Impaired kidney function, as measured by reduced estimated glomerular filtration rate (eGFR), has been associated with increased risk of coronary heart disease (CHD) in observational studies, but it is unclear whether this association is causal or the result of confounding or reverse causation. In this study we applied Mendelian randomisation analysis using 17 genetic variants previously associated with eGFR to investigate the causal role of kidney function on CHD. We used 13,145 participants from the UCL-LSHTM-Edinburgh-Bristol (UCLEB) Consortium and 194,427 participants from the Coronary ARtery DIsease Genome-wide Replication and Meta-analysis plus Coronary Artery Disease (CARDIoGRAMplusC4D) consortium. We observed significant association of an unweighted gene score with CHD risk (odds ratio = 0.983 per additional eGFR-increasing allele, 95% CI = 0.970–0.996, p = 0.008). However, using weights calculated from UCLEB, the gene score was not associated with disease risk (p = 0.11). These conflicting results could be explained by a single SNP, rs653178, which was not associated with eGFR in the UCLEB sample, but has known pleiotropic effects that prevent us from drawing a causal conclusion. The observational association between low eGFR and increased CHD risk was not explained by potential confounders and there was no evidence of reverse causation, therefore leaving the remaining unexplained association as an open question.
Chronic kidney disease (CKD) occurs in 15–20% of the general population aged 65 years or older. Markers of CKD, such as a low estimated glomerular filtration rate (eGFR) and/or elevated Urine Albumin Creatinine Ratio (UACR), are major independent risk factors for cardiovascular and all-cause mortality1,2. In particular, people with lower eGFR are at a higher risk of developing coronary heart disease (CHD).
Although it has been proposed that CKD may be causally associated with CHD, the precise biological pathways for this association are not well understood and hypotheses regarding this finding are varied, broadly involving inflammatory pathways and/or vascular calcification3,4,5,6. Currently, people with CKD are treated with blood pressure lowering drugs and statins to prevent poor outcomes. However, many genetic variants that are associated with high blood pressure are not associated with eGFR or CKD7, as might be expected if they were on the same causal pathway to CHD. Also, there is little overlap in association between CKD and CHD genetic variants and early markers of cardiovascular disease, which would be expected if kidney function were causally related to CHD8.
People with CKD tend to have had adverse early life circumstances that predispose them to cardiovascular disease9,10,11 and individuals who are overweight at earlier age are more likely to have CKD at older age suggesting that CKD is a marker or consequence of a cumulative adverse life style. Hence, it is not yet clear whether low kidney function directly causes CHD or whether the observed association is due to other shared risk factors/confounders, in particular socioeconomic status, early life risk factors and unhealthy lifestyle.
Here we applied Mendelian Randomisation (MR) to determine whether lower eGFR per se has a causal role to contribute to later CHD. To our knowledge, this is the first MR study to explicitly investigate the influence of eGFR on CHD. Previously, Olden et al. studied whether several genetic variants affect both eGFR and CHD8. They identified one SNP, rs653178 that was associated with both; but while this is consistent with a causal effect, they did not consider whether it met the assumptions required for MR nor did they estimate a causal effect size. Here we go further by combining 17 SNPs into a single gene score, which can improve the power of MR studies when individual SNPs are not significantly associated with the outcome12 and we use a richly phenotyped data set, the UCLEB consortium13 to thoroughly examine the possibility of violating the MR assumptions through confounding or pleiotropy.
Research design and methods
We used data from two consortia, UCLEB13 and CARDIoGRAMplusC4D14,15,16. The former allowed assessment of the MR assumptions for eGFR, namely that the gene score is associated with eGFR, not associated with potential confounders of the eGFR-CHD association and not associated with other biomarkers of CHD that might represent alternative pathways than that through eGFR. In UCLEB, we used individual patient data from 2,249 cases of CHD and 10,896 controls from 7 cohort studies: the British Regional Heart Study (BRHS), British Women’s Heart and Health Study (BWHHS), Caerphilly Prospective Study (CAPS), Edinburgh Type-2 Diabetes Study (ET2DS), Edinburgh Artery Study (EAS), English Longitudinal Study of Ageing (ELSA) and Whitehall II Study (WHII). Four studies (BRHS, BWHHS, CAPS and ET2DS) have eGFR available. UCLEB has 116 phenotypes related to cardiovascular function, allowing SNPs to be thoroughly assessed for validity of the MR assumptions.
