Between September 2009 and February 2014, a total of 1500 individuals aged 18 to 88 were recruited at the primary prevention unit of the S. Gerardo Hospital (Monza, Italy). Patients were both enrolled from the arterial hypertension outpatient services and from the blood donor list of the same hospital.

Exclusion criteria were: age below 18, pregnancy or lactation, previous acute cerebrovascular and cardiac events (myocardial infarction, angina pectoris, heart failure, stroke, transient ischemic attacks and claudication), medical conditions that would modify the reliability of heart and vessels HMOD evaluation (atrial fibrillation, very frequent ventricular or supraventricular extrasystoles and previous stent placement at carotid, aortic or femoral levels), chronic obstructive pulmonary disease, substance dependency and malignancies.

Three patients that were not resident in the catchment area of the ATS of Milan or Brianza and were excluded due to the impossibility of collect follow-up data.

Finally, Lp(a) measurements were available in 747 patients on which we conduct the present analysis. There were some significant differences between patients with available Lp(a) and patients without it as shown in Supplementary Table 1. Briefly, the first ones presented lower PWV values (8.0 ± 1.8 vs 8.3 ± 1.9 m/s, p=0.001) and a lower prevalence of vascular HMOD (10.3% vs 13.5%, p=0.047) while showing higher IMT values (0.66 ± 0.19 vs 0.63 ± 0.18 mm, p<0.001) but similar prevalence of IMT HMOD (7.2% vs 6.8%, p=0.768). Carotid plaques were more frequent in patients with available Lp(a) (19.8% vs 14.8%, p=0.005).

Despite these differences at baseline, no differences were seen regarding follow-up CV events and mortality between these two groups. Patients with available Lp(a) experienced 71 CV events (9.5%) while the ones without Lp(a) experienced 53 events (7.0%), without significant differences among them (log-rank test p=0.371 to account for different follow-up). For mortality, included patients experienced 34 events (4.6%), while the ones without Lp(a) had 44 events (5.8%, log-rank test p=0.107). Finally, regarding the composite event, included patients experienced 93 events (12.4%), while the ones without Lp(a) experienced 84 events (11.2%, log-rank test p=0.928).

At baseline a comprehensive medical history was obtained, and physical examinations were conducted for all participants. Blood Pressure (BP) was measured using a semi-automated device (OMRON Healthcare Europe, Hoofddorp, Netherlands) following a five-minute seated rest with the arm positioned at heart level; the mean of two readings was utilized for the analysis. Subjects were then placed in the supine position and CV HMOD assessment began. The within-operator and between-operator intra-session coefficient of variability, as well as the inter-session one, of all the HMOD evaluated were reported in Supplementary Table 2.

Body Mass Index (BMI) was calculated using measured height and weight, while Waist Circumference (WC) was recorded at the midpoint between the lower ribs and iliac crest.

We measured fasting serum glucose, serum total cholesterol, High-Density Lipoprotein (HDL) cholesterol, LDL cholesterol, serum triglycerides and creatinine levels. Diabetes was defined as a fasting plasma glucose > 126 mg/L in two occasions or as the use of antidiabetic drugs while CKD was defined as an estimated glomerular filtration rate (eGFR) < 60 mL/min. Class of antihypertensive drugs were recorded and uncontrolled hypertension was defined as values higher than 140/90 mmHg. Finally, dyslipidaemia was defined as an LDL cholesterol > 116 mg/dL or as the use of statins therapy.

The study protocol was approved by institutional ethics review committees of the institution involved and all participants provided informed written consent after being informed of its nature and purpose.

Biochemical analyses were conducted using an automated Modular Analytics SWA system (Roche Diagnostics), employing enzymatic colorimetric assays for glucose (GOD-PAP method), total cholesterol (CHOD-PAP method), HDL cholesterol (third-generation enzymatic colorimetric method), triglycerides (GPO-PAP method). For patients with TG levels above 400 mg/dL, direct LDL cholesterol measurement was used; otherwise, Friedewald’s formula was applied to calculate: Total cholesterol - [HDL cholesterol + (triglycerides/5)].

Creatinine levels were measured via the Jaffé kinetic colorimetric method, and GFR was calculated using the Modification of Diet in Renal Disease (MDRD) equation. Renal HMOD was defined as a GFR < 60 mL/min.

Lp(a) concentrations were quantified using an immunoturbidimetric method on a Roche Cobas® 8000 c502 module (Roche Diagnostics GmbH, Mannheim, Germany), calibrated against the World Health Organization/International Federation of Clinical Chemistry reference material SRM 2B. The assay is isoform-insensitive, thus minimizing the size-related analytical bias due to variable numbers of kringle IV-2 repeats in the apolipoprotein(a) moiety. All samples were run in duplicate, and internal quality control materials at two levels (low and high) were analyzed daily in accordance with CLSI EP05-A3 and EP28-A3c guidelines. The limit of detection was 1.0 mg/dL, and the functional sensitivity was 3.5 mg/dL. Inter-assay and intra-assay coefficients of variation were <5%.

The performance of the method and its analytical imprecision were verified locally through validation runs prior to study initiation. All samples were measured in a single analytical batch to minimize inter-run variability.

