OBJECTIVE

Severe hypoglycemia is associated with an increased risk of dementia. We examined if the association is consistently present in mid- and late-life hypoglycemia.

RESEARCH DESIGN AND METHODS

Using health care data from Population Data BC, we created a base cohort of patients age ≥40 years with incident type 2 diabetes. Exposure was the first occurrence of severe hypoglycemia (hospitalization or physician visit). We assessed exposure versus no exposure in mid- (age 45–64 years) and late-life (age 65–84 years) cohorts. Index date was the later of the 45th birthday (midlife cohort), 65th birthday (late-life cohort), or diabetes diagnosis. Those with hypoglycemia or dementia before the index date were excluded. Patients were followed from index date until dementia diagnosis, death, emigration, or 31 December 2018. Exposure was modeled as time dependent. We adjusted for confounding using propensity score weighting. Dementia risk was estimated using cause-specific hazards models with death as a competing risk.

RESULTS

Of 221,683 patients in the midlife cohort, 1,793 experienced their first severe hypoglycemic event. Over a median of 9.14 years, 3,117 dementia outcomes occurred (32 among exposed). Of 223,940 patients in the late-life cohort, 2,466 experienced their first severe hypoglycemic event. Over a median of 6.7 years, 15,997 dementia outcomes occurred (158 among exposed). The rate of dementia was higher for those with (vs. without) hypoglycemia in both the mid- (hazard ratio 2.85; 95% CI 1.72–4.72) and late-life (2.38; 1.83–3.11) cohorts.

CONCLUSIONS

Both mid- and late-life hypoglycemia were associated with approximately double the risk of dementia, indicating the need for prevention throughout the life course of those with diabetes.

Type 2 diabetes has been reported as a potential modifiable risk factor for dementia (1,2). Beyond hyperglycemia and insulin resistance, diabetes is a heterogeneous condition with several chronic complications, some of which have been associated with vascular dementia, including stroke and hypertension (3). Additionally, severe hypoglycemia has been associated with an increased risk of all-cause dementia among patients with diabetes (411). This acute complication, characterized by a symptomatic decrease of blood glucose <3.9 mmol/L, primarily results from diabetes management. Specifically, it is an adverse effect of concern for exogenous insulin and insulin secretagogues, including sulfonylureas and meglitinides. The risk is heightened with intensive management and longer duration of use.

Several observational studies have linked hypoglycemia requiring hospitalization among patients with type 2 diabetes to dementia, using different data sources and methodologies (411). However, most of these studies have focused on the occurrence and number of hypoglycemic episodes, without considering the effect of age, except as a potential confounder. The potential for age to affect the association between hypoglycemia and dementia among adults with type 2 diabetes is an important knowledge gap (411). For example, among children and adolescents, studies have found that hypoglycemia had quantitatively different effects on cognitive function, depending on the age of exposure (12,13). The timing of hypoglycemic episodes might be of great relevance in children because of the critical role of glucose in providing energy needed for brain development (13). Nevertheless, similar work investigating if the effect of hypoglycemia on cognition varies by age of occurrence or time during an individual’s life course has not been conducted among adults with type 2 diabetes.

Plausibly, the pathophysiological damage and consequences resulting from hypoglycemia might not be equivalent across all adults because of aging-related variations in cerebral glucose metabolism, brain vulnerability, ability to compensate, and adaptation cascades (1417). Research on the association between risk of dementia and modifiable risk factors, such as depression or diabetes onset, has further assessed if the risk differs based on age at exposure (18,19). These studies concluded that the risk is not constant and tends to differ based on when the exposure occurs in an individual’s lifetime (18,19). For example, the risk of dementia seems to be higher when diabetes onset or depression occurs earlier in life (18,19). Therefore, it can be hypothesized that the risk of dementia differs based on the age of occurrence of hypoglycemia among adults.

Given the aforementioned uncertainties, understanding if the risk of dementia differs based on the age of occurrence of severe hypoglycemia could provide clinicians with insights to guide therapeutic management, including pharmacological options, drug doses, and HbA1c targets. Conversely, evidence of a consistent risk of dementia, independent of the age of hypoglycemia occurrence, could contribute to future work aimed at understanding the complex risk factors of dementia, in addition to clinical and public health implications. Specifically, identifying if earlier exposure to a hypoglycemic insult poses a similar risk of dementia could help direct clinical, educational, lifestyle, and public health measures targeting younger patients with diabetes.

Therefore, the objective of this study was to examine the association of mid- and late-life hypoglycemia with dementia.

Study Design and Data Source

We conducted a retrospective population-based cohort study using British Columbia (BC) health care data from 1 January 1996 to 31 December 2018 obtained from the health administrative databases within Population Data BC (https://www.popdata.bc.ca/data). Population Data BC provides data on health system encounters for nearly all the BC population, which receives universal health care coverage through the provincial government (2024).

We linked six data sets through deidentified personal health identification numbers to obtain several variables. Specifically, we obtained demographic variables, including sex, date of birth, and health care coverage from the population registry (Consolidation File) (20). We used variables related to outpatient prescriptions, including dispensation date and drug identification number, from the PHARMANET program (21). This database is a comprehensive source of all nonhospital prescription drugs dispensed by community pharmacies to BC residents regardless of insurance coverage (government sponsored, private, or out of pocket). We also obtained variables on physician visits, including service date ICD-9–Clinical Modification diagnosis code, from the Medical Services Plan database (22). Hospital admission–related variables, including up to 25 ICD-10–Canadian Adaptation (CA) diagnosis codes (ICD-9–Clinical Modification before 2002) and dates of admission and discharge, were captured from the Discharge Abstract Database (23). We accessed dates of death from the Vital Events Deaths database (24). Last, we acquired an area-level measure of socioeconomic status (SES) based on the first three characters of the postal code and aggregated neighborhood-level income data from Census Geodata (25).

Study Population

First, we identified a base cohort of patients with incident type 2 diabetes between 1 January 1998 and 31 December 2016. Incident diabetes was defined based on the diabetes case-defining algorithm from the Canadian Chronic Disease Surveillance System, whereby diabetes is defined as the earliest occurrence of two physician claims (ICD-9 codes) within a 2-year period or one hospitalization (ICD-10-CA) (26). This definition has been validated (positive predictive value 81.9%; negative predictive value 98.7%) and used in diabetes research using administrative data (26).

Second, we identified patients meeting the following inclusion criteria: 1) age ≥40 years at the date of diabetes onset, 2) continuously registered in the population registry for at least 2 years before diabetes onset, 3) not dispensed any antihyperglycemic agents before diabetes onset, 4) no history of a diagnostic code indicating type 1 diabetes and not dispensed insulin monotherapy as first-line treatment, 5) no previous record of diagnostic codes indicating dementia or cognitive impairment or dispensation record for a cholinesterase inhibitor before diabetes diagnosis, and 6) not diagnosed with Down syndrome, wherein there is a higher risk of diabetes and dementia with genetic variation that we are unable to assess. ICD codes used to identify diabetes and inclusion criteria are reported in Supplementary Table 1.

Exposure Assessment and Life-Course Subcohorts

Our exposure of interest was the occurrence of one or more severe hypoglycemic episodes, defined as at least one hospitalization or a physician claim indicating hypoglycemia without previous hospitalization or claim in the entire period of available data. ICD codes used to identify hypoglycemia are reported in Supplementary Table 1.

