OUP user menu

Genotypes and haplotypes predisposing to myocardial infarction: a multilocus case-control study

Martin D Tobin, Peter S Braund, Paul R Burton, John R Thompson, Richard Steeds, Kevin Channer, Suzanne Cheng, Klaus Lindpaintner, Nilesh J Samani
DOI: http://dx.doi.org/10.1016/j.ehj.2003.11.014 459-467 First published online: 2 March 2004


Aim To identify polymorphisms and haplotypes in candidate genes that predispose to myocardial infarction (MI) using a multilocus approach.

Methods and results 1052 subjects, comprising 547 acute MI cases and 505 controls were studied. The association between MI and 58 SNPs in 35 candidate genes (generating 61 016 individual genotypes), and between MI and estimated haplotypes at 14 loci encompassing 16 genes was investigated. Two individual gene variants and haplotypes at two loci showed statistical association with MI. The α-adducin 460trp variant (OR 0.73, 95% CI 0.59–0.91, Math) and the cholesteryl ester transfer protein –629A variant (OR 0.82, 95% CI 0.68–0.97, Math) were both associated with a significant protective effect on MI, as was the paraoxonase 1/paraoxonase 2 haplotype comprising met55 and gln192 in paraoxonase 1 and cys311 in paraoxonase 2 (OR 0.52, 95% CI 0.39–0.77, Math). The apolipoprotein C III haplotypes CCTTCG and ATCCCG at positions –641*–482*–455*1100*3175*3206 were associated with an increased risk of MI, odds ratios 1.41 (95% CI 1.06–1.76, Math) and 1.71 (95% CI 1.28–2.14, Math), respectively.

Conclusions We report associations of two polymorphisms and haplotypes at two loci with risk of MI that warrants testing in future studies. Furthermore, we demonstrate the application of a multilocus assay in the setting of a large association study and the additional benefit gained from the study of haplotypes to identify variants influencing risk of coronary heart disease.

  • Genetics
  • Myocardial infarction
  • Risk factors

See page 451 for the editorial comment on this article1


Myocardial infarction is a complex disorder resulting from the combined effects of multiple environmental and genetic factors.1 Given that myocardial infarction is a leading cause of death in the Western world, its prevention is a public health priority. One approach to preventing myocardial infarction is to use observational genetic studies to identify disease-susceptibility genes that will further our understanding of the etiological pathways and thus indicate new potential targets for intervention. Genetic linkage and association studies have already identified several candidate genes that may predispose to coronary artery narrowing, such as cholesteryl ester transfer protein (CETP),2 and myocardial infarction, such as SERPINE1 (PAI 1), connexin 37, and stromelysin-1.3 However, the inability to replicate many of these findings has cast doubt on the implications of many of these studies. The power of most genetic association studies has been low, and few studies have been published with more than 500 cases and 500 controls.3,4

Haplotype-based analysis could prove a more powerful approach to dissect the genetic architecture of complex diseases.5 The human genome is comprised of haplotype blocks, which are chromosomal segments that are preserved intact over many generations, interspersed by recombination hotspots.6 Alleles within such ancestrally preserved haplotype blocks tend to be in strong linkage disequilibrium with one another, i.e., alleles at different sites within the haplotype block are strongly associated with one another at the level of the population. It is becoming clear that over small regions relatively few haplotypes may account for the vast majority of diversity7 and variations in haplotypes may be powerful predictors of disease.

Recently developed assays capable of simultaneously genotyping multiple loci allow the efficient investigation of the role of many genetic polymorphisms within a single study. Furthermore, the development of methods to estimate haplotype frequencies from unrelated cases and controls now allows the study of the combined effects of a number of polymorphisms that are in linkage disequilibrium with one another.8–12 Our aim was to identify individual polymorphisms and haplotypes that predispose to myocardial infarction using a multilocus approach13 to interrogate 63 single nucleotide polymorphisms (SNPs) in 35 candidate genes.


