Development of a Multivariable Improvement Measure for Gout

Naomi Schlesinger; N. Lawrence Edwards; Anthony E. Yeo; Peter E. Lipsky


Arthritis Res Ther. 2020;22(164) 

In This Article


Design of Pegloticase Clinical Trials

The design of the two identical RCTs of pegloticase and their open-label extension (OLE) that provided the data analyzed in this study have been described in detail previously[26,35] and are only briefly summarized here. Both studies received institutional review board approval for each site, and written informed consent and Health Insurance Portability and Accountability Act assurances were completed for each participant prior to enrollment. These trials included adults with chronic refractory gout with one or more of the following: sU ≥ 8.0 mg/dL and ≥ 3 self-reported gout flares during the previous 18 months; ≥ 1 tophus; chronic gouty arthritis; and failure to respond to the maximum medically appropriate allopurinol dose, as determined by the treating physician, or a contraindication to this drug. Patients were randomized to 6 months of intravenous infusions of either pegloticase 8 mg every 2 weeks (q2w), every 4 weeks (q4w), or placebo.[26] The primary endpoint for the RCTs was reduction of sU (sU < 6.0 mg/dL) ≥ 80% of the time during month 3 (extending from the week-9 infusion to just before the week-13 infusion) and month 6 (extending from the week-21 infusion to the week-25 final study visit). Any patient not achieving this goal or who did not complete the trial was classified as a nonresponder. Patients from the RCTs could continue into the OLE in which they were given the choice of receiving pegloticase q2w or q4w.[35]

These trials had a large number of secondary endpoints which made results particularly useful for development of a single composite outcome measure, including tophus resolution; reduction in gout flares; decreases in TJC and SJC; PGA of disease activity; and patient-reported changes in pain, physical function (HAQ–Disability Index), and quality of life (36-Item Short Form Health Survey).[26] Patients were assessed at baseline and at the week 13 and 19 visits as well as at the week 25 final visit for secondary endpoints. In addition, patients were evaluated for up to 27 months of the OLE.

Construction of the Gout Multivariable Improvement Measure (GMIM)

A total of 85 subjects were treated with biweekly pegloticase. Overall, 42% (n = 36) of the patients who received pegloticase 8 mg every 2 weeks were responders (persistently lowered sU) to treatment in the RCTs. In previous work, responders were evaluated to determine whether they met combined criteria for remission defined by a previous Delphi exercise.[33] In addition, the clinical data from the responders was employed to develop criteria for a complete response. This was derived by employing a repeated measures mixed effects model with backward elimination that related clinical and laboratory changes observed throughout the trial.[34] In this exercise, numerous outcome measures were considered, but sU, PGA, numbers of tender and swollen joints, and the degree of tophus resolution controlling for repeated measures were found to contribute the most information to the complete response model.[34] Based upon the previous criteria for remission and complete response, the final set of criteria employed in GMIM were identified and included sU < 6 mg/dL, PGA scores, visual analog scale (VAS) pain levels, TJC and SJC, the number of flare episodes, and the degree of tophus area resolution. Improvement was defined as sU < 6 mg/dL and the absence of flares during the preceding 3 months plus 20%, 50%, or 70% improvement in ≥ 3 of the other 5 clinical criteria. Because improvement in sU and a decrease in flare frequency were uniformly used in clinical trials and were felt to be mandatory components of response, these were required along with a variety of clinical features that might vary in different patients. Since sU can vary with time, evaluation of clinical responses was also carried out with and without the requirement for achievement of sU < 6 mg/dL. Sensitivity analysis was done by removing one variable at a time from the analysis.

Statistical Analysis

All comparisons were carried out with Fisher's exact test to determine whether there was a nonrandom association between two variables. Chi-square with correction for continuity was used when the frequency of events was small. Statistical significance was defined as p < 0.05. Components that contributed the most information to the GMIM model were determined by multiple linear regression with backward elimination of the least statistically significant term to model the relationship between the dependent variable, the GMIM response, and various independent variables.

All calculations were carried out with SAS version 9.4 (Cary, NC).