Determinants of Cognitive Change Over 1 Year in Older Adults With Type 1 Diabetes

Abstract

OBJECTIVE

Older adults with type 1 diabetes (T1D) have a greater risk for cognitive decline, highlighting the need to identify risk factors for meaningful cognitive change using brief screening tools suitable for annual wellness examinations.

RESEARCH DESIGN AND METHODS

This study examined determinants of 1-year cognitive change in 74 older adults with T1D (mean age 70 ± 4 years; 47% female). Cognitive outcomes included the Montreal Cognitive Assessment (MoCA) and the Processing Speed Index (PSI) from the Wechsler Adult Intelligence Scale–Fourth Edition (WAIS-IV). Bivariate correlations and multiple regression analyses evaluated demographic, vascular, and diabetes-related predictors.

RESULTS

The mean MoCA change score was 0.1 (range −5 to 7; 23% with >1.73-point lower scores after 12 months), and the mean change on the WAIS-PSI was 2.5 (range −14 to 28). Greater MoCA decline was seen in participants with neuropathy (b = −1.19; SE_b = 0.55; 95% CI −2.29 to −0.09; P = 0.04), partnered status (b = −1.15; SE_b = 0.50; 95% CI −2.15 to −0.16; P = 0.02), and hypertension (b = −1.12; SE_b = 0.50; 95% CI −2.13 to −0.12; P = 0.03), each reflecting an ∼1.1- to 1.2-point greater change relative to those without. Interval severe hypoglycemia (b = −6.73; SE_b = 2.75; 95% CI −12.23 to −1.24; P = 0.02) and hypertension (b = −5.74; SE_b = 2.02; 95% CI −9.77 to −1.70; P = 0.01) were associated with WAIS-IV PSI change scores.

CONCLUSIONS

Partnered status, older age, hypertension, neuropathy, and interval severe hypoglycemia predicted 1-year cognitive decline detectable with brief cognitive screening and processing speed measures, underscoring the importance of cognitive assessment and monitoring in older adults with T1D who present with these risk factors.

Download audio file ^{(63.8MB, mp3)}

Graphical Abstract

Introduction

Older adults with type 1 diabetes (T1D) show higher rates of clinically significant cognitive impairment (1,2) and dementia (3,4). Cognitive deficits in T1D are most often in the domains of psychomotor speed, information processing, attention, and executive functioning, with a general preservation of learning and memory (2,5–7). Demographic (8–10), diabetes-related (5,11), and vascular (12–14) factors are associated with cognitive impairment, decline, and/or dementia in older adults with T1D.

In response to this known cognitive risk, the American Diabetes Association “Standards of Care in Diabetes” recommends annual cognitive screening for people with diabetes starting at age 65 years (15). Few studies have investigated the use of cognitive screening measures in this population (7), however, with none reporting data from annual assessments (9,16). Available longitudinal studies of older adults with T1D have provided important information about potential risk factors for cognitive change and the need to better understand the timing of this change.

The Diabetes Control and Complications Trial/ Epidemiology of Diabetes Interventions and Complications (DCCT/EDIC) has followed a cohort of adults with T1D for >32 years, starting in early adulthood (8). When the cohort was approaching older adulthood (mean age 59 years), microvascular complications, systolic blood pressure, severe hypoglycemia (SH) events, and HbA_1c were the strongest independent predictors of cognitive decline, with those in the most at-risk group for HbA_1c, SH, and systolic blood pressure having the equivalent of 9 years of accelerated cognitive aging (8). However, the interval between assessments was 14 years. Longitudinal studies over shorter time intervals have had mixed results (9,16). A study of older Japanese adults with T1D followed for 3 years found no significant global change on Japanese language versions of the Montreal Cognitive Assessment (MoCA) and Mini-Mental State Examination, although longer diabetes duration was associated with greater MoCA decline (16). Generalizability of this study is limited by its small sample size (n = 28) and single-center design (16). In secondary analyses from the Wireless Innovation for Seniors with Diabetes Mellitus (WISDM) trial using the National Institutes of Health Toolbox Cognition Battery at baseline and after 1 year, nearly 50% of participants were classified as having mild cognitive impairment (MCI) at baseline, and 20% converted from normal to impaired cognitive status after 1 year (9). However, 44% of participants with cognitive impairment reverted to normal cognitive status at follow-up, raising concerns about the stability of cognitive impairment classification over this timeframe (9).

