Anakinra for the Treatment of Acute Gout Flares

A Randomized, Double-Blind, Placebo-Controlled, Active-Comparator, Non-Inferiority Trial

Carly A. Janssen; Martijn A. H. Oude Voshaar; Harald E. Vonkeman; Tim L. Th. A. Jansen; Matthijs Janssen; Marc R. Kok; Bea Radovits; Caroline van Durme; Hetty Baan; Mart A. F. J. van de Laar


Rheumatology. 2019;58(8):1344-1352. 

In This Article


Study Design and Patient Population

This randomized, double-blind, double-dummy, active comparator placebo-controlled trial, was conducted at the rheumatology departments of seven hospitals in the Netherlands, between 2016 and 2018. The study was approved by an Ethical Review Board (METC Twente, Enschede, the Netherlands), the institutional review board or ethics committee of each participating centre, and performed in accordance with the principles outlined in the Declaration of Helsinki and Dutch legislation.

Eligible subjects were recruited and screened by the attending rheumatologist and an instructed specialized rheumatology nurse at the outpatient clinics. All adult (≥18 years) patients, with a diagnosis of an acute flare of gouty arthritis, confirmed by microscopic identification of intracellular MSU crystals in the primary joint were eligible for participation. The primary joint was defined as the joint that was most affected by acute gouty arthritis, according to the rheumatologist. Patients with current use of urate lowering therapy (ULT), as well as those experiencing only no to mild gout-related pain, were not included. Other non-inclusion criteria included concurrent use of other IL-1 inhibitors; known history of allergy or sensitivity to latex; absolute contraindication for all available types of ULT; absolute contraindication to anakinra (e.g. neutropenia and severe renal impairment defined as a creatinine clearance rate < 30 ml/min); absolute contraindication for all three conventional treatment options; presence of liver disease that according to the treating rheumatologist precluded participation in the study; an active or recurrent bacterial, fungal or viral infection; use of tumour necrosis factor inhibitors; pregnancy or lactation; women who planned on becoming pregnant during the study period, and insufficient command of the Dutch language. Following screening, all patients were given a maximum of 24 h to decide on participation in the study and to provide written informed consent prior to randomization and study initiation at baseline.

Randomization, Treatment and Dosing

The attending rheumatologist, together with the patient, decided on the treatment as usual (colchicine, naproxen or prednisone) that would suit the patient best. Subsequently, patients were randomly allocated (1: 1) to either a five-day treatment with anakinra (subcutaneous injection 100 mg once daily) plus oral placebo up to three times daily (colchicine), two times daily (naproxen) or one time daily (prednisone); or to a treatment with oral standard of care in line with the assigned standard treatment by the caregiver (0.5 mg up to three times daily for colchicine; 500 mg up to twice daily for naproxen; 35 mg once daily for 5 days for prednisone) plus subcutaneous injection placebo once daily for 5 days. For treatment as usual, the dosages and duration were in line with national acute gout treatment guidelines.[4] The placebo injections and placebo pills were identical in appearance to the anakinra injections and treatment as usual pills, respectively. Considering the short duration of anakinra treatment given to patients, no pre-study screening for latent tuberculosis was done. The randomization allocation sequence list was generated using a computer randomization application, based on atmospheric noise. Patients, caregivers, local pharmacies, and trial investigators had no knowledge of the allocation sequence during the entire course of the study. Study medication was stored at the hospital pharmacy and released in sequential order to patients. Patients received instructions on the use of study medication from a blinded study nurse, and the first dosages of study medication, both oral pills and injection, were taken by patients under supervision of a blinded study nurse during the baseline visit. During the first 7 days of the study, no prescription-based, rescue medication was available to patients, but the use of over-the-counter pain-relieving agents (NSAIDs and aspirin) was allowed. Patients also initiated ULT at baseline with allopurinol, febuxostat or benzbromarone, at the discretion of the treating rheumatologist.

Study Assessments

For this study, the OMERACT recommendations for acute gout were followed.[30] Starting at baseline (day 1), patients were asked to fill in a gout flare diary for seven consecutive days, wherein levels of pain (1 = none; 2 = mild; 3 = moderate; 4 = a lot; 5 = extreme), tenderness (1 = none; 2 = slightly; 3 = fairly; 4 = very; 5 = extremely) and swelling (1 = none; 2 = somewhat; 3 = fairly; 4 = very; 5 = extremely) of the primary joint were recorded using five-point rating scales. Patients were also asked to report their level of pain in the primary joint (0 = absolutely no pain, 10 = unbearable pain) and their global assessment of overall wellbeing (0 = very bad, 10 = very good) on a 10-point numeric rating scale (NRS), and their level of treatment response on an eight-point rating scale (1 = completely disappeared; 2 = very much improved; 3 = much improved; 4 = somewhat improved; 5 = unchanged; 6 = slightly worse; 7 = much worse; 8 = very much worse). Finally, experienced side-effects and the intake of any other painkillers and/or anti-inflammatory medication were to be reported in the flare diary daily.

Physical examination, medical history and gout status were assessed at baseline, and patients returned for a clinic visit at day 7. Laboratory measurements (serum uric acid, CRP) were performed at baseline and at day 7. Any (serious) adverse events ((S)AE)) were reported and evaluated at day 7.

