Evaluating the Long-term Effect of Allopurinol Use in Gout Using Marginal Structural Models: A Primary Care Electronic Health Records Study

Background: Allopurinol is intended to be taken for life in the treatment of gout. In UK primary care, treatment is suboptimal with up to 30% of indicated patients being treated within five years of diagnosis and once prescribed allopurinol, treatment adherence is poor. Studies evaluating effect of allopurinol have ignored the time-varying propensity for treatment which may be due to changing patient health and risk of poor outcome over time.

Objectives: To estimate the effect of allopurinol on a range of outcomes in people with gout accounting for time-varying confounding by indication.

Methods: Cohort study of primary care medical records from the UK Clinical Practice Research Datalink (CPRD) included adults aged ≥18 years consulting for gout between 1997 and 2002, and not prescribed urate-lowering drugs two years prior. Allopurinol prescription ≥3 months and covariates (demographics, comorbidities, other medication usage, and number of years previously prescribed allopurinol) were measured yearly from the date of first gout consultation.

Marginal structural models estimated the effect of allopurinol on time to reaching target serum urate (SU) level ≤360μmol/L, mortality, hospitalisation due to gout, hip or knee joint replacement, and various comorbidities. Confounding was accounted for using inverse probability of treatment weights. Weights were derived in each year of follow up by estimating the propensity of initiating and continuing allopurinol given covariates via logistic regression models. Subsequently, weighted Cox regression models estimated the effect of allopurinol on outcome.

Results: 16,876 adults (mean age 62 years, 77% male) consulted for gout and were followed up for a median 10.7 years. 46% were prescribed allopurinol, of which 40% discontinued treatment and within this group 44% resumed treatment.

Allopurinol users were more likely to reach target SU level (hazard ratio 5.00 (95% CI 4.00, 6.23)) and had increased risk of hospitalisation due to gout (2.22 (1.91, 2.58)), coronary heart disease (1.11 (1.01, 1.23)) and renal disease (1.27 (1.13, 1.43)) compared to non-users. There was no increased risk of mortality (0.93 (0.83, 1.06)), joint replacement (0.93 (0.76, 1.13)), cerebrovascular disease (1.01 (0.82, 1.25)) and peripheral vascular disease (1.13 (0.87, 1.46)) for allopurinol users compared to non-users.

Conclusions: This is the first primary care population-based study to model changes in allopurinol use and patient health over time in the evaluation of the effect of allopurinol in gout. Although allopurinol use was associated with reduced SU levels below target, it increased the risk of gout hospitalisation and some comorbidities which may be due to inadequate allopurinol dosing or poor adherence to allopurinol.