Identifying demographic, glycemic, and clinical factors associated with longitudinal cognitive decline may improve dementia risk stratification and guide cognitive screening efforts in older adults with T1D. In the general population, higher educational attainment is associated with lower risk for cognitive decline and dementia (17), although education as a protective factor has been understudied in T1D. Evidence for sex differences are mixed; some studies suggested that older men with T1D perform worse than women in some cognitive domains (18), but overall impairment rates do not differ by sex (9,10). Diabetes-related factors, including HbA_1c, diabetic ketoacidosis (DKA), and SH have been associated with cognitive impairment and decline in older adults with T1D (1,2,8). Continuous glucose monitoring (CGM)-derived glycemic metrics are understudied in relation to cognitive outcomes in older adults, though one cross-sectional study found that higher average CGM nocturnal glucose and HbA_1c were independently associated with increased risk for cognitive impairment (2). Vascular comorbidities common in long-standing T1D, including microvascular complications, hypertension, and cardiovascular disease, have similarly been linked to poor cognitive outcomes (8,13,14,19). While age-related multimorbidity often leads to polypharmacy (20) and greater exposure to anticholinergic medications, the impacts of anticholinergic burden (ACB) and polypharmacy on cognitive outcomes in older adults with T1D remain largely unexplored, despite evidence linking high ACB to cognitive dysfunction and dementia in the general population (21,22).

Collectively, these findings suggest that cognitive decline and impairment are common and associated with known risk factors in older adults with T1D, yet the nature of and risk for cognitive change over clinically relevant periods are not well characterized. The current study investigated cognitive change over a 1-year period using a cognitive screening tool (MoCA) recommended for use in older adults with diabetes (7,15), as well as a measure of psychomotor processing speed (due to concerns that the MoCA alone may not be sensitive to changes in psychomotor efficiency [7]). We leveraged a sample of older adults with T1D who participated in a crossover clinical trial of automated insulin delivery that included robust baseline assessment of demographics, diabetes-related characteristics, medical comorbidities, and medication use, as well as detailed CGM data and medical monitoring over a 12-month interval (23,24). The testing interval and use of brief cognitive screening measures align with routine clinical care in this population.

Research Design and Methods

Participants and Procedures

This study used data from the Automated Insulin Delivery in Elderly With Type 1 Diabetes (AIDE T1D) trial (23,24). This was a multicenter, randomized, crossover trial during which participants each completed 12 weeks of hybrid closed-loop, predictive low-glucose suspend, and sensor-augmented pump insulin delivery in a randomized order, followed by a 3-month extension phase during which participants used the insulin delivery method of their choice (91% chose hybrid closed-loop delivery). Individuals with T1D aged ≥65 years were recruited at four clinical sites (Joslin Diabetes Center at SUNY, AdventHealth, University of Pennsylvania, and Mayo Clinic Rochester) that were coordinated by the Jaeb Center for Health Research between September 2020 and December 2022. Participants met inclusion criteria if they spoke English, had a clinical diagnosis of T1D for ≥1 year, and had an HbA_1c <10.0% obtained within the past 6 months. Individuals with a clinical diagnosis of dementia or suspected severe cognitive impairment, as evidenced by a score of ≤6 on the 5-min MoCA, were excluded from participation as the severity of cognitive impairment may preclude the ability to understand the study or use devices safely. See Kudva et al. (24) for greater detail regarding study methodology, eligibility criteria, and outcomes. Trial outcomes did not differ by cognitive status at baseline. Participants were included in the current analyses if the above criteria were met, and cognitive testing was completed at both the initial study visit and 12-month follow-up extension phase visit.

