|Home | About | Journals | Submit | Contact Us | Français|
Genome-wide association studies (GWAS) in European Americans have reported several SNPs in the lipoprotein lipase (LPL) gene associated with plasma levels of HDL-C and triglycerides. However, the influences of the LPL SNPs on longitudinal changes of these lipids have not been systematically examined.
Based on data from 2045 African American and 2116 European American young adults in the CARDIA Study, we investigated cross-sectional and longitudinal associations of lipids with 8 LPL SNPs, including two that have been reported in GWAS. Plasma levels of HDL-C and triglycerides were measured at seven examinations during 20 years of follow-up. In European Americans, rs328 (Ser447Stop), rs326 and rs13702 were significantly associated with cross-sectional interindividual variations in triglycerides and HDL-C (p<0.005) and with their longitudinal changes over time (p < 0.05). The minor alleles in rs326, rs328, and rs13702 that predispose an individual to lower triglycerides and higher HDL-C levels at young adulthood further slow down the trajectory increase in triglycerides and decrease in HDL-C during 20 years of follow-up. In African Americans, these 3 SNPs were significantly associated with triglycerides but only rs326 and rs13702 associated with HDL-C (p<0.008). Rs328 showed a stronger association in European Americans than African Americans and adjustment for it did not remove all of the associations for the other SNPs. Longitudinal changes in either trait did not differ significantly by SNP genotypes in African Americans.
Our data suggest that aging interacts with LPL gene variants to influence the longitudinal lipid variations and there is population-related heterogeneity in the longitudinal associations.
Lipoprotein lipase (LPL), associated with the luminal endothelial surface of arteries and capillaries of peripheral tissues,1 is a key enzyme in the metabolism of lipoproteins. It hydrolyzes plasma lipoprotein triglycerides into free fatty acids and glycerol, converts very-low-density lipoprotein (VLDL) to low-density lipoprotein (LDL),1,2 and enhances lipoprotein-receptor interactions.3 Since the hydrolysis of triglyceride-rich lipoproteins and production of VLDL contribute to high-density lipoprotein cholesterol (HDL-C) metabolism, LPL activity also plays a role in the regulation of HDL cholesterol levels.1 Plasma LPL activity and mass are associated with triglycerides and HDL-C in the population at large.4,5 Several functional polymorphisms in the LPL gene have been studied in numerous studies with regard to their association with lipid profile and risk of coronary heart disease (CHD).6 Among those, the LPL Ser447Stop variant (rs328) is most common with a frequency of 5.6%–21.1% in various populations.6 It involves a C to G transversion that converts the serine 447 codon to a premature termination codon, shortening the LPL molecule at the C-terminal. In a meta-analysis, the LPL Ser447Stop variant was consistently associated with baseline triglyceride and HDL-C levels and predicted CHD risk.6 Furthermore, additional variants in the LPL gene have been identified in recently published whole genome-wide association studies (GWAS).7–10
Since triglycerides and HDL-C change with age,11 it is important to investigate whether the genetic polymorphisms also influence the longitudinal changes in triglycerides and HDL-C within the same individuals: understanding the evolution of these blood lipids over time can enhance information used in risk prediction and intervention for CHD. Evidence from two previous studies suggests differential influence of the LPL Ser447Stop genotype on the longitudinal patterns of triglycerides and/or HDL-C. 12,13 Most of the participants in these 2 studies were European Americans and only one LPL variant was examined. In this study, we investigated the associations of 8 LPL SNPs and their haplotypes, including the Ser447Stop variant, with longitudinal trends of triglycerides and HDL-C in a large cohort from the Coronary Artery Risk Development in Young Adults (CARDIA) Study including both African and European Americans, spanning 20 years and 7 examinations.
A total of 5,115 black and white adults aged 18–30 years attended the baseline CARDIA exam in 1985–1986; 4,593 (89.8%) of the surviving cohort were re-examined in 1987-88 (year 2); 4,279 (83.7%) in 1990-91 (year 5); 4012 (78.4%) in 1992–1993 (year 7), 3,950 (77.2%) in 1995-96 (year 10); 3672 (74.0%) in 2000–2001 (year 15), and 3549 (73.0%) in 2005–2006 (year 20). The cohort was recruited mostly by telephone, randomly sampled from populations in Birmingham, AL; Chicago, IL; and Minneapolis, MN; and through membership of a large prepaid health plan in Oakland, CA. Approximately equal numbers of blacks and whites, men and women, persons aged 18–24 and 25–30, and those with high school or greater education or less than high school education were recruited at each site. This study was approved by the institutional review board at participating Universities and all participants gave informed consent.
