Article Text

Download PDFPDF
Original article
ABCB1 gene variants, digoxin and risk of sudden cardiac death in a general population
  1. Maartje N Niemeijer1,
  2. Marten E van den Berg2,
  3. Jaap W Deckers3,
  4. Adrianus L H J Aarnoudse4,
  5. Albert Hofman1,
  6. Oscar H Franco1,
  7. André G Uitterlinden1,5,
  8. Peter R Rijnbeek2,
  9. Mark Eijgelsheim1,5,
  10. Bruno H Stricker1,5,6
  1. 1Department of Epidemiology, Erasmus MC—University Medical Center Rotterdam, Rotterdam, The Netherlands
  2. 2Department of Medical Informatics, Erasmus MC—University Medical Center Rotterdam, Rotterdam, The Netherlands
  3. 3Department of Cardiology, Erasmus MC—University Medical Center Rotterdam, Rotterdam, The Netherlands
  4. 4Department of Internal Medicine, Catharina Hospital, Eindhoven, The Netherlands
  5. 5Department of Internal Medicine, Erasmus MC—University Medical Center Rotterdam, Rotterdam, The Netherlands
  6. 6Inspectorate of Health Care, Utrecht, The Netherlands
  1. Correspondence to Professor Bruno H Stricker, Department of Epidemiology, Erasmus C—University Medical Center Rotterdam; PO Box 2040, Rotterdam 3000CA, The Netherlands; b.stricker{at}


Objective The ATP-binding cassette B1 (ABCB1) gene encodes P-glycoprotein, a transport protein, which plays an important role in the bioavailability of digoxin. We aimed to investigate the interaction between variants within the ABCB1 gene and digoxin on the risk of sudden cardiac death (SCD).

Methods Within the Rotterdam Study, a population-based cohort study in persons 45 years of age and older, we used Cox regression to analyse the association between three polymorphisms that have been associated with digoxin bioavailability, extracted from 1000-Genomes imputed ABCB1 genotypes and the risk of SCD, stratified by digoxin use.

Results In a total study population of 10 932 persons, 419 SCDs occurred during a median follow-up of 9.8 years. In non-users of digoxin, the risk of SCD was not different across genotypes. In digoxin users, homozygous T allele carriers of C1236T (HR 1.90; 95% CI 1.09 to 3.30; allele frequency 0.43), G2677T (HR 1.89; 95% CI 1.10 to 3.24; allele frequency 0.44) and C3435T (HR 1.72; 95% CI 1.03 to 2.87; allele frequency 0.53) had a significantly increased risk of SCD in a recessive model. Interaction between the ABCB1 polymorphisms and digoxin use was significant for C1236T and G2677T in the age-adjusted and sex-adjusted model.

Conclusions In this study, we showed that in digoxin users variant alleles at each of the three loci in the ABCB1 gene were associated with an increased risk of SCD compared with digoxin users with none or one T allele. If replicated, the findings imply that the ABCB1 genotype modifies the risk of cardiac digoxin toxicity.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


The ATP-binding cassette B1 (ABCB1) gene, formerly known as the MDR1 gene, encodes P-glycoprotein.1 This transport protein influences resistance to and bioavailability of several drugs,1–3 including digoxin.4 ,5 Digoxin is the oldest cardiac drug still in use today6 and is prescribed for atrial fibrillation and mild-to-moderate heart failure.7 Digoxin has a narrow therapeutic range, and its use is, therefore, associated with a high risk of intoxications.8 ,9 Digoxin causes shortening of the heart rate corrected QT (QTc) interval10 and especially in high concentration digoxin can lead to arrhythmias and sudden cardiac death (SCD).8 ,9 From a meta-analysis in 2005, no conclusions could be drawn on the potentially modifying effect of ABCB1 single-nucleotide polymorphisms (SNPs) on digoxin serum concentrations.11 However, more recent studies did show that variant allele carriers have higher serum concentrations.5 ,12 Preliminary results showed that genetic variants within the ABCB1 gene have a modifying effect on the risk of SCD among digoxin users.13 The objective of this study was to further investigate whether ABCB1 SNPs modify SCD risk during use of digoxin in a general population of middle-aged and elderly persons.


