Original research••

Modifiable risk factors for dementia in India: a cross-sectional study revisiting estimates and reassessing prevention potential and priorities

•,,,,,,,,,.

...

Abstract

Background About 16% of worldwide dementia cases are in India. Evaluating the prospects for dementia prevention in India requires knowledge of context-specific risk factors, as relationships between risk factors and dementia observed in high-income countries (HICs) may not apply.

Methods We computed population attributable fractions (PAFs) for dementia in India by estimating associations between risk factors and dementia, their prevalence and communality, within the same nationally representative sample of 4096 Indians aged 60 and older, surveyed through the Harmonised Diagnostic Assessment of Dementia for the Longitudinal Ageing Study in India.

Results The risk factor with the largest PAF (>20%) was no education, followed by vision impairment (14%), physical inactivity (12%) and social isolation (8%). According to our estimates, eliminating exposure to risk factors significantly associated with dementia would potentially prevent up to 70% of dementia cases in India.

Discussion Previous estimates, based on samples limited to specific geographic areas and using risk factors’ definitions and relative risks from HICs, may not correctly estimate the real opportunities for preventing dementia in India or identify the most critical areas for intervention.

What is already known on this topic

Dementia is a growing public health concern, particularly in low-income and middle-income countries (LMICs) like India. Prior research from high-income countries (HICs) has identified several modifiable risk factors for dementia, but it is unclear if these same risk factors apply to the Indian population.

What this study adds

This study is the first to compute population-attributable fractions for dementia risk factors in India based exclusively on Indian-specific inputs (ie, associations between dementia and risk factors, prevalence, and communality of risk factors) obtained from a large, nationally representative sample of Indian adults. It shows that dementia risk factors in India differ in magnitude and importance from those in HICs.

How this study might affect research, practice or policy

This study emphasises the need for dementia prevention strategies that are guided by empirical evidence from representative data in LMICs, such as India, and tailored to their specific context.

Introduction

As of 2023, approximately 55 million people globally have dementia, with over 60% living in low and middle-income countries (LMICs).1 India, the largest LMIC and home to 18% of the world’s population, is experiencing rapid ageing, leading to a growing dementia burden. Recent estimates indicate that 8.8 million Indians have dementia, representing 16% of cases worldwide.2 With current demographic trends, the number of individuals with dementia in India is projected to reach 17 million by 2036. However, this estimate assumes a constant dementia prevalence and overlooks how the evolution of India-specific dementia risk factors might affect the increase in the number of dementia cases.

Recent evidence from high-income countries (HICs) shows a decline in age-specific dementia incidence, linked to changes in social, behavioural and medical factors associated with dementia risk.3 4 However, it is unclear if these associations apply to LMICs, given differences in cultural practices and health determinants. To reliably assess India’s future dementia burden and prevention potential, knowledge of which factors exhibit a robust relationship with dementia in the Indian context, along with their prevalence and trends, is crucial.

The Lancet Commission on dementia prevention, intervention and care initially identified nine modifiable risk factors causally linked to dementia. These factors included lower levels of education, hearing loss, hypertension, obesity, diabetes, smoking, depression, physical inactivity and social isolation. The Commission suggested that eliminating these risk factors through targeted public health policies could prevent approximately 35% of dementia cases.5 In 2020, the Commission expanded its list to include excessive alcohol consumption, traumatic brain injury and air pollution, estimating that the 12 identified risk factors could account for 40% of dementia cases globally.6

Dementia prevention potential may be higher in LMICs due to a greater prevalence of most risk factors. Mukadam et al (2019) documented that the nine risk factors initially identified by the Lancet Commission could account for 40% of dementias in China, 41% in India and 56% in Latin America.7 However, their estimated population attributable fractions (PAFs) were produced using geographically limited, non-representative samples.8 Also, they were based on the implicit assumption that the same associations between risk factors and dementia observed in HICs hold true in LMICs, which is often at odds with empirical evidence, especially for India.9 10 Belessiotis-Richards et al (2021) used three cross-sectional Indian samples to estimate the relationship between potentially modifiable risk factors and cognitive performance.11 Their findings did not always confirm the directionality and the magnitude of the relationships between risk factors and cognitive impairment observed in HICs. Hence, they cast doubt on the appropriateness of using relative risk estimates from HICs to calculate PAFs of dementia risk factors in India and call for further analyses to better assess the potential for dementia prevention within the Indian context.