The CARDIoGRAMplusC4D consortium provides a large sample for assessing genetic association with CHD. The Consortium released summary statistics from 3 meta-analyses of coronary artery disease: CARDIoGRAM genome-wide association study (GWAS) with 22,233 CHD cases and 64,762 controls16, C4D GWAS with 15,420 CHD cases and 15,062 controls15 and the combined data of these two GWAS with additional cohorts, CARDIoGRAMplusC4D Metabochip, with 63,746 CHD cases and 130,681 controls14. These consortia data allow the assessment of genetic association with CHD in very large samples, but do not provide information on eGFR.
Selection of SNPs and construction of gene score
We identified 32 SNPs previously reported to be associated with eGFR (S1 Table). Seventeen of them, which are not in linkage disequilibrium (LD) and have been genotyped in UCLEB, were shortlisted as potential components of an instrument for MR (S2 Table) because they could be investigated for validity of the MR assumptions using the UCLEB data. Due to the different genotyping platforms used in CARDIoGRAM GWAS, C4D GWAS and CARDIoGRAMplusC4D Metabochip, our 17 selected SNPs are not always available in all 3 meta-analyses. Thus the total sample sizes vary across SNPs used to investigate association with CHD (S1 Fig).
These selected SNPs were then combined into gene scores. Two types of gene score were used: 1) unweighted gene score, defined for subject i as where gij is the count of eGFR-increasing alleles for subject i at SNP j with the direction of effect taken from the UCLEB data and 2) weighted gene score, , which includes estimates of effects on eGFR, wj, at each SNP based on the UCLEB data. More precisely, the wj are the estimates from linear regression analyses of eGFR on SNP j, under additive genotypic coding.
Exposure and outcome variables
Our primary outcome is CHD (defined as fatal or non fatal myocardial infarction, or revascularization, from both diagnosis validation and self-report) and the exposure is the level of kidney function measured by eGFR. In the UCLEB data, we derived eGFR based on creatinine using the modification of diet in renal disease (MDRD) model17
eGFR was normally distributed in our data, therefore no further transformation was applied.
Potential confounding factors
From the total of 116 traits available in UCLEB, we systematically checked for potential confounding factors across 94 traits measured in up to 7 studies. We identified potential confounders by testing each of these traits for both eGFR-trait (available in 4 studies) and CHD-trait associations. As the presence of both associations is necessary but not sufficient to establish confounding, we considered those associated traits as potential confounding factors, unless there is additional external evidence that a particular biomarker is renally cleared, i.e. potentially on the causal pathway between kidney function and CHD. For the eGFR-trait and CHD-trait associations we controlled the false discovery rate at 0.05 by the Benjamini-Hochberg procedure. We used a permutation procedure to check whether correlation among traits would cause this procedure to be conservative, finding that the overall level of correlation is negligible in this context (see Supplementary Methods).
SNP instrument validity
We used UCLEB data to test the validity of our gene scores as instrumental variables under the three MR assumptions: 1) association between gene scores and eGFR, 2) absence of association between gene scores and common causes of eGFR and CHD and 3) absence of pathways between gene scores and CHD other than through eGFR. We tested assumption 1 using linear regression of eGFR on the gene scores, assumption 2 using linear regression of each identified potential confounder on the gene scores (except gender where logistic regression was applied) and assumption 3 using linear regression of each trait in UCLEB (whether or not a potential confounder) on the gene scores. Each regression adjusted for study as a categorical covariate. We controlled the false discovery rate at 0.05 by the Benjamini-Hochberg procedure.