Follow-up of all-cause mortality and CV events was performed through record linkage with administrative health data of the ATS of Milan and Brianza, were the patients live at baseline. The median follow-up was 10 years (I-III quartiles 9.6, 11.1). Vital status was updated to June 30th 2021 using the database of patients registered with the regional health system in each ATS (NAR: Nuova Anagrafe Regionale Assistiti). Patients emigrated outside the region were censored at emigration date. The following CV events were included and assessed using the database of hospital discharge records: acute myocardial infarction or equivalents (ICD-9-CM codes 411.X, 413.X, 414.X as principal diagnosis or 410.X, 423.0, 429.5, 429.6, 429.71, 429.79, 429.81 in any diagnostic filed), stroke (430.XX-437.XX in any diagnostic field), coronary revascularization (00.66, 36.03-36.07, 36.09-36.17, 36.19, 36.2, 36.31-36.34, 36.9 in any procedure field), carotid revascularization (00.61-00.65, 38.2, 38.10- 38.12, 38.31, 38.32, 38.41, 38.42, 39.72, 39.74 in any procedure field), and lower extremity revascularization (39.25, 39.29, 39.50, 39.90 in any procedure field). Furthermore, the following drugs were included and assessed using the database of drug prescriptions: prasugrel (ATC code B01AC22) and ticagrelor (B01AC24).

The evaluated endpoints were the time to: (i) first CV events, (ii) all-cause mortality and (iii) composite outcome (CV event or death).

Aortic stiffness was assessed by measuring PWV between the carotid and femoral arteries on the same side, with participants in a supine position. A validated automatic device (Complior, Colson; Alam Medical, Paris, France) was used to simultaneously capture arterial waveforms at both sites. The distance between measurement points was determined using a rigid ruler from the hip to the neck through a rigid ruler with sliding vertical extension has been used in order to overcome problem with obese patients. Vertical extensions have been placed on the carotid and femoral point of evaluation so that the distance can be measured without the encumbrance of the patient’s belly or breast. The values were corrected by a 0.8 factor accordingly to the PWV measurement methods consensus documents. This correction factor has been introduced to adjust for the contemporary travel of the pulse wave to the carotid artery and to the femoral artery through the aortic arch. In fact, measuring the direct distance between the two points lead to an overestimation of the real path length traveled by the pulse wave [21]. Each subject underwent two measurements, and the mean value was used for statistical evaluation. A PWV higher than 10 m/s was considered suggestive of arterial stiffness (vascular HMOD).

A trained cardiologist performed two-dimensional echocardiographic evaluations using a SONOS 5500 ultrasound system (Philips Healthcare, Andover, Massachusetts, USA) with a 2.5 MHz transducer. High-frame-rate grayscale loops were recorded from four-chamber, two-chamber, and three-chamber views at an average frame rate of 90 frames per second.

Left ventricular end-diastolic diameter (LVEDD), interventricular septal thickness, and posterior wall thickness were measured. EF was assessed using the Simpson method, while left ventricular mass (LVM) was calculated using the Devereux formula [22]: LVM (g) = 0.8 x 1.04 x [(LVEDD (cm) + septal thickness + posterior wall thickness)3 - LVEDD3] + 0.6.

LVM was indexed to both body surface area (BSA) and to height2.7 to yield the left ventricular mass indexed (LVMI). BSA was estimated using the DuBois formula: BSA (m2) = 0.007184 x height (cm)^0.725 x weight (kg)^0.425.

LVH was defined when LVMI exceeded 115 g/m2 for men and 95 g/m2 for women when indexed to BSA and 50 g/h2.7 for men and 47 g/h2.7 for women when indexed to h2.7.

Carotid ultrasonography was performed with participants in a supine position, with the neck slightly extended. A SONOS 5500 ultrasound system (Philips Healthcare) was used to examine the right carotid artery. The transducer was positioned perpendicularly to the arterial axis under B-mode guidance. Intima-media thickness (IMT) was measured on the posterior wall of the common carotid artery, 2 cm below the bifurcation, by identifying the hypoechoic and anechoic layers.

Two measurements per subject were recorded, and the mean value was used for statistical analysis. A pathological IMT was defined as >0.9 mm while carotid plaques were identified when IMT surpassed 1.2 mm.

The distribution of continuous variables was summarized using mean and Standard Deviation (± SD) (when normally distributed) or median and interquartile (IQR) (when skewed). Categorical variables were expressed as counts and percentages. T-test (or Mann-Whitney test for skewed variables) was used to compare continuous variables while Chi-square test was used to compare categorical variables between groups.

In particular, the association between Lp(a) and HMOD binary outcomes was assessed using Mann-Whitney test while Spearman correlation analyses was used for continuous outcomes. Multivariable logistic regression analysis adjusted by age, sex, diabetes, LDL cholesterol, statins, anti-hypertensive therapies and GFR was also performed for binary outcomes while linear regression was used for continuous outcomes.

The crude incidence of CV events was estimated using the Aalen-Johansen method considering death without CV event as a competing risk. The curves were compared between patients with Lp(a)>50 vs ≤50 using the Gray test. The cumulative incidence of death and of the composite outcome (CV event or death) was estimated using the Kaplan-Meier method and compared using the Log-rank test. Finally, unadjusted Cox models for CV events, death and the composite outcome were fitted using a restricted cubic spline function to model the Hazard Ratio (HR) of Lp(a). Multivariable Cox models adjusted by age, sex, diabetes, LDL cholesterol, statins, anti-hypertensive therapies and GFR were also fitted.

All statistical analyses were performed using R, with significance set at a two-tailed p-value < 0.05.