Data Collection

Baseline Variables

Demographic and diabetes-related variables were collected at baseline, including marital status, age, sex, education, diabetes duration, and age of onset. Medical records were reviewed for microvascular complications (retinopathy, neuropathy, and/or nephropathy), vascular risk factors (hypertension, hyperlipidemia, and cardiovascular disease), and current medications. Of 53 participants with a hypertension diagnosis, only 1 was not taking medication. Polypharmacy was defined as taking five or more total noninsulin medications on a fixed regimen concurrently (20). ACB of fixed medications was calculated using an online calculator (25) and dichotomized, with combined scores ≥3 indicating high burden (26). HbA_1c values were obtained using point-of-care testing or extraction from medical records. Participants self-reported any prior episodes of SH and DKA, dichotomized as none versus one or more. Hypoglycemia awareness was assessed using the Gold survey, with scores ≥4 indicating impaired awareness (27). Participants wore a Dexcom G6 CGM throughout the study; metrics included time in target range (70–180 mg/dL), time spent in hyperglycemia (>180 mg/dL), time spent in hypoglycemia (<70 mg/dL), and glucose coefficient of variation (CV).

Study Period Variables

Interval SH and DKA events were collected during the study period. Systolic blood pressure was assessed at each of the three crossover visits and averaged to obtain a single study period value. CGM metrics were averaged across all three crossover periods.

End-of-Study (12-Month Follow-up) Variables

Polypharmacy at the end of the study was analyzed. CGM data were collected for 2 weeks prior to the end- of-study visit, and HbA_1c values were again obtained.

Cognitive Outcome Measures

Assessments were administered in person during the baseline visit, as well as during the extension phase visit ∼12 months later. The MoCA, a widely used, easily implementable MCI screening tool, was used as a measure of global cognitive functioning and includes items assessing visuospatial/executive function, memory, attention, language, abstraction, and orientation (28). The total raw score after education correction (1 point for high school education or less) was used.

Processing speed was evaluated using the Processing Speed Index (PSI) composite score from the Wechsler Adult Intelligence Scale–Fourth Edition (WAIS-IV), which includes the coding and symbol search subtests. The PSI age-corrected standard score (mean 100 ± 15) was used for analysis. These measures assess visuomotor integration, sustained attention, and rapid visual scanning (29).

Statistical Analyses

Data analysis and organization were conducted using SPSS version 31. Tests for normality and homogeneity were conducted, and distributional assumptions were evaluated visually via Q-Q plots and histograms. Individual cognitive change scores were computed for the MoCA and the WAIS-IV PSI by subtracting baseline scores from 12-month extension phase scores. Bivariate correlations and independent samples t tests were conducted to examine associations of diabetes characteristics, medical comorbidities, and demographics with cognitive change scores to determine which factors would be included in final regression models. An inclusion threshold of P < 0.10 was used to reduce the risk of excluding important covariates. Multiple linear regression analyses were then performed to predict change in each cognitive outcome measure. Collinearity diagnostics were examined prior to the interpretation of regression coefficients. Continuous variables are presented as mean ± SD.

Post Hoc Analyses

A series of post hoc analyses were conducted following findings from preliminary regression analyses. After first establishing which predictors were associated with change scores on the MoCA and WAIS-IV PSI, simple binary logistic regression analyses were then conducted to examine whether each significant predictor was associated with clinically meaningful change (defined as change greater than ±1.73 based on previously defined thresholds) (30). Second, as insulin resistance has been noted as a risk factor for vascular changes and, subsequently, a potential mechanism for cognitive decline, correlational analyses examined whether the relationship between vascular risk factors and cognitive change scores may be accounted for by insulin resistance. The estimated glucose disposal rate (eGDR) was used as a measure of insulin resistance and calculated based on available data. eGDR incorporates waist circumference, hypertension, and HbA_1c to calculate estimated insulin resistance and was calculated using the following equation: eGDR (mg/kg/min) = 21.158 − (0.09 × waist circumference [in cm]) − (3.407 × hypertension [yes = 1, no = 0]) − (0.551 × HbA_1c [%]) (31). Third, to further characterize the role of recent hypoglycemia in relation to cognitive performance, the association between percentage of time spent <54 mg/dL during each interval (i.e., baseline, study period, extension phase) and cognitive change scores was examined using Spearman correlations.