We tested for a causal effect by testing the association between each gene score and CHD and appealing to the MR principle. As only summary odds ratios and their standard errors were available from the CHD consortia, we applied the Johnson formula described by Burgess et al.12, in which the regression coefficient of the gene score on CHD is
with standard error , where is the estimated log odds ratio of SNP j on CHD, sYj is its standard error and wj is a predetermined weight. The unweighted score has wj = 1 for all SNPs, whereas the weighted score sets wj to the linear regression coefficient of SNP j for eGFR, here estimated from the UCLEB data (weights and odds ratios were calculated for the eGFR increasing alleles). This is an example of two-sample Mendelian randomization18 in which the exposure and outcome associations are measured in different samples. Two-sided tests of association were performed by comparing the ratio of and its standard error to a standard normal distribution.
Using the weighted score, is an estimate of the causal effect of eGFR on CHD, but we avoid this interpretation as our aim here was merely to detect the presence of a causal effect.
Reverse Mendelian Randomisation
We considered whether a causal effect exists in the opposite direction, with CHD acting to cause reduced kidney function. We constructed an weighted gene score from 51 SNPs associated with CHD at genome-wide significance, as reported by CARDIoGRAMplusC4D14 and tested this score for association with eGFR in the UCLEB data using summary odds ratios as above.
A summary of CHD cases and controls in UCLEB is shown in Table 1. As expected, we observed that eGFR is significantly negatively associated with CHD (OR = 0.984 per ml/min/1.73 m2 increase, 95% CI = 0.979–0.988, p = 1.98 × 10−12) across its entire range and particularly when comparing subjects with reduced (<60 ml/min/1.73 m2) eGFR to those in the intermediate (60–90 ml/min/1.73 m2) and normal/high range (≥90 ml/min/1.73 m2, Fig. 1), consistent with previous studies of the MDRD eGFR-CHD association19.
From 94 traits available in 7 UCLEB studies, we identified 28 potential confounding factors of the eGFR-CHD association (Fig. 2, S4 Table). These are generally concordant with previously observed correlates of eGFR and risk factors for CHD19. Some discrepancies, for example blood pressure and smoking (ever/never), which were not associated with eGFR in UCLEB are likely due to the cross-sectional nature of the data and have been previously seen in other data20. One of the potentially confounding factors was NTproBNP, which is known to be renally cleared6. NTproBNP was therefore considered to be a variable on the causal pathway between eGFR and CHD and not a confounding variable. We also excluded percentage body fat because this was only available in the ET2DS study, which did not have many of the other confounders measured. Due to missing data in the remaining 26 potential confounders, we then investigated the observational association between eGFR and CHD in a reduced sample size of 1547 (OR = 0.977 per 1 ml/min/1.73 m2 increase, 95% CI = 0.965–0.989, p = 3.38 × 10−4). After adjusting for 26 potential confounders, the association between eGFR-CHD was slightly attenuated and remained borderline significant (OR = 0.868 per 10 ml/min/1.73 m2 increase, 95% CI = 0.760–1.000, p = 0.05).
In UCLEB, the combination of 17 SNPs explained 1.5% of the variation in eGFR. The corresponding F statistics were 59 and 91 for unweighted and weighted scores respectively, indicating that the gene score is a strong instrument. In testing for association between gene scores and all traits available in UCLEB (Fig. 3), our gene scores showed an exclusive association with kidney function measurements, including eGFR (adjusted p = 2.0 × 10−12 and p = 3.3 × 10−19 for unweighted and weighted gene scores respectively), creatinine (adjusted p = 6.0 × 10−9 and p = 6.8 × 10−16) and serum urea concentration (adjusted p = 9.8 × 10−3 and p = 3.1 × 10−5). This confirms the first MR assumption that our gene scores are indeed a good proxy for kidney function. Secondly, associations between our gene scores and 26 potential confounding factors and NTproBNP were not significant, meeting the second MR assumption that our gene scores are not associated with common causes of eGFR and CHD. Although we cannot be sure that our gene scores are not associated with unmeasured confounders, these results (including many of the strongest biomarkers for CHD) suggest that any such confounding is weak. Lastly, association between our gene scores and other traits available in UCLEB do not reach significance. Therefore our gene scores appear specific to eGFR and other kidney-related traits.
