Lipoprotein(a) and the association with atrial fibrillation

Díaz‑Peromingo et al²⁴ recruited 101 patients with AF and an equivalent number of patients without AF, matched for sex and age. They compared Lp(a) concentrations between the groups and found no significant differences. Nevertheless, they included a relatively small number of patients, which might have potentially influenced the results. Naji et al²⁵ identified Lp(a) as a potential biomarker predicting the recurrence of AF after electrical cardioversion (EC). They analyzed retrospective medical data of 79 AF patients undergoing EC and followed‑up for 2 years after the procedure. The patients were divided into 2 groups of those with and without AF recurrence. No difference in Lp(a) concentration was found between the groups. In addition, Lp(a) levels did not exhibit any association with AF recurrence.²⁵ Li et al²⁶ investigated the significance of Lp(a) in AF occurrence in 679 patients with chronic HF (CHF) divided into 2 groups based on the AF status at baseline. The patients with AF had higher Lp(a) concentration than those without AF at baseline. Furthermore, the patients without AF were followed‑up for new‑onset AF, demonstrating that Lp(a) was an independent predictor of new‑onset AF in a multivariable model.²⁶ Aronis et al²⁷ analyzed if high Lp(a) concentration increases the risk of AF incidence in a large prospective community‑based cohort from the ARIC (Atherosclerosis Risk in Communities) study. They included 9908 participants free of AF at baseline and measured their Lp(a) concentration. The researchers analyzed the incidence rate of AF in the groups with increasing Lp(a) concentration: 1) up to 10 mg/dl, 2) above 10 to 20 mg/d, 3) above 20 to 30 mg/dl, 4) above 30 to 50 mg/dl, and 5) above 50 mg/dl. No association was found between Lp(a) concentration and the rate of AF, regardless of the adjustment model.²⁷ On the contrary, Tao et al²⁸ obtained results that contradicted those of other researchers. In a group of 13 533 patients (4511 with AF and 9022 without AF) divided into 4 groups based on the quartiles of Lp(a) concentrations, they found that Lp(a) concentrations were lower in the patients with AF than those without AF. The authors hypothesized that this observation might have been influenced by the ethnicity of the patients, that is, Han ethnicity, which is known to have lower Lp(a) levels than other ethnicities. Moreover, this inverse correlation was observed only in women and was confounded by coronary heart disease and type 2 diabetes mellitus. Potential mechanisms responsible for this inverse correlation, as indicated by the authors, could be cell membrane stabilization preventing the abnormal discharge of cardiomyocytes, or inflammation status, as the patients with AF had higher levels of C‑reactive protein, a biomarker of inflammation. The authors argued that during inflammation total cholesterol and LDL cholesterol levels are reduced and potentially the same may apply to Lp(a).²⁸

Li et al²⁹ studied the interplay between Lp(a), AF, and coronary artery disease (CAD). They included 2048 patients and divided them into 4 subgroups based on the AF and CAD status: 1) no CAD or AF, 2) AF without CAD, 3) CAD without AF, and 4) CAD and AF. The authors found that patients with CAD and AF had higher levels of Lp(a) than those with CAD but without AF. This result was not observed in patients without CAD. Moreover, in a multivariate logistic regression, Lp(a) was an independent risk factor for AF in CAD patients.²⁹ Toprak et al³⁰ examined the association between several biomarkers and the incidence of AF in a large cohort of 42 280 individuals without AF at baseline. They found no association between the Lp(a) level and the occurrence of AF. Zhang et al³¹ performed a retrospective case‑control study with 89 patients newly diagnosed with nonvalvular AF (NVAF) and 88 sinus rhythm patients as controls. Their findings indicated that Lp(a) concentration was higher in the patients with NVAF than in the controls. The authors also found that Lp(a) was an independent risk factor for the incidence of newly diagnosed NVAF; however, the sensitivity and specificity were relatively low (53.9% and 73.9%, respectively). Manla et al³² compared patients with Lp(a) levels above and below 125 nmol/l (approximately 50 mg/dl). They found no difference in the incidence of AF between the groups. In contrast, Burzyńska et al¹⁹ recruited 2475 patients referred to outpatient cardiology clinics and divided them based on Lp(a) levels adopting a cutoff of 30 mg/dl. They found that AF was independently related to Lp(a) levels equal to or above 30 mg/dl.¹⁹

Of the 10 studies discussed, 5 did not identify any difference in Lp(a) levels between AF patients and healthy controls; 4 found that a high Lp(a) level was associated with the incidence of AF; and 1 concluded that a low Lp(a) level was associated with the incidence of AF. Noteworthy, 2 of the 4 studies that identified high Lp(a) levels as a potential AF risk factor included a specific cohort of patients, that is, patients with CHF and those with CAD. A different study design (retrospective vs prospective) can partially explain discrepancies within the existing evidence. However, studies based on the largest cohorts did not find an association between high Lp(a) level and AF incidence.^27,30 All studies discussed in this section, along with supplementary information, are presented in Table 1.

