Risk of conversion from mild cognitive impairment to dementia in low‐ and middle‐income countries: A systematic review and meta‐analysis

Abstract Introduction With no treatment for dementia, there is a need to identify high risk cases to focus preventive strategies, particularly in low‐ and middle‐income countries (LMICs) where the burden of dementia is greatest. We evaluated the risk of conversion from mild cognitive ompairment (MCI) to dementia in LMICs. Methods Medline, Embase, PsycINFO, and Scopus were searched from inception until June 30, 2020. The search was restricted to observational studies, conducted in population‐based samples, with at least 1 year follow‐up. There was no restriction on the definition of MCI used as long as it was clearly defined. PROSPERO registration: CRD42019130958. Results Ten thousand six hundred forty‐seven articles were screened; n = 11 retained. Of the 11 studies, most were conducted in China (n = 7 studies), with only two studies from countries classified as low income. A qualitative analysis of n = 11 studies showed that similar to high‐income countries the conversion rate to dementia from MCI was variable (range 6.0%–44.8%; average follow‐up 3.7 years [standard deviation = 1.2]). A meta‐analysis of studies using Petersen criteria (n = 6 studies), found a pooled conversion rate to Alzheimer's disease (AD) of 23.8% (95% confidence interval = 15.4%–33.4%); approximately one in four people with MCI were at risk of AD in LMICs (over 3.0–5.8 years follow‐up). Risk factors for conversion from MCI to dementia included demographic (e.g., age) and health (e.g., cardio‐metabolic disease) variables. Conclusions MCI is associated with high, but variable, conversion to dementia in LMICs and may be influenced by demographic and health factors. There is a notable absence of data from low‐income settings and countries outside of China. This highlights the urgent need for research investment into aging and dementia in LMIC settings. Being able to identify those individuals with cognitive impairment who are at highest risk of dementia in LMICs is necessary for the development of risk reduction strategies that are contextualized to these unique settings.

A meta-analysis of studies using Petersen criteria (n = 6 studies), found a pooled conversion rate to Alzheimer's disease (AD) of 23.8% (95% confidence interval = 15.4%-33.4%); approximately one in four people with MCI were at risk of AD in LMICs (over 3.0-5.8 years follow-up). Risk factors for conversion from MCI to dementia included demographic (e.g., age) and health (e.g., cardio-metabolic disease) variables.
Conclusions: MCI is associated with high, but variable, conversion to dementia in LMICs and may be influenced by demographic and health factors. There is a notable absence of data from low-income settings and countries outside of China. This highlights the urgent need for research investment into aging and dementia in LMIC settings. Being able to identify those individuals with cognitive impairment who are at highest risk of dementia in LMICs is necessary for the development of risk reduction strategies that are contextualized to these unique settings. annually). [5][6][7] Although some cases remain stable, others can revert to normal, with studies suggesting reversion ranges of 4% to 15% in clinicbased samples [8][9][10][11] and 29% to 55% in population-based samples. [12][13][14][15] In the absence of a cure for dementia, understanding the likelihood of, and risk factors associated with, conversion to dementia among MCI cases is important to help identify strategies for dementia risk reduction and prevention. 16 The definition of MCI can be a difficult concept to disentangle.
One of the most widely applied set of criteria, in clinical and research practice, are those defined by Petersen et al., describing patients with subjective memory loss verified by neuropsychological testing, with no significant impairment in other cognitive domains, no functional impairments, and no dementia. 17 Other similar criteria have also been developed and applied including, for example, from the International Working Group, 18 National Institute on Aging-Alzheimer's Association (NIA-AA), 19

METHODS
This systematic review and meta-analysis was conducted according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines (Appendix A.1 in supporting information). 27 The study protocol was registered on the PROSPERO database (registration number CRD42019130958).

Screening process
Two reviewers independently assessed potentially relevant articles for eligibility (AMM and EP). The decision to include or exclude studies was hierarchical and initially made based on the study title and abstract to eliminate obviously irrelevant studies (see Figure 1). When a study's title/abstract could not be rejected with certainty, the full text of the article was obtained for evaluation. Discrepancies between reviewers were resolved by a third reviewer (BCMS). Next, full-text articles were searched. In addition, the reference lists of all included articles were checked for any potentially missing papers. to dementia in population-or community-based studies from low and middle-income countries (LMICs).
The search strategy included terms that encompassed "dementia," "mild cognitive impairment," "incidence," and "conversion. population-based data, to determine the best strategies for identifying those individuals at highest dementia risk to inform the development of dementia risk reduction plans in these settings.

Data extraction
A standardized form was used to extract data from the included studies for assessment of study quality and evidence synthesis.

Risk of bias assessment
Quality assessment was guided by the Newcastle-Ottawa Scale for cohort studies. 33 Risk of bias was assessed on three main categories: selection, comparability, and outcome. The maximum possible score was 8; three stars for selection, two stars for comparability, and two stars for outcome. Two authors (AMM and EP) independently assessed risk of bias, with any disagreement resolved by discussion with a third assessor if required.