Information on clinical measurements, including demographic characteristics, lifestyle habits, anthropometric variables, physical activity, and medical history, are provided in additional file S. Demographic and lifestyle characteristics are also presented in Table 1. Blood samples were drawn after an overnight fast. Plasma was stored at −70°C. Triglycerides was measured enzymatically within six weeks of collection.14 HDL-C was determined after precipitation with dextran sulfate/magnesium chloride of lipoproteins containing low-density lipoprotein cholesterol.15 Comparability of the lipid assays was scrupulously assessed across the examinations.
Genotyping of rs328 was performed on the participants who attended the year 10 examination using the TaqMan assay (Applied Biosystems, Foster City, CA) and valid results were obtained in 3,817 participants. Genotype data for the following 7 LPL SNPs were obtained on participants who were examined at years 15 (94%) or 20 (6%) depending on sample availability: rs34309063, rs13266204, rs10104051, rs248, rs326, rs13702, and rs9644636. Genotyping for these SNPs was performed as part of the Modeling DNA Diversity in Reverse Cholesterol Transport study, an ancillary study to CARDIA. The 7 SNPs were also genotyped using the TaqMan method. Valid results for these SNPs were obtained from 4,039 participants. This study included 4,162 participants who had triglycerides and HDL-C measurements from at least two examinations and genotype data for at least one of the 8 LPL SNPs.
The distributions of HDL-C and triglyceride were evaluated by skewness and kurtosis statistics and visualization of histogram. A skewness or kurtosis of larger than 1.5 was considered departure from the normal distribution. Triglyceride values were naturally log-transformed to normalize their distributions and the transformed values were used in all statistical tests. During quality control analysis, we found that plasma cholesterol measurements were systematically elevated at year 2, thus HDL-C levels at year 2 were multiplied by 0.9681 prior to analysis, as recommended by the CARDIA Quality Control Committee.
All genetic analyses were conducted in European Americans and African Americans separately. The χ2 goodness-of-fit test was used to evaluate the Hardy–Weinberg equilibrium. Among the 8 SNPs, rs34309063, rs13266204, rs248, rs328, and rs9644636 had < 30 counts of homozygosity for the minor allele in one or both race groups. For these SNPs, participants with minor allele homozygosity and heterozygosity were pooled into one group in genetic association analysis. Genotypes for the other 3 SNPs were analyzed as 0, 1 or 2 copies of the minor allele in an additive genetic model. Analysis of variance implemented in PROC GLM procedure of SAS was used to assess the associations of the LPL SNPs with triglycerides and HDL-C at each examination and with changes in triglycerides and HDL-C between baseline and follow-up examinations. Both types of analyses were adjusted for age, gender, and field center. To reduce the extent of multiple testing, the analysis of lipid changes was only conducted for examination years 10, 15 and 20 and for SNPs that showed significant associations in the overall association analysis as discussed below.
To investigate the overall associations of the LPL SNPs with the lipids across the 7 examinations, lipid measurements from these examinations were analyzed simultaneously using repeated measures regression analysis (SAS PROC MIXED), adjusting for age, examination year, gender and field center. The analysis accounts for within-person correlation between exams by assuming compound symmetry, that is, a constant correlation for these lipid fractions between pairs of examinations.
For SNPs that showed significant associations in the overall association analysis, mean levels of the lipids at each examination by genotype group were calculated in PROC MIXED models after adjusting for baseline age, gender, and field center. The adjusted means were used to plot the longitudinal pattern of the lipids by genotypes in Figures 1 and S1. When a linear trend was evident, tests of interaction between genotype and follow-up time as a continuous variable were conducted in PROC MIXED models that adjusted for baseline age, gender, and field center. The testing for SNP-by-follow-up-time interactions was used as a proxy for the testing for SNP-by-age interactions because follow-up time provides more information than longitudinal age for this type of data (see Discussion for details).