In this paper, we estimated PAFs for dementia in India using the Harmonised Diagnostic Assessment of Dementia for the Longitudinal Ageing Study in India (LASI-DAD), a large-scale, nationally representative population survey.12 13 This study includes rich background information on demographics, health and socioeconomic factors, alongside unprecedented high-quality data on cognitive performance, informant reports of everyday functioning and clinical consensus rating of dementia. Exploiting these measures, we obtained India-specific associations between risk factors and dementia as well as prevalence and communality of risk factors within the same sample of 4096 Indian adults aged 60 and older. We then combined these estimates and calculated, to the best of our knowledge, the first PAFs for dementia risk factors based exclusively on Indian-specific inputs (ie, associations between dementia and risk factors, prevalence and communality of risk factors) obtained from a large, nationally representative sample of Indian adults. For comparison with the existing literature, we also computed PAFs using relative risk estimates from meta-analyses based on HIC data while estimating the prevalence and communality of risk factors in our LASI-DAD sample. Our findings provide new insights into the role of potentially modifiable risk factors for dementia in India. This knowledge can help effectively direct and rank public health interventions aimed at averting or postponing dementia in the Indian context.

Methods

Data

The Longitudinal Ageing Study in India (LASI) is an ongoing national cohort study of about 73 000 Indian adults aged 45 and older, representative of each state and union territory.12 LASI collects extensive data on demographics, health and socioeconomic status. The LASI-DAD is a substudy of n=4096 LASI respondents.13 LASI-DAD administered a detailed neuropsychological battery based on the Harmonised Cognitive Assessment Protocol, which was designed within the Health and Retirement Study to measure key cognitive domains, including attention, memory, executive function, language and visuospatial function.14 The protocol, translated into local languages and adapted to accommodate relatively low levels of literacy and numeracy in India, featured in-person interviews with both participants and informants.15 LASI-DAD also developed a web-based platform for clinicians to rate participants on the Clinical Dementia Rating (CDR) scale.16 The validity of this approach was demonstrated elsewhere.17 Our measure of dementia status came from LASI-DAD, while all other relevant variables, including dementia risk factors, are obtained from LASI.

Dementia status

In LASI-DAD, clinical consensus dementia ratings were available for n=2528 participants. CDR-trained Indian clinicians assigned participants a score based on the CDR scoring algorithm.16 Dementia was defined as a CDR score of 1 or higher. For the 1568 LASI-DAD respondents without a CDR score, we used multiple imputations (provided by LASI-DAD) to ensure that standard errors accurately reflect the uncertainty due to missingness. This imputation procedure, originally applied to the larger LASI sample of 60+ Indian adults to estimate dementia prevalence in India, was thoroughly discussed in earlier work.2 In the context of this study, its soundness is further enhanced by two key factors. First, the fraction of 60+ respondents with a missing CDR score is significantly lower in the LASI-DAD (28%) than in the LASI (92%) sample. Second, participants in the LASI-DAD underwent a more extensive cognitive assessment than in LASI (including the Informant Questionnaire on Cognitive Decline in the Elderly18 and the Blessed Dementia Rating Scale,19 ensuring a more accurate estimation of dementia status for individuals lacking a CDR score. Our primary analysis used the full LASI-DAD sample of 4096 Indian adults aged 60 and older. We evaluated the robustness of our findings by repeating the analyses on the subsample of 2528 LASI-DAD participants whose dementia status was ascertained by CDR-trained clinicians (see online supplemental material). Using sample weights scaled to the population of adults aged 60 and over in India, we estimated the prevalence of dementia to be 6.6% and 6.8% in the full LASI-DAD sample and the subsample of study participants with a CDR score, respectively.