From the base cohort, we created two subcohorts. In the first, referred to as the midlife cohort, the cohort entry date was the date of a patient’s 45th birthday. We used an open-cohort design, wherein patients diagnosed with diabetes before their 45th birthday (i.e., prevalent diabetes) and patients diagnosed after their 45th birthday (i.e., incident diabetes) were included. The index date was the date of the 45th birthday or date of diabetes onset, whichever occurred last, to account for the delayed entry of patients with incident diabetes. Patients with diagnostic codes indicating history of hypoglycemia or cognitive impairment before the index date were excluded. Exposure to hypoglycemia was assessed from the index date until their 64th birthday. Specifically, patients who experienced one or more hypoglycemic episodes from the index date until their 64th birthday were considered exposed from the date of their first hypoglycemic episode, whereas patients who did not experience any hypoglycemic episode from the index date until their 64th birthday were considered unexposed. To avoid immortal time bias, person-time between the index date and occurrence of hypoglycemia was considered unexposed time.

In the second cohort, the late-life cohort, the cohort entry date was the date of a patient’s 65th birthday. All design elements were similar to those of the midlife cohort, including eligibility criteria of being hypoglycemia free and dementia free before the index date (65th birthday for prevalent diabetes or diabetes onset for delayed entry with incident diabetes). Therefore, the late-life cohort was not independent of the midlife cohort. Specifically, the late-life cohort consisted of those with incident diabetes after age 65 years in addition to those with prevalent diabetes who did not experience hypoglycemia, were not diagnosed with dementia, or were censored before age 65 years, therefore including individuals from the midlife cohort who met these criteria (Fig. 1). Exposure to hypoglycemia was assessed from the index date until their 84th birthday. Therefore, patients who experienced one or more hypoglycemic episodes from the index date to their 84th birthday were considered exposed from the date of their first hypoglycemic episode, whereas patients who did not experience any hypoglycemic episode from the index date until their 84th birthday were considered unexposed. Person-time was handled similarly to the midlife cohort to avoid immortal time bias.

Figure 1

Flowchart of cohort study. aBased on the Canadian Chronic Disease Surveillance System (one hospitalization or two physician claims within 2 years), without a code indicating type 1 diabetes. bAt any time before diabetes diagnosis, with a minimum of 2 years; patients (n = 204) may belong to more than one group. cLate-life cohort consists of those with incident diabetes age ≥65 years and those without hypoglycemia who did not have dementia or were censored before age 65 years.

Figure 1

Flowchart of cohort study. aBased on the Canadian Chronic Disease Surveillance System (one hospitalization or two physician claims within 2 years), without a code indicating type 1 diabetes. bAt any time before diabetes diagnosis, with a minimum of 2 years; patients (n = 204) may belong to more than one group. cLate-life cohort consists of those with incident diabetes age ≥65 years and those without hypoglycemia who did not have dementia or were censored before age 65 years.

Close modal

Outcome Definition

A diagnosis of incident all-cause dementia was defined using a validated algorithm (positive predictive value 80.4%; negative predictive value 99.0%) that requires one hospitalization code, three physician claims codes (at least 30 days apart in a 2-year period), or a prescription filled for a cholinesterase inhibitor (27). We restricted the outcome to all-cause dementia, including all subtypes, because of the difficulty in ascertaining dementia subtypes using administrative data. ICD codes used to identify all-cause dementia are reported in Supplementary Table 1.

Covariates

Our covariate assessment period was 1 year before the index date. Specifically, we used inverse probability of treatment weighting (IPTW) to create balance between those exposed and unexposed on an array of important confounders. To calculate the weights, we used a logistic model to estimate the propensity score or probability of exposure. The logistic model included 40 predefined variables that are potential confounders based on our clinical knowledge as well as previous observational studies assessing hypoglycemia and dementia (411), including 1) demographic variables (age, sex, and SES, defined as area-level income quintiles based on postal codes); 2) diabetes duration; 3) proxies for diabetes severity, such as macrovascular complications (ischemic heart disease, heart failure, hypertension, dyslipidemia, stroke, and peripheral vascular disease), microvascular complications (nephropathy, neuropathy, and retinopathy), antidiabetic treatment (metformin, sulfonylurea, thiazolidinedione, glucagon-like peptide 1 receptor agonist (GLP-1–RA), dipeptidyl peptidase 4 (DPP-4) inhibitor, sodium–glucose cotransporter 2 (SGLT-2) inhibitor, insulin, meglitinide, and acarbose), and treatment for macrovascular complications (ACE inhibitor, angiotensin receptor blocker, loop diuretic, thiazide diuretics, β-blocker, calcium channel blocker (CCB), and other antihypertensive); 4) other morbidities (Parkinson disease, Huntington disease, delirium, and anxiety/mood disorder); 5) other prescription drug use (antidepressant, antipsychotic, opioid, migraine medication, Parkinson medication, and antacid); and 6) index calendar year to account for any time trends. Additionally, we used the high-dimensional propensity score algorithm to identify a total of 500 empirical variables using five dimensions (hospitalizations, procedures, medical diagnoses, medical services, and prescription medication records). We stabilized the weights to reduce the variance associated with any extreme weights. Last, we assessed the balance of baseline covariates after weighting using absolute standardized differences (ASDs), with ASD >0.10 considered a significant imbalance.

Statistical Analysis

Patients were followed from the index date until the date of dementia diagnosis, death, emigration, end of provincial health coverage, or end of study period (31 December 2018), whichever occurred first. Exposure was treated as time varying, wherein person-time from the index date until exposure to hypoglycemia was considered unexposed person-time. Additionally, to minimize any reverse time causality, those diagnosed with dementia within 2 years after the index date or hypoglycemic episode were censored. This approach and lag period have been used previously in studies assessing risk factors for dementia (9,28).

We calculated the crude and weighted incidence rates of dementia by dividing the number of incident dementia cases over the total person-time in each cohort. We estimated the hazard ratios (HRs) of dementia and 95% CIs using a weighted cause-specific hazards model with death as a competing risk, which may have provided conservative estimates of the association between hypoglycemia and dementia. Failure to account for the competing risk of death when assessing diseases in older adults can lead to biased overestimated associations (29). We assessed model assumptions, including the proportional hazards assumption using Schoenfeld residuals. We used robust variance (sandwich estimator) to calculate the CIs.

In October 2007, the BC government introduced a reimbursement policy for cholinesterase inhibitors, which affected the number of physician visits with a diagnosis of Alzheimer disease in administrative data. To address this, we created and adjusted for a before/after variable to indicate if follow-up ended before or after October 2007 (30).

All analyses were conducted using SAS software (version 9.4; SAS Institute, Inc., Cary, NC).

Secondary and Sensitivity Analyses

As a secondary analysis, we assessed potential effect modification of the association between hypoglycemia and dementia in each cohort by sex, SES, and presence of diabetes micro- or macrovascular complications at baseline. Specifically, we ran three additional models that included interaction terms added to the main effect terms. For the SES model, those with unknown income quintile (roughly 2–3%) were excluded from the analysis. For each model, we tested the statistical significance (P < 0.05) of the interaction effect using the Wald test. If the interaction term was a significant predictor, we calculated the HR of dementia for each level of the potential effect modifier using a linear combination of parameters.

We conducted three types of sensitivity analyses. In the first sensitivity analysis, we increased the lag period, during which we censored dementia cases occurring after hypoglycemia, from 2 to 4 years. In the second sensitivity analysis, we broadened the hypoglycemia definition to include four additional ICD-10-CA codes, including nondiabetic hypoglycemia and hypoglycemia captured using five-digit level codes (Supplementary Table 1). In the third sensitivity analysis, we changed the age cutoff period used to define the mid- and late-life cohorts. Specifically, we changed the age cutoff period for the midlife cohort from age 45–64 to 50–64 and 55–64 years, and we changed it for the late-life cohort from 65–84 to 65–79 and 65–74 years. Raising the lower age limit at cohort entry requires less follow-up time to capture incident dementia. Conversely, lowering the upper limit helps account for delayed diagnoses and minimizes potential reverse causality. For example, an individual can have prodromal dementia that is not yet diagnosed at age 84 years, which could lead to a hypoglycemic event; lowering the upper age limit can help minimize this issue.