We carried out a case-control association study in two UK centres. The methods have been described previously.14 Briefly, we recruited 549 consecutive Caucasian cases Math75 years who had survived admission to the coronary care units in Leicester and Sheffield. All cases satisfied the World Health Organisation criteria for myocardial infarction.15 The 505 Caucasian controls, recruited contemporaneously, were healthy adult visitors to patients with noncardiovascular illnesses at the two hospitals. This provided an epidemiologically robust control population derived from the same geographic area as the cases. Control subjects with a history of MI or angina were excluded. The study had been designed to have 90% power to detect a 1.5-fold increase in the risk of myocardial infarction associated with the angiotensin converting enzyme DD genotype.16

Subjects completed a standard questionnaire about their personal history, which included data on age, sex, smoking status, history of hypertension and history of diabetes. We measured height and weight to calculate body mass index (BMI), and collected blood samples for the measurement of serum total cholesterol and DNA analysis. The study was approved by local research ethics committees and informed consent was obtained from participants.

DNA was extracted from whole blood.14 A multilocus assay was used to genotype 63 polymorphisms within 35 candidate genes previously implicated in cardiovascular disease.13,17 The extracted genomic DNA was amplified by two separate multiplex PCRs with biotinylated primers. The amplified PCR products were detected colorimetrically with immobilised sequence-specific oligonucleotide probes arrayed in a line array on nylon membrane strips. Correct allele identification was confirmed by direct sequencing or RFLP analysis of genomic DNA samples carrying the different genotypes for each polymorphism. We excluded data on five SNPs (apolipoprotein B (APOB) (arg3500gln), cholesteryl ester transfer protein [CETP] (asp442gly), CETP intron 14 G(+1)-to-A (Int14A), CETP intron 14 (+3)T ins (Int14T) and tumor necrosis factor (TNF superfamily, member 2) [TNF] (G(-244)A)) from our analysis, as fewer than five individuals possessed the variant allele. To check for possible laboratory errors and possible relatedness of cases and controls, we plotted the mean allele sharing identical by state across all polymorphisms and its standard deviation for all possible pairs of individuals in the study. This procedure, implemented using the GRR package (Graphical Relationship Representation),18 demonstrated identical genotypes across all 58 polymorphisms from two supposedly unrelated cases, which were therefore excluded from the study. No other outliers indicating errors were identified.

Statistical analysis was carried out using STATA 7.0. In the first stage, the association between each of the 58 polymorphisms with adequate information and myocardial infarction was assessed using logistic regression to adjust for the effects of age and sex, under an additive genetic model (genetic models are explained in Table 3). Other covariates were not adjusted for at this stage so that we could assess their possible involvement in the causal pathway. Furthermore, because genotype is effectively randomly allocated at birth (a process termed `Mendelian randomisation') the effects of genetic variants are unlikely to be confounded by lifestyle risk factors.19 Maximum likelihood estimates of odds ratios and their 95% confidence intervals were obtained and the significance of each polymorphism was assessed using a likelihood ratio test. In the second stage, polymorphisms that showed significant association with myocardial infarction (Math) under an additive model were also assessed under dominant and recessive models, and the influence of other covariates was assessed. A significance level of Math was adopted for this exploratory study.

A number of software programs have now been developed for estimating haplotype frequencies from genotype data on unrelated cases and controls.8–12 We estimated haplotype frequencies using software implemented in STATA 7.0,8,20 that uses a log-linear model embedded within an Expectation–Maximization (EM) algorithm.21,22 To describe the relationship between haplotypes and disease status, we performed maximum likelihood estimation of the log-linear model using the iterative proportional fitting algorithm23 and obtained profile likelihood estimates of the confidence intervals of haplotype odds ratios.8,20,24 To compare the consistency of our findings, we also estimated haplotype frequencies using Arlequin, a software commonly employed for this purpose that is also based on an EM algorithm.9 We tested for evidence of hidden population admixture or stratification using the methods described by Pritchard and Rosenberg25, utilising data from unlinked markers.