Lipoprotein(a) as a causal risk factor of atrial fibrillation in Mendelian randomization studies

Zanetti et al³³ conducted a Mendelian randomization study assessing the causal role of Lp(a) alongside other biomarkers in several cardiovascular diseases, including AF. The authors used data from the United Kingdom (UK) Biobank, including up to 451 933 participants. AF occurred in 13 312 cases. Although several causal associations were observed, including the causal role of Lp(a) in CAD, such a relationship was not noted in the case of Lp(a) in AF.³³ Satterfield et al³⁴ conducted another Mendelian randomization study, including 804 507 participants from the Electronic Medical Records and Genomics Network, UK Biobank, and Million Veteran Program cohorts. The authors mainly focused on analyzing the association between Lp(a) levels and the risk of ASCVD subtypes. Nevertheless, using the Mendelian randomization‑phenome‑wide association analyses, they found a novel association between genetically increased Lp(a) levels and the risk of AF.³⁴

Jiang et al³⁵ also performed a Mendelian randomization study using the UK Biobank resources. They included 377 600 participants with measured Lp(a) concentrations and known single nucleotide polymorphisms (SNPs) associated with Lp(a). The analysis found that genetically high Lp(a) levels were associated with an increased risk of AF. This was also observed in multivariable models adjusted for other lipid profile components. The effect size was a 1.001 odds ratio (OR) for each 1‑SD increase in the Lp(a) measure. The authors concluded that Lp(a)-lowering therapies might benefit the primary prevention of AF.³⁵ Mohammadi‑Shemirani et al³⁶ also performed a Mendelian randomization study using data from the UK Biobank. They assessed the occurrence of AF (20 432 cases) within the studied population (435 579 participants) in the context of measured and genetically predicted Lp(a) levels. Consistent with Jiang et al,³⁵ their findings indicated that an increase in Lp(a) concentration was associated with an increased risk of AF (for a 23‑mg/dl increase in Lp[a] hazard ratio was 1.03 for measured Lp[a], and OR was 1.03 for genetically predicted Lp[a] level). Notably, although ASCVD mediated some of this risk, Lp(a) was a risk factor for AF independent of its established effects mediated via ASCVD.³⁶

Wang et al³⁷ conducted another Mendelian randomization study evaluating the relationship between Lp(a) and a risk for various CVDs. The authors based their research on a different Lp(a) dataset, specifically employing data from Neale Lab, which comprised 377 590 individuals of the European ancestry. They found no causal association between Lp(a) and AF.³⁷ Xia et al³⁸ recruited 1256 individuals and measured their Lp(a) serum concentrations. For Mendelian randomization analyses, they used published data from genome‑wide association studies. However, their findings were in contrast with the previously mentioned ones; they found that genetically increased Lp(a) level was inversely associated with the risk of AF. The possible discrepancies between these studies may result from the inclusion of different ethnic groups, which is known to affect Lp(a) concentration.³⁸ The last discussed Mendelian randomization study was conducted by Tao et al.³⁹ The authors found that genetically predicted high levels of Lp(a) are associated with an increased risk of AF in the European population. They also showed that inhibiting the LPA gene may have a potential protective effect partially mediated by lowering coronary heart disease risk.³⁹

In summary, of the 7 identified Mendelian randomization studies, 4 found high levels of Lp(a) being causally associated with the increased risk of AF. The other 2 studies found no causal association between Lp(a) and AF, while 1 revealed an inverse association between these variables. This controversial inverse association was obtained in a study with the smallest sample size, which might have caused some bias. Different or even contradictory results from the same biobank analysis can be surprising. However, it is important to note that the populations presented in Table 2 refer to the population used to obtain Lp(a) levels, and Mendelian randomization needs other studies (not presented in Table 2) to derive genetic associations with AF from an independent validation cohort, often from publicly available genome‑wide association studies (GWAS). Even if the same biobank was used for Lp(a) level analyses, using different GWAS explains the heterogeneity in the obtained results. Overall, the Mendelian randomization studies advocate for the causal association of Lp(a) with AF, contrary to the studies described in the previous section. All studies discussed in this section, along with supplementary information, are presented in Table 2.