Definitions of risk factors

We considered 11 of the 12 risk factors identified in the Lancet Commission’s 2020 report.6 We excluded traumatic brain injury because neither the LASI nor the LASI-DAD survey instrument captured this information. We added vision impairment to the list of risk factors, as it has been shown to strongly affect the risk of dementia in the USA and is highly prevalent among Indian adults.20 21 Visual loss has been recently included in the Lancet Commision’s list of potentially modifiable risk factors for dementia.22 We also added indoor air pollution, which was found to adversely affect cognitive test performances in India in previous investigations.23 We used objective measurements for five risk factors—hypertension, Body Mass Index (BMI), diabetes, vision impairment and outdoor air pollution—and self-reports for the remaining eight—education, hearing impairment, physical inactivity, social isolation, depression, smoking, excessive alcohol consumption and indoor air pollution. We adhered to the Lancet Commission’s definitions of risk factors, originally developed for HICs. In a few instances, we purposefully modified these definitions to obtain measures that are better tailored to and suited for the Indian context. Specifically, given the low educational attainment in our cohort, we considered having no formal education as a risk factor for dementia rather than having less than secondary education—the typical definition of ‘less education’ in HICs. There is evidence that the BMI cutoffs used to classify individuals as underweight, normal, overweight or obese in HICs are not ideal for the Indian population.24 Therefore, we adopted an alternative classification recommended by the Association of Physicians in India: BMI <18 for underweight, 18≤BMI<23 for normal weight, 23≤BMI<25 for overweight and BMI ≥25 for obese.24 We also modified the definition of social isolation often used in HICs to account for living arrangements that are specific to the Indian context. In particular, we created a Social Isolation Index including not only the frequency of contact with friends, relatives and neighbours and attendance at social clubs/events—the factors typically defining social isolation in HICs—but also marital status, living arrangements and quality of relationships with household members.25 We then defined an individual as socially isolated if the value of their Social Isolation Index was at least two-thirds of the maximum score (the details of how this Social Isolation Index was constructed are in the online supplemental material). Due to the high levels of air pollution across India, applying the WHO’s standard guidelines26—which recommend that the annual average PM_2.5 concentrations should not surpass 5 µg/m³—would categorise almost all LASI-DAD respondents as being exposed to detrimental outdoor air pollution levels. As a result, we adopted India’s national ambient air quality standards, which set the acceptable PM_2.5 concentration threshold at 40 µg/m³ per year.27 Accordingly, we created an indicator taking value 1 if the average annual PM_2.5 concentration at the respondent’s address exceeded 40 µg/m³ in the 5 years leading up to the conclusion of the LASI-DAD data collection phase in 2019. Finally, since biomass combustion, which is relatively common in India, can produce very high indoor PM_2.5 concentrations, we added an indoor-air-pollution risk factor, defined as an indicator taking value 1 if the respondent reported using highly polluting fuel such as kerosene, charcoal, lignite, cola, wood or dung cake. The definition of each risk factor and its prevalence in the Indian population of adults aged 60 and older are reported in table 1. Whenever alternative, India-specific definitions of risk factors were considered; Table 1 provides their definition and prevalence alongside those of their HIC counterparts.

Table 1

Risk factors definitions, prevalence and associations with dementia

Statistical analysis

We computed PAFs adopting a standard procedure commonly used in the literature5–7 (the technical details are in the online supplemental material). Unlike previous studies focusing on India that have taken relative risk estimates from meta-analyses based on data from HICs and used Indian samples to obtain the prevalence and communality of risk factors,7 we estimated all three components of the PAF formula—relative risk, prevalence and communality—within the same sample of 4096 Indian adults aged 60 and older. We computed the population-level prevalence of each risk factor using sample weights scaled to the population of adults aged 60 and over in India. We ran weighted logistic regressions separately for each risk factor to obtain relative risk estimates, using 10 imputations of dementia status for respondents without a CDR score. In these models, we controlled for sex, marital status, age, caste, rural or urban area residence, socioeconomic position (per-capita household consumption, food insecurity and household wealth) and state of residence. We selected statistically significant risk factors (p value <0.05), among which we computed communality. Finally, we obtained PAFs adjusted for communality for each risk factor and overall. We estimated PAFs’ standard errors by bootstrap, whereby in each of 250 replications, the three-step procedure described above was repeated on a different sample of n=4096 drawn with replacement from the original sample. This allowed us to consider sampling error in all the PAF formula components. As described above, for some risk factors (namely, education, BMI thresholds, social isolation and air pollution), we adopted Indian-specific definitions, which depart from standard HIC definitions. To evaluate the impact of these departures on estimated PAFs, we ran parallel analyses using standard HIC and Indian-specific definitions and only standard HIC definitions. We also compared our estimates to estimates of PAFs using the approach adopted by previous Indian studies of taking relative risks from (HIC-based) meta-analyses and computing only the prevalence and communality of risk factors within our sample. Though more complex methods permit alternative computations of PAFs in the presence of multiplicative and additive interactions,28 we chose to adopt a standard procedure lacking strong evidence guiding the choice of assumption regarding the form of interaction among risk factors. Our approach enhances comparability with existing Indian estimates and allows us to focus on the influence of using solely India-specific inputs (ie, relative risks, prevalence rates and communality of risk factors) on dementia prevention potential. All analyses were performed using Stata 18.29

Patient and public involvement

Respondents were not actively involved in the design of this study or the data analysis. During the consent process, they were informed of how their participation would enable scientific research. The authorship team is composed of academic researchers at the University of Southern California, the University of Michigan, Johns Hopkins, and partner institutions in India, who have collaborated closely in the design and implementation of the LASI and LASI-DAD studies.

Results

As can be seen in table 1 (where the rows referring to Indian-specific definitions departing from standard HIC definitions are shaded in grey), the risk factors with the highest prevalence were exposure to high outdoor air pollution (73%), no education (56%)—the standard HIC risk factor of less education had a much higher prevalence of 79%—vision impairment (46%), use of biomass fuel for heating and cooking (45%), hypertension (35%), physical inactivity (35%) and social isolation (34%)—the standard HIC definition of social isolation leads to a much higher prevalence of 49%. Depression (11%), hearing impairment (9%) and excessive alcohol consumption (2%) had the lowest prevalence.

