| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
(Hypertension. 2007;50:557.)
© 2007 American Heart Association, Inc.
Original Articles |
From the Center for Clinical Evidence Synthesis (E.Z.) and Center for Cardiovascular Health Services Research (D.K.), Institute for Clinical Research and Health Policy Studies, Department of Medicine, Tufts-New England Medical Center, Tufts University School of Medicine, Boston, Mass; Department of Biomathematics (E.Z., G.K.), University of Thessaly School of Medicine, Larissa, Greece; Division of Genetic Epidemiology, Department of Biomedical Informatics (N.J.C.), and Cardiovascular Genetics Division, Department of Internal Medicine (P.N.H., S.C.H.), University of Utah School of Medicine, Salt Lake City; and the Division of Epidemiology (L.A.), University of Minnesota, Minneapolis.
Correspondence to Elias Zintzaras, Center for Clinical Evidence Synthesis, Institute for Clinical Research and Health Policy Studies, Tufts University School of Medicine, Tufts-New England Medical Center, 750 Washington St, Tufts-NEMC #63, Boston, MA 02111. E-mail ezintzaras{at}nemc-tufts.org
| Abstract |
|---|
|
|
|---|
0.05) by either unweighted or weighted analyses, 4 of which (bins 21.2: 21q22.11 to 21q22.3, 18.3: 18q12.2 to 18q21.33, 18.4: 18q21.33 to 18q23, and 6.2: 6p22.3 to 6p21.1) were significant by both. In subjects of European descent, 3 bins (22.1: 22q11.1 to 22q12.3, 22.2: 22q12.3 to 22q13.3, 10.4: 10q22.1 to 10q23.32) had Prank
0.05 with both unweighted and weighted analyses. Bin 10.4 showed low heterogeneity (PQ=0.04). None of the bins showed low heterogeneity (PQ>0.05), indicating variation in the strength of association. Further investigation of these regions may help to direct the identification of candidate genes for pulse pressure variation.
Key Words: genome search meta-analysis heterogeneity HEGESMA pulse pressure
| Introduction |
|---|
|
|
|---|
Recent data support a genetic contribution to PP variation in the population. Twin and family based studies have shown that genetic factors play a role in PP, but the exact inheritance pattern is still unknown.8,9 The heritability estimates in various studies range from 0.13 to 0.63, suggesting that PP has a moderate or significant heritable component.9–16 It has been also suggested that the genetic etiology of PP is the same as for systolic blood pressure.14 Genetic association studies have shown that a significant number of gene polymorphisms could modulate increases of PP.17–20 However, the results from genetic association and family based studies require independent replication and identification of the functional basis of the association. In addition, the candidate gene studies on PP have produced inconsistent results so far.
Genomewide linkage scan (search) is an alternative, comprehensive, and unbiased approach for identifying chromosomal loci that may be linked to multifactorial traits such as PP.21 However, genome scans on PP9,14–16 have thus far produced inconsistent results, in part because the linkage signals were rather low.22
Thus, we sought to summarize the existing evidence on regions linked to PP by applying a heterogeneity-based genome-scan meta-analysis (HEGESMA) to available genome-scan data on PP.23–26 HEGESMA has already been applied to genome scans of several diseases24,25,27–29 including myocardial infarction.30
| Materials and Methods |
|---|
|
|
|---|
The meta-analysis included genome scans fulfilling the following inclusion criteria: (1) subjects were humans; (2) PP was defined as the mean difference between systolic blood pressure and diastolic blood pressure obtained at each examination; (3) DNA markers at 30 cM or finer density were typed throughout the genome; (4) the required linkage score data were published on the Web, provided by the authors, or were extractable from published graphs; and (5) there was no sample overlapping. Partial scans and candidate region studies were excluded. In studies with overlapping cases, the largest studies with extractable data were included in the meta-analysis. Subgroup analysis by ethnicity was conducted.
The search strategy was distributed and approved by all of the authors, the literature search was performed by G.K., and the article retrieval and evaluation were performed by 2 investigators (E.Z. and G.K.).
Databases
For each eligible study, the following information was extracted: first author, journal, year of publication, country of recruitment, ethnicity of study population, age of participants, number of families, subjects, definition of phenotype, number of microsatellite markers, linkage statistic, type of statistical analysis, and software of linkage analysis. Two investigators (E.Z. and G.K.) independently extracted data, discussed all of the disagreements, and reached consensus on all of the items.
