Lipoprotein (a) Testing in Patients With Atherosclerotic Cardiovascular Disease in 5 Large US Health Systems.
Nov 4, 2024DOI: 10.1161/jaha.124.035610Journal ArticleAHA
Written by: Nishant P ShahHillary MulderElizabeth LydonKaren ChiswellXingdi HuZachary LampronLauren CohenManesh R PatelSusan TaubesWenliang SongSuresh R MulukutlaAnum SaeedDaniel P MorinSteven M BradleyAdrian F HernandezNeha J Pagidipati
Find these authors on Roon
NS
Dr. Nishant Shah
Preventive Cardiology · Duke Hospital
MP
Dr. Manesh Patel
Interventional Cardiology · Duke University
Methods
A retrospective analysis was conducted using electronic health record data from 5 large health systems within the CardioHealth Alliance, including Allina Health, Duke University Medical Center, University of Pittsburgh Medical Center, Vanderbilt University Medical Center, and Ochsner Health System. Each health system participated in the Patient‐Centered Clinical Research Network, and electronic health record data were mapped into a common data model for analysis. Data elements include diagnosis and procedure codes, laboratory data, demographics, health system encounters, and medication data. The data used for analysis in this study are available from the corresponding author upon reasonable request.For the current analysis we included adults aged ≥18 years with established ASCVD defined as acute coronary syndrome (myocardial infarction and unstable angina), stable angina, transient ischemic attack, ischemic heart disease, peripheral artery disease (PAD), or revascularization procedures (coronary revascularization such as percutaneous coronary interventions or coronary artery bypass grafting, peripheral and cerebrovascular revascularization procedures). The diagnosis and procedure codes used for the analysis can be seen in Tables S1 and S2, respectively.Patients must have also had an ASCVD event within 5 years before an index date. For patients with a lipoprotein (a) test, the index date was defined as date of the first lipoprotein (a) test result in 2019 to 2021. Patients without any lipoprotein (a) test from 2015 to 2021 formed the non‐lipoprotein (a) test group. To permit comparisons of change in lipid‐lowering therapy (LLT) following lipoprotein (a) testing, the index date for this group was defined as the date of a randomly selected outpatient encounter in 2019 to 2021. Additionally, patients were required to have at least 2 encounters of any kind within the health system in the 2 years before the index date to capture patients who are active within the health system. The follow‐up period ranged from the index date to 6 months after until the end of study period (June 30, 2022).Variables collected for the study included age, sex, self‐identified race, ethnicity, vitals, medical history, medications, and laboratory data (lipoprotein (a) level, LDL‐C, total cholesterol, high density lipoprotein cholesterol, triglyceride levels, glycosylated hemoglobin, and creatinine). LLT was defined as either a statin at any dose, high‐intensity statin (rosuvastatin 20–40 mg daily or atorvastatin 40–80 mg daily), monoclonal antibody–based proprotein convertase substilisin/kexin type 9 inhibitor (PCSK9i), ezetimibe, bempedoic acid, bile acid sequestrants, and fenofibrates. Medical history, medication prescriptions, laboratory tests, and vitals were assessed on the basis of electronic health records in the 12 months before the index date.
Baseline characteristics for the overall cohort, including demographics, medical history, lab values, and concomitant medications, were stratified by the presence and absence of a lipoprotein (a) test result. Continuous variables were summarized as median (25th–75th percentile), and categorical variables were presented as frequencies (percentages). A robust Poisson multivariable regression model was used to determine factors associated with obtaining a lipoprotein (a) test. Factors were prespecified for inclusion in the model on the basis of clinical relevance. The effects shown from the multivariable model included all the variables shown in Table S3 as well as site. All effects are adjusted for all the other variables in the model. Missing values for covariates that had ≤10% missingness were imputed via a single random draw from regression‐based chained equations. Missingness >10% for important covariates resulted in the creation of a “Missing” category and categorization of the continuous covariate when applicable. Relative risks with 95% CIs and P values were presented for all variables in the model except for site to maintain site confidentiality. Continuous variables were checked for linearity with respect to the outcome; all relationships were linear.