Characteristics of the 1052 participants included in the study are shown in Table 1. Myocardial infarction cases were more likely to be male, older, hypertensive, diabetic, and smokers. Genotype frequencies in cases and controls are shown in Table 2. No evidence for population stratification was found for any of the polymorphisms tested. Under an additive genetic model, two polymorphisms showed statistically significant evidence of association with myocardial infarction: α-adducin [ADD1] (gly460trp) and cholesteryl ester transfer protein [CETP] (C-629A). These polymorphisms were also assessed under dominant and recessive models (Table 3). The ADD1 460trp variant was associated with a significant protective effect on myocardial infarction under a dominant, recessive or additive model (Table 3), which remained significant under each model after adjusting for smoking, history of hypertension or history of diabetes (in addition to age and sex). The CETP –629A variant was associated with a significant protective effect under a recessive or additive model (Table 3), which remained significant after adjustment for other risk factors.

View this table:
Table 1

Characteristics of the 1052 participants in the study

CharacteristicControls (Embedded Image)MI cases (Embedded Image)
Age in years (sd)58.6 (10.7)61.9 (9.2)
Male (%)62.068.0
Body mass index (sd)25.7 (3.6)25.9 (3.9)
Current smoker (%)17.040.2
Hypertension (%)16.831.0
Diabetes (%)2.08.7
View this table:
Table 2

Genotype frequency in cases and in controls and association with myocardial infarction under an additive genetic model