Setting and study population

The Rotterdam Study is a prospective population-based cohort study in the city of Rotterdam, the Netherlands. Details regarding design, objectives and methods of the Rotterdam Study have been described elsewhere.14 ,15 In short, all inhabitants of the Ommoord district, ≥55 years, were invited to participate. At baseline (1990–1993), 7983 participants (response rate 78%) were included. In 2000, an additional 3011 participants were enrolled (response rate 67%). This extension consisted of all persons living in the study district who had become 55 years of age or had moved into the study district. A second extension of the cohort was initiated in 2006, in which 3932 participants (response rate 65%), ≥45 years, were included. Follow-up examinations were conducted periodically, which consist of a home interview and an extensive set of tests at a research facility. Participants are continuously monitored for morbidity and mortality through linkage of general practitioners and municipality records to the study base.

The study population for this analysis consisted of all participants who gave informed consent for follow-up monitoring and had pharmacy, genetic and covariable data available. At every date an SCD case occurs, that is, the index date, all participants still in the cohort are sampled as controls and exposure is assessed for cases and controls on that date.16


Medication-dispensing data were obtained from all seven fully computerised pharmacies in the study district. Information on all filled prescriptions from 1 January 1991 onwards was available and included information on the product name, the Anatomical Therapeutic Chemical (ATC) code,17 the dispensing date, the prescribed dosing regimen and the amount dispensed. For every filled prescription of digoxin (ATC code C01AA05), dispensing episodes were calculated by dividing the total number of tablets/capsules by the daily prescribed number with a carryover period of 7 days. A person was considered exposed if the index date fell within a dispensing episode. The defined daily dose (DDD) of digoxin is 0.25 mg.17

Genotyping and imputation

A total of 12 453 subjects were genotyped with Illumina 500(+duo) and Illumina 610 quad and 11 496 subjects passed genotyping quality control. Exclusion criteria were a call rate <98%, Hardy–Weinberg p value <10−6, minor allele frequency <0.01%, excess autosomal heterozygosity >0.336, sex mismatch and outlying identity-by-state clustering estimates. Data were imputed with the 1000-Genomes reference panel (phase 1, V.3) using MACH V.1.0.15/1.0.16. We selected three SNPs within the ABCB1 gene, which have been associated with digoxin bioavailability: C1236T, G2677T/A and C3435T.5 ,18 Imputation quality for all three SNPs was high (>0.99). The SNPs are in relatively strong linkage disequilibrium: D′=0.930, R2=0.865 between C1236T and G2677T, D′=0.891, R2=0.449 between C1236T and C3435T, and D′=0.946, R2=0.506 between G26777T and C3435T.19 Since the allele frequency of the A allele within the triallelic G2677T/A is low in the general population (around 2%19), this SNP was imputed as a biallelic SNP with G as the major and T as the minor allele. We rounded the imputed SNP dosages to 0 (CC for C1236T and C3435T/GG for G2677T), 1 (CT for C1236T and C3435T/GT for G2677T) and 2 (TT).

Outcome definition

SCD was defined according to Myerburg's definition endorsed by the European Society of Cardiology: “a natural death due to cardiac causes, heralded by abrupt loss of consciousness within one hour from onset of acute symptoms; pre-existing heart disease may have been known to be present, but the time and mode of death are unexpected”.20 ,21 Identification of SCD cases was done independent of the current research question by two research physicians and confirmed by an experienced cardiologist after reviewing the medical files as described in more detail previously.22 ,23 Follow-up was complete for 96% of deaths until 1 January 2011.


Information on smoking status was ascertained through structured interviews. Participants were divided into three categories: never, past or current smokers. Standard, 12-lead ECGs were made during the centre visits and QT and RR interval were measured using the modular ECG analysis system.24 ,25 To correct for heart rate, Bazett's formula, QTc=QT/√RR, was used.26 On each index date, data on these covariables from the most recent examination or interview were included as time-dependent covariable.16 Heart failure diagnosis was adjudicated in accordance with the guidelines of the European Society of Cardiology and included typical signs or symptoms of heart failure confirmed by objective evidence of cardiac dysfunction.22 ,27 A history of coronary heart disease was defined as a history of myocardial infarction or a coronary revascularisation procedure.22 Atrial fibrillation was ascertained through ECG measurements and medical records.22 Serum digoxin concentrations were determined through linkage to the ‘Star-Medisch Diagnostisch Centrum’, which performs all outpatient laboratory assessments for the general practitioners in the area of Rotterdam from 1 April 1997 onwards, and were thus only available in a subset of participants.

Data analysis

To estimate the effect of ABCB1 SNPs on digoxin concentration in our study, we used linear regression adjusted for age, sex and digoxin dose. We used only the first available serum measurement of each subject to limit the effect of digoxin dose titration.