The genetic analysis program Haploview 16 was used to investigate LD patterns and calculate pairwise LD statistics (D’ and r2) among the SNPs. To reduce multiple testing, haplotype-based analysis was only performed on haplotype blocks containing significant SNPs from the single-SNP analysis. Haplotypes were inferred using PHASE version 184.108.40.206,18 Please see additional file S for details about the probability distribution of haplotypes. For each individual, the most likely haplotype pairs were used. The overall associations of the lipid variables with haplotypes across the 7 examinations were assessed in the mixed analysis models in which the probability assignment for corresponding haplotype pairs was used as a weight in the analysis. Haplotypes with frequency >5% were analyzed simultaneously in the same model. More rare haplotypes with frequency ≤ 5% were pooled. Haplotype analysis of longitudinal change in triglycerides and HDL-C at year 20, which does not involve correlated observations, was conducted using a generalized linear regression model implemented in the program haplo.stats for R,19,20 in which haplotype ambiguity was taken into account.
In the above single-SNP and haplotype analyses of longitudinal changes, a multivariable adjustment was performed by additionally adjusting for BMI (at baseline and change from baseline), smoking status (at baseline and current examination), alcohol consumption (ml/day, at baseline and change from baseline), and physical activity score (at baseline and change from baseline).
Since rs328 is the only known functional SNP among the 8 SNPs, we additionally adjusted for rs328 in the analyses of other SNPs that showed significant associations to evaluate whether the associations for other SNPs were driven by LD with rs328.
In a subset of 1622 African Americans for whom percentage of African ancestry was estimated,21 we repeated the analyses for significant genetic associations with additional adjustment for the percentage of African ancestry. We also repeated the analyses of longitudinal lipid changes, including the test of interactions between follow-up time and SNPs, after excluding participants who were on lipid-lowering medications (n=338) or lost to follow-up (n=1646). Please see additional file S for details.
Table 1 presents demographic and lifestyle characteristics at baseline in African Americans and European Americans, respectively. Table 2 presents mean levels of triglycerides and HDL-C by examination and race. Pooling all 7 examinations yielded 29134 possible records for both triglycerides and HDL-C.
Table 3 presents information on the 8 LPL SNPs stratified by race. Genotype distributions for all of the 8 SNPs agreed with Hardy-Weinberg equilibrium in both African Americans and European Americans (p>0.05). Based on pair-wise D’ measurement, the last 4 SNPs (rs326, rs328, rs13702, and rs9644636) formed a distinct haplotype block in both groups (Figure S2). Among the SNPs, the magnitude of correlations (r2) was low, except that SNPs rs34309063 with rs13266304 and rs326 with rs13702 were at least moderately correlated (r2 > 0.50) (Figure S3). The correlations in European Americans were higher than in African Americans.
P-values for the association tests between each of the 8 LPL SNPs and the two lipids at each examination are summarized in Table S1. In both groups, rs326 and rs13702 were consistently associated with triglycerides and HDL-C across the seven examinations (p<0.05). In contrast, rs328 was consistently associated with triglycerides in both groups and with HDL-C in European Americans only. Table 4 presents overall associations of the LPL SNPs with triglycerides and HDL-C in the pooled analysis of the 7 examination data using mixed analysis models. Consistent with the overall pattern of associations across individual examinations, rs326, rs328, and rs13702 were significantly associated with triglycerides in both groups and with HDL-C in European Americans; the G, G, and C alleles in rs326, rs328, and rs13702, respectively, were associated with lower levels of triglycerides and higher levels of HDL-C. In African Americans, the G and C alleles of rs326 and rs13702 showed significant and positive associations with HDL-C. The association between rs328 and HDL-C in African Americans was in the same direction as that in European Americans, but the magnitude of the association was weaker and did not reach statistical significance (p=0.21). We repeated the analysis for rs326 and rs13702 after additionally adjusting for rs328, the variant encoding for the 447 stop codon in LPL. The additional adjustment modestly attenuated, but did not abolish, the associations of rs326 and rs13702 with triglycerides or HDL-C in European Americans (table 4). In African Americans, the additional adjustment reduced the associations between both SNPs and triglycerides to below statistical significance but did not result in noticeable changes in the associations for HDL-C.