We found significant association of the unweighted score with CHD (OR = 0.983 per additional eGFR increasing allele, 95% CI = 0.970–0.996, p = 0.008), suggesting a causal effect of eGFR on CHD. This was in the direction to that seen in the observational data. However, the association was not significant when using a weighted score (OR = 0.993, 95% CI = 0.984–1.002, p = 0.11). Inspection of the weights given to the individual SNPs suggested that these results could be explained by rs653178 alone. Previously, it has been shown that there is little overlap between well-validated SNPs for kidney disease and CHD, where only rs653178 was highly associated with both cystatin-C-based eGFR and CHD21. However this was not the case in UCLEB, in which rs653178 was not associated with eGFR and received the smallest weight among all 17 instrument SNPs (S6 Table). Therefore, using weights from UCLEB diminished evidence of a causal association observed with the unweighted score.
The unweighted gene score was further investigated to assess the sensitivity of the observed nominal evidence of a causal association. We first excluded rs1260326, which has known pleiotropic effects22 and observed the same association signal remaining (OR = 0.991, 95% CI = 0.981–1.001, p = 0.003). Secondly we excluded rs653178 which could have stronger effect through cystatin-based eGFR instead of creatinine-based eGFR22 from the gene score and found that the evidence of causal effect disappeared (OR = 0.994, 95% CI = 0.982–1.005, p = 0.291). This implies that rs653178 is driving the significant association of the gene score with CHD. To confirm this, we excluded each of the 17 SNPs at a time and observed that only the exclusion of rs653178 removed the association signal. In addition, instead of identifying the eGFR increasing alleles from the UCLEB data, we incorporated external information on the direction of effect for 9 SNPs that were available in Olden et al., to allow more accurate estimates due to their larger sample size. While the eGFR increasing alleles in UCLEB are retained for the remaining SNPs, for 3 out of the 9 SNPs from Olden et al. a different eGFR increasing allele was identified in comparison to UCLEB (S6 Table). After adjusting the unweighted score accordingly, the evidence of a causal effect remained (OR = 0.971, 95% CI = 0.950–0.992, p = 0.007).
We further investigated the case in which eGFR < 60 ml/min/1.73 m2 (stage 3 nephropathy) which has been significantly associated with increased CHD events23,24. If there is a causal influence of eGFR on CHD, it may be stronger in those for whom an increased association is observed. By conditioning on eGFR, a biased estimate of causal effect may result25. However, we observed little evidence of a causal association between eGFR and CHD with the unweighted score (OR = 0.982, 95% CI = 0.963–1.002, p = 0.083) and the weighted score (OR = 0.991, 95% CI = 0.974–1.008, p = 0.286) when restricting analysis to subjects with eGFR < 60.
Since the gene scores were not specific to eGFR but also associated with other kidney function traits, we adjusted the regression of CHD on eGFR for creatinine and serum urea concentration in the UCLEB data alone. Neither gene score was significantly associated with CHD in the UCLEB data, with or without adjustment for creatinine and serum urea concentration.
Finally, the reverse MR did not show an association of the CHD gene score with eGFR (β = −0.120, 95% CI = −1.271–0.872, p = 0.715).
We applied Mendelian Randomisation to determine whether lower eGFR has a causal role in CHD. To improve the power of MR, we combined 17 SNPs into a single gene score. By using a richly phenotyped data set in the UCLEB consortium, we thoroughly examined whether our gene score meets the assumptions of Mendelian Randomisation studies, in that it is associated with eGFR, not associated with potential confounders of the eGFR-CHD association and not associated with other biomarkers of CHD that might represent alternative pathways than that through eGFR. The gene scores appeared to meet these assumptions although we noted associations with additional kidney function traits that might represent alternative pathways through kidney function to that measured by eGFR.
We found significant evidence of a causal effect on CHD using an unweighted score. However, the weighted score revealed that this result is driven by rs653178, which was not associated with eGFR in UCLEB (and therefore unreliably down-weighted the significance of the unweighted score). Potential reasons that could cause the discrepancy of rs653187 between UCLEB and the previous study by Olden et al. are random sampling, low power due to the smaller sample size in UCLEB and a substantially weaker association between rs653178 and creatinine-based eGFR (p = 1 × 10−4) compared to cystatin-C-based eGFR (p = 3.5 × 10−11)22. Further analyses, including taking the direction of effect from external sources rather than the UCLEB data and selecting study participants based on low eGFR, did not yield any stronger evidence of a causal effect.