We used Cox proportional hazard models to analyse the association between digoxin use and the occurrence of SCD and the association between the genotypes and risk of SCD, stratified by time-dependent digoxin use.16 We fitted an additive, recessive or dominant model, based on the outcome from the general genotype models, that is, HRs for GT/CT and for TT with CC/GG as reference. We tested for interaction between digoxin use and genotypes using the Wald test. Follow-up time was calculated from 1 January 1991 until date of death, loss to follow-up (n=93) or the end of the study period (1 January 2011), whichever came first. Subjects who entered the study after 1 January 1991 were counted at risk only during the time at which they were under observation. Age at baseline and sex were included in model 1. The other covariables were added in model 2. Digoxin dose was subsequently added (model 3).

A two-sided p value <0.05 was considered statistically significant. Measures of association were presented as mean differences in digoxin serum concentration or HRs for the risk of SCD with 95% CIs. Data were analysed using IBM SPSS Statistics V.21.0 (IBM, Somers, New York, USA).

Subsample and sensitivity analyses

Haplotype data were available in a subsample of the study population. Genotyping was done using TaqMan allelic discrimination assays as previously described.5 Haplotypes were estimated using HaploStats 1.3.0 package for R 2.5.0 using haplo.em.28 ,29 We excluded haplotypes with a posterior probability <0.95. The association between the main variant diplotype TTT-TTT and the occurrence of SCD was tested similar to the genotype analyses, with the CGC-CGC diplotype as reference. The other diplotypes were included as a dummy variable.

As digoxin elimination is mainly by renal excretion and involves P-glycoprotein, kidney function could be an important covariable. However, kidney function was only available at baseline for the second and third subcohorts and on two visits for the first subcohort. Therefore, we included kidney function as covariable in a second subsample analysis. Kidney function (estimated glomerular filtration rate) was calculated according to the abbreviated modification of diet in renal disease formula.30 For the first subcohort, the baseline measurement (if available) was used for index dates until March 1997 (start of the visit with a new kidney function measurement). After that, the kidney function of that visit was used (if available).

To account for possible limitations of Bazett's formula, in a sensitivity analysis we included QT and RR together in model 2 instead of using QTc. We also analysed model 2 without adjustment for QTc or QT interval as this could be an intermediate factor.


General characteristics

The study population consisted of 10 932 participants (figure 1). General characteristics of the total study population are shown in table 1. The baseline characteristics stratified on ever digoxin use are shown in online supplementary table S1. The mean age was 65.2±9.6 years, and 42% of the participants were male. During a median follow-up of 9.8 years (IQR 4.1–17.0), 419 SCDs occurred; 351 during non-use of digoxin, 68 during digoxin use. The mean digoxin dose in the total study population of digoxin users was 0.55±0.25 DDD.

Table 1

Baseline characteristics of the total study population

Figure 1

Selection of the study population. RS, Rotterdam Study.

Serum digoxin concentration and ABCB1

Dosage information and serum digoxin measurement were available in 175 people, who used a mean DDD of 0.59±0.33 and had a mean serum concentration of 1.0±0.6 μg/L. ABCB1 gene SNPs were associated with an increased digoxin serum concentration (table 2). The results of the C1236T and G2677T were most consistent with an additive model. Per additional T allele for C1236T, the mean serum digoxin concentration was 0.20 μg/L (95% CI 0.08 to 0.33) higher, for G2677T 0.19 μg/L (95% CI 0.06 to 0.32) and for C3435T 0.12 μg/L (95% CI 0.00 to 0.24).

Table 2

Mean difference in digoxin serum concentration (μg/L; with 95% CI) per ABCB1 genotype according to a genotype, additive, dominant and recessive model

ABCB1 and risk of SCD

Digoxin use was associated with an increased risk of SCD in models 1 and 2 (HR 3.81; 95% CI 2.92 to 4.98 and HR 1.95; 95% CI 1.44 to 2.64, respectively). Results for the analyses on the effect of ABCB1 variants stratified on digoxin use are presented in table 3. In non-users of digoxin, the risk of SCD was not associated with the genetic variants. In digoxin users, homozygous minor allele carriers for the C1236T and G2677T SNPs had a significantly higher risk of SCD in model 1 (HR 1.92; 95% CI 1.03 to 3.58 and HR 1.92; 95% CI 1.04 to 3.55, respectively). In models 2 and 3, this was no longer significant.