Eight risk factors were consistently and robustly associated with dementia status regardless of the adopted definitions (last column of table 1): no (less) education, hypertension, hearing impairment, being underweight, depression, physical inactivity, social isolation and vision impairment. The risk factor with the largest relative risk (RR) was no education, indicating that older Indians without any formal education were about 2.8 times as likely to have dementia as their counterparts with any schooling. The relative risk associated with the standard HIC definition of less education was higher at 3.6. The second strongest associations with dementia, in order of magnitude, were found for physical inactivity (RR 2.11, CI 1.41 to 2.81) and vision impairment (RR 2.01, CI 1.23 to 2.74). Social isolation also showed a notable association with dementia (RR 1.65, CI 1.09 to 2.22), which appeared stronger (RR 2.21, CI 1.38 to 3.03) when adopting the standard HIC instead of our Indian-specific definition. Hearing impairment, depression and hypertension were also associated with dementia, with estimated relative risks of 1.54 (CI 1.01 to 2.10), 1.54 (1.00–2.08) and 1.36 (CI 1.02 to 1.75), respectively. Being underweight was positively and significantly associated with dementia both when we used the Indian-specific BMI threshold (<18; RR 1.55, CI 1.04 to 2.05) or the standard HIC threshold (<18.5; RR 1.64, CI 1.12 to 2.16). Obesity was significantly and negatively associated with dementia (RR 0.34, CI 0.03 to 0.64) when we defined it according to the standard BMI threshold (≥30) but not according to the Indian-specific BMI threshold (≥25). We found that diabetes, smoking, and outdoor and indoor air pollution were linked to an increased risk of dementia, but their estimated relative risks were smaller in magnitude and not statistically different from one. Excessive alcohol consumption, the risk factor with the smallest prevalence of 2%, had a negative but insignificant association with dementia (RR 0.90, CI 0.09 to 1.89). Estimated associations for all demographic variables included in the regressions are included in Table SM1 in the Supplemental Material.

In table 2, we calculated PAFs for risk factors significantly linked to dementia (p value <0.05). The observed communality among statistically significant risk factors is reported in Table SM2 in the Supplemental Material. On the left side of table 2, risk factors follow the standard definitions used in HICs, except education, underweight and social isolation, which are based on our alternative definitions tailored to the Indian context (shaded in grey). On the right side, all risk factors adhere to the standard HIC definitions. Both panels present separate results for two sets of risk factors: the one recommended by the Lancet Commission in 2020 (without traumatic brain injury, not observed in our data) and an augmented one, which also includes vision impairment (as in the recent 2024 Lancet Commission’s report)22 and indoor air pollution.

Table 2

Estimated PAFs for dementia risk factors (I)

The risk factor with the largest PAF was no education: the proportion of dementia cases theoretically preventable if every Indian adult aged 60 and older had some formal education was about 22% (table 2, left panel). The risk factor with the second largest PAF was vision impairment (14%), followed by physical inactivity (12%), and social isolation (8%). Hypertension and being underweight exhibited lower but sizeable PAFs of about 5% and 4.5%, respectively. Depression (2.5%) and hearing impairment (2%) showed smaller PAFs. If our estimated associations were causal, the proportion of dementia cases potentially preventable by eliminating all significant risk factors would be substantial: 57.2% (CI 49.1 to 64.5) without vision impairment and 70.3% (CI 62.4 to 77.2) when adding vision impairment.

The right panel of table 2 indicates that adopting only standard HIC definitions results in higher PAFs. Specifically, the PAF associated with less education was estimated to be between 28% and 29%; for social isolation, it ranged between 15% and 16%; and for being underweight, it was approximately 6%. There was a negative PAF of 2% for obesity when applying standard BMI thresholds. The overall proportion of dementia cases that could potentially be prevented by eliminating all significant risk factors, as defined by HIC standards, was 71.8% (CI 62.1 to 78.3) without vision impairment and 80.7% (CI 71.9 to 86.4) with vision impairment.

We assessed the robustness of these results when limiting the analysis to the subsample of 1568 LASI-DAD respondents with a CDR score, for whom no imputation of dementia status was needed. As reported in Tables SM3 and SM4 in the Supplemental Material, our findings remained consistent. Considering the most comprehensive set of risk factors (including vision impairment), the total estimated PAF was 71.4% (CI 55.5 to 77.2) using definitions specific to the Indian context for education, weight and social isolation. This fraction increased to 84.7% (CI 70.7 to 91.7) when exclusively adopting the standard HIC definitions.

