BACKGROUND
SGLT2 inhibitors are diabetes drugs linked to a lower risk of dementia in large population studies.¹
They produce ketones (unlike other diabetes drugs) and have also been linked to a lower risk of depression.²
Lab studies show they improve mitochondrial function and reduce oxidative stress in neurons. People with psychiatric illness face higher dementia risk—2–3× higher in mood disorders, and up to 10× higher in schizophrenia.³
OBJECTIVE
We tested whether SGLT2 inhibitors are associated with lower rates of dementia and markers of psychiatric severity in people with psychiatric disorders, using a target trial emulation design.⁴

SAMPLE

Veterans Health Administration (VA) patients with:

Major depressive disorder (MDD)
Bipolar disorder (BIP)
Schizophrenia spectrum disorders (SCZ)
Eligibility conditions in Figure 1.

METHODS

Exposure: SGLT2-inhibitor use, either initiation (Intention to treat: ITT) or adherence (per-protocol: PP)

Outcomes:

I. All-cause dementia: Alzheimer's Disease (AD), AD-related dementias (RD), or non-specific dementias (NSD)
II. Psychiatric Emergency Department visit (PED), with primary psychiatric symptoms as the cause of admission
III. Psychiatric-related hospitalization (PHS), either admission or transfer.

Statistical Analysis: Marginal structural models for person-period level data. Each patient is represented as a sequence of 3-month periods. Time-varying covariates (Table 1) were adjusted per period. Inverse probability weights for treatment (IPTW) and censoring (IPCW) were used towards a pooled logistic regression towards outcome.

PP analysis focused on participants who continuously adhered to their initial treatment assignment throughout the follow-up period. ITT analyzed patients according to their initial treatment assignment regardless of subsequent changes.

RESULTS
Cohort: 112,725 patients (median age 74.1; 93.6% male; median follow-up 3.3 yrs); 7,631 initiated SGLT2i and 4,616 (4.1%) developed dementia.
DEM (Outcome I): SGLT2i initiation significantly lowered dementia risk (ITT OR 0.61 [0.52–0.73]; PP OR 0.54 [0.40–0.73]); 1.3% absolute 5-yr risk reduction (1.6% vs. 2.9%).
PED (Outcome II): Significant reduction in psychiatric ED visits (ITT OR 0.80 [0.66–0.97]; PP OR 0.74 [0.53–1.05]); 0.7% absolute 5-yr risk reduction.
PHS (Outcome III): Non-significant trend in ITT (OR 0.68 [0.44–1.04]); significant in PP (OR 0.56 [0.31–1.00]); 0.2% absolute 5-yr risk reduction.

CONCLUSIONS AND FUTURE WORK
In a target trial emulation using VA data, starting an SGLT2 inhibitor was linked to a significantly lower risk of dementia in people with major depression, bipolar disorder, or schizophrenia.
The risk of psychiatric emergency visits and hospitalization rates (among adherent users) also reduced. Future studies should confirm these findings in other psychiatric populations.
This work will be published iin JAMA Network Open as of July 1, 2026

Table 1 – Cohort Characteristics at Baseline

Variable	Type	Temporality	Did not use SGLT2-i (N=105,094)	Used SGLT2-i (N=7,631)
Sex (female), n (%)	Demographics	Static	7907 (7.5%)	273 (3.6%)
Age, median [IQR]	Demographics	Dynamic	73.9 [69.4, 77.5]	74.6 [71.1, 77.2]
Race (Black), n (%)	Demographics	Static	20557 (19.6%)	1422 (18.6%)
Ethn. (Hispanic), n (%)	Demographics	Static	6974 (6.6%)	652 (8.5%)
Obese, n (%)	Vital	Dynamic	50204 (47.8%)	5313 (69.6%)
Hgb. A1C, median [IQR]	Vital	Dynamic	5.8 [5.6, 5.9]	6.5 [5.8, 6.9]
Tobacco use, n (%)	Lifestyle	Dynamic	33209 (31.6%)	2235 (29.3%)
Alcohol use, n (%)	Lifestyle	Dynamic	3817 (3.6%)	139 (1.8%)
Benzo. use, n (%)	Other Drugs	Dynamic	34188 (32.5%)	3195 (41.9%)
Antidep. use, n (%)	Other Drugs	Dynamic	83408 (79.4%)	6647 (87.1%)
DPP4-I use, n (%)	Other Drugs	Dynamic	4230 (4.0%)	2043 (26.8%)
GLP1-a use, n (%)	Other Drugs	Dynamic	2693 (2.6%)	2074 (27.2%)
SCZ, n (%)	Psychiatric dx at baseline	Both	5801 (5.5%)	272 (3.5%)
BIP, n (%)	Psychiatric dx at baseline	Both	8089 (7.7%)	646 (8.5%)
MDD, n (%)	Psychiatric dx at baseline	Both	91204 (86.8%)	6713 (88.0%)
Psychotic MDD, n (%)	(MDD subtype)	Both	2243 (2.1%)	169 (2.2%)
Diabetes, n (%)	Comorbidities	Dynamic	41859 (39.8%)	7038 (92.2%)
Hypertension, n (%)	Comorbidities	Dynamic	86594 (82.4%)	7380 (96.7%)
PVD, n (%)	Comorbidities	Dynamic	13254 (12.6%)	1662 (21.8%)
CAD, n (%)	Comorbidities	Dynamic	47416 (45.1%)	5261 (68.9%)
KD, n (%)	Comorbidities	Dynamic	24724 (23.5%)	3429 (44.9%)
TBI, n (%)	Comorbidities	Dynamic	4321 (4.1%)	367 (4.8%)
PTSD, n (%)	Comorbidities	Dynamic	43144 (41.1%)	3516 (46.1%)
Neurodeg., n (%)	Comorbidities	Dynamic	7050 (6.7%)	539 (7.1%)
Follow-up time (years), median [IQR]	Follow-up	—	3.0 [1.3, 5.3]	4.8 [3.2, 6.4]
All-Cause Dementia, n (%)	Outcome	—	4485 (4.3%)	131 (1.7%)
Death, n (%)	—	—	17387 (16.5%)	504 (6.6%