Genome-Scan Meta-Analysis and Heterogeneity Testing
The genome-scan meta-analysis starts by splitting the chromosomes into bins of approximately equal length; usually each bin has a width of 30 cM, giving 120 bins in total for the whole genome.22,24,26 Each bin is conventionally symbolized as "chromosome"."number of bin", ie, bin 3.2 symbolizes the second bin of chromosome 3. For each genome scan, the most significant result of the test statistic obtained within the bin is recorded. Then, for each scan, the bins are ranked according to the significance of their results, and the ranks for each bin are summed across scans. The significance of the average rank of each bin is assessed empirically against the distribution of average ranks. When a bin has a high average (or summed) rank, then this is considered as evidence for linkage. Equal test statistics for several bins within a study were assigned as tied ranks.
Heterogeneity between studies for each bin was assessed using the Q statistic. The Q statistic is defined as the sum of the squared deviations of each studys bin rank from the mean of the ranks. In genome-scan meta-analysis, low between-study heterogeneity indicates consistency of study results for the same bin. Then, the presence of low heterogeneity for a specific bin with high ranks in all of the studies can be interpreted as further supportive evidence for the importance of this bin. The statistical significance of the average rank and the Q metric was assessed using a Monte Carlo method.24,25 In the Monte Carlo method, the ranks of each study are randomly permuted and the simulated average rank and Q metric is calculated; this procedure is repeated for 50 000 cycles to produce a null distribution for both the average rank and the Q metric. The significance level (Prank) of the average rank of bins against the null distribution of average ranks is the percentage of simulated average ranks greater or equal than the observed. The statistical significance level (PQ) for low heterogeneity is the percentage of simulated metrics less or equal than the observed; thus, the test for Q is left sided.24 In addition, the probability of observing a given average rank for a bin by chance in bins with the same "place" in the ascending order of average ranks in the runs (ordered ranks; Porder) is calculated.26 Prank assesses the significance of each bin independently, whereas Porder is based on the distribution of average ranks across all of the bins simultaneously.25 Moreover, a Monte Carlo test has been performed generating null distributions separately for each bin, considering only the simulated distributions of the Q metric (Qadjusted) for bins with the neighboring simulated average rank (±2) as the bin being considered each time.24 In the present study, genome-scan meta-analysis (Prank and Porder) and heterogeneity testing was performed unweighted and weighted. In weighted analysis, the ranks of the bins in each study have been weighted by
(subjects), and, then, the weights were scaled to sum up to 1.
This meta-analysis consisted of the main analysis including all of the available data and a subgroup analysis based on population ethnicity. When the analysis identifies significant adjacent bins, then an analysis based on 60 bins is performed to further verify the significance of the bins, because the separation of bins is distance sensitive.31 The evaluation of the significance of average ranks and the significance of heterogeneity was performed using HEGESMA software (http://biomath.med.uth.gr).25
We sought to investigate whether known candidate genes for PP variation are harbored in the identified chromosomal regions. The search for clinically relevant genes followed a narrow and a broad approach. Under the narrow approach, the search was limited to genes investigated in case-control studies for PP or other arterial stiffness indices. Under the broad approach, the search was extended to known candidate genes related to "vascular disease," including phenotypes such as coronary artery disease, stroke, peripheral occlusive arterial disease, intracerebral hemorrhage, and hypertension. We selected any genetic association study that evaluated an association with PP (See the search strategy shown in Table 6). Moreover, these data were supplemented by the differentially expressed genes from the 1 microarray study of arterial stiffness that we could identify in the literature.32
|
| Results |
|---|
|
|
|---|
1 marker existed within the boundaries of every bin in every study. The mean breadth of the bins used was 29.38±2.7 cM.
|
|
Briefly, the study by Atwood et al14 consisted of 10 pedigrees with 440 subjects (Mexican Americans) with ages ranging from 40 to 60 years. The linkage analysis was adjusted for sex, age, and body mass index (BMI). No correction for medication was performed, because a Monte Carlo test in a separate set of nuclear families showed no effect of medication on PP.14 In the study by Camp et al,9 the genome scan was performed in 26 pedigrees with 1454 subjects from a population of European descent with an average age of 27.8 years. PP for unmedicated subjects was adjusted for sex, age, and BMI. For the 125 subjects taking antihypertensive medications, a PP value of 55 mm Hg was assigned for analysis, unless their observed medicated value was higher, in which case that value was used. This substitution approach for managing medicated subjects was used to increase power.37 In DeStefano et al,15 345 pedigrees containing 1584 subjects with an average age 47.5 years from a population of European descent were evaluated. PP values were adjusted for age, time period of examination, and BMI. For subjects receiving antihypertensive treatment, a nonparametric algorithm to adjust PP for treatment effect was used, as described previously.38 In Bielinski et al,16 the study in blacks, East Asians, Latinos, and subjects of European descent was performed using 3962 subjects (1292 pedigrees), 1557 subjects (535 pedigrees), 1612 subjects (389 pedigrees), and 3667 subjects (1104 pedigrees), respectively. In the 4 studies, the average age ranged from 51 (in blacks and East Asians) to 55 years (in Latinos). The linkage analysis for PP was adjusted for the effects of sex, age, age-by-sex interaction, BMI, and field center. The antihypertensive medication status was not considered in this analysis.