Data and Resource Availability

We obtained ethics approval from the University of Waterloo (Waterloo, Ontario, Canada). All data were deidentified, and no personal information was available at any point of the study. Access to data provided by the data steward(s) is subject to approval but can be requested for research projects through the data steward(s) or their designated service providers. All inferences, opinions, and conclusions drawn in this publication are those of the authors and do not reflect the opinions or policies of the data steward(s).

A total of 358,090 patients met the inclusion criteria and were included in the base cohort (Fig. 1).

Midlife Cohort

The midlife cohort included 221,683 patients, of whom 0.81% (n = 1,793) experienced one or more severe hypoglycemic episodes during the study period (Fig. 1). Baseline characteristics before weighting are reported in Supplementary Table 2. After weighting, patients in the midlife cohort who experienced hypoglycemia (exposed) and those who did not (unexposed) were comparable across all characteristics, including several expected to confound the association between hypoglycemia and dementia, such as insulin and sulfonylurea use, polypharmacy, and a range of macro- and microvascular complications (Table 1).

Table 1

Baseline characteristics of exposure groups after propensity score weighting in mid- and late-life cohorts

CharacteristicMidlife cohortLate-life cohort
Hypoglycemia (n = 1,793)No hypoglycemia (n = 219,890)ASDHypoglycemia (n = 2,466)No hypoglycemia (221,474)ASD
Age, years, mean (SD) 54.66 (5.60) 54.68 (6.17) 0.004 70.24 (5.43) 69.90 (5.78) 0.061 
Diabetes duration, years, mean (SD) 0.24 (0.91) 0.25 (0.88) 0.006 2.12 (3.85) 2.18 (3.71) 0.017 
Female sex, n (%) 791.09 (45.11) 95,314.30 (43.35) 0.040 1,108.90 (45.56) 102,073.00 (46.09) 0.001 
Socioeconomic status quintile, n (%)   0.087   0.106 
 1 (lowest) 401.44 (22.89) 47,841.50 (21.76)  554.18 (21.96) 48,617.20 (21.95)  
 2 390.08 (22.25) 47,197.00 (21.46)  463.44 (19.04) 46,738.70 (21.10)  
 3 322.31 (18.38) 43,911.90 (19.97)  475.28 (19.53) 43,212.80 (19.51)  
 4 293.69 (16.75) 40,859.30 (18.58)  508.07 (20.87) 40,787 (18.42)  
 5 (highest) 309.75 (17.66) 36,020.70 (16.38)  404.46 (16.62) 38,913.60 (17.57)  
 Missing 36.29 (2.07) 4,061.19 (1.85)  28.70 (1.18) 3210.27 (1.45)  
Health care use, n (%)       
N of distinct drugs in 100 days before index date   0.068   0.048 
  0 190.35 (10.86) 20,479.00 (9.31)  106.59 (4.38) 10,019.50 (4.52)  
  1 178.97 (10.21) 22,951.90 (20.44)  109.45 (4.56) 11,842.10 (5.35)  
  2 202.31 (11.54) 25,450.30 (11.57)  176.23 (7.24) 16,031.10 (7.24)  
  ≥3 1,181.93 (67.40) 151,010 (68.67)  2,041.42 (83.87) 183,587.00 (82.89)  
N of physician visits in 100 days before index date   0.001   0.014 
  0 21.27 (1.21) 2,716.90 (1.24)  29.89 (1.23) 2,725.67 (1.23)  
  1 8.99 (0.51) 1,391.65 (0.63)  12.75 (0.52) 955.44 (0.43)  
  2 17.64 (1.01) 1,500.80 (0.68)  10.87 (0.45) 995.89 (0.45)  
  ≥3 1,705.67 (97.27) 214,282.00 (97.45)  2,380.63 (97.74) 216,803 (97.89)  
N of hospitalizations in 100 days before index date   0.153   0.127 
  0 1,425.17 (81.27) 187,136.00 (85.10)  1,750.36 (71.91) 168,424.00 (76.05)  
  1 250.33 (14.28) 23,076.20 (10.49)  450.77 (15.31) 33,914.30 (15.31)  
  2 38.77 (2.21) 6,110.34 (2.78)  146.07 (6.00) 11,975.60 (5.41)  
  ≥3 39.29 (2.24) 3,569.25 (1.62)  86.93 (3.57) 7,165.48 (3.24)  
Comorbidities in year before index date, n (%)       
 Parkinson disease 230.16 0.035 11.17 (0.46) 1,029.52 (0.46) 0.001 
 Huntington disease 5.95 (<0.01) 0.007 8.90 (<0.01) 0.009 
 Delirium 9.75 (0.56) 419.39 (0.19) 0.060 12.87 (0.53) 1,010.22 (0.46) 0.010 
 Anxiety/mood disorder 640.10 (36.50) 70,883.00 (32.24) 0.089 1,009.99 (41.49) 78,779.30 (35.57) 0.121 
 Hypertension 632.49 (36.07) 82,354.40 (37.45) 0.029 1,258.27 (51.69) 112,408 (50.75) 0.018 
 Ischemic heart disease 129.08 (7.36) 18,748.80 (8.53) 0.043 388.10 (15.94) 35,276.40 (15.93) 0.001 
 Dyslipidemia 212.02 (12.09) 30,130.70 (13.70) 0.048 309.39 (12.71) 28,989.10 (13.09) 0.011 
 Heart failure 45.62 (2.60) 4,337.48 (1.97) 0.042 159.87 (6.57) 13,872.70 (6.26) 0.012 
 Stroke 29.77 (1.70) 3,048.91 (1.39) 0.025 94.69 (3.89) 8,392.92 (3.79) 0.005 
 Nephropathy 38.61 (2.20) 3,390.44 (1.54) 0.049 103.36 (4.25) 11,699.20 (5.28) 0.048 
 Neuropathy 26.41 (1.51) 3,085.81 (1.40) 0.009 46.66 (1.92) 3,687.58 (1.66) 0.018 
 Retinopathy 23.84 (1.36) 2,157.65 (0.98) 0.035 28.49 (1.17) 3,790.49 (1.71) 0.045 
 Peripheral vascular disease 32.30 (1.84) 3,435.02 (1.56) 0.022 75.02 (3.08) 7,144.62 (3.23) 0.008 
Use of medications in year before or on index date, n (%)       
 Antidepressants 357.74 (20.41) 40,364.30 (18.36) 0.052 467.06 (19.19) 35,541.80 (16.05) 0.082 
 Antipsychotics 353.02 (20.13) 38,523.80 (17.52) 0.067 562.93 (23.13) 45,811.80 (20.68) 0.059 
 Opioids 406.55 (23.18) 47,719.20 (21.70) 0.036 542.95 (22.31) 47,997.00 (21.67) 0.015 
 Migraine medications 18.78 (1.07) 2,621.91 (1.19) 0.011 9.06 (0.37) 1,191.74 (0.54) 0.025 
 Parkinson disease medications 31.55 (1.80) 1,953.56 (0.89) 0.079 36.07 (1.48) 2,537.63 (1.15) 0.029 
 Antacids 341.95 (19.50) 39,916.20 (18.15) 0.034 698.97 (28.72) 54,469.30 (24.59) 0.093 
 Metformin 427.32 (24.37) 55,622.1 (25.30) 0.021 735.16 (30.20) 61,733.80 (27.87) 0.051 
 Sulfonylureas 99.12 (5.65) 10,926.50 (4.97) 0.030 223.82 (9.20) 19,506.80 (8.81) 0.013 
 Thiazolidinediones 9.63 (0.55) 1,240.71 (0.56) 0.002 23.31 (0.96) 2,307.75 (1.04) 0.008 
 GLP-1–RAs 413.50 (0.19) 0.004 7.06 (0.29) 824.06 (0.37) 0.014 
 DPP-4 inhibitors 692.27 (0.31) 0.012 38.80 (1.59) 2,916.42 (1.32) 0.023 
 SGLT-2 inhibitors 328.19 (0.15) 0.053 11.63 (0.48) 799.15 (0.36) 0.018 
 Insulin 8.42 (0.48) 670.81 (0.31) 0.028 56.79 (2.33) 3,920.92 (1.77) 0.040 
 Meglitinides 200.62 (0.09) 0.016 396.43 (0.18) 0.007 
 Acarbose 218.28 (0.10) 0.025 400.82 (0.18) 0.001 
