# Statin therapy for primary prevention in older adults with type 2 diabetes: target trial emulation from Hong Kong EHRs
Older people with type 2 diabetes are frequently excluded from randomized trials that define preventive cardiovascular care.
That gap leaves clinicians and policymakers uncertain about the balance of benefit and harm when starting statins for primary prevention in patients aged 75 years and older.
This study used territory-wide electronic health records to emulate a sequence of pragmatic target trials and estimate associations between initiating statins and cardiovascular outcomes, mortality, and key safety endpoints in this understudied group.
In this emulation of nested target trials using Hong Kong Hospital Authority records, initiating statin therapy in people with type 2 diabetes aged 75 years or older was associated with lower risks of cardiovascular events and all-cause death, without a marked increase in reported muscle-related adverse events or liver dysfunction.
Evidence guiding statin use for primary prevention in very old adults with diabetes is limited because such individuals are underrepresented in randomized controlled trials.
The clinical question addressed here was whether starting statin treatment among older adults with type 2 diabetes and elevated LDL cholesterol reduces cardiovascular events or mortality, and whether it increases the risk of statin-related harms.
The investigators leveraged a comprehensive clinical database to emulate randomized trial conditions within observational data, attempting to approximate causal comparisons between initiators and noninitiators of statin therapy across age strata, including those aged 75–84 and those older than 85 years.
Design and data source: The study used the Hospital Authority Clinical Management System in Hong Kong, a territory-wide electronic health record repository, to construct a series of nested target trials.
Eligible patients were identified on a rolling monthly basis between January 2009 and December 2015, thereby producing an emulation of multiple sequential trials covering that enrollment period.
Eligibility and exclusions: Each monthly emulated trial included adults with type 2 diabetes aged 60 years or older who had low-density lipoprotein cholesterol at or above 2.6 mmol/L during the baseline month.
Individuals with prior type 1 diabetes, established cardiovascular disease, cancers, muscle-related disorders, or liver dysfunction were excluded to focus on primary prevention and to reduce confounding by prevalent conditions that might influence both treatment choice and outcomes.
Treatment assignment emulation: Within each monthly trial, persons were classified at baseline as statin initiators if they began statin therapy at enrollment, or as noninitiators if they did not.
The investigators grouped participants into age bands (60–74, 75–84, and ≥85 years) with the 60–74 group serving as a benchmark to assess the validity of the emulation approach.
Follow-up and outcomes: Participants were followed until the first occurrence of the study outcome, death, or the administrative cutoff in December 2018.
Outcomes of interest included incident cardiovascular disease events, all-cause mortality, muscle-related adverse events, and liver dysfunction as recorded in the EHR.
Analytical approach to confounding and adherence: To emulate randomization and adjust for confounding, the authors applied propensity score matching at the baseline of eligible person-trials, using demographic factors, clinical and laboratory measures, comorbidities, prior medications, and healthcare utilization variables as matching covariates.
For longitudinal analysis and to account for time-varying confounding related to treatment adherence, pooled logistic regression models were fitted, and inverse probability weighting was used to adjust for these dynamic confounders under the assumption that no important unmeasured confounders remained.
Study scale and main effect estimates: The emulation produced a large analytic sample across the nested trials.
Among the matched person-trials aged 75–84 years, there were more than thirty thousand observations included in the matched comparisons.
In this age group, initiation of statins was associated with a statistically significant reduction in the incidence of cardiovascular disease, with an estimated hazard ratio below unity and a confidence interval excluding no effect.
The reduction in risk was consistent with a protective association when comparing initiators to noninitiators in this older cohort.
Effect on mortality and safety outcomes: Statin initiation in participants aged 75 years and older was associated with a lower risk of all-cause death, and similar associations were observed even among those older than 85 years.
Regarding safety outcomes captured in the records, the analyses did not detect substantially increased risks of muscle-related adverse events or liver dysfunction among statin initiators compared with noninitiators.
Benchmark and validity check: The investigators used the 60–74 years group as a benchmark to examine whether the emulated target trial produced effect estimates that comport with expectations derived from existing trial data and clinical knowledge.
The use of this younger benchmark group was intended to test whether the emulation approach and analytic adjustments produced plausible results before interpreting findings in the older age bands.
Limitations inherent to the design: The emulation strategy relied on the assumption of no unmeasured confounding.
As with all observational emulations, residual confounding cannot be ruled out.
Outcome ascertainment depended on recorded diagnoses and laboratory values within the administrative EHR, and some adverse events might be underrecorded or misclassified.
The source material did not present randomized evidence; rather, it reported associations estimated under strong assumptions about measured confounders and adherence processes.
Clinicians and health services researchers should interpret these observational findings as associations derived from a sophisticated emulation of randomized trials within routine-care data rather than as definitive causal proof.
The results suggest potential protective associations of starting statins in older adults with type 2 diabetes and elevated LDL cholesterol, and they indicate that, within the limits of the recorded data, substantial increases in coded muscle or hepatic events were not observed.
However, uncertainty remains because of the observational design, potential residual confounding, and the reliance on administrative recording of adverse events and outcomes.
The authors call for randomized trials and additional research to confirm effectiveness and safety, to explore dosing strategies, and to compare different statin agents in this age group.
Q: What population did this study address?
A: The analysis focused on adults with type 2 diabetes, LDL cholesterol at or above 2.6 mmol/L, and no prior cardiovascular disease, cancer, muscle disorders, or liver dysfunction, enrolled across monthly emulated trials from 2009–2015.
Age groups included 60–74, 75–84, and 85 years or older.
Q: How were comparisons between statin users and nonusers made?
A: The investigators emulated treatment start in each monthly trial and used propensity score matching at baseline, followed by pooled logistic models with inverse probability weighting to account for time-varying confounding related to adherence, aiming to approximate randomized comparisons.
Q: What outcomes were evaluated and what were the results?
A: Outcomes included incident cardiovascular events, all-cause mortality, muscle-related adverse events, and liver dysfunction.
In the older age bands, initiation of statin therapy was associated with lower rates of cardiovascular events and death, and no sizable increases in muscle or liver adverse events were observed in the recorded data.
Q: Should these findings change practice immediately?
A: The study authors recommended further research, including randomized trials, to confirm these observational associations, determine optimal dosing, and compare statin types for primary prevention in very old adults with diabetes.
The emulation provides important observational evidence but does not replace randomized data.
Using a target trial emulation approach within a comprehensive Hong Kong EHR system, investigators found that initiating statin therapy in people with type 2 diabetes aged 75 years and older was associated with reduced cardiovascular events and lower all-cause mortality, and that no marked increases in reported muscle or liver adverse effects were detected.
These results add to observational evidence addressing a gap left by randomized trials, yet they rest on assumptions inherent to nonrandomized emulations.
Randomized studies and further comparative work were recommended to validate these associations and to refine therapeutic choices and dosing for older adults with diabetes.