Initiation of a medication was based on the presence of a prescription within 6 months following the index date and was assessed among patients without a corresponding prescription in the prior 12 months. Initiation of LLT within 6 months after the index date was described and compared for patients with versus without lipoprotein (a) testing and for those with lipoprotein (a) testing, elevated values (defined as lipoprotein (a) ≥50 mg/dL or ≥125 nmol/L) versus nonelevated, as well as by lipoprotein (a) level (<50, 50–100, or >100 mg/dL, among those with lipoprotein (a) measured in mg/dL). P values for comparisons between groups were calculated using χ2 or Fisher's exact tests, as appropriate.
A 2‐sided P value of <0.05 was considered nominally statistically significant. All analyses were performed by the Duke Clinical Research Institute (Durham, NC) using SAS version 9.4 (SAS Institute, Inc., Cary, NC). The data were extracted through SAS queries that were distributed to health systems and executed against the most recent Patient‐Centered Clinical Research Network Common Data Model. Limited data sets were delivered through site‐approved secure file transfer methods. The study was approved by the Duke University Institutional Review Board under a waiver of Health Insurance Portability and Accountability Act authorization and informed consent, as no informed consent was required for this observational analysis.
Results
Among 707 212 patients with ASCVD in the 5 health systems, 3437 (0.5%) had a lipoprotein (a) test between 2019 and 2021. Of the patients with a lipoprotein (a) test, 2781 had an ASCVD event or diagnosis in the 5 years before the lipoprotein (a) test and at least 2 encounters in the health system 2 years before the lipoprotein (a) test. After excluding patients with a prior lipoprotein (a) test between 2015 and 2018, 2587 (0.4% of the study cohort) remained in the lipoprotein (a) testing cohort. For the cohort of patients without lipoprotein (a) testing, 670 138 had at least 1 outpatient encounter between 2019 and 2021. Among these patients without lipoprotein (a) testing, 593 097 had a diagnosis of ASCVD in 5 years before the outpatient encounter and at least 2 encounters in the health system in the 2 years before the outpatient encounter. This created an overall study cohort of 595 684 patients with ASCVD active in the health systems with and without a lipoprotein (a) test (Figure 1).
sASCVD indicates atherosclerotic cardiovascular disease; and Lp(a), lipoprotein (a).
Baseline characteristics of the overall study cohort stratified by presence or absence of lipoprotein (a) testing are presented in Table 1. The median age of the overall cohort was 70 (quartiles 1–3:62–78) years. The cohort consisted of 45.0% women and 84.5%, 13.0%, and 1.4% White, Black, and Hispanic individuals, respectively. The majority of the study population had established coronary artery disease (75.4%), followed by PAD (31%), and stroke or transient ischemic attack (20.4%). In terms of risk factors, 84.0% had hypertension, 78.0% had hyperlipidemia (with 3.7% diagnosed with FH), and 35.9% had diabetes. The median LDL‐C was 81 (quartiles 1–3:62–107) mg/dL, and only 48.9% of the study cohort were on any statin therapy. Furthermore, 3.0% were on ezetimibe, 0.9% were on PCSK9i, 1.8% were on fenofibrate, and <0.01% were on bempedoic acid.
Baseline Characteristics of Patients With ASCVD With and Without a Lipoprotein (a) Test
Units: blood pressure, mm Hg; body mass index, kg/m2; creatinine, mg/dL; eGFR, mL/min per 1.73 m2; total cholesterol, mg/dL; HDL, mg/dL; triglycerides: mg/dL; LDL, mg/dL. Elevated lipoprotein (a) is defined as lipoprotein (a) ≥50 mg/dL or ≥125 nmol/L. ACE/ARB indicates angiotensin‐converting enzyme/angiotensin receptor blocker; eGFR, estimated glomerular filtration rate; GLP1 RA, glucagon‐like peptide 1 receptor agonist; HDL, high‐density lipoprotein cholesterol; LDL, low‐density lipoprotein cholesterol; LLT, lipid‐lowering therapy; PCSK9i, monoclonal antibody based proprotein convertase subtilisin/kexin type 9 inhibitor; and SGLT2i, sodium glucose cotransporter 2 inhibitor.