 Statins 366.53 (20.90) 54,334.70 (24.71) 0.091 984.99 (40.47) 93,983.60 (42.43) 0.040 
 ACE inhibitors 376.27 (21.46) 49,951.40 (22.72) 0.030 882.40 (36.25) 78,567.70 (35.47) 0.016 
 ARBs 159.52 (9.10) 18,988.20 (8.64) 0.016 370.73 (15.23) 34,679.00 (15.66) 0.011 
 Loop diuretics 60.14 (3.43) 6,367.64 (2.90) 0.031 224.94 (9.24) 17,280.60 (7.85) 0.050 
 Thiazide diuretics 254.23 (14.50) 32,553.60 (14.80) 0.009 519.20 (21.36) 50,856.60 (22.96) 0.038 
 β-blockers 217.58 (12.41) 28,922.80 (13.15) 0.022 604.95 (24.85) 53,966.70 (24.37) 0.011 
 CCBs 146.20 (8.34) 21,636.60 (9.84) 0.052 487.85 (20.04) 47,108.90 (21.27) 0.030 
 Other antihypertensives 22.93 (1.31) 1,926.96 (0.88) 0.042 24.33 (1.00) 3,080.56 (1.39) 0.036 
CharacteristicMidlife cohortLate-life cohort
Hypoglycemia (n = 1,793)No hypoglycemia (n = 219,890)ASDHypoglycemia (n = 2,466)No hypoglycemia (221,474)ASD
Age, years, mean (SD) 54.66 (5.60) 54.68 (6.17) 0.004 70.24 (5.43) 69.90 (5.78) 0.061 
Diabetes duration, years, mean (SD) 0.24 (0.91) 0.25 (0.88) 0.006 2.12 (3.85) 2.18 (3.71) 0.017 
Female sex, n (%) 791.09 (45.11) 95,314.30 (43.35) 0.040 1,108.90 (45.56) 102,073.00 (46.09) 0.001 
Socioeconomic status quintile, n (%)   0.087   0.106 
 1 (lowest) 401.44 (22.89) 47,841.50 (21.76)  554.18 (21.96) 48,617.20 (21.95)  
 2 390.08 (22.25) 47,197.00 (21.46)  463.44 (19.04) 46,738.70 (21.10)  
 3 322.31 (18.38) 43,911.90 (19.97)  475.28 (19.53) 43,212.80 (19.51)  
 4 293.69 (16.75) 40,859.30 (18.58)  508.07 (20.87) 40,787 (18.42)  
 5 (highest) 309.75 (17.66) 36,020.70 (16.38)  404.46 (16.62) 38,913.60 (17.57)  
 Missing 36.29 (2.07) 4,061.19 (1.85)  28.70 (1.18) 3210.27 (1.45)  
Health care use, n (%)       
N of distinct drugs in 100 days before index date   0.068   0.048 
  0 190.35 (10.86) 20,479.00 (9.31)  106.59 (4.38) 10,019.50 (4.52)  
  1 178.97 (10.21) 22,951.90 (20.44)  109.45 (4.56) 11,842.10 (5.35)  
  2 202.31 (11.54) 25,450.30 (11.57)  176.23 (7.24) 16,031.10 (7.24)  
  ≥3 1,181.93 (67.40) 151,010 (68.67)  2,041.42 (83.87) 183,587.00 (82.89)  
N of physician visits in 100 days before index date   0.001   0.014 
  0 21.27 (1.21) 2,716.90 (1.24)  29.89 (1.23) 2,725.67 (1.23)  
  1 8.99 (0.51) 1,391.65 (0.63)  12.75 (0.52) 955.44 (0.43)  
  2 17.64 (1.01) 1,500.80 (0.68)  10.87 (0.45) 995.89 (0.45)  
  ≥3 1,705.67 (97.27) 214,282.00 (97.45)  2,380.63 (97.74) 216,803 (97.89)  
N of hospitalizations in 100 days before index date   0.153   0.127 
  0 1,425.17 (81.27) 187,136.00 (85.10)  1,750.36 (71.91) 168,424.00 (76.05)  
  1 250.33 (14.28) 23,076.20 (10.49)  450.77 (15.31) 33,914.30 (15.31)  
  2 38.77 (2.21) 6,110.34 (2.78)  146.07 (6.00) 11,975.60 (5.41)  
  ≥3 39.29 (2.24) 3,569.25 (1.62)  86.93 (3.57) 7,165.48 (3.24)  
Comorbidities in year before index date, n (%)       
 Parkinson disease 230.16 0.035 11.17 (0.46) 1,029.52 (0.46) 0.001 
 Huntington disease 5.95 (<0.01) 0.007 8.90 (<0.01) 0.009 
 Delirium 9.75 (0.56) 419.39 (0.19) 0.060 12.87 (0.53) 1,010.22 (0.46) 0.010 
 Anxiety/mood disorder 640.10 (36.50) 70,883.00 (32.24) 0.089 1,009.99 (41.49) 78,779.30 (35.57) 0.121 
 Hypertension 632.49 (36.07) 82,354.40 (37.45) 0.029 1,258.27 (51.69) 112,408 (50.75) 0.018 
 Ischemic heart disease 129.08 (7.36) 18,748.80 (8.53) 0.043 388.10 (15.94) 35,276.40 (15.93) 0.001 
 Dyslipidemia 212.02 (12.09) 30,130.70 (13.70) 0.048 309.39 (12.71) 28,989.10 (13.09) 0.011 
 Heart failure 45.62 (2.60) 4,337.48 (1.97) 0.042 159.87 (6.57) 13,872.70 (6.26) 0.012 
 Stroke 29.77 (1.70) 3,048.91 (1.39) 0.025 94.69 (3.89) 8,392.92 (3.79) 0.005 
 Nephropathy 38.61 (2.20) 3,390.44 (1.54) 0.049 103.36 (4.25) 11,699.20 (5.28) 0.048 
 Neuropathy 26.41 (1.51) 3,085.81 (1.40) 0.009 46.66 (1.92) 3,687.58 (1.66) 0.018 
 Retinopathy 23.84 (1.36) 2,157.65 (0.98) 0.035 28.49 (1.17) 3,790.49 (1.71) 0.045 
 Peripheral vascular disease 32.30 (1.84) 3,435.02 (1.56) 0.022 75.02 (3.08) 7,144.62 (3.23) 0.008 
Use of medications in year before or on index date, n (%)       
 Antidepressants 357.74 (20.41) 40,364.30 (18.36) 0.052 467.06 (19.19) 35,541.80 (16.05) 0.082 
 Antipsychotics 353.02 (20.13) 38,523.80 (17.52) 0.067 562.93 (23.13) 45,811.80 (20.68) 0.059 
 Opioids 406.55 (23.18) 47,719.20 (21.70) 0.036 542.95 (22.31) 47,997.00 (21.67) 0.015 
 Migraine medications 18.78 (1.07) 2,621.91 (1.19) 0.011 9.06 (0.37) 1,191.74 (0.54) 0.025 
 Parkinson disease medications 31.55 (1.80) 1,953.56 (0.89) 0.079 36.07 (1.48) 2,537.63 (1.15) 0.029 
 Antacids 341.95 (19.50) 39,916.20 (18.15) 0.034 698.97 (28.72) 54,469.30 (24.59) 0.093 
 Metformin 427.32 (24.37) 55,622.1 (25.30) 0.021 735.16 (30.20) 61,733.80 (27.87) 0.051 
 Sulfonylureas 99.12 (5.65) 10,926.50 (4.97) 0.030 223.82 (9.20) 19,506.80 (8.81) 0.013 
 Thiazolidinediones 9.63 (0.55) 1,240.71 (0.56) 0.002 23.31 (0.96) 2,307.75 (1.04) 0.008 
 GLP-1–RAs 413.50 (0.19) 0.004 7.06 (0.29) 824.06 (0.37) 0.014 
 DPP-4 inhibitors 692.27 (0.31) 0.012 38.80 (1.59) 2,916.42 (1.32) 0.023 
 SGLT-2 inhibitors 328.19 (0.15) 0.053 11.63 (0.48) 799.15 (0.36) 0.018 
 Insulin 8.42 (0.48) 670.81 (0.31) 0.028 56.79 (2.33) 3,920.92 (1.77) 0.040 
 Meglitinides 200.62 (0.09) 0.016 396.43 (0.18) 0.007 
 Acarbose 218.28 (0.10) 0.025 400.82 (0.18) 0.001 
 Statins 366.53 (20.90) 54,334.70 (24.71) 0.091 984.99 (40.47) 93,983.60 (42.43) 0.040 
 ACE inhibitors 376.27 (21.46) 49,951.40 (22.72) 0.030 882.40 (36.25) 78,567.70 (35.47) 0.016 
 ARBs 159.52 (9.10) 18,988.20 (8.64) 0.016 370.73 (15.23) 34,679.00 (15.66) 0.011 
 Loop diuretics 60.14 (3.43) 6,367.64 (2.90) 0.031 224.94 (9.24) 17,280.60 (7.85) 0.050 
 Thiazide diuretics 254.23 (14.50) 32,553.60 (14.80) 0.009 519.20 (21.36) 50,856.60 (22.96) 0.038 
 β-blockers 217.58 (12.41) 28,922.80 (13.15) 0.022 604.95 (24.85) 53,966.70 (24.37) 0.011 
 CCBs 146.20 (8.34) 21,636.60 (9.84) 0.052 487.85 (20.04) 47,108.90 (21.27) 0.030 
 Other antihypertensives 22.93 (1.31) 1,926.96 (0.88) 0.042 24.33 (1.00) 3,080.56 (1.39) 0.036 