Results

Participant Characteristics

Participant characteristics are reported in Tables 1 and 2. Of the 82 participants enrolled in the parent AIDE study, 74 were included in the current analysis, and 8 were excluded due to not completing the 12-month follow-up cognitive measures. The average MoCA change score was 0.1 (range −5 to 7), and the average WAIS-IV PSI change score was 2.5 (range −14 to 28). We defined clinically meaningful change on the MoCA as greater than ±1.73 (30). Forty-seven percent of the sample exceeded this threshold, with 23% having significantly lower scores and 24% having higher scores after 12 months. At baseline, 20 participants (27%) scored in the MCI range (total score <26), while 19 (26%) scored in the MCI range at 12 months. Of note, 11 of 20 participants (55%) who scored below the cutoff at baseline continued to score below the cutoff at follow-up, while 9 of 20 (45%) were no longer below the cutoff (reverted to normal cognitive status). There were eight participants (10.8%) who scored below cutoff at both time points. For the WAIS-IV PSI score, 8.1% of participants remained stable (change score of 0), while 35.1% had lower scores and 56.8% had higher scores at the 12-month follow-up.

Table 1.

Participant characteristics (N = 74)

Characteristic	Value
Age at enrollment, years
Mean ± SD	70 ± 4
Range	65–86
Sex, n (%)
Female	35 (47)
Male	39 (53)
Race and ethnicity, n (%)
White, non-Hispanic	72 (98)
Black or African American, non-Hispanic	1 (1)
American Indian or Alaska Native	1 (1)
Education, n (%)
Less than high school	0 (0)
High school	10 (13)
Some college, AA, technical, or vocational	17 (23)
College graduate (bachelor’s or equivalent)	28 (37)
Advanced degree (e.g., master’s, PhD, MD)	20 (27)
Marital status, n (%)
Married or partnered	53 (72)
Single, widowed, separated, or divorced	21 (28)
Age at diagnosis of T1D, years
Mean ± SD	29 ± 16

AA, associate of arts.

Table 2.

Participant study characteristics (N = 74)

Characteristic	Baseline	Study perioda	12 Months
HbA1c (%) [mmol/mol]
Mean ± SD	7.2 ± 0.9 [55 ± 10]	7.0 ± 0.7 [53 ± 8]	6.8 ± 0.7 [51 ± 8]
Range	5.5–11.0 [37–97]	5.5–9.2 [37–77]	5.2–9.1 [33–79]
CGM metrics,b mean ± SD
Time in range (%)	67.4 ± 13.9	68.8 ± 11.6	74.3 ± 10.3
<70 mg/dL (%)	2.5 ± 2.3	1.8 ± 1.2	1.7 ± 1.0
>180 mg/dL (%)	30.1 ± 14.6	29.3 ± 12.0	21.4 ± 10.5
%CV	33.9 ± 5.3	33.9 ± 4.0	33.1 ± 4.5
DKA,c n (%)
0	45 (61)	61 (82)	41 (55)
≥1	29 (39)	13 (18)	33 (45)
SH events,c n (%)
0	20 (27)	65 (88)	18 (24)
≥1	54 (73)	9 (12)	56 (76)
Impaired awareness of hypoglycemia (Gold score)
Mean ± SD	3 ± 2	—	3 ± 2
Range	1–7	—	1–7
Score >4, n (%)	27 (33)	—	15 (20)
eGDR
Mean ± SD	6.0 ± 2.4	—	—
Range	1.1–10.3	—	—
eGDR >8, n (%)	15 (20)	—	—
Number of noninsulin medications
Mean ± SD	7 ± 4	1 ± 2	8 ± 4
Range	1–18	0–9	2–26
Polypharmacy >5, n (%)	58 (74)	4 (5)	64 (87)
ACB score
Mean ± SD	1 ± 1	0 ± 1	1 ± 1
Range	0–6	0–3	0–6
ACB score >3, n (%)	12 (16)	2 (3)	15 (20)
Microvascular complications, n (%)	43 (58)	2 (3)	45 (61)
Retinopathy	34 (46)	0 (0)	34 (46)
Neuropathy	21 (28)	0 (0)	21 (28)
Nephropathy or chronic kidney disease	12 (16)	2 (3)	13 (17)
Vascular risk factors
Blood pressured
Systolic
Mean ± SD	136 ± 15	132 ± 14	134 ± 16
Range	95–165	108–162	100–180
Diastolic
Mean ± SD	71 ± 9	68 ± 8	71 ± 8
Range	51–90	54–84	53–91
Hyperlipidemia, n (%)	64 (87)	0 (0)	64 (87)
Hypertension, n (%)	53 (72)	0 (0)	53 (72)
Cardiovascular disease, n (%)	29 (39)	3 (4)e	31 (42)
MoCAf
Mean ± SD	27 ± 2	—	27 ± 2
Range	20–30		18–30
Normal (≥26), n (%)	54 (73)		55 (74)
MCI (18–25), n (%)	20 (27)	19 (26)
WAIS-IV PSI
Mean ± SD	100.4 ± 12.7	—	102.9 ± 14.3
Range	76–132		71–140
WAIS-IV symbol search score
Mean ± SD	10 ± 3		11 ± 3
Range	5–16	—	4–19
WAIS-IV coding score
Mean ± SD	10 ± 3		10 ± 3
Range	5–17	—	4–17