Gene and polymorphismFrequency in controls (%), Embedded ImageFrequency in cases (%), Embedded ImageP-value adjusted for age and sex
Lipoprotein Lp(a) (LPA) (C93T)71.2926.731.9875.0923.441.470.164
Lipoprotein Lp(a) (LPA) (G121A)68.8527.183.9768.3827.394.230.902
Apolipoprotein A-IV (APOA4) (thr347ser)66.3429.903.7664.7232.542.740.801
Apolipoprotein A-IV (APOA4) (gln360his)81.9817.620.4083.7315.360.910.501
Apolipoprotein B (APOB) (thr71ile)44.9544.3610.6944.6142.4112.980.498
Apolipoprotein C-III (APOC3) (C(-641)A)41.4746.0312.5037.3252.5710.110.716
Apolipoprotein C-III (APOC3) (C(-482)T)56.0438.225.7453.2042.604.200.790
Apolipoprotein C-III (APOC3) (T(-455)C)42.3845.3512.2838.5751.929.510.788
Apolipoprotein C-III (APOC3) (C1100T)58.6134.067.3354.4838.217.310.262
Apolipoprotein C-III (APOC3) (C3175G)83.5615.840.5983.3615.900.730.973
Apolipoprotein C-III (APOC3) (T3206G)43.9643.7612.2838.9447.9013.160.141
Apolipoprotein E (APOE) (no. of epsilon 4 alleles)72.8225.791.3972.0624.823.130.495
Peroxisome proliferative activated receptor, γ (PPARG) (pro12ala)75.4523.760.7979.3418.831.830.492
Adrenergic, β-3-, receptor (ADRB3) (trp64arg)85.7413.860.4087.5712.430.000.521
Lipase, hepatic (LIPC) (C(-480)T)66.1430.892.9762.1632.725.120.149
Lipoprotein lipase (LPL) (T(-93)G)97.622.380.0097.442.560.000.675
Lipoprotein lipase (LPL) (asp9asn)97.622.380.0097.622.380.000.785
Lipoprotein lipase (LPL) (asn291ser)97.032.970.0096.343.660.000.434
Lipoprotein lipase (LPL) (ser447term)79.6019.600.7980.4419.010.550.614
Paraoxonase 1 (PON1) (met55leu)13.0746.5340.4015.7243.8840.400.372
Paraoxonase 1 (PON1) (gln192arg)51.6841.786.5353.2037.669.140.875
Paraoxonase 2 (PON2) (ser311cys)56.2437.825.9458.3236.934.750.314
Low density lipoprotein receptor (LDLR) (exon 18 NcoI +/−)45.1344.7310.1446.8942.6710.440.885
Cholesteryl ester transfer protein, plasma (CETP) (C(-631)A)84.7515.050.2083.5516.090.370.485
Colesteryl ester transfer protein, plasma (CETP) (C(-629)A)23.9648.3227.7226.1453.5620.290.025*
Colesteryl ester transfer protein, plasma (CETP) (ile405val)44.3643.3712.2844.0645.3410.600.707
5,10-Methylenetetrahydrofolate reductase (MTHFR) (C677T)42.3845.5412.0843.1444.9711.881.000
Nitric oxide synthase 3 (endothelial cell) (NOS3) (A(-922)G)38.6147.9213.4737.1148.2614.630.498
Nitric oxide synthase 3 (endothelial cell) (NOS3) (C(-690)T)83.7615.051.1979.8919.200.910.090
Nitric oxide synthase 3 (endothelial cell) (NOS3) (glu298asp)42.3844.7512.8743.3844.3012.320.758
Angiotensin I converting enzyme (peptidyl-dipeptidase A) 1 (ACE) (intron 16 Ins/Del)24.1647.9227.9221.0248.9929.980.217
Angiotensin II receptor, type 1 (AGTR1) (A1166C)52.8738.818.3247.8043.049.160.229
Angiotensinogen (AGT) (met235thr)39.0144.7516.2438.7646.0715.170.953
Natriuretic peptide precursor A (NPPA) (G664A)88.5111.490.0091.228.590.180.207
Natriuretic peptide precursor A (NPPA) (T2238C)70.6326.592.7869.0628.912.030.908
Adducin 1 (α) (ADD1) (gly460trp)60.9933.665.3567.6429.802.560.006*
Sodium channel, nonvoltage-gated 1 α (SCNN1A) (trp493arg)94.465.540.0094.705.300.000.958
Sodium channel, nonvoltage-gated 1 α (SCNN1A) (ala663thr)43.5643.7612.6744.6943.5911.720.849
Guanine nucleotide binding protein (G protein), β polypeptide (GNB3) (C825T)45.9444.369.7047.6242.679.710.698
Adrenergic, β-2-, receptor, surface (ADRB2) (arg16gly)11.8847.5240.5911.5849.8238.600.793
Adrenergic, β-2-, receptor, surface (ADRB2) (gln27glu)29.3151.4919.2127.4753.1119.410.484
Matrix metalloproteinase 3 (stromelysin 1, progelatinase) (MMP3) (5A/6A)26.9348.3224.7524.6851.1924.130.723
Selectin E (endothelial adhesion molecule 1) (SELE) (ser128arg)82.3816.630.9981.7217.730.550.881
Selectin E (endothelial adhesion molecule 1) (SELE) (leu554phe)92.676.730.5989.7610.240.000.264
Intercellular adhesion molecule 1 (CD54), human rhinovirus receptor (ICAM1) (gly214arg)77.6220.791.5877.5121.391.100.799
Tumor necrosis factor (TNF superfamily, member 2) (TNF) (G(-376)A)97.432.570.0096.343.660.000.276
Tumor necrosis factor (TNF superfamily, member 2) (TNF) (G(-308)A)66.7328.914.3666.6729.853.480.742
Tumor necrosis factor (TNF superfamily, member 2) (TNF) (G(-238)A)87.1312.280.5988.1211.880.000.504
Lymphotoxin α (TNF superfamily, member 1) (LTA) (thr26asn)43.5644.1612.2841.3245.1613.530.446
Coagulation factor II (thrombin) (F2) (G20210A)96.633.370.0097.812.190.000.205
Coagulation factor V (proaccelerin, labile factor) (F5) (arg506gln)95.454.550.0095.254.750.000.983
Coagulation factor VII (serum prothrombin conversion accelerator) (F7) ((-323) 0/10bp ins)77.6221.780.5976.7822.300.910.645
Coagulation factor VII (serum prothrombin conversion accelerator) (F7) (arg353gln)80.1619.440.4078.9820.290.730.595
Serine proteinase inhibitor (plasminogen activator inhibitor type 1), member 1 (SERPINE1) (5G/4G)20.9946.9332.0819.7451.1929.070.662
Serine proteinase inhibitor (plasminogen activator inhibitor type 1), member 1 (SERPINE1) (G11053T)20.5945.9433.4718.6850.0031.320.930
Fibrinogen, B β polypeptide (FGB) (G(-455)A)67.1329.313.5669.7827.472.750.175
Integrin α-2 precursor (Platelet membrane glycoprotein Ia, GPIa) (ITGA2) (G873A)32.8749.3117.8236.7046.7916.510.158
Integrin, β 3 (platelet glycoprotein IIIa, GPIIIa) (ITGB3) (leu33pro)65.7431.682.5769.7827.662.560.170
  • A refers to baseline allele (usually the commonest allele) and B to variant allele, AA refers to baseline homozygote, AB to heterozygote and BB to variant homozygote.

