Study design
This is register-based cohort study of persons with dyslipidemia who received treatment with an anti-PCSK9 drug.
Study cohort
The study included all persons who received anti-PCSK9 treatment in the Capital Region of Denmark from January 1, 2016, until July 26, 2022. Alirocumab and evolocumab were the only two anti-PCSK9 treatment used in the study period. Anti-PCSK9 treatment is provided to persons at no cost by hospitals in Denmark. The use of anti-PCSK9 treatment is limited by the national healthcare reimbursement policies and persons who receive anti-PCSK9 treatment are required to meet specific treatment requirements. The Danish Medicines Council has divided the requirements for reimbursement for hypercholesterolemia treatment into two categories, namely familial hypercholesterolemia, and non-familial hypercholesterolemia. To be eligible for treatment, persons with hypercholesterolemia are required to meet the predefined criteria outlined by the council 7.
Study variables
In this study, we obtained data variables such as sex, age, vital status, laboratory values, prescriptions, administrations of medicines, allergies, and diagnosis and procedural codes describing the type and degree of the ASCD from the Sundhedsplatformen program (EPIC EHR) (Appendix). Familial hypercholesterolemia was defined using ICD-10 codes (Appendix). Additionally, we defined all persons suspected of familial hypercholesterolemia (with an ICD-10 code) and any LDL-C measurement greater than 5 mmol/L as having familial hypercholesterolemia.
Outcome definitions
We considered persons who had stopped any anti-PCSK9 treatment for 30 days or more as having discontinued treatment; this included those who switched to another anti-PCSK9 drug after such a pause. MACE was defined as either a myocardial infarction, a stroke or death. Death was determined using the vital status from the SAP Web Intelligence (WebI).
Statistical analysis
A generalized additive model (GAM) was used to examine the effect of a treatment initiation and switch in treatment on LDL-C levels, with each person treated as a random intercept. We fitted two models, one using all LDL-C measurements to evaluate the effect of anti-PCSK9 treatment and the effect of dosage. In another model, we only included LDL-C measurements within the range of 180 days prior to the mandatory switch and from 40 days post-switch up to 180 days, excluding measurements taken during treatment pauses. Covariates including age, sex, dose of anti-PCSK9, and other lipid-lowering medications were included in both models. To account for individual-level random effects, the "re" option of the gam function was utilized. Penalized spline terms were also added for age in the model. For further details, please refer to the appendix.
In a sensitivity analysis, we employed the t-test to assess the impact of transitioning from evolocumab to alirocumab and vice versa, both pre- and post the mandated switch date. For each group, we identified the most recent (closest to switch date) LDL-C measurement taken within a maximum of 180 days before the switch during ongoing treatment with an anti-PCSK9, as well as the LDL-C measurement taken between 40 and 180 days after the baseline and ongoing treatment with a PCSK9.
In the second part of the study, we assessed treatment retention, ASCVD risk, and survival by setting up a cohort study and following individuals from the start of anti-PCSK9 treatment until death or end of follow-up, whichever occurred first. Using a Poisson model, we modeled the number of first MACE and deaths, with person-days as an offset. For discontinuation, we included a random effect for each person and allowed for multiple events per person. All models were adjusted for variables such as the type of switch (first treatment initiation, switch before mandated date, switch after mandated date), calendar date, days on anti-PCSK9 treatment, sex, age, and presence of any comorbidity (Appendix). Continuous variables were included as penalized splines.
In the final part of the study, we examined all persons taking alirocumab on February 1, 2021, who were alive on December 31 of the same year. We used chi-squared and Wilcoxon rank sum tests to compare the group of persons who switched to evolocumab with the group who did not switch. Continuous variables are presented as means with standard deviations, and categorical data is presented as frequencies and percentages, unless otherwise stated. A two-tailed P value less than 0.05 was considered statistically significant, and 95% confidence intervals are provided. The study was conducted and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) recommendations 8. Data management and statistical analysis were conducted using R version 4.0 9. This study was approved by regional council of Copenhagen R-22034085.