Other indicates those who do not self‐identify as White or Black.
The population with a lipoprotein (a) test were younger (median age, 61 versus 70 years), more frequently men (45.0% versus 41.1%), and less frequently Black individuals (11.6% versus 13.0%) compared with patients without a lipoprotein (a) test. In terms of medical comorbidities, patients with a lipoprotein (a) test had higher rates of ischemic stroke (25.2% versus 20.3%), hyperlipidemia (87.0% versus 78%), and FH (7.7% versus 3.7%) compared with those without a lipoprotein (a) test. In contrast, patients with a lipoprotein (a) test had lower rates of atrial fibrillation (17.7% versus 23.0%), hypertension (75.1% versus 84.1%), and diabetes (30.2% versus 35.9%) compared with those without a lipoprotein (a) test. LDL‐C was higher in those with lipoprotein (a) testing (87 versus 81 mg/dL), and a greater proportion were on statin therapy (70% versus 48.8%). Additionally, more patients with a lipoprotein (a) test were on ezetimibe (12.1% versus 2.9%) and PCSK9i (6.9% versus 0.8%) (Table 1).
Multivariable analysis suggested that several factors were independently associated with the likelihood of lipoprotein (a) testing, including diagnosis of coronary artery disease, ischemic stroke/transient ischemic attack, PAD, or heart failure (Figure 2, Table S3). A diagnosis of FH or hyperlipidemia, along with use of any prior LLT, were also positively associated with lipoprotein (a) testing. Of note, LDL‐C level appeared to be associated with lipoprotein (a) testing in a graded relationship, such that lower levels <130 mg/dL were not associated with lipoprotein (a) testing, but higher levels were increasingly associated with higher likelihood of receiving a lipoprotein (a) test (Figure 2). In contrast, older age, Black race, higher body mass index, current smoking status, and diagnosis of hypertension or diabetes were associated with a lower likelihood of lipoprotein (a) testing. Missing glycosylated hemoglobin and lipid laboratory values in the 12 months before the index date were also associated with lower likelihood of lipoprotein (a) testing.
The majority of lipoprotein (a) testing was reported in mg/dL (74.4%) compared with nmol/L (25.6%). Those with elevated values, defined as lipoprotein (a) ≥50 mg/dL or ≥125 nmol/L, (n=821, 31.7%) were more frequently Black individuals (18.2% versus 8.6%), with higher LDL‐C values (91 versus 85 mg/dL) and higher rates of LLT use (78.1% versus 73.5%) than those without elevated lipoprotein (a) (Table S4). Specifically, ezetimibe (16.3% versus 10.1%) and PCSK9i (8.5% versus 6.1%) were used at index more frequently in those with elevated lipoprotein (a).
With regard to initiation of LLT, those with a lipoprotein (a) test, regardless of the lipoprotein (a) value, more often initiated LTT within 6 months of the index date (Table 2). Specifically, patients with a lipoprotein (a) test were more frequently initiated on any statin therapy (30.3% versus 10.6%, P < 0.001), or high‐intensity statin therapy (14.6% versus 5.1%, P < 0.001) compared with those without a test. Initiation of PCSK9i (6.7% versus 0.3%, P < 0.001) and ezetimibe (7.65 versus 0.8%, P < 0.001) were more often in the group with lipoprotein (a) testing as well.