One explanation of our results is that, if a causal effect exists, it may act only through some pathways contributing to measured eGFR, marked by rs653178, whereas other pathways marked by the other 16 SNPs do not have a causal effect. However, rs653178 has been previously shown to be associated with a number of phenotypes, including mean arterial pressure26, blood pressure27, celiac disease28 and peripheral artery disease, a known complication of both CHD and CKD29. Therefore the known pleiotropic effect of rs653178 restricts inference of a direct causal effect of eGFR on CHD.
Our study has some limitations. Only 1.5% of the variation in eGFR was explained by the gene scores comprising 17 SNPs. In the largest consortium data available, CARDIoGRAMplusC4D including approximately 200,000 individuals, we would have 74% and 25% power to detect causal odds ratios of 0.9 and 0.95 respectively30, both in excess of the observational odds ratio of 0.98 in the UCLEB consortium. Although the study was likely underpowered over the full range of eGFR, a stronger causal effect may exist among those with eGFR < 60, but our gene scores were not associated with CHD among those subjects. A score including more eGFR associated SNPs would have more power to detect a causal effect, though at greater risk of violating the MR assumptions. Another limitation concerns the validation of the MR assumptions for our gene scores. While UCLEB provides a richly phenotyped and large data set, other phenotypes not measured in those studies may be confounding factors or show pleiotropic effects of our gene scores. Furthermore our power to detect confounding and pleiotropy was reduced by control for multiple testing, which we deemed necessary to avoid falsely inferring that our gene scores are invalid instruments.
Observational studies have mainly shown increased risk of CHD among those with low MDRD eGFR compared to those in the normal range. Studies using newer biomarkers provide clear evidence that there is association even at higher eGFR31. We have treated eGFR as a continuous variable in MR analysis, which may have reduced power if the causal effect is restricted to subjects with low MDRD eGFR. We took this approach because the associated SNPs apparently influence eGFR over its entire range so that a standard MR analysis would correspond to a population-wide intervention on eGFR levels. Again however, we found no association between the gene scores and CHD among subjects with low eGFR. More robust methods for MR analysis in discrete exposure strata have recently become available25,32 and could be applied here. However, we believe our current results are not sufficiently encouraging to warrant these approaches.
We are not the first to observe that inflammatory and thrombotic markers associated with CHD are raised in people with CKD5. These potential confounding biomarkers may be a result of other life style associated risk factors leading to CKD, in particular overweight at younger age and subsequent obesity33,34,35. In total we identified 27 potential confounders but upon adjustment the association between eGFR and CHD remained significant with a similar odds ratio. Our power to detect confounders was limited by the number of cases of CHD in UCLEB, the range of phenotypes considered (in particular excluding socio-economic status), the limited sample size for some particular phenotypes and the multiple testing burden and further confounders may exist that attenuate or abolish the eGFR-CHD association. We used reverse MR to address the possibility of reverse causation, but did not observe significant association between a gene score for CHD and levels of eGFR.
In conclusion, this study observed weak evidence for a causal effect of low eGFR on CHD risk, while the observational association was not explained by potential confounders nor by reverse causation. However, this result was highly influenced by rs653178, which has known pleiotropic effects, therefore restricting any inference of a direct causal effect of eGFR on CHD. Our results leave the remaining unexplained association between eGFR and CHD as an open question.
How to cite this article: Charoen, P. et al. Mendelian Randomisation study of the influence of eGFR on coronary heart disease. Sci. Rep. 6, 28514; doi: 10.1038/srep28514 (2016).
Chronic Kidney Disease Prognosis Consortium et al. Association of estimated glomerular filtration rate and albuminuria with all-cause and cardiovascular mortality in general population cohorts: a collaborative meta-analysis. Lancet 375, 2073–2081, doi: 10.1016/S0140-6736(10)60674-5 (2010).
van der Velde, M. et al. Screening for albuminuria identifies individuals at increased renal risk. J Am Soc Nephrol 20, 852–862, doi: 10.1681/ASN.2008060655 (2009).