To compare our results with existing evidence for India,7 we calculated PAFs using standard HIC definitions and relative risk estimates from meta-analyses based on HIC data. We relied on the LASI-DAD sample only to obtain the prevalence and communality of risk factors. These calculations are reported in table 3. The first column lists the relative risk values underlying these estimates, all of which are derived from Livingston et al (2020),7 except for vision impairment, which comes from the work of Ehrlich et al (2023).20 In the second column, we considered the same set of risk factors as Mukadam et al (2019), who provided PAFs of dementia risk factors for three LMICs, including India. In the third column, we expanded the analysis to include the set of risk factors recommended by the Lancet Commission (the estimated communality of all these factors is in Table SM5 in the online supplemental material). The estimated PAFs presented in table 3 are generally lower, owing to lower relative risks for some of the risk factors with the highest prevalence. However, the estimated magnitude of PAFs had a similar ordering compared with those calculated using relative risks specific to the Indian context. The risk factor of less education showed the highest PAF, followed by social isolation, vision impairment, hypertension and lack of physical activity. The overall PAF stood at 59.3% (CI 47.8 to 62.3) when using Mukadam et al (2019)’s set, and it increased to 64.1% (CI 54.9 to 67.0) with the most recent Lancet Commission’s set.22

Table 3

Estimated PAFs for dementia risk factors (II)

Discussion

Relying on newly collected, nationally representative data, we computed PAFs of dementia risk factors in India. Unlike previous studies, we did not take relative risk estimates from meta-analyses based on HICs’ data but estimated associations between potentially modifiable risk factors and dementia in a nationally representative sample of Indian adults over the age of 60. Within the same sample, we computed the prevalence and communality of risk factors, which we then combined with relative risk estimates to obtain India-specific PAFs. We documented strikingly consistent patterns across different estimation procedures, with no education, vision impairment, physical inactivity and social isolation identified as the highest-impact risk factors in India.

Our results differed from those previously published in the literature. Risk factors’ prevalence rates were significantly different from those in HICs, partly due to differences in measurement and context.6 They also deviated from those found in geographically limited Indian samples, even when measures and definitions were identical.7 11 Discrepancies between prior results and our findings from a large, nationally representative sample of Indian adults indicate that selectivity introduced by restricted geographic coverage could be crucial in shaping empirical outcomes. Such selection effects may hamper the comparability of results across studies and the generalisability of findings.

Six risk factors (no education, hypertension, hearing impairment, depression, physical inactivity and social isolation) were significantly and robustly associated with dementia status in India. The direction of these associations was the same as in HICs, although the magnitudes were larger for no education, physical inactivity and social isolation. Vision impairment appeared to be positively and strongly associated with dementia in India, consistent with its recent inclusion in the Lancet Commision’s list.22 In contrast with findings from HICs, obesity, diabetes, smoking, excessive alcohol consumption and outdoor air pollution did not exhibit statistically significant associations with dementia. The direction of the estimated associations of dementia with diabetes and smoking was positive, as in HICs. Overall, our results provided strong empirical evidence against the assumption commonly adopted in the literature that a certain set of risk factors may have the same impact on dementia in India as in HICs.

Results for BMI exhibited the largest differences when comparing findings from HICs and India. Obesity correlated negatively but insignificantly with dementia, while being underweight exhibited a positive and significant correlation with dementia. Our findings could stem from reverse causality (see below for a detailed discussion of this issue) and/or contextual differences between HICs and India, where malnutrition, which is relatively prevalent in the population, may adversely impact cognitive functioning, and obesity, which is concentrated among better-off groups, may proxy for other protective factors.

Differences were smaller but notable for several other risk factors. The estimated association between excessive alcohol consumption and dementia was too imprecise to be meaningful due to the extremely low prevalence of individuals who drink alcohol in the LASI-DAD cohort. High exposure to outdoor air pollution showed a positive but weak and insignificant association with dementia. It is possible that, given the extremely high and widespread levels of ambient air pollution in India and our definition of ‘high exposure,’ there is insufficient variability in the data to identify the adverse effect of high PM_2.5 concentrations on the likelihood of having dementia. In previous investigations, indoor air pollution was found to affect cognitive test performances in India negatively.23 Based on this finding, we considered indoor air pollution an additional Indian-specific dementia risk factor, but we did not observe a significant association with dementia (although the estimated relative risk was positive, as expected).

According to our preferred estimates based on Indian-specific risk factors’ definitions (when appropriate) and assuming observed associations are causal, eliminating all significant risk factors would potentially prevent up to 70% of dementia cases in India. This figure is substantially larger than previously reported in the literature for HICs. For instance, recent work for the USA estimated that including vision impairment as a risk factor might account for 60% of dementia cases,20 while the latest Lancet Commission’s report offers a more conservative global estimate of 45%.22