^aBaseline DKA, SH, and medical conditions represent the total number of participants who reported the variable prior to the study. Study period variables reflect diagnoses that were first reported during the study period; 12-month variables represent the total participants who reported the variable prior to, during, or at the end of the study period.

^bCGM data were collected for 2 weeks prior to baseline and each crossover follow-up period. CGM data collected during each crossover phase were averaged for the study period values.

^cEpisodes of DKA and SH during the study period consisted of self-reported events. These events were not independently verified.

^dEighteen participants had missing data for study period blood pressure values due to the coronavirus disease 2019 pandemic restrictions on in-person visits.

^eThe cardiovascular complications acquired during the study included atrial flutter, deep vein thrombosis, and vascular calcification.

^fScores range from 0 to 30, with lower scores reflecting greater cognitive impairment. A score of ≥26 is considered normal cognitive status, 18–25 MCI, 10–17 moderate cognitive impairment, and <10 severe cognitive impairment.

Initial Associations With Change Scores

Tests for normality and graphical inspection indicated that most variables were normally distributed. A positive skew was observed for time spent in the hypoglycemic range across all time points; thus, nonparametric tests were used in correlational analyses using this variable. Age was associated with MoCA change scores, such that younger individuals showed greater improvement after 12 months (r = −0.23; P = 0.045). Independent samples t tests revealed that the mean MoCA change scores differed for participants with neuropathy (−1.14 ± 1.62) compared with those without (0.59 ± 2.08) (t[72] = 3.41; P = 0.001; d = 0.88; 95% CI 0.35–1.40), those with hypertension at baseline (−0.32 ± 2.13) and those without (1.15 ± 1.62) (t[72] = 2.83; P = 0.006; d = 0.73; 95% CI 0.21–1.25), and those with cardiovascular complications (−0.55 ± 2.25) and those without (0.51 ± 1.91) (t[72] = 2.18; P = 0.03; d = 0.52; 95% CI 0.04–0.99). Participants with an ACB of three or more at the end of the study showed a greater change on the MoCA (−1.00 ± 1.80) compared with those with a low ACB (0.36 ± 2.12) (t[71] = 2.21; P = 0.03; d = 0.66; 95% CI 0.06–1.25). Women demonstrated an overall mean increase in scores (0.60 ± 1.94), while men showed a decline (−0.36 ± 7.52) (t[72] = −2.0; P = 0.049; d = −0.47; 95% CI −0.93 to −0.001). Participants who reported being single improved (1.24 ± 2.23) compared with their partnered counterparts (−0.36 ± 1.88) (t[72] = 3.12; P = 0.003; d = 0.80; 95% CI 0.28–1.32). Using the criterion of P < 0.10 for inclusion for subsequent regression analyses, total medications were also included as a potential covariate (r = −0.21; P = 0.07).

