single-question method may limit the interpretation of these studies.
Cutoff point> 3 Applied questions score 0-2 (0-14). Cutoff points for memory complaints: Mild (3-6), Moderate (7-10), Severe (11)(12)(13)(14) Applied question: Episodic memory (5), attention (1), language (1) Episodic memory (16), attention (5) Episodic memory (5), attention (2), language (2), orientation (2), processing speed (1) Episodic memory (6), attention (1) addressed the evaluation of tools according to their appropriate demographic characterization, validation process and content. To our knowledge, this is the first study to systematically organize and compare the questionnaires on SCD that have been used for the Portuguese-speaking population. Appropriate demographic characterization is a key feature of evaluation tools that address SCD 10 . Therefore, this review only included studies from Portuguese-speaking populations. However, only a small number of studies evaluating SCD in this population could be retrieved, precisely totaling 30 studies. We correlated the low number of studies on this population, in both Brazil and Portugal, with educational factors. There is only limited capacity for identifying SCD in both countries, since the majority of studies have focused on later stages of AD (Figure 1). In Brazil, another important factor was low awareness of cognitive decline, among family members, general practitioners and other healthcare professionals 33,34 . In Portugal, dementia is also underdiagnosed 35 . Thus, a greater number of studies addressing SCD will provide better identification of demographic and cultural aspects of the Portuguese-speaking population.
The majority of the tools reviewed in this study showed limited psychometric validation, even though this is an important factor recommended by the SCD-I 10 . Four questionnaires and seven different single questions were found in this review for classifying individuals as SCD positive ( Table 2). Only two instruments had been validated for the Portuguese-speaking population (PRMQ and MCS), and both of these tools showed high internal consistency 19,36 . However, PRMQ evaluated not only prospective memory (as in all questionnaires) but also retrospective memory, which becomes diminished through mild cognitive impairment 37 . Moreover, the MCS was the only tool presenting a questionnaire for the patient's companion (i.e. an SCD-plus feature), and had better predictive value for further cognitive impairment than the other tools reviewed. The SMC-scale was used in five studies, but its validity and translation have not been described in the literature. The MAC-Q was the most used questionnaire and has been validated as a screening tool for Portuguese speakers. However, this was not described in the literature and it has not been formally translated into Portuguese. The first study mentioning this tool was from 2003 21 . Validation for a questionnaire is essential in order to be able to determine its psychometric properties 38 . Tools that have not been adequately translated and tested before usage are susceptible to fundamental flaws of construct and consistency 39 , which ultimately restrict their accuracy for making measurements. Further studies need to carefully address non-validated or untranslated questionnaires, so as to be able to adequately classify individuals as SCD-positive.
A total of seven studies used a single-question method to classify individuals as SCD-positive. This method of classification has been described as potentially useful in clinical practice for a variety of screening scenarios 21,40 . The low level of literacy presented by Brazilian people causes bias because of poor understanding of complex questionnaires 32 . Simple, concise and direct questions are preferred for this elderly population, especially those with yes/no answers. However, previous studies suggested that a single-question method showed worse predictive values than brief questionnaires 41,42 . In a research setting, small differences in a sentence or a word may lead to different outcomes and might result in low reliability 43 . For example, "What do you think of your memory?" and "Do you have memory problems?" were used in two studies and presented different outcomes. Hence, interpretating these outcomes together may be inaccurate. Because these questions measure SCD slightly differently, it dilutes the rigor of science in grouping all the results. Thus, we suggest that classifying an individual as SCD using a single question needs to be carefully addressed in the research setting. Instead, a questionnaire using short questions with yes/no answers is preferable for older adults with low literacy.
Currently, there is no single gold standard tool to differentiate between SCD-positive and negative individuals 3 . The SCD-I has established that the research priority is to harmonize existing measures and develop a validated SCD scale 10 . A variety of tools is available for classifying individuals as SCD. Item-response theory and computerized adaptive test modeling were performed to identify reliable SCD questions among a variety of questions 8 . A set of nine questions were selected, including global memory functioning, temporal comparisons and everyday activities. Some of these questions were associated with amyloid positivity 44 and medial temporal atrophy 45 . It has been reported that functional cognitive decline is a major confounding factor of SCD 46 , and that this includes affective symptoms and negative self-evaluation. Hence, it is recommended that novel tools should be designed to evaluate SCD that combine existing measures to improve the diagnostic accuracy of the instrument.
Ideally, an instrument with high reliability and predictive value for progressive cognitive decline should include items both positively and negatively associated with AD 10 . The features that define SCD-plus features were not described in the questionnaires retrieved, except the MCS. A screening tool that includes a brief evaluation of affective symptoms would also be useful for distinguishing individuals at higher risk of cognitive decline. For example, a two-question screen is a valuable tool for rapidly identifying symptoms of depression, and both questions can easily be included in a screening questionnaire 47 . Similarly, a brief tool for anxiety has been developed using only two questions 48 . Both short screenings of affective symptoms can improve the specificity of an SCD questionnaire for pathological cognitive decline, instead of functional decline 46 . Thus, an adequate questionnaire should include SCD-plus features and enable screening for affective symptoms, validated through confirmatory factor analysis in a short and concise manner. This instrument hypothetically would present increased predictive value and specificity for progressive cognitive decline.
Despite our efforts to avoid bias, several key limitations of our review need to be considered. Because we focused on the Portuguese-speaking population, the conclusion of this study is valid only for this community. We tried to minimize publication bias using four major databases and a major search engine (Google Scholar). The latter resulted in inclusion of nine studies through using a manual search of the literature. Thus, we believe that a few other tools may have been used, though with a small impact on SCD. Some of the tools retrieved were not found online, which limited the utility of these tools.
In summary, this study retrieved four questionnaires and seven different single questions that have been used to evaluate SCD among Portuguese speakers. The MAC-Q was the most used tool, despite its lack of formal translation and validation for the Portuguese-speaking population. Further studies should be conducted to validate a screening tool for SCD that would include investigation of SCD-plus features and affective symptoms and would provide increased predictive value with regard to future cognitive decline.