ACE, angiotensin converting enzyme; ARB, angiotensin receptor blocker; ASD, absolute standardized difference; CCB, calcium channel blocker; DPP-4, dipeptidyl-peptidase 4; GLP1-RA, glucagon-like peptide-1 receptor agonist; SGLT, sodium-glucose cotransporter; s, suppressed number <5 as per data provider requirements to ensure patient confidentiality/health privacy is maintained.

Over a median (interquartile range [IQR]) follow-up of 4.90 (6.70) years, 32 patients were diagnosed with dementia among those who experienced hypoglycemia, whereas 3,085 were diagnosed with dementia over 9.18 (7.81) years among those who did not experience hypoglycemia (Table 2). The weighted incidence rate of all-cause dementia was higher among those who experienced hypoglycemia (2.41; 95% CI 1.47–3.96 per 1,000 person-years) compared with those who did not (1.45; 95% CI 1.39–1.50 per 1,000 person-years). Results from survival models show that the occurrence of hypoglycemia in midlife was associated with a higher risk of dementia compared with occurrence in those without an episode of hypoglycemia (weighted HR 2.85; 95% CI 1.72–4.72) (Table 2).

Table 2

Association of first hypoglycemic episode in mid- and late life and incidence of dementia

ExposureTotal patientsN of eventsMedian (IQR) follow-up time, yearsIncidence rateHR (95% CI)P
CrudeaWeightedaCrudeWeightedbAdjustedcInteraction term with sexdInteraction term with SESdInteraction term with complicationsd,e
Lag period of 2 years (primary analysis)            
 Midlife cohort         0.3166 0.3399 0.3160 
  Hypoglycemia 1,793 32 4.90 (6.70) 3.11 (2.21–4.39) 2.41 (1.47–3.96) 3.30 (2.32–4.70) 2.85 (1.72–4.72) 2.88 (1.72–4.81)    
  No hypoglycemia 219,890 3,085 9.18 (7.81) 1.45 (1.40–1.50) 1.45 (1.39–1.50) 1.00 1.00 1.00    
 Late-life cohort         0.6445 0.2139 0.6630 
  Hypoglycemia 2,466 158 2.90 (5.02) 16.36 (14.06–19.03) 13.79 (10.63–17.88) 2.40 (2.04–2.82) 2.38 (1.83–3.11) 2.63 (2.01–3.45)    
  No hypoglycemia 221,474 15,839 6.80 (6.53) 9.45 (9.31–9.60) 9.48 (9.34–9.63) 1.00 1.00 1.00    
Lag period of 4 years (sensitivity analysis)            
 Midlife cohort         0.4355 0.3025 0.2450 
  Hypoglycemia 1,793 25 4.90 (6.70) 2.43 (1.65–3.59) 1.46 (0.88–2.41) 3.04 (2.04–4.53) 2.10 (1.26–3.51) 2.11 (1.24–3.57)    
  No hypoglycemia 219,890 2,833 9.18 (7.81) 1.33 (1.28–1.38) 1.33 (1.28–1.38) 1.00 1.00 1.00    
 Late-life cohort         0.6466 0.6157 0.8049 
  Hypoglycemia 2,466 107 2.90 (5.02) 11.08 (9.23–13.30) 9.08 (6.56–12.57) 2.25 (1.84–2.74) 2.35 (1.68–3.30) 2.56 (1.82–3.60)    
  No hypoglycemia 221,474 13,028 6.80 (6.53) 7.78 (7.65–7.91) 7.80 (7.67–7.93) 1.00 1.00 1.00    
Broad hypoglycemia definition (sensitivity analysis)            
 Midlife cohort         0.0685 0.7639 0.7296 
  Hypoglycemia 2,392 44 4.35 (6.58) 3.47 (2.58–4.65) 2.81 (1.82–4.35) 3.89 (2.88–5.26) 3.55 (2.28–5.53) 3.69 (2.36–5.77)    
  No hypoglycemia 219,284 3,067 9.16 (7.8) 1.44 (1.39–1.49) 1.44 (1.39–1.49) 1.00 1.00 1.00    
 Late-life cohort         0.9523 0.2655 0.8622 
  Hypoglycemia 3,960 222 2.33 (4.76) 16.22 (14.28–18.43) 12.35 (9.98–15.28) 2.52 (2.20–2.89) 2.22 (1.78–2.77) 2.53 (2.02–3.16)    
  No hypoglycemia 219,842 15,696 6.78 (6.52) 9.39 (9.25–9.54) 9.45 (9.30–9.59) 1.00 1.00 1.00    
ExposureTotal patientsN of eventsMedian (IQR) follow-up time, yearsIncidence rateHR (95% CI)P
CrudeaWeightedaCrudeWeightedbAdjustedcInteraction term with sexdInteraction term with SESdInteraction term with complicationsd,e
Lag period of 2 years (primary analysis)            
 Midlife cohort         0.3166 0.3399 0.3160 
  Hypoglycemia 1,793 32 4.90 (6.70) 3.11 (2.21–4.39) 2.41 (1.47–3.96) 3.30 (2.32–4.70) 2.85 (1.72–4.72) 2.88 (1.72–4.81)    
  No hypoglycemia 219,890 3,085 9.18 (7.81) 1.45 (1.40–1.50) 1.45 (1.39–1.50) 1.00 1.00 1.00    
 Late-life cohort         0.6445 0.2139 0.6630 
  Hypoglycemia 2,466 158 2.90 (5.02) 16.36 (14.06–19.03) 13.79 (10.63–17.88) 2.40 (2.04–2.82) 2.38 (1.83–3.11) 2.63 (2.01–3.45)    
  No hypoglycemia 221,474 15,839 6.80 (6.53) 9.45 (9.31–9.60) 9.48 (9.34–9.63) 1.00 1.00 1.00    
Lag period of 4 years (sensitivity analysis)            
 Midlife cohort         0.4355 0.3025 0.2450 
  Hypoglycemia 1,793 25 4.90 (6.70) 2.43 (1.65–3.59) 1.46 (0.88–2.41) 3.04 (2.04–4.53) 2.10 (1.26–3.51) 2.11 (1.24–3.57)    
  No hypoglycemia 219,890 2,833 9.18 (7.81) 1.33 (1.28–1.38) 1.33 (1.28–1.38) 1.00 1.00 1.00    
 Late-life cohort         0.6466 0.6157 0.8049 
  Hypoglycemia 2,466 107 2.90 (5.02) 11.08 (9.23–13.30) 9.08 (6.56–12.57) 2.25 (1.84–2.74) 2.35 (1.68–3.30) 2.56 (1.82–3.60)    
  No hypoglycemia 221,474 13,028 6.80 (6.53) 7.78 (7.65–7.91) 7.80 (7.67–7.93) 1.00 1.00 1.00    
Broad hypoglycemia definition (sensitivity analysis)            
 Midlife cohort         0.0685 0.7639 0.7296 
  Hypoglycemia 2,392 44 4.35 (6.58) 3.47 (2.58–4.65) 2.81 (1.82–4.35) 3.89 (2.88–5.26) 3.55 (2.28–5.53) 3.69 (2.36–5.77)    
  No hypoglycemia 219,284 3,067 9.16 (7.8) 1.44 (1.39–1.49) 1.44 (1.39–1.49) 1.00 1.00 1.00    
 Late-life cohort         0.9523 0.2655 0.8622 
  Hypoglycemia 3,960 222 2.33 (4.76) 16.22 (14.28–18.43) 12.35 (9.98–15.28) 2.52 (2.20–2.89) 2.22 (1.78–2.77) 2.53 (2.02–3.16)    
  No hypoglycemia 219,842 15,696 6.78 (6.52) 9.39 (9.25–9.54) 9.45 (9.30–9.59) 1.00 1.00 1.00    
a