  • * Statistically significant at Embedded Image.

View this table:
Table 3

Comparison of dominant, recessive and additive genetic models for polymorphisms associated with myocardial infarction in the screening analysis

Gene and polymorphismDominant modelbRecessive modelbAdditive modelb
Odds ratio (95% CIa)Embedded Image-valueOdds ratio (95% CI)P-valueOdds ratio (95% CI)Embedded Image-value
CETP (C-629A)0.87 (0.66–1.16)0.350.67 (0.50–0.90)0.0060.82 (0.68–0.97)0.025
ADD1 (gly460trp)0.74 (0.57–0.96)0.0210.46 (0.23–0.89)0.0220.73 (0.59–0.91)0.006
  • The dominant genetic model compares individuals with one or more polymorphic alleles (AB and BB genotypes combined in Table 2) with a baseline group with no polymorphic alleles (AA).

    The additive genetic model assumes that there is a linear gradient in risk between the AA, AB and BB genotypes (BB genotype baseline). This is equivalent to a comparison of the B allele versus the A allele (baseline).

    The recessive genetic model compares the BB genotype with the combined AB and AA genotypes (combined AA and AB genotypes form the baseline group).

  • a CI denotes confidence interval.

  • b Each model was adjusted for age and sex.

Haplotypes at two loci showed statistically significant evidence of association with myocardial infarction (Table 4). The apolipoprotein C III (APOC3) haplotypes CCTTCG and ATCCCG at positions –641*–482*–455*1100*3175*3206 were associated with an increased risk of myocardial infarction: CCTTCG (OR 1.41 (95% CI 1.06–1.76), Math) and ATCCCG (OR 1.71 (95% CI 1.28–2.14), Math). The paraoxonase 1 (PON1)/paraoxonase 2 (PON2) haplotype comprising met55 and gln192 in PON1 and cys311 in PON2 was associated with a significant protective effect on myocardial infarction risk (OR 0.52 (95% CI 0.39–0.77), Math). Haplotype frequencies estimated using Arlequin9 were consistent with the frequencies we obtained and presented in Table 4.

View this table:
Table 4

Estimated haplotype frequencies and odds ratios for association with myocardial infarction

