Table 1 Specification and emulation of a target trial of antidiabetic drug metformin vs. sulfonylureas on the risk of death and dementia, using observational data from Electronic Health Records of the US RPDR and the UK CPRD
Target trial specification | Emulation (US RPDR) | Emulation (UK CPRD) |
|---|---|---|
Eligibility criteria | ||
Age ≥ 50 | Same | |
No hypoglycemics | No recorded prior exposure to any hypoglycemic agents | |
No MCI*, dementia, or prescription of dementia drugs; normal cognitive testing | No recorded diagnosis of dementia or MCI*, or use of dementia-specific drugs (see Extended Data Tables 10–11) | No recorded diagnosis of dementia (MCI* diagnoses not available in CPRD) or use of dementia-specific drugs (see Extended Data Tables 12–13) |
No chronic kidney disease (metformin contraindication) | No ICD*-9/10 code for chronic kidney disease or eGFR* <45 (Extended Data Table 1) | No diagnosis of chronic kidney disease at or prior to baseline (Extended Data Table 2) |
Trial with 1-year run in period conducted for a specified duration with history obtained at baseline and ongoing monitoring of outcomes | • PCP* within Mass General Brigham Health Care system EHR* system • At least one visit during the 18 months preceding baseline • At least 1 year of follow-up • No dementia or death in first year (1 year washout period) | • At least 1-year registration in CPRD practices before the first prescription • At least 1 year of follow-up • No dementia or death in first year (1-year washout period) |
Treatment strategies | ||
Treatment arm: metformin monotherapy Control arm: sulfonylurea monotherapy | Initiation of metformin or sulfonylurea from 1/2007-9/2017 (see Extended Data Fig. 8 for the number of new prescriptions per year) | Initiation of metformin or sulfonylurea from 1/2001-5/2017, with ≥2 monotherapy prescriptions for first 12 months (see Extended Data Fig. 9 for the number of new prescriptions per year) |
Treatment assignment | ||
Double-blind, randomized treatment assignment | Emulated randomization by balancing baseline confounders using IPTW* for treatment choice | |
Outcomes | ||
Diagnosis of MCI* or dementia | Diagnosis of MCI/Dementia by: ICD*-9/10 codes (Extended Data Table 10) OR at least one dementia-specific drug prescription (Extended Data Table 11) | Diagnosis of dementia by: Medcodes in CPRD or ICD*-9/10 codes in linked HES* or ONS* database (Extended Data Table 12) OR at least one dementia-specific drug prescription (Extended Data Table 13) |
Time to death | Time to death recorded in EHR* | |
Follow-up | ||
From baseline and ends at dementia onset, death, lost to follow-up, or end of study | From the date of initial prescription of drug until the date of dementia incidence, death, last encounter date, 9/2018 (US RPDR) or 5/2018 (UK CPRD), whichever occurred first | |
Causal contrast | ||
Intention-to-treat effect | Observational analog of intention-to-treat effect | |
Statistical analysis | ||
Intention-to-treat analysis of primary outcomes (dementia and death) using Cox PH | Intention-to-treat analysis using Cox Proportional Hazards (PH) regression model and a competing risks framework accounting for death prior to dementia Subgroup analyses by age, sex, and BMI* level at baseline | |