Per 1,000 person-years.

b

IPTW.

c

IPTW adjusted for impact of policy change in cholinesterase inhibitor coverage in British Columbia.

d

Wald test.

e

Presence of diabetes micro- or macrovascular complications (nephropathy, neuropathy, retinopathy, ischemic heart disease, heart failure, hypertension, dyslipidemia, stroke, and peripheral vascular disease).

Late-Life Cohort

There were 223,940 patients who entered the late-life cohort, of whom 1.10% (n = 2,466) experienced one or more severe hypoglycemic episodes (Fig. 1). Baseline characteristics before weighting are reported in Supplementary Table 2. After weighting, patients who experienced hypoglycemia (exposed) and those who did not (unexposed) were comparable in the late-life cohort across all characteristics, including several important potential confounding variables, such as insulin and sulfonylurea use, polypharmacy, and a range of macro- and microvascular complications (Table 1).

Over a median follow-up of 2.90 (5.02) years, 158 patients were diagnosed with dementia among those who experienced hypoglycemia, whereas 15,839 were diagnosed with dementia over 6.80 (6.53) years among those who did not experience hypoglycemia (Table 2). The weighted incidence rate of all-cause dementia was higher among those who experienced hypoglycemia (13.79; 95% CI 10.63–17.88 per 1,000 person-years) compared with incidence in those who did not (9.48; 95% CI 9.34–9.63 per 1,000 person-years). Results from survival models show that the occurrence of the first hypoglycemic episode in late life was associated with a higher risk of dementia compared with occurrence in those without an episode of hypoglycemia (weighted HR 2.38; 95% CI 1.83–3.11) (Table 2).

Secondary and Sensitivity Analyses

Results from the secondary analyses did not indicate potential effect modification of the association of hypoglycemia and dementia by sex, SES, or presence of diabetes micro- or macrovascular complications at baseline in the mid- or late-life cohort (Table 2).

The overall findings of the primary and secondary analyses were consistent across a range of sensitivity analyses, which included changing the hypoglycemia definition, lag period, or age cutoff used to define the cohorts, although some estimates had wider CIs because of a smaller sample size (Table 2 and Fig. 2).

Figure 2

Weighted HRs and 95% CIs of dementia using different age cutoffs to define mid- and late-life cohorts.

Figure 2

Weighted HRs and 95% CIs of dementia using different age cutoffs to define mid- and late-life cohorts.

Close modal

Findings from this cohort study show that the risk of all-cause dementia is higher among those with at least one episode of severe hypoglycemia, regardless of the timing of the first serious hypoglycemic episode after age 45 years. Specifically, the risk of dementia is consistently more than doubled in patients with diabetes who experience hypoglycemia in mid- or late life compared with those who do not experience any severe hypoglycemia during that period. Moreover, this increased risk does not seem to differ for men compared with women or by residence in lower- compared with higher-income areas.

Because of the complex, paradoxical, and bidirectional pathophysiological relationship between hypoglycemia and cognitive impairment, we hypothesized that the association between severe hypoglycemia and dementia is not consistent across an individual’s life course. The association may not be significant, or may at least be attenuated, if hypoglycemia occurs during midlife. This is potentially due to higher brain reserve (brain structural or physiological premorbid capacity) or resistance to injury in midlife compared with in late life, as well as reduced compensatory and adaptive mechanisms of the aging brain (16,18,31,32). Findings from multiple studies show that older stroke patients have higher morbidity, including cognitive deficits, and poorer functional recovery than younger stroke patients (33). Similar to stroke, serious hypoglycemia is an acute event disrupting brain structure, hemostasis, and function; therefore, a similar scenario is plausible. However, our results did not support this hypothesis. In fact, our findings indicate that the long-term cognitive consequences of hypoglycemia are similar regardless of whether hypoglycemia occurs earlier or later in life, potentially indicating long-lasting damage.

Previous work has shown that earlier onset of chronic conditions, such as diabetes, depression, or hypertension, is associated with higher risk of dementia (19,20,34). This may be explained by the longer duration spent under chronic stress and the complications induced by these conditions. Severe hypoglycemia, however, is an acute insult and can lead to irreversible damage, including neuronal death, loss of gray matter volume, and cortical atrophy in areas involved with cognitive functions (12,35). Additionally, multiple mechanisms have been proposed linking glucose deprivation to an increase in amyloidogenesis and tau phosphorylation, both of which are important hallmarks in the pathophysiology of Alzheimer disease (36,37). It is also worth noting that severe hypoglycemia can lead to a proinflammatory state, characterized by increased platelet activation and decreased fibrinolysis, ultimately leading to a prothrombotic state in addition to increased blood pressure and changes in cardiac output and rhythm as well as stroke (38). Thus, beyond a direct insult to the brain, hypoglycemia can lead to long-term changes that might contribute to the pathophysiology of dementia. Therefore, additional neuropathology and imaging studies are needed to better understand both the immediate and delayed consequences of hypoglycemia for brain structure and function, while taking into account several concepts that relate to the aging brain process, including brain reserve, resistance, and compensation.