Table 3

Effect of ABCB1 genotype on the risk of sudden cardiac death, stratified on digoxin use

Based on the results from the genotype model, we additionally analysed the SNPs according to a recessive model (table 4). According to model 2, the risk of SCD was significantly increased in homozygous minor allele carriers of C1236T and G2677T. In model 3, all three SNPs were associated with a significantly increased risk of SCD; C1236T HR 1.90 (95% CI 1.09 to 3.30), G2677T HR 1.89 (95% CI 1.10 to 3.24) and C3435T HR 1.72 (95% CI 1.03 to 2.87). When testing interaction between digoxin use and genetic variants according to the recessive model, the interaction was significant only for C1236T and G2677T in model 1. We did not adjust for serum digoxin concentration as we had concentration measurements available in only 12 SCD cases.

Table 4

Effect of ABCB1 genotype on the risk of sudden cardiac death according to the recessive model, stratified on digoxin use

Subsample and sensitivity analyses

In 5572 participants, haplotypes were estimated, and after exclusion of participants with a posterior probability <0.95, 5089 remained with a mean posterior probability of 0.99. The haplotype analyses showed comparable results with slightly higher HRs than the genotype analyses, especially in the recessive model (table 5).

Table 5

Subsample analysis: risk of sudden cardiac death associated with ABCB1 diplotype, according to the diplotype and recessive model, stratified on digoxin use (n=5089)

Additional adjustment for estimated glomerular filtration rate gave comparable results (see online supplementary table S2). Leaving QTc interval out of the model or adjusting for QT and RR interval together instead of using QTc did not change the estimates substantially (see online supplementary table S3).


In this study, we showed that among users of digoxin homozygous T allele carriers of ABCB1 gene SNPs C1236T, G2677T and C3435T have an increased risk of SCD. In non-users of digoxin, this effect is not present. These findings might be useful for informed decision-making in personalised pharmacotherapy.

We showed an increase in digoxin serum concentration of 0.12–0.20 μg/L per T allele of all three SNPs or an increase of 0.27–0.39 μg/L in homozygous T allele carriers. As for most people a serum concentration between 0.8 and 2.0 μg/L is considered therapeutic,7 this is a substantial increase.

A remaining question is whether the SNPs have an effect independent of serum concentrations in digoxin users. The results for the association between ABCB1 SNPs and serum digoxin concentration suggest an additive or dominant model, while the association of the SNPs with SCD suggests a recessive effect of the variant alleles. Given these different results, it is possible that the effect on the risk of SCD is not completely explained by the higher serum concentrations. Another possibility is that the serum concentration has to exceed a certain threshold before it increases the risk of SCD, which is possibly mainly reached in persons homozygous for the T allele. Unfortunately, we could not adjust for serum concentrations as only a limited number of participants had serum concentration measurements available. The low number of available serum concentration measurements could be explained by the fact that we only have these measurements if they are requested by the general practitioners and not during hospital or outpatient clinic visits. Therefore, further studies are needed to establish whether the effect of the ABCB1 SNPs on SCD is fully explained by higher serum digoxin concentrations or that the SNPs also have an independent effect on the risk of SCD in digoxin users. If this is the case, serum concentration measurements alone will probably not be the most adequate and safe monitoring method. As genotyping has become less expensive over the past years, genotyping new digoxin users might become a clinically useful addition to enhance safe use of digoxin. The p values for interaction between digoxin and C1236T and G2677T were significant only in the age-adjusted and sex-adjusted model. For C3435T, there was no significant interaction. This could be due to the small number of exposed cases who were homozygous major allele carriers. However, this limits the possibility to draw firm conclusions on the modifying effect of ABCB1 variants. The finding of a significant association and interaction for C1236T and G2677T probably reflect the same association signal given the high degree of linkage disequilibrium. C3435T has been used in earlier studies that used haplotype analyses that showed stronger associations compared with single SNP analysis. In our subsample with haplotype data available, participants with a TTT-TTT diplotype had a slightly higher risk compared with the genotype analyses.

Our study has strengths and limitations. The high level of completeness of follow-up in this cohort is an important asset. We had access to detailed information on morbidity and mortality through access to the medical records. We also had data available on a large number of time-dependent covariables, limiting the confounding in our analyses. Unfortunately, kidney function measurements were not available for all participants, and for participants who did have measurements available, the majority only had a baseline measurement, which could introduce misclassification in the time-dependent analyses. Therefore, we adjusted for this covariable in subsample analyses, which showed comparable results to the main analyses. Within the Rotterdam Study, data are collected according to standardised protocols. We stratified on digoxin use since digoxin is used in subjects who already have a higher risk of SCD and in such patients confounding by indication is usually difficult to exclude.