Tables 5 and and66 present mean changes in log-transformed triglycerides and HDL-C from baseline to years 10, 15, and 20 by genotypes for rs326, rs328, and rs13702 after adjustment for baseline age, gender, and field center. In African Americans, the changes in log-transformed triglycerides (table 5) and HDL-C (table 6) from baseline to the 3 follow-up examinations did not differ significantly across genotype groups (p>0.05). In European Americans, there was a trend toward less increase in triglycerides from baseline to each of the 3 latter examinations associated with the G, G, and C alleles in rs326, rs328, and rs13702 compared to the reference homozygosity groups; this genotype-related difference reached statistical significance at all of the 3 examinations for rs326, at year 20 examination for rs328, and years 10 and 20 examinations for rs13702. Moreover, the genotype-related gap became mostly widened at year 20 examination and was only slightly attenuated in the multivariable adjustment (table 5). For HDL-C in European Americans (table 6), the G, G, and C alleles in rs326, rs328, and rs13702 tended to be associated with smaller reductions in HDL-C at years 10 and 15 and larger increases at year 20 compared to the reference homozygous groups. These effects were statistically significant at year 20 for all of the three SNPs. The associations at year 20 were not attenuated for rs326 and rs13702 after the multivariable adjustment. To evaluate whether the above observed longitudinal changes associated with rs326 and rs13702 in European Americans were explained by LD with rs328, we additionally included rs328 as a covariate for the analysis of lipid changes at year 20. The additional adjustment did not abolish the associations: for triglycerides change, p=0.02 for rs326 and 0.04 for rs13702; for HDL-C change, p=0.04 for rs326 and 0.06 for rs13702.
For illustrative purposes, the longitudinal patterns of log-transformed triglycerides and HDL-C by rs13702 and rs328 genotypes are presented in Figure 1 for European Americans and Figure S1 for African Americans. As an example to illustrate the interactions of longitudinal time with genotype in European Americans, at the baseline exam, CC homozygotes for rs13702 had on average 8.0 mg/dL lower triglycerides and 1.9 mg/dL higher HDL-C than TT homozygotes, this gap was widened to 21.2 mg/dL for triglycerides and 4.2 mg/dL for HDL-C twenty years later. Since the log-transformed triglycerides across the 7 examinations exhibited a linear trend, we tested the interactions between the 3 SNPs and follow-up time (as a continuous variable) on log-transformed triglycerides using mixed analysis models. Consistent with the tests of lipid changes at individual follow-up examinations, there was a significant interaction between each of the 3 SNPs and follow-up time in European Americans, independent of the influence of baseline age, gender, field center, BMI, smoking status, alcohol consumption, and physical activity (Table 5). In African Americans, there was no significant interaction between the 3 SNPs and follow-up time (data not shown).
Haplotype analysis was performed for haplotypes inferred based on rs326, rs328, rs13702, and rs9644636. Four and five haplotypes with frequency > 5% were reconstructed for European Americans and African Americans, respectively (Table S2).
Cross-sectional associations: In both race groups, two haplotypes (GGCT and GCCT) were significantly associated with triglycerides and HDL-C in the pooled analysis using mixed analysis models (Table S2). For triglycerides, GGCT had a stronger effect than GCCT in both race groups; for HDL-C, GGCT had a stronger effect in European Americans while its effect was not substantially different from GCCT in African Americans. This race-related difference in haplotype association was likely due to different effect of rs328 between the two groups.
Longitudinal changes at year 20 compared to baseline: In European Americans, only GGCT was significantly associated with changes in triglycerides and HDL-C (Table S3). One copy of this haplotype was associated with 0.08 ± 0.03 mg/dL (p=0.007) less increase in triglycerides and 1.35 ± 0.67 mg/dL (p=0.0457) more increase in HDL-C than the common haplotype ACTT. Similar results were observed in the multivariate adjustment models (data not shown). There was no significant haplotype association in African Americans.
In both race groups, additional adjustment for the 4 SNPs that were used to reconstruct the haplotypes abolished all the haplotype associations in both cross-sectional and longitudinal analyses (data not shown), indicating that the haplotype analysis did not contribute additional information beyond the single SNP approach.
In African Americans, additional adjustment for the percentage of African ancestry did not materially change the significant associations reported in table 4 (data not shown). Also, excluding participants who were on lipid-lowering medications or lost to follow-up did not result in material changes in the associations between SNPs and longitudinal lipid changes (data not shown).
