The antibiotic flucloxacillin is an established rare cause of serious liver injury, with a number of studies showing risks of around 8 per 100 000 general population users. More recently, analyses looking specifically at different age groups and variations in duration of therapy have shown that the risk increases substantially by age and number of prescriptions, with 110 per 100 000 people over the age of 70 who receive consecutive flucloxacillin prescriptions experiencing jaundice. While genetic studies have identified that people with a specific genetic characteristic (the HLA-B*5701 genotype) are much more likely to experience flucloxacillin-induced liver injury than those without it, further characterisation of the risk is needed to guide clinicians' prescribing decisions. One important unknown is whether flucloxacillin-induced liver injury only occurs as a result of first ever exposure to the drug, or if people who have had an initial course of therapy without suffering an injury remain at risk. In this study, we will compare the risk of liver injury in people initiating a second course of flucloxacillin therapy with the risk of liver injury within people treated with oxytetracycline (an antibiotic used to treat similar conditions that is not associated with liver injury).
This study aims to quantify the association between being prescribed flucloxacillin and serious drug-induced liver injury within people who have already previously completed a course of flucloxacillin therapy. Two different liver injury outcomes will be assessed: a symptom-defined outcome (jaundice) and a laboratory confirmed outcome (any of a number of symptoms indicating liver injury along with a liver test result indicative of drug-induced liver injury). The 1-45 day risk for each outcome in the flucloxacillin-exposed group will be calculated and compared to the 1-45 day risk in a group of patients consisting of people prescribed oxytetracycline, a comparator antibiotic that is prescribed for similar indications. Relative effects will be calculated as odds ratios, which will be interpreted as risk ratios given the rarity of the outcome under study (likely to be less than 8 per 100 000 in the general population of users). Logistic regression will then be used to estimate a risk ratio adjusted for important confounders of the association. Variables considered as confounders of the association will include age, gender, smoking, ethnicity, BMI, alcohol intake, socio-economic status, use of other drugs known to cause liver injury and calendar period.
Health Outcomes to be Measured:
- Drug-Induced liver injury