All of the studies used similar number of microsatellites, ranging from 36615 to 405,9 and apart from 1 study,14 the remaining studies used multipoint analysis. Linkage statistics was logarithm of odds score in all of the studies except for 1,9 where a quantitative nonparametric linkage score was used. In weighting, the study by Atwood et al14 produced the least weight with a weight factor equal to 0.07, and the study by Bielinski et al16 in blacks and subjects of European descent produced the most weight (0.21 and 0.20, respectively). In the remaining studies, the weight factor was equal (w=0.13). The chromosome regions with suggestive or significant linkage identified from each individual genome scan are shown in Table 2.
The Figure shows the average ranks for each bin from the 5 genome scans using 120 bins. The bins with significant Prank in unweighted analysis are located above the line at P
0.05. The significant bins (P
0.05), the observed ranks, and the HEGESMA statistics for each study are shown in Table 3
.
|
|
|
Five bins were found to have Prank
0.05 by either unweighted or weighted analyses (bins 21.2: 21q22.11 to 21q22.3, 18.3: 18q12.2 to 18q21.33, 18.4: 18q21.33 to 18q23, 6.2: 6p22.3 to 6p21.1, and 3.3: 3p23–3p14.2), and 4 of them (bins 21.2, 18.3, 18.4, and 6.2) had Prank
0.05 with both methods (Table 3
). These bins were not significant in the order statistics for the unweighted or weighted analysis (Porder>0.43). None of the bins showed low heterogeneity (PQ>0.09 and PQadjusted>0.07), indicating variation in their rankings between scans. Bin 21.2 ranked at the top ranks in Atwood et al14 and in Bielinski et al16 for subjects of European descent and Latinos but did not rank that high in DeStefano et al.15
In subgroup analysis for subjects of European descent,9,15,16 4 bins were found to have Prank
0.05 by either unweighted or weighted analyses (bins 22.1: 22q11.1 to 22q12.3, 22.2: 22q12.3 to 22q13.3, 10.4: 10q22.1 to 10q23.32, and 8.1: 8p23.3 to 8p23.1), and 3 of them (bins 22.1, 22.2, and 10.4) had Prank
0.05 with both methods (Table 4). None of the bins was significant in order statistics for the unweighted or weighted analysis (Porder>0.10). Bin 10.4 showed low heterogeneity that was statistically significant in both unweighted and weighted analysis (PQ=0.01 and PQ=0.01, respectively). Even when the Monte Carlo distributions were adjusted for the average rank of each bin, significantly low heterogeneity was seen in bin 10.4 for the unweighted and weighted analysis (PQadjusted=0.04 and PQadjusted=0.03, respectively), indicating consistency between scans. Thus, in the subgroup analysis for subjects of European descent, bin 10.4 provided consistent evidence of linkage to PP. Although 2 bins (21.1 and 21.2) showed evidence for significant between-study heterogeneity (PQadjusted>0.05), these bins ranked at the top ranks in 2 of the 3 studies (Table 4). However, none of the bins identified in subjects of European descent were found in the overall analysis (7 studies), indicating an ethnic diversity regarding PP.
|
Because the analysis identified significant bins that are adjacent (bins 18.3 and 18.4 [overall analysis] and bins 22.1 and 22.2 [subgroup analysis in subjects of European descent]), a posthoc analysis restricted to 60 bins was performed. This analysis could verify further the significance of these bins (Table 5). Indeed, overall, bins 18.3 to 18.4 and 21.1 to 21.2 were found to have Prank
0.05 in both unweighted and weighted analyses. The analysis revealed another bin (8.5 to 8.6) not identified previously, but the average rank is marginally significant in the unweighted analysis (P=0.04) and nonsignificant in the weighted analysis (P=0.12). In subjects of European descent, bins 22.1 to 22.2 and 10.3 to 10.4 were found to have Prank
0.05 in both unweighted and weighted analyses. However, none of the bins approached significance at the 5% level when adjustments for multiple comparisons were made; then, a genome-wide significance bin was tested at P
0.0004, after accounting for 119 comparisons.
|
When we searched for clinically relevant genes that investigated in case-control studies for PP or other arterial stiffness indices (narrow approach), only bin 18.3 includes a candidate gene for PP, the neural precursor cell expressed, developmentally downregulated 4-like gene (NEDD4L). The candidate genes identified in the significant bins when the search was extended to known candidate genes related to "vascular disease" (broad search) are shown in Table 6.
| Discussion |
|---|
|
|
|---|
Heterogeneity testing revealed low unadjusted and adjusted heterogeneity only at bin 10.4 in both weighted and unweighted analysis in subjects of European descent. Regions with significant average rank and low heterogeneity imply that there is consistency of linkage across scans. Extended studies are needed to investigate these regions further. However, these results do not exclude the possibility that other chromosomal regions affecting PP, not shown in the present study, exist and that regions with linkage may exist in one or few populations.