Clinically, these findings have implications that relate to diabetes care. Hypoglycemia is a concern for patients with diabetes receiving insulin or insulin secretagogues across all age groups; however, this concern is heightened among older adults and frail individuals. This heightened concern is attributed to lowered awareness of hypoglycemia among older adults and more severe consequences of hypoglycemia, including falls that are more debilitating for older adults (39). However, findings from this study provide insight into the equal necessity of preventing the occurrence of severe hypoglycemic episodes in adults in midlife. Additionally, this study highlights the future need to assess frailty across an individual’s lifetime and how frailty can affect diabetes management, the probability of experiencing hypoglycemia, and the risk of dementia in all ages.

Ultimately, identifying individuals who are at increased risk of hypoglycemia irrespective of their age can help clinicians determine an individualized tradeoff between the intensity of diabetes therapy and avoidance of the risk of a severe hypoglycemic event. This can be achieved through treatment optimization, reduction of inappropriate medication use, and lifestyle and educational interventions, with an overarching goal of preserving cognitive function and preventing, or at least delaying, dementia.

Strengths and Limitations

We believe our study to be one of the first to explore the association of mid- and late-life hypoglycemia with dementia using population-level data that span >20 years. The available literature suggests that the association between hypoglycemia and dementia is bidirectional and that hypoglycemia is not associated with dementia but rather is a prodromal symptom or an early manifestation of cognitive impairment. This argument is especially of concern in previous studies that included older adults only. However, results from our midlife cohort, in which the maximum age at which to assess hypoglycemia was set at 64 years, show this group was less prone to this issue and therefore provide a more definitive answer to establish an association between hypoglycemia and dementia. We also used a lag period approach, where we censored those diagnosed with dementia within 2 years of experiencing hypoglycemia. This approach minimized potential overestimation, as a result of reverse causality, of the risk of hypoglycemia leading to dementia. Moreover, we conducted secondary analyses to explore the role of two social determinants, sex and SES, as well as of the presence of diabetes complications.

Our study has limitations. First, we used health administrative data; therefore, misclassification of type 2 diabetes and dementia is possible. Such a misclassification would be expected to be nondifferential, and we used validated algorithms, when available, to minimize any misclassification bias. Second, our exposure was limited to very severe hypoglycemic episodes that would require hospitalization or a physician visit. Findings from the In-Hypo DM study showed the 1-year incidence proportion of severe hypoglycemia among patients with type 2 diabetes to be 38% (40). Therefore, we expect the number of those exposed in our study to have been underestimated, given that less severe hypoglycemia that could have been managed independently was not captured. However, misclassification of potential exposed individuals as unexposed would bias estimates toward the null. Importantly, we conducted a sensitivity analysis, in which we used a broader definition of hypoglycemia. Nevertheless, large population-level cohort studies in mid- and late life that include reporting of minor and moderate hypoglycemic episodes are warranted. Third, our outcome was limited to all-cause dementia, and we were not able to accurately differentiate between subtypes. Fourth, our median follow-up time among those who experienced hypoglycemia was roughly 5 and 3 years in the mid- and late-life cohorts, respectively. Given the long prodromal period of dementia that can last up to decades, we are unable to fully rule out the possibility of hypoglycemia as an early manifestation of dementia. Although we used multiple approaches to minimize reverse causality, including a lag period and an upper-age limit, reverse causality remains possible, given the bidirectional nature of the relationship between hypoglycemia and dementia. This issue of reverse causality would be more relevant to the late-life cohort than to the midlife cohort. Last, despite adjusting for >500 variables, we were not able to include important clinical indicators, such as HbA1c, although we included several indicators for diabetes severity, including duration, macro- and microvascular complications, and diabetes therapies. Moreover, data on genetic determinants of dementia, including apolipoprotein E, were not available. Additionally, data did not include a variable to indicate frailty, which has an impact on both hypoglycemia and dementia (39). Our data also lacked information on education, race/ethnicity, and lifestyle-related covariates, such as smoking and alcohol consumption. Thus, residual and unmeasured confounding remains possible, despite the use of a high-dimensional propensity score and IPTW.

Conclusion

Both mid- and late-life hypoglycemia were associated with a higher risk of dementia in this population-based cohort study using data spanning >20 years. These findings support the need to prevent hypoglycemia throughout the life course of patients with type 2 diabetes. Additionally, this finding indicates a possible long-lasting effect that can direct future research aimed at understanding the pathophysiological mechanisms by which severe hypoglycemia increases the risk of dementia.

This article contains supplementary material online at https://doi.org/10.2337/figshare.21561945.

Funding. This project was funded by the Mike & Valeria Rosenbloom Foundation Research Award through the Alzheimer’s Society of Canada.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. W.A. conducted all analyses and wrote the first draft of the manuscript. W.A. and J.-M.G. conceived the study idea. All authors contributed to the study design and contributed to and approved the final version of the article. W.A. and J.-M.G. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

