Effectiveness of treat-to-target cholesterol-lowering interventions on cardiovascular disease and all-cause mortality risk in the community-dwelling population: a target trial emulation

Abstract

Little is known about the long-term effectiveness of risk-based treat-to-target cholesterol-lowering interventions on cardiovascular risk. Here, we show the emulated effectiveness of guideline-recommended low-density and non-high-density lipoprotein cholesterol-lowering interventions using the absolute risk reduction (ARR) and the restricted mean event-free time-based number needed to treat (NNT). With 5,375 participants, the 29-year risks for cardiovascular disease (CVD), all-cause mortality, and atherosclerotic CVD were 18.6%, 25.6%, and 17.7%, respectively. Long-term treat-to-target interventions showed significant reductions in CVD (ARR -2.3%, 95%CI -3.4% to -0.8%; NNT 115), all-cause mortality (-3.0%, -4.3% to -1.8%; 95), and atherosclerotic CVD (-2.6%, -3.5% to -1.2%; 104). Such effects appear more pronounced in women, smokers, and those with body mass index < 24 kg/m² or higher adherence rates.

Fig. 1. Flowchart of study participants’ selection.

Flowchart of selecting study participants from the Chinese Multi-provincial Cohort Study 1992–2020.

Fig. 2. Smoothed observed cumulative risk versus simulated risk under no intervention.

The smoothed observed cumulative risk (orange) compared to simulated risk (dark blue) under the natural course of no intervention for cardiovascular disease (a), all-cause mortality (b), and atherosclerotic cardiovascular disease (c).

Fig. 3. Estimated cumulative risk curves under hypothetical interventions.

Estimated cumulative risk curves comparing treat-to-target hypothetical interventions (i.e., per 1 mmol/L LDL-C reduction (orange), treat-to-target (cyan), and feasible interventions (dark red) with the “natural course” of no interventions (dark blue) for cardiovascular disease (a), all-cause mortality (b), and atherosclerotic cardiovascular disease (c) based on the Chinese Multi-provincial Cohort Study 1992-2020. Note: ^† Treat-to-target cholesterol-lowering intervention is based on cholesterol-lowering targets recommended by the Chinese Society of Cardiology in 2020 on LDL-C and non-HDL-C levels, i.e., for participants with diabetes at high cardiovascular risk, lower the LDL-C to < 1.8 mmol/L (70 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 1.8 mmol/L) or LDL-C reduction to > 50% from baseline whichever is the lowest and non-HDL-C to < 2.6 mmol/L (100 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 2.6 mmol/L); for participants without diabetes who are at moderate-to-high cardiovascular risk lower the LDL-C to < 2.6 mmol/L (100 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 2.6 mmol/L) and non-HDL-C to < 3.4 mmol/L (130 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 3.4 mmol/L); for participants at low cardiovascular risk, lower LDL-C to < 3.4 mmol/L (130 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 3.4 mmol/L) and a non-HDL-C < 4.2 mmol/L (160 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 4.2 mmol/L). ^‡ Feasible treat-to-target cholesterol-lowering intervention, defined as 80% of eligible participants receiving the intervention at the follow-up examination over the study period.

Fig. 4. Forest plots of the estimated effects under target-to-treat interventions.

Forest plot of the estimated parametric g-formula risk (%), risk difference (RD, %), and restricted mean event-free based number needed to treat (NNT) to prevent one cardiovascular, all-cause mortality, and atherosclerotic cardiovascular event under natural course and treat-to-target interventions recommended by the 2020 Chinese Society of Cardiology after 29 years of follow-up from the Chinese Multi-provincial Cohort Study 1992-2020. Note: ^† Treat-to-target cholesterol-lowering intervention is based on cholesterol-lowering targets recommended by the Chinese Society of Cardiology in 2020 on LDL-C and non-HDL-C levels, i.e., for participants with diabetes at high cardiovascular risk, lower the LDL-C to < 1.8 mmol/L (70 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 1.8 mmol/L) or LDL-C reduction to > 50% from baseline whichever is the lowest and non-HDL-C to < 2.6 mmol/L (100 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 2.6 mmol/L); for participants without diabetes who are at moderate-to-high cardiovascular risk lower the LDL-C to < 2.6 mmol/L (100 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 2.6 mmol/L) and non-HDL-C to < 3.4 mmol/L (130 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 3.4 mmol/L); for participants at low cardiovascular risk, lower LDL-C to < 3.4 mmol/L (130 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 3.4 mmol/L) and a non-HDL-C < 4.2 mmol/L (160 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 4.2 mmol/L).

Fig. 5. Forest plots of the estimated effects under feasible interventions.

Forest plot of the estimated parametric g-formula risk (%), risk difference (RD, %), and restricted mean event-free based number needed to treat (NNT) to prevent one cardiovascular, all-cause mortality, and atherosclerotic cardiovascular event under natural course and feasible interventions (unless stated) after 29 years of follow-up from the Chinese Multi-provincial Cohort Study 1992-2020. Note: ^† Feasible treat-to-target cholesterol-lowering intervention, defined as 80% of eligible participants receiving the intervention at the follow-up examination over the study period, where the treat-to-target cholesterol-lowering intervention is based on cholesterol-lowering targets recommended by the Chinese Society of Cardiology in 2020 on LDL-C and non-HDL-C levels, i.e., for participants with diabetes at high cardiovascular risk, lower the LDL-C to < 1.8 mmol/L (70 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 1.8 mmol/L) or LDL-C reduction to > 50% from baseline whichever is the lowest and non-HDL-C to < 2.6 mmol/L (100 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 2.6 mmol/L); for participants without diabetes who are at moderate-to-high cardiovascular risk lower the LDL-C to < 2.6 mmol/L (100 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 2.6 mmol/L) and non-HDL-C to < 3.4 mmol/L (130 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 3.4 mmol/L); for participants at low cardiovascular risk, lower LDL-C to < 3.4 mmol/L (130 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 3.4 mmol/L) and a non-HDL-C < 4.2 mmol/L (160 mg/dL, i.e., a fixed level drawed from a uniform distribution with a upper bound of 4.2 mmol/L).

Fig. 6. Causal diagrams of target trial specification and emulation.

Causal graphs illustrate the effects of the treat-to-target cholesterol-lowering interventions ($A$) on cardiovascular disease, all-cause mortality, and atherosclerotic cardiovascular disease ($Y$) in the presence of competing events ($C$) using the sequentially randomized trial (the upper panel, i.e., target trial specification, a in which the intervention strategy $A_k$ only depends on prior cardiovascular risk profiles ($L_k$), cholesterol levels and intervention history ($A_{k-1}$) at each time $k\left(k=\mathrm{0,\; 1},\dots ,28\right)$ and the hypothetical pragmatic trial (the bottom panel, i.e., target trial emulation, b in which $A_k$ depends on prior $L_k$, $A_{k-1}$, and the potential unmeasured confounders $U_k$. $C_{k-1}$ and $Y_k$ indicate the competing events at time interval $k-1$ and the outcome of interest at time interval $k$, respectively.