HaplotypeControls Embedded ImageCases Embedded ImageOdds ratio (95% CI)
NPPA G664A, T2238C
GC162.11180.821.02 (0.76–1.27)
AT57.8149.000.79 (0.59–1.19)
APOA (LPA): C93T, G121A
CA177.34196.291.00 (0.75–1.50)
TG155.00144.350.84 (0.63–1.26)
ADRβ2: arg16gly, gln27glu
arg gln360.00400.081.00
gly gln196.00191.160.88 (0.66–1.33)
gly glu454.00502.771.00 (0.75–1.50)
APOAIV: thr347ser, gln360his
thr gln728.00792.011.00
thr his188.99207.991.01 (0.76–1.26)
ser gln92.9993.990.93 (0.70–1.40)
APOC3: C(-641)A, C(-482)T, T(-455)C, C1100T, C3175G, T3206G
CCTCCG66.4457.490.83 (0.63–1.25)
CCTTCG98.88144.011.41 (1.06–1.76)*
ACCCCT72.1580.181.07 (0.80–1.34)
ACCCCG32.3831.650.94 (0.71–1.41)
ATCCCT106.07111.471.00 (0.75–1.26)
ATCCCG19.9136.261.71 (1.28–2.14)*
ATCTCG34.4835.151.00 (0.76–1.26)
ATCTGG80.5289.021.07 (0.80–1.33)
CETP: C(-631)A, C(-629)A, ile405val
CC ile338.05408.601.00
CC val69.9578.400.83 (0.62–1.25)
CA ile265.55252.280.77 (0.57–1.15)
CA val258.45262.720.83 (0.62–1.25)
AC ile63.4069.120.82 (0.62–1.23)
AC val14.6022.881.67 (0.84–4.60)
SELE: ser128arg, leu554phe
ser leu876.08935.001.00
ser phe39.9256.001.31 (0.98–2.30)
arg leu93.92103.001.03 (0.77–1.28)
SCNN1A: trp493arg, ala663thr
trp thr326.51341.940.95 (0.71–1.42)
arg thr22.4924.711.00 (0.75–1.50)
F7: arg353gln, (-323) 10bp ins/del
arg ins894.00959.971.00
gln ins13.0615.031.08 (0.54–2.42)
gln del102.94116.971.07 (0.80–1.33)
LPL: T(-93)G, asp9asn, asn291ser, ser447term
T asp asn ser878.02951.761.00
T asp asn term104.98109.000.96 (0.72–1.44)
T asn ser ser13.6319.251.30 (0.65–2.93)
G asn asn ser11.3512.250.99 (0.50–2.49)
NOS3: A(-922)G, C(-690)T, glu298asp
AC glu525.99555.011.00
AC asp106.01114.991.03 (0.77–1.28)
GC glu128.01159.741.19 (0.89 –1.48)
GC asp161.99149.260.88 (0.66–1.31)
GT asp88.00112.351.22 (0.91–1.52)
SERPINE1: 5G/4G, G11053T
4G G424.00456.091.00
5G G15.9921.931.28 (0.64–2.88)
5G T433.00474.071.02 (0.76–1.28)
4G T136.99141.910.97 (0.72–1.45)
PON1 met55leu, gln192arg, PON2 ser311cys
met gln ser308.24370.561.00
met gln cys55.1234.180.52 (0.39–0.77)*
leu gln ser260.42276.280.88 (0.66–1.32)
leu gln cys109.22106.980.81 (0.61–1.22)
leu arg ser190.34193.160.84 (0.63–1.27)
leu arg cys83.02105.591.06 (0.79–1.32)
TNF G(-376)A, G(-308)A, G(-238)A, LTA: thr26asn
GGG thr595.00634.011.00
GGG asn157.00193.311.16 (0.87–1.44)
GGA thr55.0045.000.77 (0.58–1.15)
GAG asn190.00201.680.99 (0.75–1.50)
GAG thr13.0019.991.47 (0.72–3.25)
  • Haplotypes with frequencies of Embedded Image2% not shown in table.

  • * Significant at Embedded Image.


We examined the relationship between 58 polymorphisms in 35 candidate genes and myocardial infarction in 1052 subjects (generating 61016 individual genotypes). Our study, which is one of the largest such multilocus genetic association studies of myocardial infarction involving healthy controls published to date, revealed two polymorphisms and haplotypes at two loci significantly associated with myocardial infarction risk. To our knowledge, only two other studies have examined simultaneously the association of multiple candidate gene polymorphisms with risk of coronary heart disease. Topol et al.26 reported associations between three thrombospondin gene family variants and myocardial infarction in their study of 72 individual SNPs in 352 premature coronary artery disease cases and 418 controls recruited from a population of white Americans. Yamada et al.3 reported associations between the C1019T polymorphism in the connexin 37 gene and the 4G-668/5G polymorphism in the plasminogen-activator inhibitor type 1 gene in their study of 112 polymorphisms in 2819 Japanese myocardial infarction cases and 2242 controls. Although large, the power of Yamada's study was modest, due to the similarity of cases and controls (controls were selected who had at least one `risk factor', such as diabetes or hypertension), and by their adjustment for many potential intermediates on the causal pathway in all of their analyses. Notably, haplotype analysis was not undertaken in either study.3,26