The overall PAF increased to 80% if we adopted only standard HICs definitions while still estimating relative risks, prevalence and communality in the LASI-DAD sample. This increase is largely explained by a much higher prevalence of education and social isolation risk factors under HICs definitions. When we only obtained prevalence and communality of risk factors—defined according to HIC standards—from our sample and took relative risk estimates from meta-analyses based on HIC data, the overall PAF was 64%. In this case, the higher prevalence of education and social isolation risk factors was offset by the adoption of lower relative risk values compared with those found in the LASI-DAD sample. Importantly, when using standard HIC definitions and relative risk estimates, the ranking of risk factors by PAF magnitude changed. While education remained the risk factor with the largest PAF, the second most important risk factor was social isolation instead of vision impairment. Physical inactivity fell from being the third most important risk factor to being the fifth. Smoking, which did not have a significant correlation with dementia in our sample, was the sixth most important risk factor. Our findings highlight that PAFs for India based on HIC definitions and/or relative risk estimates from HICs could lead to skewed assessments of dementia prevention potential and priority intervention areas in the Indian context.

When we used the same set of risk factors and applied a similar calculation method to that of Mukadam et al (2019) in table 3, our analysis yielded an overall PAF of 59%, compared with the 41% reported by Mukadam et al.7 This difference may be due to minor variations in the measurement of risk factors between the two studies and changes over time in both the prevalence of risk factors and dementia in India (with Mukadam et al’s data collected from 2003 to 2006, and ours in 2018 and 2019). However, considering that Mukadam et al focused on a sample of older individuals (aged 65 and above) from Tamil Nadu, the significant 18-percentage-point discrepancy highlights how using data from a geographically limited sample might lead to selectivity concerns that affect the outcome and hinder the generalisability of the results to the broader Indian population (eg, within the LASI sample, which is representative of each Indian state, respondents residing in Tamil Nadu are significantly less likely to live in a rural area: 41% vs 66% in the rest of the sample; similar figures are observed in the LASI-DAD sample).

Our study is subject to limitations. First, given the cross-sectional nature of our data, we are unable to establish a temporal relationship between the observed risk factors and dementia. Hence, caution should be exercised in interpreting estimated relative risks as causal, as they may be affected by reverse causation and survivor bias. This limitation is especially pertinent when considering weight-related measures. The positive association we found between being underweight and dementia might reflect reverse causality, where cognitive decline leads to weight loss or sarcopenia, rather than underweight being a direct causal factor for dementia. Similarly, the inverse association between obesity and dementia might partly be attributable to reverse causality, where dementia-related weight loss could lead to erroneously concluding that obesity is protective. This interpretation aligns with prior literature showing that BMI in middle age, but not older age, is associated with cognitive outcomes and that a steeper decline in BMI is linked to lower memory scores at subsequent time points.30 31 If the observed association between underweight and cognition were to stem entirely from reverse causation, our total PAF would likely be overestimated. However, the impact on overall results is modest: even if the causal association were null (that is, assuming we estimated a completely spurious relationship), the total PAF would be reduced by only about four percentage points. In our preferred estimates, based on Indian-specific definitions of risk factors, obesity was not statistically significant and, therefore, was not included in the final PAF calculation. This mitigates the potential bias of total PAF due to the questionable protective role of obesity observed in our cross-sectional analysis. If obesity is a true causal risk factor for dementia in India, its omission in the present analysis would lead to conservative estimates. A second limitation related to our cross-sectional data is the inability to identify whether the effects of risk factors vary at different points in the life course and, thus, to determine when preventive interventions may be most effective. Though such complexity is not typically considered in PAF analyses, there is some accumulating evidence that the effects of risk factors differ across life stages,32 and future studies should seek to integrate this aspect when assessing dementia prevention potential. Despite these issues with temporality and reverse causality due to the cross-sectional nature of our data, this study removes the critical assumption made in prior research that relative risks estimated using data from HICs apply to India. With time and the continuation of existing data collection efforts, future research will be able to use relative risks from longitudinal data collected in India. Third, using an older cohort’s prevalence rates can bias the estimated future prevention potential. However, the direction of this bias is unclear as it depends on whether younger/future cohorts exhibit increasing (eg, cardiovascular conditions) or decreasing (eg, no education) prevalence of risk factors and the relative contribution of these risk factors to the total PAF. Since the strongest association is between dementia and low education, and the prevalence of no formal education is higher in older cohorts, our total PAF likely overestimates future prevention potential in India. However, it is reasonable to assume that raising education levels should remain a top priority for dementia prevention, at least for the next decade. Fourth, although we used objective measures of hypertension, BMI, diabetes and vision impairment, typically not available in a large-scale national survey in India, we relied on self-reports for hearing impairment and physical inactivity (the use of self-reports for these outcomes, especially physical activity, is a limitation shared with most other studies in this literature). Hearing impairment based on previous diagnosis of hearing or ear-related problems/conditions can be problematic in the Indian context, where medical care utilisation is very unevenly spread across the population. An objective and reliable assessment of hearing impairment will be available in the second wave of LASI-DAD. Self-reports of physical inactivity are known to be subject to reference bias and may not necessarily reflect actual differences in physical activity across population groups.