As the WAIS-IV PSI score calculation is age-adjusted, age was not included as a covariate for PSI analyses. Independent samples t tests demonstrated that participants with hypertension at baseline showed less improvement (0.90 ± 7.90) in WAIS-IV PSI scores over the 1-year period compared with those without hypertension (6.62 ± 8.21) (t[35.38] = 2.74; P = 0.01; d = 0.72; 95% CI 0.20–1.24), women (4.51 ± 8.81) improved more than men (0.72 ± 7.52) (t[72] = −2.0; P = 0.049; d = −0.47; 95% CI −0.93 to −0.001), and those who had an episode of SH during the study period (−3.33 ± 6.52) declined, while those without SH during the study improved (3.32 ± 8.26) (t[72] = 2.32; P = 0.02; d = 0.82; 95% CI 0.11–1.53). There was a small association between CGM %CV at baseline and WAIS-IV PSI change scores (r = −0.21; P = 0.07), which while not statistically significant, met criteria for inclusion as a potential covariate in final analyses (P < 0.10).

Multiple Regression Analyses

Multiple linear regressions were conducted to determine which variables with significant bivariate associations uniquely predicted change in each cognitive outcome measure. For MoCA change scores, the model included baseline age, sex, marital status, cardiovascular complications, hypertension, neuropathy, end-of-study ACB, and total medications (Table 3). There was no indication of multicollinearity, with all variance inflation factors <10 (range 1.13–1.44) and tolerance values >0.1 (range 0.69–0.90). The model accounted for 29% of the variance in MoCA change scores (adjusted R² = 0.29). Significant predictors included age (unstandardized regression coefficient [b] = −0.11; SE_b = 0.05; 95% CI −0.21 to −0.00; standardized regression coefficient [β] = −0.21, t[64] = −2.03; P = 0.046), partnered status (b = −1.15; SE_b = 0.50; 95% CI −2.15 to −0.16; β = −0.25; t[64] = −2.32; P = 0.02), hypertension (b = −1.12; SE_b = 0.50; 95% CI −2.13 to −0.12; β = −0.24; t[64] = −2.24; P = 0.03), and neuropathy (b = −1.19; SE_b = 0.55; 95% CI −2.29 to −0.09; β = −0.26; t[64] = −2.16; P = 0.04).

Table 3.

Regression analyses for predictors of MoCA and WAIS-IV PSI change scores

Predictor	b	SEb	95% CI	β	t	P
MoCAa
Neuropathy	−1.19	0.55	−2.29 to −0.09	−0.26	−2.16	0.04
Hypertension	−1.12	0.50	−2.13 to −0.12	−0.24	−2.24	0.03
Married or partnered	−1.15	0.50	−2.15 to −0.16	−0.25	−2.32	0.02
Age	−0.11	0.05	−0.21 to −0.00	−0.21	−2.03	0.05
Cardiovascular complications	−0.52	0.48	−1.47 to 0.43	−0.12	−1.11	0.28
ACB	−0.61	0.64	−1.88 to 0.67	−0.11	−0.95	0.34
Total medications	0.02	0.06	−0.10 to 0.13	0.03	0.27	0.79
Sex (female)	0.11	0.47	−0.83 to 1.05	0.03	0.23	0.82
WAIS-IV PSIb
Hypertension	−5.74	2.02	−9.77 to −1.70	−0.31	−2.84	0.01
SH during study	−6.73	2.75	−12.23 to −1.24	−0.27	−2.45	0.02
CGM CV	−0.26	0.17	−0.60 to 0.08	−0.17	−1.55	0.13
Sex (female)	2.09	1.82	−1.54 to 5.72	0.13	1.15	0.26

Boldface indicates significance at P < 0.05.

^aMoCA model statistics: R² = 0.37; adjusted R² = 0.29; F(8,64) = 4.64; P < 0.001.