In our study, the ADD1 460 trp allele was found to be protective against myocardial infarction. α-Adducin is a ubiquitously expressed cytoskeleton protein involved in the formation of actin-spectrin lattice, actin polymerisation and cell signal transduction,27–28 including an effect on Na-KATPase. There is considerable experimental and clinical evidence that the ADD1 (gly460trp) polymorphism is associated with functional effects related to sodium and water homeostasis. Paradoxically, it is also the 460trp allele that is associated with salt retention and tendency to hypertension,29 effects that one might expect to increase the risk of myocardial infarction. Consistent with this, Psaty et al.30 have reported a protective effect of the 460Trp allele against myocardial infarction and stroke in hypertensive patients taking diuretics. If further studies confirm that the 460Trp allele is also primarily protective against myocardial infarction, then this is likely to be through a cellular mechanism different from its effect leading to hypertension; alternatively, the observed association may be indicative of linkage disequilibrium to a protective gene variant.

We found that the CETP –629 A allele was protective against myocardial infarction. The corresponding haplotype analysis was also consistent with this finding, with a trend towards a protective effect of haplotypes containing the CETP –629 A allele (Table 4). CETP is a hydrophobic glycoprotein and has a key role in the metabolism of high density lipoproteins (HDL). CETP mediates the transfer of cholesterol esters from HDL to other lipoproteins and uptake of cholesterol by the liver. High plasma CETP is associated with reduced HDL, which is a strong risk factor for atherosclerosis.31 Recent data strongly suggest that the C-629A polymorphism is functional, with the –629A allele associated with reduced promoter activity, decreased CETP mass and increased HDL levels.32,33 Therefore, the finding of a protective effect of the –629A allele against the risk of myocardial infarction is consistent with its postulated functional effect. However, it should be noted that although the CETP locus has consistently been shown to be associated with reduced risk of myocardial infarction, there are discrepant findings as to which of the polymorphisms is associated or responsible for the effect33–36 and it is possible that –629A allele is a marker for some other variant at the locus.

We also found a significant protective effect on myocardial infarction risk of the PON1/PON2 haplotype comprising met55 and gln192 in PON1 and cys311 in PON2. The paraoxonases are glycoproteins located on the surface of HDL. They are believed to play a pivotal role in preventing low density lipoprotein (LDL) oxidation,37 which may be one mechanism by which HDL is protective against atherosclerosis. The paraoxonase genes are located as a tightly linked multigene family (also comprising PON3) at 7q21-q22.38 There is evidence that genetic variation in the paraoxonase genes influences paraoxonase activity. Thus, compared with the 192arg and the 55leu alleles, both the PON1 192gln and 55met alleles have been shown to be associated with higher paraoxonase activity and to be more efficient at inhibiting the oxidation of LDL.39,40 Correspondingly, several studies have shown these alleles to be associated with decreased risk of coronary heart disease.41,42 However, the findings have not always been consistent43,44 and recent studies have sought to identify haplotypes that may be more predictive.45,46 Sanghera et al. found a significant interaction between the PON1 gln192arg and the PON2 cys311ser polymorphisms in predicting risk of coronary heart disease in Asian Indians.41 Increased risk was restricted to carriers of both the PON1 arg and the PON2 ser alleles. Our findings are consistent with this and indicate that a haplotype that also includes the met55 allele provides the best discrimination in identifying risk associated with the PON locus.

Haplotype analysis for the polymorphisms at positions –641, –482, –455, 1100, 3175 (often referred to as the SstI S1/S2 site), and 3206 in the APOC3 gene identified two haplotypes associated with increased risk for myocardial infarction. APOC3 is a glycoprotein synthesised in the liver and intestine that is physically associated with very low density lipoprotein (VLDL), chylomicron remnants and high density lipoprotein (HDL).47 Previous studies have associated single polymorphisms in APOC3 with risk of coronary heart disease48 and altered plasma triglyceride levels.49,50 Dammerman et al.49 found a 3-fold increased risk of hypertriglyceridemia associated with a haplotype that included the –482T and SstI S2 variants, while Surguchov et al.50 reported a 4-fold increased risk of hypertriglyceridemia associated with a haplotype that included –641A, –482T, –455C and SstI S2 variants. This latter haplotype shares the same variants at the first three positions as the ATCCCG haplotype that we found to be associated with myocardial infarction risk, but we did not observe association of the more common ATCTGG (i.e., also S2) haplotype with risk. Therefore, some APOC3 haplotypes may influence myocardial infarction risk via their effect on plasma triglyceride levels while other haplotypes may play a role via other mechanisms.

