This study utilized data from the Korea National Health Insurance Service (KNHIS) database, which is a comprehensive healthcare information system that provides healthcare coverage to all South Korean individuals. The database contains health-related information for approximately 50 million individuals and includes sociodemographic data; lifestyle questionnaires; anthropometric measurements; laboratory test results; medical diagnoses based on the International Statistical Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM); and treatment data for the Korean population [15]. This study was approved by the Seoul St. Mary’s Hospital Institutional Review Board, Korea (approval number: KC21ZNSI0448) and adhered to the principles outlined in the Declaration of Helsinki, along with other applicable regulations and guidelines. Since the research involved the analysis of publicly accessible data that had been anonymized and de-identified, the requirement for obtaining informed consent was waived.

This retrospective case–control cohort study utilized the Korean National Health Insurance Service (KNHIS) database. We systematically identified individuals between January 1, 2009, and December 31, 2020. The case group consisted of patients diagnosed with MM based on the specific ICD-10 code (C90) as either a primary or secondary diagnosis. For the non-MM control group, which lacked a prior history of MM, we employed 1:10 age/sex-matching from the KNHIS database’s sample cohort. This sample cohort selects approximately 2% of samples directly from the entire Korean population database. This approach minimizes non-sampling errors.

Initially, we established a primitive case cohort (n = 15,402), excluding patients with MM for reasons, such as missing values (n = 332), age under 19 years (n = 52), MM as a secondary diagnosis (n = 13,432), exclusive diagnosis in 2009 or only once (n = 6396), or prior diagnosis of CVD before MM diagnosis (n = 2220). Similarly, we constructed a primitive control cohort (n = 123,216) without MM by 1:8 matching with the case cohort based on birth year and sex. This cohort was further refined by excluding individuals with missing values (n = 2405), those who died prior to the index date (n = 119), and those diagnosed with CVD before the index date (n = 13,442), resulting in a mother control cohort (n = 107,610). To balance the baseline characteristics of both cohorts, we applied 1:1 propensity score matching (PSM), adjusting for factors such as birth year, sex, index year, socioeconomic status, and comorbidities. Consequently, the final study cohort paired one MM patient with each non-MM control, resulting in 15,402 participants in both the MM (case cohort) and non-MM (control cohort) groups.

Data on baseline characteristics were collected at the index date. We utilized the Charlson Comorbidity Index (CCI) to collect comorbidity details up to 6 months prior to the index date [16]. To remove potential symptoms of MM, a 6-month washout period was set. Additionally, for each comorbidity, the primary diagnosis or at least two entries for sub-diagnoses before the index date were required, with the initial diagnosis date being that of the comorbidity. Each comorbidity was defined using ICD-10 codes. For example, congestive heart failure (I09.9, I11.0, I13.0, I13.2, I25.5, I42.0, I42.5–I42.9, I43.x, I50.x, or P29.0); peripheral vascular disease (I70.x, I71.x, I73.1, I73.8, I73.9, I77.1, I79.0, I79.2, K55.1, K55.8, K55.9, Z95.8, or Z95.9); dementia (F00.x–F03.x, F05.1, G30.x, or G31.1); chronic pulmonary disease (I27.8, I27.9, J40.x–J47.x, J60.x–J67.x, J68.4, J70.1, or J70.3); autoimmune disease (M05.x, M06.x, M31.5, M32.x–M34.x, M35.1, M35.3, or M36.0); peptic ulcer disease (K25.x–K28.x); diabetes with (E10.2–E10.5, E10.7, E11.2–E11.5, E11.7, E12.2–E12.5, E12.7, E13.2–E13.5, E13.7, E14.2–E14.5, or E14.7) or without chronic complications (E10.0, E10.1, E10.6, E10.8, E10.9, E11.0, E11.1, E11.6, E11.8, E11.9, E12.0, E12.1, E12.6, E12.8, E12.9, E13.0, E13.1, E13.6, E13.8, E13.9, E14.0, E14.1, E14.6, E14.8, or E14.9); hemiplegia or paraplegia (G04.1, G11.4, G80.1, G80.2, G81.x, G82.x, G83.0–G83.4, or G83.9); renal disease (I12.0, I13.1, N03.2–N03.7, N05.2–N05.7, N18.x, N19.x, N25.0, Z49.0–Z49.2, Z94.0, or Z99.2); any malignancy excluding skin cancer (C00.x–C26.x, C30.x–C34.x, C37.x–C41.x, C43.x, C45.x–C58.x, C60.x–C76.x, C81.x–C85.x, C88.x, or C91.x–C97.x); moderate or severe hepatic disease (I85.0, I85.9, I86.4, I98.2, K70.4, K71.1, K72.1, K72.9, K76.5, K76.6, or K76.7); metastatic solid tumor (C77.x–C80.x); and AIDS/HIV (B20.x–B22.x, or B24.x). In defining ‘any cancer’ for this study, we included any malignancy excluding skin cancer and considered metastatic solid tumors as separate occurrences. The study also considered socioeconomic status, represented by a numeric value derived from the average monthly insurance premium in the KNHIS database. This status was initially categorized into 11 groups, which included a medical beneficiary set and ten dual score level groups. For analytical purposes, these were then grouped into two major categories: the lower 1st to 3rd percentiles (scores 0–2) and the 4th to 10th percentiles (scores 3–20).

