With the global population aging at a substantial and increasing pace, dementia has become a leading cause of morbidity and mortality (1). As of 2019, approximately 55.2 million people worldwide were living with dementia, of whom 60% were living in low- and middle-income countries (LMICs) (1, 2). This estimate is projected to rise to 139 million people by 2050 (1). Reducing or treating modifiable risk factors by 10% per decade until 2050 may prevent approximately 8.8 million cases of dementia, underscoring the importance of addressing potentially preventable dementia risk factors that can lead to paths for primary prevention initiatives (3, 4).

Type II diabetes is a recognized risk factor for cognitive decline and dementia in older adults (5–7). Findings from prospective observational data in high-income countries have reported a 56% elevated risk of dementia among those diagnosed with type II diabetes (7). There are several mechanisms which may underlie the association between type II diabetes and dementia, including neurotoxicity resulting from hyperglycemia, alterations in amyloid metabolism, increased risk of cerebrovascular disease, and a combination of these mechanisms (6). Additionally, insulin resistance, which is characteristic of type II diabetes, has been shown to be a risk factor of dementia via a surge of generating amyloid-beta in the brain (8). The relationship between diabetes and declining cognitive health is multifaceted and complex, thus further research is needed to understand how this risk factor may differ outside of high-income countries.

There is little research on the relationship between diabetes and dementia across LMICs such as India and China; two of the most populous countries in the world (9, 10). Diabetes is highly prevalent among these countries. India is estimated to have a population of 74.2 million people with diabetes (11) while China has an estimated population of 141 million with diabetes (12). Furthermore, as socioeconomic growth and industrialization occur at a rapid pace in these developing LMICs, epidemiologic and nutrition transitions may drastically influence shifts in health behaviors and outcomes that could deviate from what is observed in Western, high-income countries (13).

Epidemiologic transition encompasses population shifts in causes of mortality and morbidity from communicable diseases (driven by malnutrition and poor maternal conditions) to noncommunicable/chronic diseases, including type II diabetes (14). It has multifactorial implications for disease and health trends (14, 15). Drivers of this trend include increased life expectancy, urbanization, globalized food production, economic growth, and unhealthy lifestyles (such as physical inactivity and poor diet) (16). Nutrition transition also plays a role in driving excessive intake of nutrient poor foods through changes in food availability and increased purchasing power (14). Consequently, nutrition patterns shift from consuming diets with minimal processed food to consuming highly processed diets (14, 17). Exploring the relationship between type II diabetes and later-life cognitive health in India and China will provide insight into diabetes-related dementia risk in populations that are at their intermediate stages of epidemiologic and nutrition transitions.

In order to assess the relationship between diabetes and cognitive function cross-nationally, cognitive assessments need to be valid across diverse contexts. This includes varying educational, social, cultural and political environments (18). The Harmonized Cognitive Assessment Protocol (HCAP) is a neuropsychological test battery that provides comparable measures across countries through the US Health and Retirement Study (HRS) and its international partner studies (19). Both India and China have ongoing longitudinal aging studies that have administered HCAP to a subset of participants (20, 21). Therefore, India and China have comparable measures that allow for cross-national evaluations to better understand the association between diabetes and cognitive function.

Previous research has investigated the relationship between diabetes and cognitive health in India using data from a partner study, the Longitudinal Aging Study in India – Diagnostic Assessment of Dementia (LASI-DAD) (22). The researchers found that associations were different from findings in high-income countries, where rural respondents with diabetes had greater cognitive scores than respondents without diabetes and urban respondents had similar scores between those with diabetes compared to those without diabetes (22). We seek to build on these findings by comparing the association in India with that observed in China. To achieve this aim, we capitalize on the use of harmonized cognitive assessments, thus allowing for direct comparisons between two countries (18). Based on previous research (22) and differences seen in health patterns between high-income countries and LMICs, we hypothesized that presence of type II diabetes is associated with higher average cognitive scores, rather than lower average cognitive scores typically seen in high-income countries.

Characteristics of participants with and without type II diabetes, for both samples, are shown in Table 1. The prevalence of type II diabetes in the LASI-DAD sample (N = 4,062) was 21.1% (n = 858). In CHARLS-HCAP (N = 9,741), the prevalence of type II diabetes was 15.6% (n = 1,506). LASI-DAD and CHARLS-HCAP samples mainly consist of participants aged 60 to 69 years (57.8 and 62.8%, respectively), were female (54.1 and 50.8%), had no education or early childhood education (62.6 and 50.3%), and resided in rural areas (62.0 and 60.7%). In India, participants with type II diabetes were more likely to live in urban areas (54.9%) compared to those without diabetes (33.5%). This pattern differed in China, where participants with type II diabetes (52.1%) and without (62.3%) resided in rural areas. In both samples, participants with abdominal obesity were more prevalent among participants with diabetes (79.2 and 79.3%).