In conclusion, our research underscores the importance of using nationally representative samples and context-specific estimates to compute PAFs for dementia risk factors in an LMIC such as India. Calculations based on geographically restricted samples and/or relative risk estimates from HICs might not accurately gauge the true potential for dementia prevention in the Indian context and prevent pinpointing the most crucial public health measures to prioritise. Estimates can be used to guide the prioritisation of resources to target risk factors, such as no education, vision impairment and physical inactivity, with the highest estimated PAFs.

Contributors: MA is the lead author involved in all aspects of the study, including literature review, study design, methodology, data, analysis and writing. EN, EM and ALG contributed to study design, data analysis, interpretation and writing. JL contributed to securing funding, study design, interpretation and writing. JE, KML and SDA contributed equally to result interpretation and writing of the manuscript. MV and AD contributed expertise about the Indian context, result interpretation and writing. MA is the guarantor and accepts full responsibility for the finished work and the conduct of the study, had access to the data and controlled the decision to publish.
Funding: We acknowledge funding from the United States National Institute on Aging, National Institute of Health (R01AG051125; U01AG058499; RF1AG055273).
Competing interests: None declared.
Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review: Not commissioned; externally peer reviewed.
Supplemental material: This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

Data availability statement

Data are available in a public, open access repository. Data are available to the public upon free registration as data users with the LASI and LASI-DAD studies.

Ethics statements

Patient consent for publication:

Ethics approval:

This study involves human participants. Written consents were obtained from each participating respondent and informant in the form of a signature or thumbprint. The LASI-DAD study received IRB approval at the University of Southern California (approval number: UP-15-00684) and the Indian Council of Medical Research (approval number: 2022-16741). Participants gave informed consent to participate in the study before taking part.