Only bin 18.3 includes a candidate gene for PP: NEDD4L. NEDD4L gene is instrumental for the regulation of the amiloride-sensitive epithelial sodium channel in the distal nephron and has been associated with PP, hypertension, and orthostatic hypotension.39–42 However, the other significant bins, except bin 8.1, include a number of candidate genes for vascular diseases. Overlapping of independent sources of information (genome scans, gene expression studies, and genetic association studies) has been termed "genomic convergence" and provides further insights into the genetics of PP.43,44
In the meta-analysis involving only subjects of European descent, any heterogeneity showed in the meta-analysis could be attributed more to genuine inconsistency of genetic effects and to differences in design and conduct of the studies and less to differences across populations.
The PP shows a small early rise in young subjects and an accelerated late rise around age of 50 years.1 This age-dependent rise in PP is in large part an indicator of arterial aging.2 However, in young subjects, ventricular ejection contributes significantly in the determination of PP and, thus, increases in PP do not reflect arterial stiffness per se.45 In the present meta-analysis, one study9 recruited subjects with relatively low mean age. The age-dependent changes of PP show a significant variability from one population to another.1 This variability is highly influenced by the association of other cardiovascular diseases and concomitant risk factors.45 Because the selection criteria of pedigrees were not aligned among the studies analyzed, the differential clustering of risk factors in the recruited subjects could also be a potential confounder. Moreover, the differential adjustment methodology for PP values among the studies analyzed could be another source of bias.
Recent data indicate that PP reflects a phenotype in which sex dependency represents an important component of phenotypic determination, and heritability is expected to be higher in males.46 The incorporation of sex-dependent models in future genome scans for PP is awaited to provide a more powerful analytical framework.47
PP provides only an approximation of arterial stiffness, because it may be influenced by other factors, such as age and ventricular ejection.45 More direct indices of arterial stiffness are pulse wave velocity and augmentation index.17 So far, only 1 published genome scan33 has performed a linkage analysis of carotid-femoral pulse wave velocity in a Framingham Study offspring cohort. The use of more direct measurements of arterial stiffness may be of value for elucidating the genetic component of arterial stiffness in future studies.17
Although conventionally HEGESMA is based on bins with widths of 30 cM, an analysis with bin width <30 cM could provide some more information on possible regions with linkage. An analysis using >120 bins was omitted because of lack of regions with very strong evidence of linkage. In this analysis, decisions on multiple tests have been made, though HEGESMA is an exploratory nonparametric procedure interested in the relative significance of the regions.48 Power analysis26 for average rank or heterogeneity testing was not considered, because there is no currently available software for these type of analysis. Power analysis, however, will be incorporated in a new version of HEGESMA.
Alternative approaches for meta-analysis of linkage data30 are to construct a combined map of the markers49 from the original genotypes for each study, to perform a new linkage analysis, and to combine P values after correcting for the size of the linkage area.50 The advantages of the approach used in this study26 are that it does not require raw data, results from several genetic analyses performed in a particular study can be maximized to produce a single set of ranks, it requires no assumptions about models of inheritance, and it provides a test of genetic heterogeneity.
Perspectives
The genome-scan meta-analysis and heterogeneity testing in PP provided some preliminary evidence of linkage for new candidate chromosomal regions, bin 6.2 (overall) and bin 10.4 (subjects of European descent), and reproduced regions already identified by individual studies: bins 18.3 to 18.4 and 21.1 to 22.2 (overall) and bin 22.1 to 22.2 (subjects of European descent). In addition, the meta-analysis verified the significance of bin 18.3, which includes a candidate gene for PP (NEDD4L). Thus, genotyping these regions with additional markers and families may identify candidate genes contributing to PP variation.
| Acknowledgments |
|---|
None.
Received March 2, 2007; first decision March 30, 2007; accepted June 19, 2007.
| References |
|---|
|
|
|---|
This article has been cited by other articles:
![]() |
D. A. Duprez Genetic Variants of Inflammatory Markers and Arterial Stiffness Hypertension, June 1, 2008; 51(6): 1472 - 1473. [Full Text] [PDF] |
||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Hypertension Home | Subscriptions | Archives | Feedback | Authors | Help | AHA Journals Home | Search Copyright © 2007 American Heart Association, Inc. All rights reserved. Unauthorized use prohibited. |