Amélioration du niveau cognitif grâce aux examens de la vision et à la réfraction (CLEVER)

The number of people living with dementia globally will rise from 50 million in 2017 to 131.5 million by 2050. The majority reside in low and middle-income countries (LMICs), where a 223% increase in the burden of dementia is expected from 2015 to 2050. The household cost of caring for demented relatives in India was £2.48 billion in 2013, where total expenses for dementia care will soon consume 0.5% of Gross Domestic Product (GDP). Despite the rapidly-growing burden of dementia in India, only 1 in 10 affected are diagnosed or treated. Proven preventive strategies are desperately needed to close the gap, and could reduce incidence by an estimated 41%, especially early, at the stage of mild cognitive impairment, which itself increases risk of dementia by 23-fold. Existing interventions have predominantly been evaluated in high-income countries (HIC), despite the acknowledged need for prevention in low resource settings. India's national dementia strategy, documented in the Alzheimer's and Related Society of India (ARDSI) Report (2018) and Dementia in India (2020), echo the WHO Dementia Global Action Plan (2017) in prioritising evidence-based approaches to reduce risk.

The elderly, at greatest risk for cognitive disorders, also have the highest prevalence of vision impairment, which affects 2.2 billion persons globally. Growing longitudinal evidence links vision impairment and risk of cognitive decline, with population studies in the United States reporting a 55% increased 9-year risk of new cognitive impairment among visually impaired persons. Longitudinal studies elsewhere in the United States, France, Singapore and China report similar results. While vision impairment is a strong predictor of future cognitive performance, cognition is relatively weakly linked to future visual decline. Both distance impairment and impairment of near vision, a nearly universal part of aging without refractive correction, have been implicated. Observational studies also suggest vision care slows rates of cognitive decline by up to 50%. The association between vision loss and cognitive decline and dementia, while consistent and strong, is not well understood, and various causal pathways have been posited. These include common neurodegenerative or micro vascular aetiology, increased cognitive load, and limited social contact and physical activity due to poor vision, each independently shown to elevate risk of dementia. Studies investigating the role of vision correction in slowing cognitive decline will have a significant impact in developing strategies to reduce the burden of dementia.

Design: single centre, open label, two-arm, parallel group, stratified, interventional randomised trial with a qualitative component Rationale: Despite considerable longitudinal observational evidence, there have been no randomised trials to investigate whether providing near and distance glasses can slow cognitive decline in the elderly. Proof of a causative association between vision correction and the slowing of cognitive decline is needed to potentiate investment in inexpensive vision care for prevention. This requires randomised trials, and none have yet been done.

Study Plan:

Eligibility assessments conducted for 5000-10000 participants. After providing the consent and meeting the eligibility criteria, 820 participants randomised to intervention and control groups.

HbA1C and information about oral health will be collected from the participants in the follow-up visits.

Annual follow-ups for 3 years for both groups. Home visits will be done for those participants who requested a home visit and who otherwise cannot participate.

Cataract follow up: Along with annual visits and the spectacle compliance contacts, follow-up would also be done for those participants who are advised for cataract surgery and efforts would be made to ensure they get the cataract surgery done. This surgery would be done at no cost to the patient.

Lost to follow ups documented with reason for both groups. Missed data for both groups will be analysed as intention to treat (ITT) analysis.

Statistical methods: The change in LASI-DAD global cognitive score between baseline and 36 months will be compared between the intervention and control groups. A linear mixed model will be used to estimate the overall change in LASI-DAD global cognitive score over-time adjusting for potential determinants of change in cognitive status, such as baseline cognitive score, age, gender, education level, occupation, quality of life, depression, social interaction, and physical activity. Restricted maximum likelihood estimation (REML) will be used as the estimation method. If the LASI-DAD global cognitive score is skewed, the log transformation will be used and if the data contains zero, the square root transformations will be done (since the log of zero is undefined). The log transformation is mainly preferred for its ease use and interpretability. The mean change score (and its 95% CI) will be reported for 12 months versus baseline and 24 months versus baseline, and 36 months versus baseline. Significance tests will be two-sided with 5% level of significance and reported using overall Wald p-values for linear regression. All continuous outcomes (INDVFQ, IPAQ, PHQ9, HMSE scores) will be analysed following the same procedure as the main outcome. SNI (categorical ) will be analysed using the multinomial logistic regression. Further details will be provided in the SAP

The trial anticipates two intercurrent events: (i) cataract surgery and (ii) cataract development with refusal of surgery. The primary estimand uses a treatment policy strategy including all patients regardless of cataract status or surgery. The secondary estimand uses a treatment policy for surgery and a hypothetical strategy for refusal, censoring at refusal.

Justification for sample size: Imputing three-year un-intervened decline in the outcome variable (LASI-DAD global score from cross-sectional, age-stratified data, at 29% effect size (from the ACHIEVE study), 502 participants give 90% power at p = 0.05 (two-tailed) based on a two-sample test. With annual follow-up loss of 13% based on estimation from previous studies conducted in India and WHO annual mortality rates in the elderly Indian population (ranging between 10.1 % and 12.7 %) a total of 760 participants is required.

To align with the estimands framework, and to account for intercurrent events that may impact outcome assessment (e.g. loss to follow-up, the development of cataract or competing risks), the sample size was further adjusted. Assuming 7% missing data for the intercurrent events and using the formula by Fang and Jin (2021), 820 total participants are needed across the two study groups. This ensures appropriate power for the primary analysis under a missing-at-random assumption.

With 60% prevalence of uncorrected distance and/or near refractive error, the total number of people needing to be screened is estimated at approximately 1370. However, based on available and other local factors as evident from pre-pilot work, 5,000-10,000 participants may have to be screened for recruiting 820 participants for the trial.