Initiation of LLT Within 6 Months After Index Date
Initiation defined by no prescription in the year before index date, but a prescription is present in the 6 mo following index date. For patients with lipoprotein (a) test, the index date was the date of the earliest lipoprotein (a) test in 2021. For patients without a lipoprotein (a) test, the index date was the date of a randomly selected outpatient visit in 2021. Any LLT indicates a patient was initiated on ≥1 of the following prescriptions: statins, PCSK9i, ezetimibe, or bempedoic acid. LLT indicates lipid‐lowering therapy; and PCSK9i, monoclonal antibody based proprotein convertase subtilisin/kexin type 9 inhibitor.
The lipoprotein (a) value itself was associated with initiation of nonstatin LLT (Figure 3). Among patients with lipoprotein (a) values, there was no difference in the initiation of any statin (32.1% versus 29.7%, P = 0.51), high‐intensity statin (16.2% versus 14.0%, P = 0.29), or in up‐titration from regular‐ to high‐intensity statin dosing (49.3% versus 41.1%, P=0.27) in those with elevated versus nonelevated lipoprotein (a) levels (Table 3). However, initiation rates of PCSK9i (10.9% versus 4.8%, P < 0.001) and ezetimibe (11.5% versus 5.9%, P < 0.001) were significantly higher in those with elevated lipoprotein (a) levels. Of note, there appeared to be a graded relationship such that those with highly elevated lipoprotein (a) levels (≥100 mg/dL) were more likely to initiate PCSK9i or ezetimibe than those with lipoprotein (a) 50–100 mg/dL or <50 mg/dL, respectively (PCSK9i: 15.5% versus 7.4% versus 5.5%, P < 0.001; ezetimibe: 15.6% versus 12.8% versus 7.3%, P < 0.001) (Table S5). This trend was not present for statin initiation.
Initiation defined by no prescription in the year before index date, but a prescription is present in the 6 mo following a lipoprotein (a) test. Above lipoprotein (a) threshold is defined as lipoprotein (a) ≥50 mg/dL or ≥125 nmol/L). Any LLT indicates a patient was initiated on ≥1 of the following prescriptions: statins, PCSK9i, ezetimibe, or bempedoic acid. BA indicates bempedoic acid; Eze, ezetimibe; LLT, lipid‐lowering therapy; Lp(a), lipoprotein (a); and PCSK9, monoclonal antibody‐based proprotein convertase subtilisin/kexin type 9 inhibitor.
Initiation of LLT Within 6 Months After Index Date by Elevated Lipoprotein (a)
Initiation defined by no prescription in the yearbefore index date, but a prescription is present in the 6 mo following a lipoprotein (a) test. Elevated lipoprotein (a) threshold is defined as lipoprotein (a) ≥50 mg/dL or ≥125 nmol/L. Any LLT indicates a patient was initiated on ≥1 of the following prescriptions: statins, PCSK9i, ezetimibe, or bempedoic acid. LLT indicates lipid‐lowering therapy; and PCSK9i, monoclonal antibody based proprotein convertase subtilisin/kexin type 9 inhibitor.
Discussion
Across 5 large US health systems within the CardioHealth Alliance, the frequency of lipoprotein (a) testing was low (0.4%) among patients with ASCVD. Disparities in lipoprotein (a) testing were apparent, as older age, Black race, higher body mass index, current smoking status, and diagnosis of hypertension or diabetes were associated with a lower likelihood of lipoprotein (a) testing. Testing for lipoprotein (a), regardless of lipoprotein (a) level, was associated with greater initiation of LLT, including statin, ezetimibe, and PCSK9i. Interestingly, an elevated lipoprotein (a) result was associated with ezetimibe and PCSK9i initiation but not statin initiation or up‐titration.