Study Endpoints

The primary end point, i.e. ΔPain, was defined as the mean change in patient-reported pain in the most affected joint, from baseline to the average of pain scores at days 2–4 on the five-point rating scale.[31,32] We reasoned a priori that anakinra could be a useful treatment alternative for gout patients with contraindications to the standard treatment options, even when not proven more effective than treatment as usual, if it could be demonstrated that anakinra has an effect greater than the effect of imputed placebo;[33] and that any potential inferiority of anakinra compared with treatment as usual would not be clinically meaningful. Therefore, a NI design was adopted with a NI margin of 0.4 points on the five-point rating scale. A difference of 0.4 points in favour of treatment as usual is slightly more stringent compared with previous NI studies in gout known to us, that have used 0.5 as a NI margin for ΔPain.[31,34,35] It is also more stringent than other studies assessing change in pain as a primary outcome, for which 10% of the scale (e.g. 10 mm on a 100-mm visual analogue scale (VAS)) is frequently used as an upper limit of acceptable difference.[36,37] Moreover, in the only available study that allowed for a placebo effect to be estimated for ΔPain, the difference between groups in ΔPain was found to be at least 0.4 points in favour of NSAID compared with placebo (i.e. 95% CI was 0.4–1.0).[32] Therefore, assuming a placebo effect of similar magnitude, the effect of anakinra in the current trial can be considered greater than imputed placebo when the upper bound of the 95% CI for the ΔPain (anakinra-treatment as usual) does not exceed 0.4 points in favour of treatment as usual.[32] A priori power calculation showed that, if there would be no difference in ΔPain between the groups, 87 patients per treatment arm would need to be included to have an 80% chance of demonstrating NI with a chance of a type 1 error of 5% or less.

Secondary outcomes included the improvement of primary joint pain (NRS scores), tenderness, swelling, treatment response and patient global assessment (PGA) of wellbeing across days 1–5. The number of patients achieving ≥ 50% decrease in NRS pain scores following baseline on days 2–5 were also compared between treatment arms. Other outcomes included the number and type of AE that occurred during the first 7 days of the study, as well as the decrease in CRP levels after 7 days. Finally, use of concomitant pain-relieving agents, both prescription-based and over the counter medication, during days 1–7 were determined.

Statistical Analysis

The primary study end point was assessed using an analysis of covariance (ANCOVA) model, with treatment received as a fixed effect, and baseline pain scores measured on the five-point rating scale as a covariate. Subsequently, we obtained a 95% CI for the baseline pain adjusted marginal mean difference in ΔPain between the treatment groups. We tested whether the upper bound of the CI of the numeric difference in changed pain scores between anakinra and treatment as usual would not exceed the NI margin of 0.4 in favour of treatment as usual (H0: ΔPainAnakinra–ΔPainTreatmentAsUsual > 0.4 vs Ha: ΔPainAnakinra–ΔPainTreatmentAsUsual ≤ 0.4). If this criterion was met in both the per-protocol (PP) and intention-to-treat (ITT) populations, NI of anakinra compared with treatment as usual would be concluded. The ITT analysis was performed on all patients who were randomized to a treatment at baseline and received at least one dose of study medication. The PP population contained patients who had no missing data for ≥ 1 of the assessments needed for the primary outcome, and who did not take any interfering concomitant pain-relieving medications during days 1–4 of the study. This included any prescription-based pain-relieving medication (e.g. opioid formulations, intake of colchicine while the patient was prescribed naproxen), but over the counter pain-relieving agents were allowed.

Analysis of secondary endpoints were performed in the ITT population, using linear mixed effects models, with time, treatment arm, and the interaction between time and treatment arm as fixed effects. For each outcome analysed, the covariance matrix was chosen that had the best fit according to the Bayesian information criterion. All statistical tests were performed at the Bonferroni corrected 0.05 level. The difference between treatment groups in achieving ≥ 50% decrease in NRS pain scores in the days following baseline was assessed using binary logistic regression analysis.

For the analyses of the ITT population, plausible values for missing observations of the seven-day flare diary and CRP values were generated using multiple imputation by chained equations. Percentages of missing data were determined for variables needed for the primary efficacy analyses. Data from the seven-day flare diary was used as input for the imputation models, as well as treatment received, age, gender and log-transformed CRP values. Following the imputation process, the latter were transformed back to normal CRP values. In total, 20 datasets were generated, and the observed and imputed distributions were visually compared for similarity. Reporting guidelines for handling analysis affected by missing data were followed.[38]

Characteristics of the study population were summarized using means and standard deviations, or median and first and third quartiles for continuous outcomes. Categorical variables were summarized using frequency counts and percentages, and compared using Pearson's χ2 statistics, as appropriate, for secondary outcomes. Other secondary outcomes were compared using t-tests or a non-parametric equivalent, as appropriate. P-values <0.05 were considered statistically significant. All analyses were done using IBM statistics SPSS version 22. The Reporting of NI and Equivalence Randomized Trials extension of the CONSORT 2010 statement was followed in reporting this study.[39]