Analysis
For each study, we report the proportion of MCI cases that converted to dementia for each definition of MCI separately. This was calculated as the ratio of those who converted to dementia over the total sample size. We also report on the key risk factors significantly associated with an increased risk of dementia.
Details of all risk factors assessed are in Table S2 in supporting information.
Where there were multiple studies using the same criteria to diagnosis MCI and dementia, a meta-analysis was undertaken. This was only possible for studies that diagnosed MCI using Petersen-type criteria with an outcome of AD (n = 6 studies). The analysis was run in Stata using the Metaprop command to compute the meta-analysis of pooled proportions. This allows computation of 95% confidence intervals (95% CI) using the score statistic and the exact binomial method and incorporates the Freeman-Tukey double arcsine transformation of proportions to compute the weighted pooled estimate for normality assumptions. The program also allows the within-study variability to be modeled using the binomial distribution. 34 Given large differences in the design and sampling across studies, the random effects model was computed. Heterogeneity was assessed using the I 2 statistic.

Study selection
From the electronic search, the titles and abstracts of 8977 publications were screened, and the full texts of 87 articles reviewed. The screening and study selection process is illustrated in the PRISMA flow diagram ( Figure 1). Nine articles met the eligibility criteria. The most common reasons for exclusion were that the study was not from a LMIC, the sample was not population-based, and the study did not report incident dementia. A second search conducted in June 2020 identified a further 1670 articles, from which two studies were eligible for inclusion. Thus, 11 articles are included in this review.

Conversion from MCI to dementia
Rates of conversion from MCI to dementia ranged from 6.0% 39  CIND (n = 1 study), 44 and CDR (n = 1 study) 43 Table S2.

Risk of bias assessment
Full details of the risk of bias assessment can be found in Table S3 in supporting information. The included studies averaged 7.3 stars out of 10 (range [5][6][7][8]. Eight studies (out of 11) scored the maximum of eight stars.

Meta-analysis-risk of conversion to AD
Six studies were included in the meta-analysis; n = 4 from China 35,36,38,40 and n = 1 each from Tanzania 45 and Brazil. 42 Across the six studies, MCI sample size ranged from n = 21 42 to n = 837 35 and age from > 55 years 35,36 to > 70 years. 45 Figure 2  is a significant research gap particularly considering the high burden of cognitive impairment and dementia in low-income country settings. 25 Of the studies included in this review, there was also variability in terms of outcome measure (all-cause dementia vs. AD vs. VaD), length of follow-up, sample size, and diagnostic criteria for MCI. In addition, the examination of MCI conversion to other dementia subtypes was limited as only AD and VaD were investigated. Therefore, there is an urgent need for future studies to attempt to standardize the methodology used to allow for better cross-study comparisons, aiming for studies to be population representative and generalizable.
Studies from HICs estimate annual conversion from MCI (irrespective of MCI definition used) to dementia at approximately 3% to 10% in community settings and 10% to 15% in clinical settings. 5,[16][17][18]54 Similar to findings from HCIs, conversion rates were found to be variable across the different LMICs sites ranging from 6.0% to 44.8%. Rates of conversion to dementia were generally higher for those definitions that capture broader impairment (e.g., range of conversion for all-MCI, CIND, and CDR: range 16.8% to 44.8% over 2.0-5.8 years follow-up) compared to more restricted definitions of single domain MCI (e.g., aMCI range 6.0% to 6.9% over an average of 3.5 years follow-up). [55][56][57][58][59][60][61] Given the high reported prevalence of MCI in LMIC settings 4 in addition to the high dementia conversion rates reported here, the development of strategies to prevent or delay dementia progression in those individuals with cognitive impairment could have a significant impact on the burden of disease associated with mental health conditions in these settings.
Similar to findings from HICs, non-modifiable risk factors for progression to dementia from MCI included age [62][63][64][65] and APOE ε4 allele status. 51,66,67 Regarding sex, while being female has been found to be associated with increased risk of prevalent MCI, 4 and has been associated with higher risk of progression to dementia in HICs, 68 only 2 studies 38,44 out of 10 that investigated sex effects found that being female was a risk factor for conversion from MCI to dementia.
Research evidence, predominantly from HICs, suggests a putative link between sex and/or educational attainment and cognition. [69][70][71][72] However, methodological weaknesses and potential of reverse causality within these studies adds limitations to their interpretation and warrants longitudinal studies with longer follow-up. 71 Furthermore, key modifiable risk factors were also similar to those reported in HIC settings, including poor cardiometabolic health, the presence of vascular risk factors, and poor neuro-psychiatric health such as the presence of depression. 23,62,73 Targeting these factors could be an early strategy for not only preventing MCI, but also reducing the burden of dementia. Research evidence suggests that up to 40% of dementia cases may be preventable through targeting 12 modifiable risk factors, 74 many of which can be influenced by diet and lifestyle practices. 75,76 Emerging evidence also indicates that non-pharmacological interventions such as cognitive training may reduce dementia risk. 77   There is an urgent need for research investment into robust, population-representative studies focused on risk of cognitive impairment and dementia in LMICs using harmonized methodology. 78 This is necessary to make it possible to campaign for prioritization of funding toward cognitive screening and risk reduction. This would also allow investment in better education and development of infrastructure in these settings to improve knowledge of diagnosis and risk factor management, but also facilitate the implementation of more population representative, robust studies, particularly in countries of low income. Care.

DATA AVAILABILITY STATEMENT
This systematic review was registered via Prospero; PROSPERO registration number: CRD42019130958. A copy of the systematic review protocol is available on request and can be provided by the corresponding author. Requests for access to the data reported in this article will be considered by the corresponding author.