This study is based on the large cohort of the CARDIA study that has collected data from 7 examinations during 20 years of follow-up. In European Americans, the SNPs rs328 (Ser447Stop), rs326 and rs13702 were significantly associated with cross-sectional variations in triglycerides and HDL-C as well as with longitudinal changes in the two traits during 20 years of life span (p < 0.05). In African Americans, these 3 SNPs were significantly associated with interindividual variation in triglycerides but only rs326 and rs13702 associated with HDL-C (p<0.008). The SNP rs328 was not the statistical variable explaining most of the observed associations and showed a stronger association in European Americans than African Americans. The associations observed in African Americans did not change significantly over time.
A meta-analysis of previous candidate gene studies of LPL functional variants reported 3 variants that showed consistent associations with triglycerides and HDL-C.6 Of the 3 significant variants, only rs328 has a minor allele frequency larger than 5%. More recently, common SNPs in the LPL gene have been reported in several GWAS of lipid profile, which have been performed predominantly in populations of European ancestry.7–10 From these studies, 4 common SNPs from the LPL gene were associated with triglycerides at the genome-wide significance level: rs328, rs2197089, rs325, and rs326;7–10 for HDL-C, 3 common SNPs were identified: rs328, rs2197089, and rs326.7,8,10 Since not all of the reported SNPs are in high LD, these multiple variants indicate the possibility of multiple independent signals from the LPL gene. Our study, which included two of the reported SNPs from the GWAS, contributes to this body of literature by showing that the associations of blood lipids with rs328 do not completely explain the observed associations for rs326 and rs13702 in European Americans.
As the majority of candidate gene studies of LPL variants and the GWAS of lipids were based on participants of European ancestry, it remains to be determined whether the variants identified in European populations have similar effects in African Americans. To the best of our knowledge, our study is one of the only two studies that specifically evaluated multiple LPL SNPs with regard to their association to triglycerides and HDL-C in African American populations. Deo et al successfully genotyped 85 out of 95 tagging SNPs in the LPL gene and adjacent segment in 3300 unrelated African Americans from the Jackson Heart Study.22 In all participants, rs10096633, rs328, and 4 other SNPs that correlated with rs328 at r2=1 in participants with European ancestral background were significantly associated with triglycerides. The SNP rs328 exhibited a significant interaction with local ancestry, with a smaller effect size in those with two copies of African ancestral chromosomes compared to those with at least one copy of European ancestral chromosome. As LD extends in general over longer stretches in Europeans than Africans, they concluded that rs328 had at most a modest effect on triglycerides and the stronger effect associated with rs328 in European populations are due to other causal variants that are in LD with rs328.22 This is in line with our findings that the effect sizes associated with rs328, rs326, and rs13702 were stronger in European Americans than in African Americans, and adjusting for rs328 abolished the associations for rs326 and rs13702 in African Americans but not in European Americans. The effects carried by rs326 and rs13702 in European Americans of this study are mediated by rs328 and additional causal variants, while those in African Americans are likely attributable to rs328. For HDL-C in that study, only rs13702 showed a significant association, which is essentially in agreement with our findings that rs13702 as well as the haplotype carrying the at-risk allele for rs13702 showed a stronger (and significant) effect than rs328. Future studies are need to determine whether it is rs13702 itself or other variants that are in LD with rs13702 that influence HDL-C levels.
Furthermore, our study shows that the LPL variants that have been confirmed to contribute to the cross-sectional inter-individual differences in triglycerides and HDL-C also influence longitudinal changes in the same individuals in European Americans. The minor alleles in rs326, rs328, and rs13702 that predispose an individual to lower triglycerides and higher HDL-C levels at young adulthood further slow down the trajectory increase in triglycerides and decrease in HDL-C or promote the increase in HDL-C during 20 years of follow-up. The genotype-related longitudinal effects became strongest at year 20, when participants were aging into middle-aged adulthood. Two previous studies, both focusing on rs328 only, have investigated the longitudinal effects of the LPL gene polymorphism on triglycerides and HDL-C.12,13 In the first study, Chen et al measured triglycerides and HDL-C in 597 white and 232 black children aged 5–18 years at baseline and then again after 14.6 to 20.3 years of follow-up.12 They found that rs328 was not associated with triglycerides or HDL-C in childhood but showed a significant association with both in adulthood, implying a SNP by longitudinal time interaction. The second study was conducted in 2864 adults of European ancestry for whom lipids were measured at 4–6 time points during 15.9–17.2 years of follow-up.13 A significant SNP-by-age interaction was detected for HDL-C, but not for triglycerides. The minor allele at rs328 was associated with an increase in HDL-C with increasing age.13 In our study, we did not model an SNP-by-age interactions. Instead, we analyzed the associations of longitudinal change with SNP genotypes as well as SNP-by-time interactions. We chose to use this slightly different approach based on the consideration that participants entered into the study at different ages and they were not followed-up annually; therefore, each participant only contributed data to certain age groups; also, the sample size for each individual age group is more limited than that for each follow-up time point. Nevertheless, the longitudinal time effects mostly reflects the age effects and findings from our study in general agree with the two studies and extend beyond those by including large European and African American populations, multiple SNPs from the LPL gene, and more follow-up examinations.