World Health Organization. Global status report on the public health response to dementia. 2021;
Available: here
Google Scholar
Lee J, Meijer E, Langa KM, et al. Prevalence of dementia in India: National and state estimates from a nationwide study. Alzheimers Dement 2023; 19:2898–912.
doi:10.1002/alz.12928•Google Scholar
Wolters FJ, Chibnik LB, Waziry R, et al. Twenty-seven-year time trends in dementia incidence in Europe and the United States: The Alzheimer Cohorts Consortium. Neurology (ECronicon) 2020; 95:e519–31.
doi:10.1212/WNL.0000000000010022•Google Scholar
Langa KM, Larson EB, Crimmins EM, et al. A Comparison of the Prevalence of Dementia in the United States in 2000 and 2012. JAMA Intern Med 2017; 177:51–8.
doi:10.1001/jamainternmed.2016.6807•Google Scholar
Livingston G, Sommerlad A, Orgeta V, et al. Dementia prevention, intervention, and care. Lancet 2017; 390:2673–734.
doi:10.1016/S0140-6736(17)31363-6•Google Scholar
Livingston G, Huntley J, Sommerlad A, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet 2020; 396:413–46.
doi:10.1016/S0140-6736(20)30367-6•Google Scholar
Mukadam N, Sommerlad A, Huntley J, et al. Population attributable fractions for risk factors for dementia in low-income and middle-income countries: an analysis using cross-sectional survey data. Lancet Glob Health 2019; 7:e596–603.
doi:10.1016/S2214-109X(19)30074-9•Google Scholar
Rodriguez JJL, Ferri CP, Acosta D, et al. Prevalence of dementia in Latin America, India, and China: a population-based cross-sectional survey. Lancet 2008; 372:464–74.
doi:10.1016/S0140-6736(08)61002-8•Google Scholar
Nair T. Challenges of hypertension and dementia in the Indian subcontinent: a review. J Hum Hypertens 2019; 33:568–74.
doi:10.1038/s41371-019-0205-z•Google Scholar
Cheng HG, Shidhaye R, Charlson F, et al. Social correlates of mental, neurological, and substance use disorders in China and India: a review. Lancet Psychiatry 2016; 3:882–99.
doi:10.1016/S2215-0366(16)30166-3•Google Scholar
Belessiotis‐Richards C, Livingston G, Marston L, et al. A cross‐sectional study of potentially modifiable risk factors for dementia and cognitive function in India: A secondary analysis of 10/66, LASI, and SAGE data. Int J Geriat Psychiatry 2022; 37.
doi:10.1002/gps.5661•Google Scholar
Bloom DE, Sekher TV, Lee J, et al. Longitudinal Aging Study in India (LASI): new data resources for addressing aging in India. Nat Aging 2021; 1:1070–2.
doi:10.1038/s43587-021-00155-y•Google Scholar
Lee J, Khobragade PY, Banerjee J, et al. Design and Methodology of the Longitudinal Aging Study in India-Diagnostic Assessment of Dementia (LASI-DAD). J Am Geriatr Soc 2020; 68 Suppl 3:S5–10.
doi:10.1111/jgs.16737•Google Scholar
Langa KM, Ryan LH, McCammon RJ, et al. The Health and Retirement Study Harmonized Cognitive Assessment Protocol Project: Study Design and Methods. Neuroepidemiology 2020; 54:64–74.
doi:10.1159/000503004•Google Scholar
Lee J, Banerjee J, Khobragade PY, et al. LASI-DAD study: a protocol for a prospective cohort study of late-life cognition and dementia in India. BMJ Open 2019; 9.
doi:10.1136/bmjopen-2019-030300•Google Scholar
Morris JC. The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology (ECronicon) 1993; 43:2412–4.
doi:10.1212/wnl.43.11.2412-a•Google Scholar
Lee J, Ganguli M, Weerman A, et al. Online Clinical Consensus Diagnosis of Dementia: Development and Validation. J Am Geriatr Soc 2020; 68 Suppl 3:S54–9.
doi:10.1111/jgs.16736•Google Scholar
Jorm AF. The Informant Questionnaire on cognitive decline in the elderly (IQCODE): a review. Int Psychogeriatr 2004; 16:275–93.
doi:10.1017/s1041610204000390•Google Scholar
Blessed G, Tomlinson BE, Roth M, et al. The association between quantitative measures of dementia and of senile change in the cerebral grey matter of elderly subjects. Br J Psychiatry 1968; 114:797–811.
doi:10.1192/bjp.114.512.797•Google Scholar
Ehrlich JR, Goldstein J, Swenor BK, et al. Vision impairment in a life course model of potentially modifiable dementia risk factors. Alzheimer's & Dementia 2023; 19.
doi:10.1002/alz.060166•Google Scholar
Ehrlich JR, Agarwal A, Young C, et al. The prevalence of vision impairment and blindness among older adults in India: findings from the Longitudinal Ageing Study in India. Nat Aging 2022; 2:1000–7.
doi:10.1038/s43587-022-00298-6•Google Scholar
Livingston G, Huntley J, Liu KY, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet 2024; 404:572–628.
doi:10.1016/S0140-6736(24)01296-0•Google Scholar
Adar S, D’Souza J, Shaddick G, et al. Higher Household Air Pollution Levels Correlate with Poorer Cognitive Function in the Longitudinal Aging Study in India (LASI). Alzheimer's & Dementia 2022; 18.
doi:10.1002/alz.059940•Google Scholar
Misra A, Chowbey P, Makkar BM, et al. Consensus statement for diagnosis of obesity, abdominal obesity and the metabolic syndrome for Asian Indians and recommendations for physical activity, medical and surgical management. J Assoc Physicians India 2009; 57:163–70.
Google Scholar
Steptoe A, Shankar A, Demakakos P, et al. Social isolation, loneliness, and all-cause mortality in older men and women. Proc Natl Acad Sci USA 2013; 110:5797–801.
doi:10.1073/pnas.1219686110•Google Scholar
World Health Organization. WHO global air quality guidelines. 2021;
Google Scholar
Central Pollution Control Board, Ministry of Environment, Forest & Climate Change, Government of India. National ambient air quality status and trends 2019. NAAQMS/45/2019-2020. 2020;
Google Scholar
Welberry HJ, Tisdell CC, Huque MH, et al. Have We Been Underestimating Modifiable Dementia Risk? An Alternative Approach for Calculating the Combined Population Attributable Fraction for Modifiable Dementia Risk Factors. Am J Epidemiol 2023; 192:1763–71.
doi:10.1093/aje/kwad138•Google Scholar
StataCorp. Stata statistical software: release 18. College Station, TX:, StataCorp LLC 2023;
Google Scholar
Suemoto CK, Gilsanz P, Mayeda ER, et al. Body mass index and cognitive function: the potential for reverse causation. Int J Obes 2015; 39:1383–9.
doi:10.1038/ijo.2015.83•Google Scholar
Crane BM, Nichols E, Carlson MC, et al. Body Mass Index and Cognition: Associations Across Mid- to Late Life and Gender Differences. J Gerontol A Biol Sci Med Sci 2023; 78:988–96.
doi:10.1093/gerona/glad015•Google Scholar
Marden JR, Tchetgen Tchetgen EJ, Kawachi I, et al. Contribution of Socioeconomic Status at 3 Life-Course Periods to Late-Life Memory Function and Decline: Early and Late Predictors of Dementia Risk. Am J Epidemiol 2017; 186:805–14.
doi:10.1093/aje/kwx155•Google Scholar

Overview

Abstract
Introduction
Methods
Results
Discussion
References
Supplementary files
Footnotes
Publication history
Rapid Responses

Article metrics

Altmetric

Dimensions