Cai, Q., Mukku, V. K. & Ahmad, M. Coronary artery disease in patients with chronic kidney disease: a clinical update. Curr Cardiol Rev 9, 331–339 (2013).
Matsushita, K. et al. Subclinical Atherosclerosis Measures for Cardiovascular Prediction in CKD. J Am Soc Nephrol 26, 439–447, doi: 10.1681/ASN.2014020173 (2015).
Shlipak, M. G. et al. Elevations of inflammatory and procoagulant biomarkers in elderly persons with renal insufficiency. Circulation 107, 87–92 (2003).
Vickery, S. et al. B-type natriuretic peptide (BNP) and amino-terminal proBNP in patients with CKD: relationship to renal function and left ventricular hypertrophy. Am J Kidney Dis 46, 610–620, doi: 10.1053/j.ajkd.2005.06.017 (2005).
International Consortium for Blood Pressure Genome-Wide Association Studies et al. Genetic variants in novel pathways influence blood pressure and cardiovascular disease risk. Nature 478, 103–109, doi: 10.1038/nature10405 (2011).
Olden, M. et al. Overlap between common genetic polymorphisms underpinning kidney traits and cardiovascular disease phenotypes: the CKDGen consortium. Am J Kidney Dis 61, 889–898, doi: 10.1053/j.ajkd.2012.12.024 (2013).
Al-Qaoud, T. M., Nitsch, D., Wells, J., Witte, D. R. & Brunner, E. J. Socioeconomic status and reduced kidney function in the Whitehall II Study: role of obesity and metabolic syndrome. Am J Kidney Dis 58, 389–397, doi: 10.1053/j.ajkd.2011.04.017 (2011).
Silverwood, R. J. et al. Early-life overweight trajectory and CKD in the 1946 British birth cohort study. Am J Kidney Dis 62, 276–284, doi: 10.1053/j.ajkd.2013.03.032 (2013).
Silverwood, R. J. et al. Association between younger age when first overweight and increased risk for CKD. J Am Soc Nephrol 24, 813–821, doi: 10.1681/ASN.2012070675 (2013).
Burgess, S., Butterworth, A. & Thompson, S. G. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol 37, 658–665, doi: 10.1002/gepi.21758 (2013).
Shah, T. et al. Population genomics of cardiometabolic traits: design of the University College London-London School of Hygiene and Tropical Medicine-Edinburgh-Bristol (UCLEB) Consortium. PLoS One 8, e71345, doi: 10.1371/journal.pone.0071345 (2013).
CARDIoGRAMplusC4D Consortium et al. Large-scale association analysis identifies new risk loci for coronary artery disease. Nat Genet 45, 25–33, doi: 10.1038/ng.2480 (2013).
Coronary Artery Disease Genetics Consortium A genome-wide association study in Europeans and South Asians identifies five new loci for coronary artery disease. Nat Genet 43, 339–344, doi: 10.1038/ng.782 (2011).
Schunkert, H. et al. Large-scale association analysis identifies 13 new susceptibility loci for coronary artery disease. Nat Genet 43, 333–338, doi: 10.1038/ng.784 (2011).
Levey, A. S. et al. Using standardized serum creatinine values in the modification of diet in renal disease study equation for estimating glomerular filtration rate. Ann Intern Med 145, 247–254 (2006).
Pierce, B. L. & Burgess, S. Efficient design for Mendelian randomization studies: subsample and 2-sample instrumental variable estimators. Am J Epidemiol 178, 1177–1184, doi: 10.1093/aje/kwt084 (2013).
Di Angelantonio, E., Danesh, J., Eiriksdottir, G. & Gudnason, V. Renal function and risk of coronary heart disease in general populations: new prospective study and systematic review. PLoS Med 4, e270, doi: 10.1371/journal.pmed.0040270 (2007).
Nitsch, D. et al. Prevalence of renal impairment and its association with cardiovascular risk factors in a general population: results of the Swiss SAPALDIA study. Nephrol Dial Transplant 21, 935–944, doi: 10.1093/ndt/gfk021 (2006).