^bWAIS-IV PS model statistics: R² = 0.23; adjusted R² = 0.19; F(4,69) = 5.22; P < 0.001.

For the WAIS-IV PSI change scores, sex, hypertension, CGM %CV, and study period SH were included as predictors in the model. There was no indication of multicollinearity, with all variance inflation factors <10 (range 1.03–1.10) and tolerance values >0.1 (range 0.91–0.97). Overall, the model accounted for 19% of the variance in change scores (adjusted R² = 0.19). Baseline hypertension (b = −5.74; SE_b = 2.02; 95% CI −9.77 to −1.70; β = −0.31; t[64] = −2.84; P = 0.01) and study period SH (b = −6.73; SE_b = 2.75; 95% CI −12.23 to −1.24; β = −0.27; t[64] = −2.45; P = 0.02) remained significant predictors in the final model (Table 3).

Post Hoc Analyses

Simple binary logistic regression analyses revealed that the odds of a meaningful decline in MoCA scores were four times higher in those with neuropathy compared with those without (β = 1.44; SE = 0.59; Wald χ² = 6.07; P = 0.014; odds ratio [OR] 4.22; 95% CI 1.34–13.27). The odds of meaningful improvement in MoCA scores were five times higher in single older adults compared with those who were married (β = −1.63; SE = 0.58; Wald χ² = 7.98; P = 0.005; OR 5.1; 95% CI 1.64–15.84). There was an 80% lower odds of MoCA improvement in participants with hypertension compared with those without (β = −1.63; SE = 0.58; Wald χ² = 7.89; P = 0.005; OR 0.20; 95% CI 0.06–0.61) and an 89% lower odds of MoCA improvement in those with neuropathy compared with those without (β = −2.25; SE = 1.07; Wald χ² = 4.44; P = 0.04; OR 0.11; 95% CI 0.01–0.86).

Waist circumference was available for 45 of the 74 participants to calculate eGDR. Correlational analyses found no significant association between eGDR (6.00 ± 2.39 mg/kg/min; range 1.05–10.25) and MoCA change scores (r = 0.22; P = 0.15) or WAIS-IV PSI change scores (r = 0.17; P = 0.25). Additionally, Spearman correlations did not demonstrate a significant association between percentage time <54 mg/dL and cognitive change scores across any of the three time points.

Conclusions

This study examined determinants of cognitive change over 1 year in a sample of older adults with T1D using the MoCA and a targeted measure of processing speed. Age, partnership status, hypertension, and neuropathy predicted significant MoCA change, whereas baseline hypertension and interval SH significantly predicted WAIS-IV PSI change. Countering expectations, CGM metrics were not associated with change scores on either measure.

At baseline, 20 participants (27%) scored in the MCI range on the MoCA, while 19 (26%) scored in the MCI range at the 12-month visit. However, substantial individual variability was observed, with 45% of those with cognitive impairment at baseline reverting to normal cognitive status after 12 months (Fig. 1). This aligns with prior literature in general (32) and T1D (9) populations, suggesting reversible impairment and varied trajectories. Conversely, 10.8% of the sample converted from normal cognitive status to MCI over 1 year, exceeding general population estimates of 2.4% for those aged 70–74 years (33). Additionally, 47% of individuals demonstrated clinically meaningful MoCA change (±1.73 points) over 12 months, with 24% declining and 23% improving. Many improved on the WAIS-IV PSI, likely reflecting practice effects. Normative data from the WAIS-IV showed ∼3-point gains over a 22-day interval for older adults aged 70–90 years (29). The literature suggests that individuals with MCI show significantly reduced practice effects, and a lack of these effects may serve as an early marker of progression (34).

Figure 1.

Change in MoCA classification over the study period.