The low observed rate of lipoprotein (a) testing among a secondary prevention population is consistent with previous reports. In a large claims data analysis performed on 4 million patient records in Germany, the frequency of lipoprotein (a) testing was similarly low at 0.34%.18 An analysis of US claims data revealed that only 0.7% of secondary prevention patients were tested for lipoprotein (a).19 Additionally, an analysis across 6 academic health systems in the University of California health system found an overall prevalence of lipoprotein (a) testing to be 0.3% in patients with ASCVD or at risk of ASCVD.20 Reasons behind the consistently low testing rates for lipoprotein (a) in patients with ASCVD are likely multifactorial and may include lack of clinician understanding or awareness of lipoprotein (a) as a prevalent and causal risk factor for ASCVD.21 Further, current clinical guidelines are inconsistent about recommendations for lipoprotein (a) testing, though most recommend testing in high‐risk patients,3, 9 if not all patients at least once in a lifetime.4 In addition, there are no available therapies that are currently indicated to lower lipoprotein (a), though several are in clinical trials (NCT04023552; NCT05581303).22, 23 Furthermore, health systems may not have resources internally to test for lipoprotein (a) and may have to send laboratory tests out, which could add to the cost of the test and therefore less motivation by providers to order the test. Additionally, providers may be unaware of local laboratories outside the health system that could test lipoprotein (a) for lower cost if cost is a concern. However, even in the absence of lipoprotein (a)–lowering therapies, testing for lipoprotein (a) is still clinically relevant for risk stratification, aggressive preventive management such as early use of statins, and testing of first‐degree relatives.6, 11
Those who did undergo lipoprotein (a) testing in our study population were different in important ways from those who did not undergo testing, revealing potential disparities in testing patterns. Older age was associated with a lower likelihood of lipoprotein (a) testing; this may be related to the National Lipid Association guidelines, which specifically recommend lipoprotein (a) testing for individuals with premature ASCVD and who are therefore young by definition.9 Black race was also associated with a 24% lower likelihood of lipoprotein (a) testing; this likely represents a disparity in care, as Black individuals are known to have generally higher lipoprotein (a) levels than White individuals, with correspondingly higher lipoprotein (a)–related ASCVD risk.19, 24 Access to health care, socioeconomic status, patient comfort, and lack of provider awareness of high‐risk subgroups are possible explanations in general for lipid screening and management disparities in underrepresented minorities.25 As far as we are aware, this disparity in lipoprotein (a) testing across health systems has not been shown before and is important for addressing health equity in ASCVD prevention and management. Unsurprisingly, individuals with missing glycosylated hemoglobin and lipid levels were less likely to undergo lipoprotein (a) testing, potentially because they were not receiving routine preventive care.
In contrast, several factors were associated with greater likelihood of lipoprotein (a) testing, including prior coronary artery disease, ischemic stroke/transient ischemic attack, PAD, or heart failure. This pattern is consistent with prior studies18, 26 likely because of guideline recommendations for testing in high‐risk populations.3, 9 Similarly, those with hyperlipidemia or FH were more likely to undergo testing, which also has been seen in prior studies and is consistent with guideline recommendations and the observation that elevated lipoprotein (a) is more common in individuals with elevated LDL‐C or FH.27, 28 Interestingly, LDL‐C level appeared to have a graded relationship with lipoprotein (a) testing, with increasing levels beyond 130 mg/dL being associated with greater likelihood of lipoprotein (a) testing. Alternatively, patients with higher body mass index, active smoking, hypertension, and diabetes were less likely to be tested. One possible explanation for this observation could be that these factors alone either represented competing markers of risk to providers or were very prevalent across the population to justify lipoprotein (a) testing on the basis of these markers alone.