As the most uniform and evident change in participants’ characteristics during the follow-up is increase in age, the most plausible interpretation for the longitudinal effects of the LPL SNPs is differential expression of the LPL variants with aging, especially after participants became middle-aged. In animal studies, a substantial reduction in LPL activity has been observed with aging in adipose tissue, heart, and postural skeletal muscles.23,24 In humans, post-heparin plasma LPL activity was reduced in the elderly compared to younger adults.25 Therefore, it is plausible to postulate that the longitudinal influence of the LPL SNPs may reflect protective effects of responsible variants against the decline of LPL activity with aging. Another explanation is that the distributions in other risk factors may change over time and interact with the LPL variants. In our study, the associations between the LPL SNPs and the lipid changes were only slightly attenuated after adjusting for BMI, physical activity, smoking, and alcohol consumption. Dietary change, which might contribute to the longitudinal changes in triglycerides and HDL-C, was not included in the study since dietary measurements were collected at baseline and years 7 and 20 examinations and there was a significant number of participants (300–400) who were missing for dietary measurements.
The associations between the LPL SNPs and longitudinal change of triglycerides and HDL-C were detected in European Americans, but not in African Americans of the study. These discrepant findings may be explained by the followings: 1) The longitudinal associations observed in European Americans likely reflect additional effects from other causal variants that are in LD with the observed SNPs. The signals from those underlying causal variants were not captured in African Americans of the study because of a shorter LD distance in African Americans; 2) Other genetic or environmental factors that interact with the LPL variants to influence the longitudinal patterns distributed differently between the two populations.
The characteristics of the mutant allele homozygotes for rs328 (Ser447Stop) are of interest because this is the only functional variant among the 8 SNPs and has been extensively investigated in the literature. Baseline mean and longitudinal changes in triglycerides and HDL-C by rs328 genotype are presented in Table S4. For the cross-sectional and longitudinal associations that were statistically significant (Tables 4, ,5,5, and and6),6), the mutant allele homozygotes (GG) tend to have a stronger effect in lowering triglycerides and increasing HDL-C than the heterozygotes (GC). This variant introduces a premature termination codon that shortens the LPL molecule at the C-terminal. While the molecular mechanisms underlying the beneficial effect on lipids have not been clearly delineated, data from the literature suggest that this mutation confers a gain-of-function effect that is possibly attributable to increased LPL activity, especially under postprandial conditions.26
We detected multiple variants in the LPL gene that are associated with inter-individual differences in triglycerides and HDL-C in both European Americans and African Americans, and with longitudinal changes in the two traits in European Americans. Findings from this study not only contribute to the current understanding of genetic influences for lipids, but also provide a basis for future finemapping studies to identify additional causal variants contributing to both cross-sectional and longitudinal variations in triglycerides and HDL-C. Identification of such variants will improve risk stratification as well as pharmaceutical intervention to reduce the risk of CHD over a long life span.
The authors thank the staff and participants of the CARDIA study for their important contributions.
This work was supported by CARDIA contracts [N01-HC-95095, N01-HC-48047, N01-HC-48048, N01-HC-48049, N01-HC-48050, N01-HC-45134, and N01-HC-05187] from the NIH/NHLBI, and by Modeling DNA Diversity in Reverse Cholesterol Transport study from the NIH [1R01HL072810], an ancillary study to CARDIA.
George Apostol is currently employed by a pharmaceutical company and owns shares in Pfizer, Bristol Myers, and Boston Scientific. The other authors do not have a real or perceived conflict of interest to report for this manuscript.