Risk of chronic kidney disease in patients with gout and the impact of urate lowering therapy: a population-based cohort study

This retrospective cohort study utilised data from the Clinical Practice Research Datalink (CPRD). The CPRD is a large database containing anonymised UK primary care medical records [22]. Approximately 98% of the population of England and Wales is registered with a general practitioner (GP), who is responsible for the majority of a patient’s medical care [23]. The CPRD covers more than 7% of the UK population and is representative of the general UK population in terms of age and gender distribution [23]. More than 58% of CPRD practices are linked to hospital episode statistics (HES). HES holds data items including admissions, diagnoses and operative procedures for all patients treated in hospitals in England [24]. The linkage is performed by a trusted third party based on National Health Service number, date of birth and gender. As HES only covers England; practices from Scotland, Wales and Northern Ireland were excluded from this analysis.

In this cohort study the exposed group consisted of individuals with a first-ever recorded diagnosis of gout and these were identified from general practice between 1998 and 2016 using previously published methods [25]. Ascertainment of gout was based on a medical (Read) code assigned by the GP. Gout diagnoses have been validated in the CPRD and have a positive predictive value of 90% [26]. Each patient with gout was assigned an index date corresponding to the date of gout diagnosis and randomly matched to one patient without a gout diagnosis or evidence of ULT, on age (± 5 years), gender, available follow-up time (± 3 years) and practice. Matching on follow up is a common approach when using the CPRD as patients with chronic illness typically have longer follow up compared to those without, and gout is associated with several comorbidities [25], it is a proxy method of minimising the potential bias this may induce. For both exposed and unexposed patients, follow up commenced from the index date. Those with evidence of CKD stage ≥ 3 or RRT before the index date or < 1 year after the index date were excluded from the study.

The primary outcome was developing CKD stage ≥ 3 and was based on two consecutive measurements of eGFR< 60 mL/min/1.73m² at least 3 months apart. eGFR was calculated using serum creatinine values recorded in patients’ medical records using the Chronic Kidney Disease Epidemiology Collaboration equation [27]. For those considered to have CKD stage ≥ 3, the date of the first eGFR measurement was taken as the first occurrence of CKD. We also identified patients with CKD stage ≥ 3 or more based on a recorded diagnosis of CKD stages 3–5, ESRD or having evidence of renal replacement therapy (RRT (kidney transplant or dialysis)) in their primary or secondary care medical record.

To assess the independent association between gout and CKD stage ≥ 3, information on various baseline characteristics was extracted. These included body mass index (BMI), smoking status, index of multiple deprivation (IMD), and specific comorbidities. The comorbidities included were; myocardial infarction, systemic lupus erythematosus (SLE), rheumatoid arthritis, congestive heart failure, cerebrovascular disease, peripheral vascular disease, hospitalisations and treated hypertension or diabetes mellitus before the index date. Information was extracted on NSAID use (two or more prescriptions) in the 6 months before gout diagnosis. In addition, baseline serum uric acid (SUA) level was adjusted for in the analyses examining risk of CKD associated with ULT prescription. Finally, for each subject we calculated the visit rate on unique calendar dates with a medical diagnosis code over the observation time to estimate how often they visited their general practitioner. The visit rate was then categorised into tertiles.

Landmark analysis is routinely used to assess the impact of treatment where there is a potential lag between disease occurrence and initiation of therapy [28]. As the timing of initiation of ULT varies after gout diagnosis, we utilised landmark analysis to examine the effect of ULT on the risk of CKD. Landmark analysis deals with the issue of immortal time bias, which biases the results in favour of the treatment under study by granting a spurious survival advantage to the treated group [28]. In the case of gout, patients receiving ULT must have at least survived from time of diagnosis to time of treatment whereas no such requirement is necessary for the unexposed group (individuals with gout not receiving ULT). Bias would be introduced by ignoring this, as ULT exposure status may be dependent on the length of follow up. In landmark analysis, a fixed time after the initiation of therapy is selected a priori for conducting survival analysis [29]. Only those alive, event-free and contributing data at the landmark time were included in the analysis. Exposure to ULT was evaluated between the index date (diagnosis of gout) and the landmark time, whereas development of CKD stage ≥ 3 was only considered after the landmark time point. Two landmark points were considered in the analysis (1 and 3 years after diagnosis) based on a previously published study [30]. Only patients initiated on and prescribed more than 6 months of ULT were considered to be exposed (Fig. 1). This was based on previous literature [30] and expert consensus, as allopurinol is started at a low dose and increased gradually and it can take several months to escalate the dose sufficiently to lower serum urate to below the biochemical target level. The duration of ULT was calculated based on quantity prescribed and numeric daily dose.

Absolute rates (ARs) of CKD stage ≥ 3 per 10,000 person-years and 95% confidence intervals (CI) were calculated for the exposed and unexposed groups. These were stratified by age, gender, IMD and time after diagnosis. Hazard ratios (HRs) were modelled using Cox proportional hazards regression adjusting for the stated confounding factors. Those with missing body mass index (BMI) status were categorised separately and included in the analysis, as BMI was assumed not to be missing at random. Similarly, we compared the risk of CKD stage ≥ 3 among those prescribed ULT within 1 and 3 years after diagnosis to patients with gout who were not prescribed ULT. The HRs were additionally adjusted for baseline serum creatinine and uric acid levels. Baseline serum creatinine and uric acid level was considered before the ULT exposure or landmark date for those not prescribed ULT. For those with missing laboratory values, an indicator variable was included in the regression analysis. All missing values were imputed using a constant to ensure that all data were included in the analysis. This study was approved by the CPRD in-house Independent Scientific Advisory Committee (ISAC) reference number 15_214RA.

Sample size calculations: based on previous literature, we anticipated at least 30,000 cases of incident gout in HES-linked CPRD matched to a similar number of unexposed individuals [31]. Given the annual incidence of stage 3 CKD is 15% (aged 65–74 years) in the UK, our sample size provided more than 99% power to detect a HR of 1.5 using Cox proportional hazards model at 5% level of significance. For the landmark analysis, assuming that 10% of patients with gout are treated with ULT within the first year, we had approximately 82% power to detect a HR of 1.35 between ULT users and non-users, using a Cox proportional hazards model at 5% level of significance.