Introduction

Adverse drug reactions (ADRs) are a significant cause of morbidity and mortality, leading not only to individual treatment failures but also to substantially increased healthcare costs. ADRs have been estimated to cause or contribute to at least 5–7% of hospital admissions [1,2,3,4], and to about 3% of all fatalities [5]. About 10% of the Swedish health-care budget has been attributed to ADRs [6].

Virtually all drugs are unsafe in a subset of patients even when used according to the approved label. There is good reason to believe that a significant part of an individual’s risk of being intolerant to a drug is explained by genetic predisposition [7]. In some cases, dose-dependent ADRs are known to be caused by mutations in genes involved in the metabolism of the drug or in the drug target. An example is the dose-dependent bone-marrow suppression that develops in patients with defective detoxification of thiopurines related to genetic variants of the key enzyme thiopurine methyltransferase [8].

Other types of serious ADRs appear to be less dependent on dose. These ADRs—so called idiosyncratic or type B reactions—can affect various organ tissues, including e.g., the heart, liver, skin, kidney, and muscle or cause generalized hypersensitivity reactions [9]. In this category of ADRs, the cause is generally unknown and there are no obvious candidate genes. Microarray-based genotyping of multiple genetic variants as well as next-generation sequencing (NGS), where rarer sequence variants can be discovered, have made it possible to perform genome-wide association studies (GWAS) of these reactions, as well as other types of association analyses. For both of these methods, large numbers of patients are usually needed due to the statistical requirement of correction for multiple testing and for replication of findings.

In the relatively few large-scale genome-wide studies performed on serious ADRs so far, immune-related genetic variants involving the human leukocyte antigen (HLA) molecules in the major histocompatibility complex (MHC) on chromosome 6 have often been implicated as risk factors [10]. Such risk factors have been shown to be drug-specific and to vary between different ethnic populations. A well-known example is the association between HLA-B*57:01 and the abacavir-induced hypersensitivity syndrome [11]. Abacavir, a nucleoside reverse transcriptase inhibitor used in the treatment of human immunodeficiency virus (HIV), is associated with hypersensitivity reactions in 5–8% of patients [12]. Introduction of HLA-B*57:01 screening prior to abacavir therapy has reduced the incidence of this ADR from up to 8 to <1% [13].

In the current project, we are collecting ADRs on a large scale in Sweden. We aim to establish a large nation-wide DNA sample collection with clinical data to enable studies of both genetic and clinical risk factors of severe ADRs in order to improve the benefit/risk balance of drug treatment. The ultimate goal is to develop predictive tests and models that minimize the risk of severe ADRs, and thus reduce patient suffering and health-care costs. Such tests can also be used for diagnostic purposes to differentiate an ADR from spontaneous disease.

Materials and methods

Patient recruitment and data collection

SWEDEGENE (www.swedegene.se) was established in 2008 and is a Swedish nation-wide DNA sample collection with phenotype data on cases of ADRs. Most patients are identified and recruited through the Swedish national database of spontaneously reported ADRs run by the Medical Products Agency since 1965. Non-fatal cases reported from 1990 and onwards are extracted from the database, and each reporter is asked whether the patient can be approached about participation in SWEDEGENE. In addition, patients can be recruited directly from collaborating clinicians at health-care facilities. When a clinician at these collaborating centers identifies a suitable patient, the patient is either directly recruited in collaboration with SWEDEGENE, or a research nurse at SWEDEGENE will approach the patient and ask for participation. Another mode of recruitment is through advertising campaigns.

Population controls are obtained from the Swedish TwinGene biobank that has genome-wide data from over 10000 twins as well as whole-genome sequencing data from 1000 individuals born 1958 and before [14, 15]. Only one twin out of each pair is selected as a control. Through linkage with the Swedish Prescribed Drug Register and the National Patient Register kept by the Swedish National Board of Health and Welfare, diagnoses and drug prescriptions are matched between cases and controls. We also collect treated controls with full phenotype information directly from Swedish health-care facilities when necessary. The mode of recruitment for controls is identical to that for cases.

A study kit is provided to each consenting patient or treated control including a questionnaire holding information about demographics, medical history, environmental factors and information about drug treatment, as well as an informed consent form. A research assistant contacts the patient by telephone and the questionnaire is completed through a telephone interview. If needed, copies of the participant’s medical and laboratory records are obtained. In addition, blood samples are drawn at the patient’s nearest health-care facility, and sent to the central laboratory at Uppsala University Hospital, where they are stored for later use. If a patient is reluctant to draw blood, saliva sampling is undertaken instead.

Phenotype data concerning the drug suspected to have caused the ADR, the indication for which the drug was prescribed, concomitant drugs and diseases, a summary code for the ADR, demographic variables (sex and age), relevant laboratory data, a brief narrative, and all information acquired through the questionnaire is compiled in a study database by a research nurse. The same questionnaire is used for all cases except for certain ADR specific questions. For treated controls, the questionnaire contains demographic variables, all drug treatments and diseases. To ensure the security of participant data, the clinical data is stored in a local encrypted database. Access to the database is limited to specific computers, and access is locked behind passwords and two-factor authentication. To further limit the potential for data breach the user is only allowed to view essential data for the user-group, with access to other parts locked behind a permission system.