There are several notable polymorphisms, which have been associated with risk of myocardial infarction in other studies that did not show any association in our study. These include the Thr26Asn polymorphism in the lymhotoxin-α gene (LTA)51 and the 5G/4G and 5A/6A polymorphisms in the plasminogen activator inhibitor type 1 and stromelysin 1 genes, respectively.3 Some of the differences may relate to the nature of the population studied. Specifically, allele frequencies are known to vary significantly between Japanese3,51 and Caucasian populations. Methods of recruitment also need to be considered. All our cases were recruited after admission into hospital. If a polymorphism has an effect, not only on risk of event but also on acute survival, then this could obscure any association. Finally, although our study was adequately powered to detect a moderate effect of a gene with a minor allele frequency similar to that of the angiotensin converting enzyme polymorphism, the power is reduced with lower allele frequencies. Thus, the lack of replication of associations between less common polymorphisms and myocardial infarction that have been reported in previous studies in our study alone would not rule out the possibility of such an association.

We adopted a range of methods to ensure that the study maintained high degree of scientific rigour, including detailed genotype and relationship error checking, testing for hidden population stratification and checking the consistency of haplotype frequency estimates. Furthermore, we minimised the number of hypotheses tested using an initial screen based on an additive genetic model. However, any genetic association study including such a large number of polymorphisms should be regarded as exploratory in character and our results must be interpreted accordingly. Hence, corrections for multiple testing, which are considered appropriate for hypothesis-testing studies, but over-conservative for exploratory studies, were not applied. However, the issue of correction for multiple testing is becoming increasingly important as multilocus analyses become popular. This is illustrated by the fact that despite our study being one of the larger studies to date, if we had applied a Bonferroni correction, no individual polymorphisms would have reached statistical significance (Math) and only one haplotype (paroxonase) would have reached borderline significance. Indeed, a study of over 5000 cases and 5000 controls would have been required to reach nominal significance (Math) after correction, to identify a polymorphism with a 50% allele frequency associated with an OR of 1.2, when 50 independent tests are being undertaken. The study sizes required increase dramatically for less common alleles. This highlights the fact that apart from much larger studies, efficient strategies for pooling of data from different studies will be required as we enter a new era of genetic association analyses for complex traits.

In conclusion, high-throughput genotyping assays will be in common use in association studies in the near future. We have shown here that such studies are feasible and can reveal genetic polymorphisms that warrant testing in further studies. Furthermore, we have demonstrated the feasibility and utility of using a number of polymorphisms in linkage disequilibrium to assess the association between specific haplotypes and myocardial infarction in a study of unrelated individuals. This may increase the power to identify disease-associated variants, a crucial step towards the eventual clinical application of new predictive markers.


We thank all participants in the study. We also thank Michael Grow and Arkadiy Silbergleit for their efforts in developing and optimising the genotyping assays. NJS and RS act as guarantors for the Leicester and Sheffield data, respectively. NJS is supported by the British Foundation.


  1. [1]
  2. [2]
  3. [3]
  4. [4]
  5. [5]
  6. [6]
  7. [7]
  8. [8]
  9. [9]
  10. [10]
  11. [11]
  12. [12]
  13. [13]
  14. [14]
  15. [15]
  16. [16]
  17. [17]
  18. [18]
  19. [19]
  20. [20]
  21. [21]
  22. [22]
  23. [23]
  24. [24]
  25. [25]
  26. [26]
  27. [27]
  28. [28]
  29. [29]
  30. [30]
  31. [31]
  32. [32]
  33. [33]
  34. [34]
  35. [35]
  36. [36]
  37. [37]
  38. [38]
  39. [39]
  40. [40]
  41. [41]
  42. [42]
  43. [43]
  44. [44]
  45. [45]
  46. [46]
  47. [47]
  48. [48]
  49. [49]
  50. [50]
  51. [51]
View Abstract