Lipoprotein(a) as a predictor of thromboembolic events in patients with atrial fibrillation

Igarashi et al⁴⁰ recruited 172 patients with NVAF. They divided them into 2 groups of thromboembolic (patients with a recent embolic event or a thrombus in the LA found on transesophageal echocardiography [TEE]) and nonthromboembolic. The patients from the thromboembolic group had higher Lp(a) concentrations than those from the nonthromboembolic group. Further analysis showed that Lp(a) concentration equal to or above 30 mg/dl was an independent predictor of thromboembolism in the NVAF patients.⁴⁰ Yan et al⁴¹ evaluated the use of Lp(a) in assessing the risk of thrombotic events among NVAF patients with low CHA₂DS₂-VASc scores. They included 595 patients and divided them into 2 groups based on the incidence of thrombotic events. The study showed that the patients who experienced thrombotic events had higher Lp(a) concentrations than those who did not experience such events. The authors also found that Lp(a) level was an independent risk factor for thrombotic events, as assessed in the multivariate logistic regression.⁴¹

Song et al⁴² conducted a large cross‑sectional study including 16 357 patients suffering from NVAF. They analyzed the association between Lp(a) levels and the risk for ischemic stroke and systemic embolism incidence in patients grouped according to Lp(a) level quartiles. An increase in Lp(a) level was found to increase the risk of both conditions, suggesting its potential role in the pathogenesis of thromboembolism. Guo et al⁴³ evaluated the possible role of Lp(a) as a predictive factor for stroke in 301 NVAF patients. The patients were divided into 2 groups based on the occurrence of acute ischemic stroke during follow‑up. The patients from the stroke group had a higher Lp(a) concentration than those from the nonstroke group. The authors constructed a nomogram model for predicting acute ischemic stroke in elderly patients with NVAF. In addition to Lp(a) concentration, neutrophil‑to‑lymphocyte ratio, red cell distribution width (RDW), systolic pressure, history of stroke, hyperlipidemia, and high‑density lipoprotein cholesterol concentration were included in the model. This model was more effective at predicting stroke than the CHA₂DS₂-VASc score, when assessing the area under the receiver operating characteristic curve.⁴³ Hou et al⁴⁴ retrospectively analyzed 313 NVAF patients, dividing them into 2 groups based on the occurrence of acute ischemic stroke during follow‑up. They found that Lp(a) concentration was higher in the patients who developed acute ischemic stroke. The authors also showed that the CHA₂DS₂-VASc‑60 score combined with Lp(a) and RDW predicted the occurrence of stroke better than the CHA₂DS₂-VASc‑60 score alone.⁴⁴ The reports by Guo et al⁴³ and Hou et al⁴⁴ were conducted in the same hospital during a similar period, and they analyzed similar research problems. Thus, caution is required when extrapolating the results to other institutions. Furthermore, Zhang et al⁴⁵ retrospectively analyzed 2258 AF patients using the propensity score matching method to create 2 groups of patients who experienced stroke and those who did not. They found that the first group had higher levels of Lp(a). Moreover, Lp(a) levels above 30 mg/dl were associated with an increased risk of stroke in the unadjusted model.⁴⁵

All presented studies found some association between the risk of thromboembolic events and the high level of Lp(a). Some works showed an increased concentration of Lp(a) in the subgroup of patients who experienced thromboembolic events. However, this does not necessarily indicate causality, and may only be an indicator. Two studies concluded that high Lp(a) levels are an independent risk factor for thromboembolism / stroke. Nevertheless, the findings in this field require further exploration to clarify the mechanisms underlying this association. All studies discussed in this section, along with supplementary information, are presented in Table 3.

Lipoprotein(a) correlation with the presence of left atrial thrombus in patients with atrial fibrillation

Igarashi et al⁴⁶ evaluated 150 patients with AF. The patients underwent TEE and were divided into 2 groups based on the presence of LAT. The patients with LAT had higher Lp(a) concentrations than those without. Moreover, in a multivariate analysis, Lp(a) concentration was found to be an independent positive predictor of LAT.⁴⁶ Saeed et al⁴⁷ obtained results that were consistent with those of Igarashi et al.⁴⁶ They included 50 AF patients divided into 2 groups based on the presence of LAT. It was found that patients with LAT had higher Lp(a) concentrations than the patients without this condition.⁴⁷ Kamili et al⁴⁸ performed a cross‑sectional study that included 481 patients suffering from NVAF. They were classified into 2 groups with and without LAT or spontaneous echo contrast (SEC) found on TEE. The study showed that Lp(a) concentrations were higher in the LAT/SEC group than the non‑LAT/SEC group. Moreover, high Lp(a) concentration was an independent predictor of LAT/SEC even after adjustment for propensity score matching. Thus, incorporating Lp(a) and LA diameter into the CHA₂DS₂-VASc score improved its predictive value for the LAT/SEC.⁴⁸

All presented studies consistently found that patients with LAT had higher Lp(a) levels. Again, this may not be a causative factor, although consistency within this subject is fascinating and merits further exploration. Table 4 presents all studies discussed in this section, along with supplementary information.