We also found that lipoprotein (a) testing regardless of lipoprotein (a) level, along with the level itself, are both associated with changes in LLT in the 6 months after testing. Lipoprotein (a) testing (compared with no testing) was associated with initiation of both statin and nonstatin therapies, while an elevated lipoprotein (a) level (compared with a nonelevated level) was associated only with initiation of PCSK9i and ezetimibe. These results suggest that clinicians may be acting on lipoprotein (a) results with more aggressive therapies, indicating that testing may have an impact on clinical management (though causality cannot be established in this observational analysis). This is consistent with a prior study from Germany, which showed increased treatment intensity after lipoprotein (a) testing18; our study extends these results to the United States. Further, our finding that statins were initiated regardless of lipoprotein (a) result may reflect overall clinician concern and/or better care in those with an lipoprotein (a) result, given that all of these individuals have a clear indication for statin therapy. Notably, however, only 30% of individuals with a lipoprotein (a) test were initiated on a statin after testing, compared with 11% in those without testing, reflective of a continued gap in guideline‐based care. In contrast with statins, PCSK9i and ezetimibe were more likely to be initiated in those with lipoprotein (a) testing than in those without, but among those with testing, were only more likely to be initiated in those with an elevated lipoprotein (a) level. This may indicate that clinicians are responding to the level itself by initiating more advanced LLT, which is consistent with guidelines that recommend more aggressive LDL‐C lowering in those with elevated lipoprotein (a).3, 9 Again, however, the overall rate of initiation of these therapies was relatively low, with 11% and 12% of individuals with elevated lipoprotein (a) initiating PCSK9i and ezetimibe, respectively.
Our study has several limitations that are important to note. First, this was a retrospective analysis that evaluated electronic health record data, so there is the potential for missing data, especially data generated outside the studied health systems; we attempted to minimize this risk by ensuring that all individuals in the study received regular care within these systems. Second, our study period overlapped with the advent of the COVID‐19 pandemic period, which could have decreased the frequency of encounters and laboratory values. Third, adherence patterns and intolerance to LLT could not be captured, which could have influenced the distribution of various LLT medications. Fourth, it is possible that patients with a lipoprotein (a) test before 2019 were not captured; however, we aimed to capture testing patterns and influences on lipid management over a contemporary guideline period.
Conclusions
Across 5 large US health systems, the frequency of lipoprotein (a) testing among patients with ASCVD remains low, at 0.4%. Additionally, disparities exist as older and Black individuals were particularly unlikely to have a lipoprotein (a) test. Those with lipoprotein (a) testing were more likely to initiate any LLT regardless of lipoprotein (a) level, including statins, compared with those without lipoprotein (a) testing. Initiation of LLT was higher among those with a lipoprotein (a) test and elevated levels, though driven mostly by nonstatin therapies like ezetimibe and PCSK9i. In general, the overall initiation of LLT remained low in this population, despite clear guideline indications for LLT in patients with ASCVD. Thus, there is a critical need for multidisciplinary and inclusive approaches to raise awareness of lipoprotein (a) testing and its implications for aggressive preventive management. Such awareness may increase if lipoprotein (a)–lowering therapies, which are currently being tested, are shown to provide clinical benefit.
Sources of Funding
This study was supported by Novartis Pharmaceutical Corporations.
Disclosures
Dr Shah has received research grants from Amgen, Janssen, and the NIH. Dr Shah also is a consultant/advisor for Esperion, Amgen, and Norvartis. Dr Hu is employed by Novartis Pharmaceuticals Corporation. Dr Patel has received research grants from Amgen, Bayer, Janssen, Heartflow, National Institutes of Health, and the Gordon and Betty Moore Foundation. Dr Patel is a consultant/advisor for Bayer, Janssen, Novartis, and Amgen. S. Taubes is employed by Novartis Pharmaceuticals Corporation. Dr Pagidipati has received research grants from Amgen, AstraZeneca, Baseline Study LLC, Boehringer Ingleheim, Duke Clinical Research Institute, Eggland's Best, Eli Lilly & Company, Novartis, Novo Nordisk Pharmaceutical Company, Sanofi‐S.A, and Verily Sciences Research Company; and is a consultant/advisor for AstraZeneca, Boehringer Ingleheim, Esperion Therapeutics, Eli Lilly & Company, and Novo Nordisk Pharmaceutical Company. The remaining authors have no disclosures to report.
Supporting information
Tables S1–S5
NS
Dr. Nishant Shah
Cardiology · Preventive CardiologyDuke University Hospitalnishant