Olden, M. et al. Common Variants in UMOD Associate with Urinary Uromodulin Levels: A Meta-Analysis. J Am Soc Nephrol 25, 1869–1882, doi: 10.1681/ASN.2013070781 (2014).
Kottgen, A. et al. New loci associated with kidney function and chronic kidney disease. Nat Genet 42, 376–384, doi: 10.1038/ng.568 (2010).
Anavekar, N. S. et al. Relation between renal dysfunction and cardiovascular outcomes after myocardial infarction. N Engl J Med 351, 1285–1295, doi: 10.1056/NEJMoa041365 (2004).
Go, A. S., Chertow, G. M., Fan, D., McCulloch, C. E. & Hsu, C. Y. Chronic kidney disease and the risks of death, cardiovascular events and hospitalization. N Engl J Med 351, 1296–1305, doi: 10.1056/NEJMoa041031 (2004).
Silverwood, R. J. et al. Testing for non-linear causal effects using a binary genotype in a Mendelian randomization study: application to alcohol and cardiovascular traits. Int J Epidemiol, doi: 10.1093/ije/dyu187 (2014).
Wain, L. V. et al. Genome-wide association study identifies six new loci influencing pulse pressure and mean arterial pressure. Nat Genet 43, 1005–1011, doi: 10.1038/ng.922 (2011).
Newton-Cheh, C. et al. Genome-wide association study identifies eight loci associated with blood pressure. Nat Genet 41, 666–676, doi: 10.1038/ng.361 (2009).
Dubois, P. C. et al. Multiple common variants for celiac disease influencing immune gene expression. Nat Genet 42, 295–302, doi: 10.1038/ng.543 (2010).
Kullo, I. J. et al. The ATXN2-SH2B3 locus is associated with peripheral arterial disease: an electronic medical record-based genome-wide association study. Frontiers in genetics 5, 166, doi: 10.3389/fgene.2014.00166 (2014).
Brion, M. J., Shakhbazov, K. & Visscher, P. M. Calculating statistical power in Mendelian randomization studies. Int J Epidemiol 42, 1497–1501, doi: 10.1093/ije/dyt179 (2013).
Waheed, S. et al. Combined association of albuminuria and cystatin C-based estimated GFR with mortality, coronary heart disease and heart failure outcomes: the Atherosclerosis Risk in Communities (ARIC) Study. Am J Kidney Dis 60, 207–216, doi: 10.1053/j.ajkd.2012.03.011 (2012).
Burgess, S., Davies, N. M., Thompson, S. G. & EPIC-InterAct Consortium. Instrumental variable analysis with a nonlinear exposure-outcome relationship. Epidemiology 25, 877–885, doi: 10.1097/EDE.0000000000000161 (2014).
Greenfield, J. R. et al. Obesity is an important determinant of baseline serum C-reactive protein concentration in monozygotic twins, independent of genetic influences. Circulation 109, 3022–3028, doi: 10.1161/01.CIR.0000130640.77501.79 (2004).
Singh, A. et al. Elevated circulating tissue factor procoagulant activity, factor VII and plasminogen activator inhibitor-1 in childhood obesity: evidence of a procoagulant state. Br J Haematol 158, 523–527, doi: 10.1111/j.1365-2141.2012.09160.x (2012).
Timpson, N. J. et al. C-reactive protein and its role in metabolic syndrome: mendelian randomisation study. Lancet 366, 1954–1959, doi: 10.1016/S0140-6736(05)67786-0 (2005).
The authors declare no competing financial interests.
Electronic supplementary material
About this article
Cite this article
Charoen, P., Nitsch, D., Engmann, J. et al. Mendelian Randomisation study of the influence of eGFR on coronary heart disease. Sci Rep 6, 28514 (2016). https://doi.org/10.1038/srep28514
This article is cited by
Nonlinear causal effects of estimated glomerular filtration rate on myocardial infarction risks: Mendelian randomization study
BMC Medicine (2022)
BMC Medicine (2018)
Behavior Genetics (2017)
IL-34 is associated with the presence and severity of renal dysfunction and coronary artery disease in patients with heart failure
Scientific Reports (2016)