Investigators reported associations between achieved LDL and non‑HDL cholesterol measured within two years of starting a statin, and subsequent risk of major adverse cardiovascular events (MACE) and all‑cause mortality. The study used new‑user cohorts assembled from electronic healthcare records in Hong Kong (HK), New Zealand (NZ) and the United Kingdom (UK). The analysis excluded people with prior atherosclerotic cardiovascular disease, heart failure or chronic kidney disease. Authors observed that, when compared with higher lipid categories, lower achieved LDL and non‑HDL were linked to lower hazards of MACE after multivariable adjustment.
The primary comparisons used categorical LDL bands with ≥2.6 mmol/L as the reference. Pooled adjusted hazard ratios for MACE were reported for multiple lower LDL bands; for example, the 1.8–2.5 mmol/L group had a lower hazard than the reference. The study team pooled region‑specific estimates using random‑effects meta‑analysis. These are the associations as documented in the source; exact numerical hazard ratios and confidence intervals are reported in the article tables and figures.
At the two‑year landmark, LDL concentrations were distributed across predefined bands: a small fraction achieved <1.4 mmol/L, a larger proportion clustered in the 1.8–2.5 and ≥2.6 mmol/L groups. The study team reported concrete counts for each band in the pooled sample. Mean duration of follow‑up for the MACE outcome differed by region: HK had the longest average follow‑up, followed by NZ and the UK, according to the article. These follow‑up differences shaped the person‑time available for event capture and influenced precision of region‑specific estimates.
Across regions, the cohort composition and the achieved lipid distributions varied. The authors noted that the NZ estimates were less consistent and more imprecise at the very low lipid concentrations compared with HK and the UK. That heterogeneity in attainment and precision is a recurring theme in the source analysis.
The study team conducted landmark analyses with lipid measurements taken within two years after statin initiation. Time‑to‑event associations were estimated using Cox proportional hazards models adjusted for a broad set of covariates. The models included age, sex and cardiovascular risk conditions such as hypertension, atrial fibrillation and diabetes mellitus. Rheumatologic and connective tissue disorders, severe mental illness, obesity indicators, smoking and alcohol misuse were also included, as were concurrent medications: antihypertensives, antidiabetics, antiplatelets and anticoagulants.
Investigators pooled adjusted HRs across regions with random‑effects meta‑analysis. The authors described potential confounding by indication: clinicians may intensify lipid‑lowering in patients judged higher risk, and such confounding could bias associations. The source also stressed residual confounding from unmeasured or imperfectly measured variables, including lifestyle interventions and disease severity.
When the study team modelled achieved LDL and non‑HDL as continuous variables, the association with MACE appeared approximately log‑linear in HK and the UK. That means the hazard of cardiovascular events declined steadily as achieved lipid concentrations decreased across the observed range in those regions. The NZ continuous estimates did not mirror that pattern as closely; at the lowest concentrations the NZ estimates were more variable and confidence intervals were wider, indicating imprecision.
The authors linked this interregional heterogeneity to differences in sample size, testing frequency, follow‑up and potential differences in treatment patterns and baseline cardiovascular risk. They therefore presented pooled and region‑specific results to let readers judge consistency and robustness.
After multivariable adjustment, very low achieved LDL cholesterol (reported in the source as <1.4 mmol/L) was not clearly associated with higher all‑cause mortality. Similarly, very low non‑HDL cholesterol did not show a definitive increase in mortality risk after adjustment. Point estimates for mortality at very low concentrations were near the null and confidence intervals included unity, which the authors interpreted as limited evidence for increased mortality.
The source cautioned that mortality estimates at very low lipid levels were imprecise and varied across regions. Readers should therefore interpret the near‑null mortality findings in the context of uncertainty and the observational design.
The authors documented several study limitations grounded in the available data. Heterogeneity across the three regional cohorts could reflect differences in clinical practice, baseline risk and data capture. Confounding by indication was identified as a likely influence — treatment intensity and lipid testing may have been guided by cardiovascular risk, which can bias associations between achieved lipids and outcomes. Residual confounding from comorbidities, lifestyle interventions and disease severity was also acknowledged.
The source explicitly noted that estimates are conditional on patients having initiated a statin, surviving to the two‑year landmark and undergoing follow‑up lipid testing; these selection conditions limit generalisability. The authors provided data access statements for each region and indicated that individual‑level linked data are not publicly shareable because of privacy protections.
Specific unanswered questions documented by the authors include the drivers of interregional heterogeneity and the causal role of very low lipid concentrations on mortality once confounding by indication and unmeasured factors are fully accounted for. The article ends by reporting those limitations and by directing readers to the region‑specific figures and tables for numerical detail and precision estimates.