Baseline hypertension predicted lower change scores on both the WAIS-IV PSI and the MoCA, with participants with hypertension more likely to decline than improve, consistent evidence linking hypertension with decline in executive functioning and processing speed (8). Participants who experienced one or more episodes interval SH, but not those who reported a lifetime history, were more likely to decline in processing speed, suggesting an association between recent SH and acute changes in processing speed. This aligns with prior research demonstrating that SH within the past year, but not lifetime events, was associated with a greater likelihood of MCI (2). Research on the cognitive impact of lifetime SH is mixed, with significant associations being more common in studies that ascertained SH exposure longitudinally or via medical records (8,35). Future research should consider prospective collection of SH events, age at the time of the event, and the recency of SH prior to testing to understand this relationship further. Neuropathy was the only microvascular complication predicting MoCA change, with more than a fourfold increased odds of significant decline, consistent with evidence of deficits in memory, attention, and executive function in individuals with neuropathy (12,19).

Single participants were more likely than partnered individuals to improve on the MoCA. While prior research in T1D links higher perceived social support to better self-management and glycemic control, poor relationship quality may increase stress and reduce self-efficacy (33). Importantly, partnership quality and levels of social support were not assessed in the current study, making it possible that poor partnership quality may explain this finding. Alternatively, recent evidence suggests that unmarried individuals may have lower dementia risk compared with their married counterparts, potentially due to not only the quality of social support but also the tendency for unmarried (specifically never married) adults to socialize more with friends and neighbors (36). Therefore, the need to actively seek social support outside the home among unpartnered individuals may strengthen cognitive reserve and serve as a protective factor against cognitive decline (36). Future investigation of the effects of marital status on cognitive change in T1D should consider relationship quality and other sources of social support to help better clarify the mechanism driving these inconsistent findings. While our study was underpowered to examine sex by marital status interactions, prior research in older adults has demonstrated that single males have a higher risk for adverse health outcomes than single females, though findings regarding dementia risk are inconsistent (36,37). Larger studies in older adults with T1D are needed to better understand the role of sex in the relationship between marital status and cognitive change.

Contrary to expectations, no associations were found between glycemic metrics (HbA_1c or CGM) and cognitive change, possibly due to the sample’s overall excellent glycemic control improvement in glycemic metrics over the course of the study. Associations may be more evident in samples with greater variability in glycemic exposure over longer periods. Neither polypharmacy nor ACB was associated with cognitive change despite prior evidence in the general population (22,38) likely due to the high prevalence of polypharmacy (87%) in our sample, limiting the ability to detect group differences.

Limitations included highly educated, predominantly non-Hispanic White participants without major medical complications who were independent with diabetes management and receiving specialized care, limiting generalizability. Studies with larger sample sizes and greater demographic diversity are needed to confirm the current findings. While the predictors in the regression models explained a significant portion of variance in MoCA and WAIS-IV PSI change scores, study design limitations should be considered. The observational nature of the analyses cannot determine causality and leaves open the possibility of unmeasured or residual confounding when interpreting the findings.

In summary, the study findings demonstrate that meaningful cognitive change can be detected over a 1-year period in older adults with T1D using brief, easily implemented, clinical assessments. Several established and emerging risk factors, including hypertension, recent SH, and neuropathy, were associated with change in MoCA and WAIS-IV PSI scores. These findings are consistent with prior literature demonstrating associations between these factors and cognitive decline in individuals with T1D over a period that aligns with routine clinical care guidelines in older adults with T1D. The current findings also raise questions about the stability of cognitive impairment in T1D, with a notable percentage of participants falling below clinical cutoff scores and reverting to normal cognitive status within 1 year. Early identification and addressing potentially modifiable risk factors for cognitive decline in routine care are essential to maintain optimal brain health in this population. Close cognitive monitoring of older adults with T1D, and especially those who present with the identified risk factors, is warranted, as well as taking steps to reduce risk (e.g., optimal management of hypertension, preventing incident SH). Future research with more diverse samples is needed to understand the mechanisms driving these associations, including how relationship quality, social support, and SH timing and frequency influence cognition.

This article contains supplementary material online at https://doi.org/10.2337/figshare.33127076.

Funding Statement

The trial was supported by National Institute of Diabetes and Digestive and Kidney Diseases grant R01 DK122603. Trial visits at the University of Pennsylvania were supported in part by National Center for Advancing Translational Sciences grant UL1 TR001878. Additional trial supplies were provided by Tandem Diabetes Care and Dexcom.