A limitation of our study is the heterogeneity of SCD cases that is introduced through the definition of SCD. However, this definition has been widely endorsed for several years and incidence rates in our study are comparable to other studies.23 Moreover, misclassification would probably be random for the variant alleles and mean that our estimates are conservative and the risk of SCD may be even higher. Second, since we do not have genotype data for all participants, we rounded imputed dosages to obtain categorical estimates of the genotypes. Although this could lead to misclassification, imputation quality was high for all three SNPs. Since the use of imputed data hampered haplotype analysis, we analysed haplotypes within a subsample of the population with genotype data available. We only studied three SNPs, but with the emergence of whole-genome and sequencing data in large cohorts, more variants within genes could be studied to find the variants with the best predictive value. Another limitation is that our results are obtained in an older and white population. Therefore, our results may not necessarily be generalisable to younger and non-white individuals. Fourth, the number of exposed cases within our study population was low, especially for serum digoxin concentrations. Finally, we did not correct for multiple testing as this is an a priori defined candidate-gene study based on previous findings on the associations between digoxin, ABCB1 gene variants and SCD, and second because the three SNPs are highly correlated. The choice for the recessive model was not pre-hypothesised, but based on the results of the genotype model. Our findings should be replicated in an independent cohort before this knowledge can be applied to clinical practice.

In conclusion, we showed that homozygous T allele carriers of the ABCB1 SNPs C1236T, G2677T and C3435T using digoxin have an increased risk of SCD compared with digoxin users with none or one T allele, while there was no effect in non-users of digoxin. As these variant alleles are common in the general population, this might influence a substantial part of the digoxin users. If these findings can be replicated in an independent cohort, testing genetic variants in new users of digoxin might add to safe use of digoxin if drug concentration monitoring is insufficient to reduce the associated risk in a specific group of patients who are at high risk of the adverse event on the basis of underlying morbidity.

Key messages

What is already known on this subject?

  • ATP-binding cassette B1 (ABCB1) gene variants modify the digoxin serum concentration, and use of digoxin has been associated with an increased risk of sudden cardiac death.

What might this study add?

  • This study shows that ABCB1 gene variants modify the risk of sudden cardiac death in digoxin users.

How might this impact on clinical practice?

  • If these findings can be replicated in an independent cohort, testing ABCB1 gene variants in new users of digoxin might add to safer use of digoxin if drug concentration monitoring alone is insufficient to reduce the associated risk in a specific group of patients.


The dedication, commitment and contribution of inhabitants, general practitioners and pharmacists of the Ommoord district to the Rotterdam Study are gratefully acknowledged.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors Conceived and designed the experiments: AH, BHS, OHF and AGU. Performed the experiments: MNN, MEB, ME, JWD, PRR, BHS and ALHJA. Analysed the data: MNN, MEB, ME, PRR, BHS and ALHJA. Wrote the first draft of the manuscript: MNN. Contributed to the writing of the manuscript: AH, BHS, OHF, PRR, ME, JWD, MEB, AGU and ALHJA. Agree with manuscript results and conclusions: all authors.

  • Funding This work was supported by grants from the Netherlands Organisation for Health Research and Development (ZonMw) (Priority Medicines Elderly 113102005 to ME and PRR; and HTA 80-82500-98-10208 to BHS). OHF works in ErasmusAGE, a centre for ageing research across the life course funded by Nestlé Nutrition (Nestec); Metagenics; and AXA. The generation and management of GWAS genotype data for the Rotterdam Study is supported by the Netherlands Organisation of Scientific Research NWO Investments (nr. 175.010.2005.011, 911-03-012). This study is funded by the Research Institute for Diseases in the Elderly (014-93-015; RIDE2), the Netherlands Genomics Initiative (NGI)/Netherlands Organisation for Scientific Research (NWO) project no. 050-060-810. The Rotterdam Study is supported by the Erasmus MC and Erasmus University Rotterdam; the Netherlands Organisation for Scientific Research (NWO); the Netherlands Organisation for Health Research and Development (ZonMw); the Research Institute for Diseases in the Elderly (RIDE); the Netherlands Genomics Initiative (NGI); the Ministry of Education, Culture and Science; the Ministry of Health Welfare and Sport; the European Commission (DG XII); and the Municipality of Rotterdam.

  • Competing interests None declared.

  • Patient consent Obtained

  • Ethics approval The Rotterdam Study was approved by the medical ethics committee according to the Wet Bevolkingsonderzoek ERGO (Population Study Act Rotterdam Study) executed by the Ministry of Health, Welfare, and Sport of the Netherlands.

  • Provenance and peer review Not commissioned; externally peer reviewed.