General cognitive score distribution plots by diabetes status, area of residence and educational attainment are shown in Figure 1 for LASI-DAD and Figure 2 for CHARLS-HCAP. For LASI-DAD, the mean general cognitive score among people with diabetes was −1.20 with a standard deviation (SD) of 0.88 and without diabetes was −1.54 (SD: 0.86). For CHARLS, the mean general cognitive score among people with diabetes and without diabetes was −1.30 (SD: 0.98) and −1.36 (SD: 0.96), respectively. For both study samples, participants with diabetes consistently had greater mean general cognitive scores across area of residence groups and education attainment groups compared to participants without diabetes.

The current study aimed to evaluate the association between type II diabetes and general cognition in India and China, the two largest LMICs. No association was found between type II diabetes and cognition in fully adjusted models for both studies. However, there was a positive association between diabetes and cognition among older adults with none or early childhood education in India. Further investigation determined that area of residence and educational attainment modified the association. In India, participants living in rural areas with diabetes had better cognitive scores compared to those without diabetes. Alternatively, there was no significant difference seen among urban residents. In both India and China, diabetes was associated with worse cognitive scores among participants who have less than lower secondary education compared to those with none/early childhood education. These relationships differ from high income countries. Although further investigation through longitudinal studies is needed, these findings reveal the possible impact of the epidemiologic and nutrition transition in LMICs. Therefore, this study highlights important implications of type II diabetes as a potentially preventable risk factor for late-life cognitive decline in two of the largest LMICs.

Studies on the relationship between type II diabetes and late-life cognition have produced various findings. One study incorporated three cross-sectional surveys, including 10/66 Dementia Research Group survey, Study of global ageing [SAGE], and LASI pilot survey, focused on modifiable risk factors for cognitive function and dementia (35). Similar to our study’s findings, diabetes was associated with better cognitive performance in LASI and SAGE but not in 10/66 data (35). Previous research using CHARLS also ultimately determined no association seen between diabetes and cognition (specifically episodic memory) among participants 60 years and older, after adjusting for cardiovascular factors (4). However, research conducted in Brazil (36) and Mexico (37) have similar associations between type II diabetes and cognition as the relationships seen in high-income countries. Therefore, as research in this area produces differing findings, it is all the more reason to continue investigating this relationship in LMICs.

The differences seen between urban versus rural residents and with varying levels of education in India and China may indicate changes in health patterns that reflect differing stages of the epidemiologic transition and nutrition transition. According to Mattei et al. (14), India and China are both considered countries that are in the ‘ongoing transition’ phase of the epidemiologic transition. This means that like countries in the ‘early transition’ phase, India and China have steep declines in age-adjusted mortality, however they also have higher lifer expectancy and higher median age since these countries possess more modern health care and stronger economic growth (14). During the ‘ongoing transition’ phase, the burden of non-communicable diseases can be predisposed by malnutrition or poor maternal conditions during fetal and early childhood development and then exacerbated through unhealthy lifestyles encouraged by factors, such as urbanization and poor diet, in adulthood (14, 38, 39). Limited health care access and resources further strengthens the burden of non-communicable diseases seen in ‘ongoing transition’ phase countries (14, 40).

In India, area of residence modified the relationship between diabetes and cognitive function, such that the difference in cognitive score between diabetes status is greater for those who live in rural areas compared to those in urban areas. This difference likely reflects the accumulated effect of factors involved in the epidemiologic and nutrition transition effect, as described earlier. Urban residents live in environments that feature economic growth, increased access to highly processed foods, and reduced physical activity, thus encouraging high levels of obesity (14, 41). In addition to a genetic predisposition for type II diabetes in the Indian population (42), living in urban areas may intensify the severity of type II diabetes and related complications, including dementia, thus resulting in worse cognitive outcomes in older adulthood. Additionally, we must consider the influence of access to health care facilities and selection effects. Health care facilities for general health as well as reliable screening and treatment resources for type II diabetes are often unavailable in rural areas (42). Thus, predominant causes of morbidity and mortality in rural areas may reflect what is typically seen in ‘early transition’ countries, where poor sanitation, food insecurity, and communicable diseases are predominant drivers of health (14). This health trend is likely represented in the study sample, as older adult participants with diabetes, especially those who live in rural areas, may be biologically or socioeconomically advantageous, thus having better cognitive outcomes.