1.
Livingston
G
,
Sommerlad
A
,
Orgeta
V
, et al
.
Dementia prevention, intervention, and care
.
Lancet
2017
;
390
:
2673
2734
2.
Rolandi
E
,
Zaccaria
D
,
Vaccaro
R
, et al
.
Estimating the potential for dementia prevention through modifiable risk factors elimination in the real-world setting: a population-based study
.
Alzheimers Res Ther
2020
;
12
:
94
3.
Bello-Chavolla
OY
,
Antonio-Villa
NE
,
Vargas-Vázquez
A
,
Ávila-Funes
JA
,
Aguilar-Salinas
CA
.
Pathophysiological mechanisms linking type 2 diabetes and dementia: review of evidence from clinical, translational and epidemiological research
.
Curr Diabetes Rev
2019
;
15
:
456
470
4.
Whitmer
RA
,
Karter
AJ
,
Yaffe
K
,
Quesenberry
CP
Jr
,
Selby
JV
.
Hypoglycemic episodes and risk of dementia in older patients with type 2 diabetes mellitus
.
JAMA
2009
;
301
:
1565
1572
5.
Yaffe
K
,
Falvey
CM
,
Hamilton
N
, et al.;
Health ABC Study
.
Association between hypoglycemia and dementia in a biracial cohort of older adults with diabetes mellitus
.
JAMA Intern Med
2013
;
173
:
1300
1306
6.
Chin
SO
,
Rhee
SY
,
Chon
S
, et al
.
Hypoglycemia is associated with dementia in elderly patients with type 2 diabetes mellitus: an analysis based on the Korea National Diabetes Program Cohort
.
Diabetes Res Clin Pract
2016
;
122
:
54
61
7.
Kim
YG
,
Park
DG
,
Moon
SY
, et al
.
Hypoglycemia and dementia risk in older patients with type 2 diabetes mellitus: a propensity-score matched analysis of a population-based cohort study
.
Diabetes Metab J
2020
;
44
:
125
133
8.
Zheng
B
,
Su
B
,
Price
G
,
Tzoulaki
I
,
Ahmadi-Abhari
S
,
Middleton
L
.
Glycemic control, diabetic complications, and risk of dementia in patients with diabetes: results from a large U.K. cohort study
.
Diabetes Care
2021
;
44
:
1556
1563
9.
Han
E
,
Han
KD
,
Lee
BW
, et al
.
Severe hypoglycemia increases dementia risk and related mortality: a nationwide, population-based cohort study
.
J Clin Endocrinol Metab
2022
;
107
:
e1976
e1986
10.
Alkabbani
W
,
Maxwell
CJ
,
Marrie
RA
,
Tyas
SL
,
Lega
IC
,
Gamble
JM
.
Hypoglycaemia and the risk of dementia: a population-based cohort study using exposure density sampling
.
Int J Epidemiol
.
1 September 2022 [Epub ahead of print]. DOI: 10.1093/ije/dyac168
11.
Huang
L
,
Zhu
M
,
Ji
J
.
Association between hypoglycemia and dementia in patients with diabetes: a systematic review and meta-analysis of 1.4 million patients
.
Diabetol Metab Syndr
2022
;
14
:
31
12.
He
J
,
Ryder
AG
,
Li
S
,
Liu
W
,
Zhu
X
.
Glycemic extremes are related to cognitive dysfunction in children with type 1 diabetes: a meta-analysis
.
J Diabetes Investig
2018
;
9
:
1342
1353
13.
Perantie
DC
,
Lim
A
,
Wu
J
, et al
.
Effects of prior hypoglycemia and hyperglycemia on cognition in children with type 1 diabetes mellitus
.
Pediatr Diabetes
2008
;
9
:
87
95
14.
Shen
X
,
Liu
H
,
Hu
Z
,
Hu
H
,
Shi
P
.
The relationship between cerebral glucose metabolism and age: report of a large brain PET data set
.
PLoS One
2012
;
7
:
e51517
15.
Park
DC
,
Reuter-Lorenz
P
.
The adaptive brain: aging and neurocognitive scaffolding
.
Annu Rev Psychol
2009
;
60
:
173
196
16.
Puente
EC
,
Silverstein
J
,
Bree
AJ
, et al
.
Recurrent moderate hypoglycemia ameliorates brain damage and cognitive dysfunction induced by severe hypoglycemia
.
Diabetes
2010
;
59
:
1055
1062
17.
Mattson
MP
,
Arumugam
TV
.
Hallmarks of brain aging: adaptive and pathological modification by metabolic states
.
Cell Metab
2018
;
27
:
1176
1199
18.
Xu
W
,
Qiu
C
,
Gatz
M
,
Pedersen
NL
,
Johansson
B
,
Fratiglioni
L
.
Mid- and late-life diabetes in relation to the risk of dementia: a population-based twin study
.
Diabetes
2009
;
58
:
71
77
19.
Barbiellini Amidei
C
,
Fayosse
A
,
Dumurgier
J
, et al
.
Association between age at diabetes onset and subsequent risk of dementia
.
JAMA
2021
;
325
:
1640
1649
20.
British Columbia Ministry of Health
.
Population Data BC: Consolidation File (MSP Registration & Premium Billing)
.
V2. Data Extract. Ministry of Health (2020). Accessed 10 May 2021. Available from https://www.popdata.bc.ca/data
21.
British Columbia Ministry of Health
.
Population Data BC: PharmaNet
.
V2. Data Extract. Data Stewardship Committee (2020). Accessed 10 May 2021. Available from https://www.popdata.bc.ca/data
22.
British Columbia Ministry of Health
.
Population Data BC: Medical Services Plan (MSP) Payment Information File
.
V2. Data Extract. Ministry of Health (2020). Accessed 10 May 2021. Available from https://www.popdata.bc.ca/data
23.
Canadian Institute for Health Information
.
Population Data BC: Discharge Abstract Database (Hospital Separations)
.
V2. Data Extract. Ministry of Health (2020). Accessed 10 May 2021. Available from https://www.popdata.bc.ca/data
24.
British Columbia Ministry of Health
.
Population Data BC: Vital Events Deaths
.
V2. Data Extract. Ministry of Health (2020). Accessed 10 May 2021. Available from https://www.popdata.bc.ca/data
25.
Wilkins
R
.
PCCF+ Version 5E User’s Guide. Automated Geographic Coding Based on the Statistics Canada Postal Code Conversion Files. Catalogue 82F0086-XDB
.
Ottawa, Ontario, Canada
,
Statistics Canada
,
2009
26.
Lipscombe
LL
,
Hwee
J
,
Webster
L
,
Shah
BR
,
Booth
GL
,
Tu
K
.
Identifying diabetes cases from administrative data: a population-based validation study
.
BMC Health Serv Res
2018
;
18
:
316
27.
Jaakkimainen
RL
,
Bronskill
SE
,
Tierney
MC
, et al
.
Identification of physician-diagnosed Alzheimer’s disease and related dementias in population-based administrative data: a validation study using family physicians’ electronic medical records
.
J Alzheimers Dis
2016
;
54
:
337
349
28.
Baek
YH
,
Lee
H
,
Kim
WJ
, et al
.
Uncertain association between benzodiazepine use and the risk of dementia: a cohort study
.
J Am Med Dir Assoc
2020
;
21
:
201
211.e2
29.
Berry
SD
,
Ngo
L
,
Samelson
EJ
,
Kiel
DP
.
Competing risk of death: an important consideration in studies of older adults
.
J Am Geriatr Soc
2010
;
58
:
783
787
30.
Fisher
A
,
Carney
G
,
Bassett
K
,
Maclure
KM
,
Dormuth
CR
.
Policy-induced selection bias in pharmacoepidemiology: the example of coverage for Alzheimer’s medications in British Columbia
.
Pharmacoepidemiol Drug Saf
2019
;
28
:
1067
1076
31.
Montine
TJ
,
Cholerton
BA
,
Corrada
MM
, et al
.
Concepts for brain aging: resistance, resilience, reserve, and compensation
.
Alzheimers Res Ther
2019
;
11
:
22
32.
Aron
L
,
Zullo
J
,
Yankner
BA
.
The adaptive aging brain
.
Curr Opin Neurobiol
2022
;
72
:
91
100
33.
Umarova
RM
,
Schumacher
LV
,
Schmidt
CSM
, et al
.
Interaction between cognitive reserve and age moderates effect of lesion load on stroke outcome
.
Sci Rep
2021
;
11
:
4478
34.
Corrada
MM
,
Hayden
KM
,
Paganini-Hill
A
, et al
.
Age of onset of hypertension and risk of dementia in the oldest-old: the 90+ Study
.
Alzheimers Dement
2017
;
13
:
103
110
35.
Suh
SW
,
Hamby
AM
,
Swanson
RA
.
Hypoglycemia, brain energetics, and hypoglycemic neuronal death
.
Glia
2007
;
55
:
1280
1286
36.
Lauretti
E
,
Praticò
D
.
Glucose deprivation increases tau phosphorylation via P38 mitogen-activated protein kinase
.
Aging Cell
2015
;
14
:
1067
1074
37.
Velliquette
RA
,
O’Connor
T
,
Vassar
R
.
Energy inhibition elevates beta-secretase levels and activity and is potentially amyloidogenic in APP transgenic mice: possible early events in Alzheimer’s disease pathogenesis
.
J Neurosci
2005
;
25
:
10874
10883
38.
Yang
S-W
,
Park
K-H
,
Zhou
Y-J
.
The impact of hypoglycemia on the cardiovascular system: physiology and pathophysiology
.
Angiology
2016
;
67
:
802
809
39.
Abdelhafiz
AH
,
McNicholas
E
,
Sinclair
AJ
.
Hypoglycemia, frailty and dementia in older people with diabetes: reciprocal relations and clinical implications
.
J Diabetes Complications
2016
;
30
:
1548
1554
40.
Ratzki-Leewing
A
,
Harris
SB
,
Mequanint
S
, et al
.
Real-world crude incidence of hypoglycemia in adults with diabetes: results of the InHypo-DM Study, Canada
.
BMJ Open Diabetes Res Care
2018
;
6
:
e000503
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/journals/pages/license.