Figure 1 – Schematic diagram of trial emulation design
Jan 2006 – Jun 2024: Medical History
Jan 2016 – Jun 2024: Trial Window
Discrete 3-month trial periods: Trial 0 (Q3 2016), Trial 1 (Q4 2016), Trial 2 (Q1 2017), ... Trial 30 (Q2 2024)
6-month run-in period
Eligible at Q1 2016 / Q2 2016
Eligibility: Trials start at latest of:

(a) Jan (Q1) 2016
(b) Second psychiatric diagnosis date (MDD, BIP, SCZ)
(c) Reaching age 65
With no prior dementia diagnosis, SGLT2-i use, personality disorder diagnosis.

Intention-To-Treat (ITT): Control (SGLT2-i non-initiation) vs. Treatment (SGLT2-i initiation, ≥3 months)

Per-Protocol (PP): Control (SGLT2-i sustained non-use) vs. Treatment (SGLT2-i initiation and adherence, ≥6 months)

SGLT2 inhibitors: Canagliflozin, Dapagliflozin, Empagliflozin, Ertugliflozin, Bexagliflozin, Sotagliflozin

Censoring events:

Outcome (Dementia, PED, PHS)
Death (censoring)
Loss to follow-up (last VA encounter)
End of study window (June 2024)
Treatment switching (PP)
Funding: This study was supported by the VA Cooperative Studies Program

Table 2 – Main Effects

Effect / Outcome	I: DEM	II: PED	III: PHS
ITT — Treatment Coefficient (− log odds ratio) [95% CI]	−0.49 ** [−0.66, −0.31]	−0.22 ** [−0.42, −0.03]	−0.38 [−0.81, 0.04]
PP — Treatment Coefficient [95% CI]	−0.61 ** [−0.92, −0.31]	−0.30* [−0.64, 0.05]	−0.58 ** [−1.16, 0.00]

** Indicates significance at 95%

Figure 2 – Estimated 5-year CIF curves for ITT effect
Outcome I: Dementia – Cumulative incidence ~0.06 at 5 years; Initiated SGLT2 (blue) lower than Did not initiate SGLT2 (red)
Outcome II: PED – Cumulative incidence ~0.02 at 5 years; Initiated SGLT2 lower than Did not initiate
Outcome III: PHS – Cumulative incidence ~0.04 at 5 years; Initiated SGLT2 lower than Did not initiate
All curves over Follow Up Time (Years) 0–5.

 

REFERENCES
1. Wium-Andersen IK, Osler M, Jørgensen MB, Rungby J, Wium-Andersen MK. Antidiabetic medication and risk of dementia in patients with type 2 diabetes: a nested case-control study. Eur J Endocrinol. 2019 Nov;181(5):499-507. doi: 10.1530/EJE-19-0259.
2. Liebers DT, Ebina W, Iosifescu DV. Sodium-Glucose Cotransporter-2 Inhibitors in Depression. Harv Rev Psychiatry. 2023 Jul-Aug 01;31(4):214-221.
3. Stroup TS, Olfson M, Huang C, Wall MM, Goldberg T, Devanand DP, Gerhard T. Age-Specific Prevalence and Incidence of Dementia Diagnoses Among Older US Adults With Schizophrenia. JAMA Psychiatry. 2021 Jun 1;78(6):632-641.
4. Hernán MA, Robins JM. Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available. Am J Epidemiol. 2016;183:758-764.