An interesting interaction effect between education attainment and diabetes was found in India and China. In both studies, diabetes was associated with better cognitive scores in the none/early childhood education group (significantly so in India) but statistically significantly worse cognitive scores in the group with less than lower secondary education attainment. Based on pattern shifts that are characteristic of the ongoing transitional phase in epidemiological and nutritional transitions (14, 16), participants who have completed up to lower secondary education attainment may be a demographic group who are most vulnerable by the ‘ongoing transition’ phase. These individuals may experience malnutrition or poor health care conditions during early childhood and adolescent years. However, during adulthood they may seek areas with increased employment opportunities, thus encouraging migration to regions with rapid urbanization and engaging in overconsumption of highly refined foods as well as a sedentary lifestyle (43–45). In contrast, older adults who had none or early childhood education likely reside in areas where lifestyles and employment are labor intensive, albeit they likely had less access to health care. While older adults who obtained upper secondary or higher education are indicative of those with higher socioeconomic status, had more access to health care and are associated to have higher late-life cognitive scores (46). Although, these individuals likely sought employment opportunities in urbanized areas that resulted in sedentary behavior and poor diet. Considering the collective effect of risk factors that encourage poor type II diabetes control as well as reduced protective effects of education attainment, general cognition of older adults with less than lower secondary education and diabetes may have poorer outcomes in contrast to those with none or early childhood education attainment. Future directions in research should further investigate these possibilities in India and China.

While associations of type II diabetes with general cognitive performance among participants with confirmed diabetes (HbA1c ≥ 6.5%) were null (Table 3), it does suggest participants who do not self-report a diagnosis have worse general cognitive outcomes compared to those with a diagnosis in India. This may be attributable to limited access to health care and could be manifested by lack of diabetes education and awareness, inaccessible treatments, maintenance of poor lifestyle behaviors, and reduced monitoring of overall health (47). Together these factors could raise the risk of type II diabetes complications, including long-term effects on cognitive impairment and dementia. Furthermore, given that only 5% of the CHARLS-HCAP sample were undiagnosed with type II diabetes and no association was found between self-reported type II diabetes and cognition, this may indicate that the lack of access to health care is less problematic in China compared to India. In comparison to China, the current associations we see in India may be more likely due to disparities in health care access. Whereas the association between type II diabetes and late-life cognitive health may be more attributable to the epidemiologic and nutrition transition in China compared to India. However, we strongly recommend further investigation to confirm these differences, which could inform health care policy and support interventions that address inaccessible health care or improve urban lifestyles within India and China.

There are several strengths to this study. First, we utilized harmonized variables to facilitate more direct comparisons of associations between countries. We also integrated self-report and blood-based biomarker (HbA1c) data to determine presence of type II diabetes, mitigating potential misclassification bias due to undiagnosed individuals. Furthermore, LASI-DAD (48) and CHARLS-HCAP (49) used venous HbA1c samples. In comparison to dried blood spot assays, using venous blood samples allows for more direct result interpretation whereas blood spot samples would require further adjustments for interpretation and various parameters can influence the accuracy of dried blood spot analysis (49, 50). This study was able to deduce associations using nationally representative samples. Both LASI-DAD and CHARLS-HCAP sampled from almost all states/regions in India and China, respectively. Another strength is that LASI-DAD and CHARLS-HCAP studies collected data recently, thus the current study provides updated analyses on type II diabetes status and general cognition.

Despite these strengths, this study also has its limitations. While incorporating blood biomarker data (HbA1c levels) is a strength, a limitation of HbA1c levels is that it can be affected by conditions such as iron deficiency anemia and may skew findings (51). Since cross-sectional data was used, these analyses were unable to capture changes in cognitive function over time, and temporality between the exposure and outcome could not be considered. Furthermore, given that participants had to survive until 60 years-old to participate in the study, this study was unable to establish and compare how the type II diabetes and cognitive score association may differ in younger age groups. In addition, data on diabetes treatments were unavailable or incomplete for many participants.

To build upon our findings, future research is needed to further understand the association between diabetes and cognition using longitudinal cognitive data. Longitudinal studies can incorporate sub-classes of diabetes status (i.e., no diabetes, untreated diabetes, controlled diabetes and treated but uncontrolled diabetes) to assess differences in treatment effects as well as differences in diabetes duration. Given the large number of diverse communities and sub-populations present in these LMICs, future research should also further examine social and health care differences that influence diabetic outcomes and consequently its association with late-life cognitive health. This might include exploring differences between specific geographic areas (ex. hukou systems in China) or by indicators of SES, including caste and household income in addition to education attainment. This will provide additional insight of how the epidemiologic transition and inequalities in health care access, or an education may influence the association between type II diabetes and cognition. Cross-national comparisons can then be determined between socioeconomically similar regions across India and China.

In conclusion, the present study identified overall positive associations between type II diabetes and general cognitive scores in India and China, which deviates from that observed in high-income countries. This difference may indicate ongoing effects of epidemiologic and nutrition transitions seen in both India and China. Area of residence and education attainment modified the relationship between type II diabetes and cognitive function. Urban and rural differences that contribute to the epidemiologic and nutrition transitions, such as urbanization, lifestyle, and possible health care access, may be underlying factors that explain these relationships in India. In both countries, diabetes may have an adverse effect on late-life cognition for socially and financially vulnerable populations who have less than lower secondary education attainment. Future longitudinal research is recommended to determine the underlying factors of these relationships in order to inform policy makers, public health interventions and educators on mitigating the risk of cognitive decline in older adults with diabetes.