Any type of genetic data is pseudonymized and stored separate from clinical data. For smaller data volumes in the range 1–20 TB, data is archived on secured encrypted drives. Data that is currently being analyzed is stored on the UPPMAX Bianca Cluster at Uppsala University (www.upmax.se). Secure archiving of larger data volumes has not yet been needed. As the genetic analyses that are conducted are not approved as clinical tests in routine health-care, genomic results are not returned to participants.

Inclusion and exclusion criteria

All patients included are at least 18 years of age and able to give informed consent. To be included in the study, the initial event should have occurred after the start of treatment and in some instances after withdrawal of the drug. Causality is assessed with the WHO standard algorithm [16]. Certain ADRs have specific inclusion and exclusion criteria and are adjudicated by clinical experts. Examples of such critera from published studies are given in Table 1.

Table 1 Examples of inclusion and exclusion criteria for cases of adverse drug reactions in published studies.
Full size table

Genomic analysis

Power calculations for GWAS using a dominant genetic model show that 50 cases and 5000 controls give us 80% power to detect an odds ratio of 3–4 with a minor allele frequency of 40%, and 80% power to detect an odds ratio of 4–5 for variants with a minor allele frequency of 20%. This is based on the conventional genome-wide significance threshold of p < 5 × 10−8 [17].

Results

To date, SWEDEGENE has DNA and curated clinical data from about 2550 individuals that have experienced specific ADRs. A list of collected ADRs as per July 2019 with at least 15 cases is presented in Table 2. We have also collected 580 drug-treated controls, and the largest group is methotrexate-treated rheumatoid arthritis patients showing no signs of liver toxicity, and individuals exposed to the swine influenza A (H1N1) vaccination Pandemrix without having developed signs of narcolepsy. However, for most ADRs comparisons are made with the 5000 populations controls with genome-wide data and 1000 with whole-genome sequencing data from TwinGene.

Table 2 Collected adverse drug reaction diagnoses with at least 15 cases presented in decreasing order of frequency as per July, 2019.
Full size table

GWAS of agranulocytosis induced by antithyroid drugs or sulfasalazine, cough induced by angiotensin-converting enzyme (ACE) inhibitors, narcolepsy induced by Pandemrix and atypical femoral fractures induced by bisphosphonates have been published [18,19,20,21,22]. SWEDEGENE has also provided cases and controls in several other collaborative studies, such as GWAS and whole  exome sequencing  of statin induced myopathy [23, 24], drug induced liver toxicity [25,26,27,28,29,30], and hypersensitivity reactions to carbamazepine [31]. Additional GWAS and whole  genome exome  sequencing studies on ADR diagnoses collected by SWEDEGENE are currently underway. In addition, 1000 selected SWEDEGENE individuals have been whole genome sequenced and are  being compared with  1000 whole  genome sequenced individuals from TwinGene. This will give us the possibility to find novel genetic associations with ADRs and to map population frequencies of known pharmacogenomic targets in the Swedish population. As for planned analyses, enrichment tests [32], pathway based analysis, and genome-wide complex trait analysis [33] will be performed beyond GWAS.

As many pharmacogenomic targets are rare variants, novel associations in a limited sample population can be hard to detect. To increase the probability there are two options; increase the sample size or decrease the number of tested associations. As rare ADRs are, by definition, rare, our option is to decrease the number of tests. This will be done by selecting variants a priori based on predicted mutation effects using software and databases as Ensembl Variant Effect Predictor (VEP) [34], Eigen- [35] or Combined Annotation-Dependent Depletion (CADD) scores [36]. To further decrease the number of tests we will use enrichment tests such as burden and non-burden tests, to test for genetic burden on an exon or pathway basis [32].

Future perspective

SWEDEGENE is an important resource for pharmacogenetic studies of ADRs. Due to our unique nation-wide collection, SWEDEGENE has the potential to discover novel genetic and clinical risk factors for rare and serious ADRs, with the ultimate goal to identify patients at risk and to improve the benefit/risk balance of drug treatment. Since pharmacogenetic variants in general have larger effect sizes than variants that increase the risk of complex diseases [37], the clinical benefit is estimated to be great. It is also easier to select an alternative drug than to modify risk factors for complex diseases. It has already been shown that hospitalizations and emergency department visits can be reduced by genotyping elderly polypharmacy patients [38], and that pre-prescription genotyping is cost-effective for certain ADRs [39]. Barriers for implementing genotype-based drug therapy will be overcome once patients have their genome readily available in the medical record [40]. This is estimated to happen within the near future through emerging Precision Medicine initiatives. SWEDEGENE is well placed to discover novel gene-ADR associations to be analyzed in these initiatives and invites collaborators for joint efforts.