In our large population-based retrospective cohort study of patients with type 2 diabetes, 40 years of age and older, we observed fewer events of incident dementia among SGLT-2i users compared with DPP-4i users; however, there was no statistical significant association between the risk of dementia and SGLT-2i use compared with DPP-4i use among people with type 2 diabetes. Our secondary analysis revealed that SGLT-2i use was associated with a reduced risk of dementia among patients aged 65 years or older and when sulfonylurea use was the comparator. Finally, our study findings also demonstrated SGLT-2i use was associated with a lower risk of MCI.

Proietti et al compared the risk of dementia among SGLT-2i users with non-users among elderly patients with mean age of 66.7 years in a retrospective cohort study. They found a 66% elevated risk of incident dementia among non-users of SGLT-2is compared with users of SGLT-2is.25 However, in this study, they included prevalent users and did not use an active comparator, which could lead to confounding and bias. The recent retrospective cohort study conducted by Wu et al, with the most comparable sample size to our study, found that the use of SGLT-2is compared with the use of DPP-4is was associated with a 20% reduction in the risk of dementia among elderly Ontario residents aged ≥66 years, over a mean follow-up of 2.8 years using an intention-to-treat approach.42 Their secondary analysis with an as-treated approach revealed a stronger association of SGLT-2is with lower dementia risk (aHR 0.66; 95% CI 0.57 to 0.76).42 This finding was consistent with our study, where we assessed the risk of dementia associated with SGLT-2i use among individuals aged ≥65 years of age. Similarly, a recent retrospective study using data from the Korea National Health Insurance Service database showed that SGLT-2i use was associated with a reduced risk of dementia compared with DPP-4i use (aHR 0.65; 95% CI 0.58 to 0.73) among individuals aged 40–69 years.43 Furthermore, in subgroup analysis of individuals by age, this study showed no significant association between SGLT-2i use and the risk of dementia among individuals aged <60 years.43 In two other retrospective cohort studies, Siao et al and Mui et al reported that the use of SGLT-2is was associated with lower risks of dementia, although their comparator group and age of the cohort were different.22 23 Similar to our study population, in Siao et al, the average age of participants was lower compared with other studies, which could have potentially influenced the study outcomes, including the relatively low HR reported when compared with prior studies.23 42 43 Mui et al observed a much larger reduction in the risk of dementia (HR 0.41, 95% CI 0.27 to 0.61) associated with SGLT-2i use when compared with DPP-4i use among elderly individuals in a Chinese population cohort (ie, average age of 61 years).23 None of these studies excluded patients with a history of MCI.

In summary, retrospective studies have found that the use of SGLT-2i is associated with a reduced risk of incidentdementia among elderly individuals with type 2 diabetes. The findings from our study are consistent with prior studies when we restrict our analysis to individuals aged ≥65 years. Although the association between SGLT-2i use and the risk of dementia compared with DPP-4i use among adults aged >40 years is not statistically significant, the low number of incident dementia cases in our study could be attributed to the exclusion of patients with a history of MCI, the inclusion of younger individuals (aged <65 years)and the relatively short follow-up period. In our study, the finding of a decreased risk of dementia associated with SGLT-2i use when sulfonylurea use was the comparator may be attributable to the cognitive decline associated with sulfonylurea use, as some studies have shown that sulfonylurea use is associated with an increased risk of dementia compared with the use of metformin and DPP-4i.44–47 Further studies are required to determine if sulfonylurea use is associated with cognitive function decline. Finally, our study assessed the risk of MCI as an outcome, which has not been assessed in prior studies. MCI is considered a prodrome of dementia. Our study finding suggests that SGLT-2i use may be associated with a reduced risk of MCI. Further studies are needed to confirm this finding.

There were a handful of prospective studies conducted on the association between SGLT-2is and cognitive function. In studies in which patients with MCI and dementia were included, the use of SGLT-2is was strongly associated with improvements in cognitive function measured by the Montreal Cognitive Assessment score.48–50 In contrast, studies conducted with patients with normal baseline cognitive function found no significant associations between the use of SGLT-2is and changes in cognitive function scores.51 52 This aligns with our study findings, as we excluded patients with a history of MCI or dementia before starting the follow-up. Indeed, in our sensitivity analysis, when the study cohort included individuals with a history of MCI, we found a significant decreased risk of dementia associated with SGLT-2i use, which is likely due to a larger sample size. However, it is also important to consider the potential of confounding by indication, where individuals with MCI are less likely to be prescribed SGLT-2is, as their use is associated with side effects such as genitourinary tract infections, volume depletion and worsening urinary incontinence, which may limit its use in this population.53

RCTs examining the association between SGLT-2i use and cognitive function are limited, with short follow-up durations, smaller sample sizes and often focusing on specific SGLT-2i molecules.50 51 54 Perna et al conducted an RCT among elderly patients with type 2 diabetes and observed that individuals treated with SGLT-2i did not have significant changes in cognitive performance during the 1 year of treatment.54 Other studies had a shorter follow-up period of 16 weeks51 and 6 months50 and did not show any significant association between SGLT-2i use and cognitive function. Thus, the absence of a cognitive effect from SGLT-2i treatment in previously conducted RCTs may be due to small sample sizes and short follow-up durations.

Our study has several strengths. A key strength of our study was that our cohort consisted of patients without any history of dementia or MCI, which allowed us to assess the effect size of the risk of developing dementia in a population with type 2 diabetes and no prior cognitive impairment. Another strength was the use of an active comparator. By using DPP-4is as a comparator, the baseline characteristics between the two user groups of these antidiabetic agents were more similar in terms of diabetes duration, comorbidities and other factors. We also rigorously adjusted for multiple potential confounders, including covariates associated with frailty, such as falls, being housebound and the presence of tremor.55 56 These covariates were balanced between the two groups after adjustment using PS fine stratification.

Our study has several limitations. The relatively short follow-up period in our study may have resulted in fewer incident dementia events, potentially leading to an underestimation of the association between SGLT-2i use and the risk of incident dementia among younger individuals (aged <65 years). Although we used an active comparator drug (ie, DPP-4i) and adjusted for 34 potential confounders to achieve good balance between SGLT-2i and DPP-4i users, there remains a possibility of residual confounding. For example, individuals using SGLT-2is were younger than those using DPP-4is, and despite adjusting for age, residual confounding related to age may still exist due to unmeasured factors. Another limitation of our study was the low number of dementia subtype events. The small number of cases of Alzheimer’s disease and vascular dementia precluded us from drawing any conclusions about their association. However, the total number of dementia events was very low among SGLT-2i users, so any stratified analysis by dementia type would have been challenging. Even if all dementias had been specified by subtype, the number of events would still have been limited. The study was conducted using UK data, and the findings may not be generalizable to other populations. Future studies are needed to assess the association between SGLT-2i use and the risk of dementia in different populations. Finally, although the MCI outcome definition used in this study has been shown to be accurate in a prior study,38 there may still be misclassification of this outcome, as MCI can be difficult to detect in the clinical setting. However, any misclassification of MCI would likely occur equally between SGLT2i and DPP-4i users, potentially biasing the results toward the null.