The primary outcome measure was the incidence of CV events among individuals who survived long-term following the index date. We determined the incidence of CV events using specific ICD-10 codes. These included codes for myocardial infarction (I21.x, I22.x, I24.x, or I25.2) and cerebrovascular disease (G45.x, G46.x, H34.0, or I60.x–I69.x). A CV event was recognized if it was recorded as either the primary or secondary diagnosis in the medical expense claims submitted to the KNHIS until the end of the follow-up period, which was on December 31, 2020.

To account for baseline characteristic differences between the case and control group participants, we employed PSM. The PS was derived using a logistic regression model, which factored in variables such as age, sex, index year, socioeconomic status, and a range of comorbidities. These comorbidities included congestive heart failure, peripheral vascular disease, chronic pulmonary disease, autoimmune disease, peptic ulcer disease, diabetes, hemiplegia or paraplegia, renal disease, hepatic disease, and any type of cancer. We only included CCI variables with a prevalence of 0.1% or higher as conditioning variables in the model. For constructing the final study population, we used 1:1 PSM without replacement, employing a greedy-matching algorithm with a caliper width of 0.25. To evaluate the effectiveness of the matching process, we calculated the standardized mean differences (SMDs) of each covariate across the groups both before and after applying PSM. The matching was considered balanced and effective when the SMD for each covariate was less than 0.1 [17].

Continuous variables are expressed as means ± standard deviations, and categorical variables are presented as frequencies or percentages. We measured the time-to-event from the baseline until a CV event diagnosis or death, whichever came first. Patients who did not experience a CV event or die during follow-up were censored at their last medical encounter. The cumulative incidence of CV events was calculated, accounting for death as a competing event, and the difference in cumulative incidence between the MM and non-MM groups was tested using the Gray test static for equality. Instead of the traditional Cox proportional hazards model, which typically treats death as censored, we employed the Fine–Gray model. This model considers death as a competing risk, providing a more conservative estimate than methods that treat patients who die as censored, assuming they would remain at risk with more extended follow-up. To account for the clustering of matched pairs, statistical inference was based on the Fine–Gray proportional hazard model with robust standard errors using the sandwich covariance matrix estimation. The hazard ratio (HR) and its 95% confidence interval (CI) were estimated using Fine–Gray models. We implemented a five-year landmark analysis to address immortal time bias in survival analysis, focusing on long-term survivors to identify CVD events. All tests were two-tailed, and statistical significance was set at P < 0.05. Analyses were conducted using SAS version 9.4 (SAS Institute Inc.) and R version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria).