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


The results from this analysis indicate that the GMIM criteria are useful as a composite outcome measure to capture response to pegloticase in chronic refractory gout patients. The measures comprising the composite outcome measure included sU, tophus area, SJC, TJC, PGA, flares, and pain. The first five were based on a repeated measures mixed effects model with backward elimination[34] using data from the pegloticase RCTs; pain was added given the importance attributed to it by patients with gout[36,37] and the fact that it was a component of the original remission criteria.[33] Inclusion of this endpoint into the composite GMIM is consistent with results from studies that have evaluated the importance of different symptoms in patients with acute or chronic gout.[16,30] Results from interviews of 30 patients with gout (10 with and 20 without clinically apparent tophi) indicated that pain was identified as being the cardinal, defining symptom of gout, leading to a range of impacts on health-related quality of life, most notably physical functioning and sleep.[16] Flares are included in the recommendations for elements in composite measures for assessment of gout treatment,[28] the GAS composite,[31] and proposed remission criteria,[33] and this prompted their inclusion in the GMIM composite.

Results from the sensitivity analysis carried out in the present study showed that flares did not contribute significantly to the composite and that the contribution of pain was modest. This is not to suggest that flares are not important to patients, but just that they did not add information to the GMIM model. This could relate to the way information was collected in the RCTs. Flares were self-reported and pain was general and non-specifically related to gout. More precise collection of these variables could add important information to a multivariable outcome model.

Evaluation of results eliminating the requirement for sU < 6 mg/dL increased the percentages of patients achieving GMIM20, 50, and 70 among the q2w subjects without persistent urate lowering and also resulted in small numbers of placebo-treated patients achieving these goals. The observation that the percentages of subjects without persistent urate lowering achieving GMIM20, 50, and 70 at 6 months is consistent with previous analyses of results from these studies indicating significant clinical improvements in these patients despite failure to achieve sustained urate lowering.[38] It has been suggested that these improvements may be related to the transient but profound reductions in urate observed in these patients.[38] Results from the OLE showed that GMIM responses were sustained in the patients who initiated pegloticase treatment in the RCT and continued in the OLE. They also demonstrated that switching from placebo to q2w pegloticase treatment when progressing from the RCTs to the OLE resulted in clinical improvement comparable to that observed for those who initially received pegloticase and responded with persistent urate lowering in the RCTs and continued on this dose in the OLE. The ability of patients to "catch up" after delayed initiation of pegloticase should not be surprising since the patients randomized to placebo in the RCTs had gout symptoms for a mean of ~ 13 years at the time of study entry.[26] Importantly, patients switching from placebo to pegloticase were not included in the generation of GMIM and, therefore, serve to validate the approach.

As noted above, a composite endpoint for assessment of gout treatment was developed before the GMIM. The GAS was based on longitudinal analysis of results from a multicenter observational cohort study (Kick-Off of the Italian Network for Gout [KING]), in which 68.7% of a cohort of 406 patients were treated with allopurinol and 13.6% were treated with febuxostat.[31] A multistep process that began with all measures in the OMERACT core domain[20] and included factor analysis, linear discriminant analysis, and linear regression resulted in inclusion of four factors: gout flares in the past 12 months, sU, pain, and number of tophi.[31] All of these measures were included in GMIM and both TJC and SJC were also added into GMIM, which differed from the GAS in that it was focused on responsivity to pegloticase treatment. Recently, Chinchilla et al. confirmed the predictive validity of the GAS, its correlation with the GIS, and its sensitivity to change.[32] It is important to note that the disease severity for patients enrolled in the KING study which provided the basis for the GAS was very different from those in the RCTs for pegloticase.[26,31] The patients in the pegloticase studies had longer duration of disease (12–16 years across treatment groups vs 3.8 years), greater numbers of tender (11.1–14.1 vs 1) and swollen (8.9–13.2 vs 0) joints, tophi (65.2% to 76.7% vs 19.5%), flares (20–43 over 18 months before treatment vs 0 over 3 months), and sU (9.4–10.4 mg/dL vs 6.3 mg/dL). In addition, patients enrolled in the pegloticase RCTs were required to have gout that was refractory to allopurinol treatment or to be intolerant of this medication whereas this was not the case in the KING study. GMIM was, therefore, tested in a very severe group of subjects with advanced gout. Whether it would be effective in subjects with less advanced disease remains to be evaluated, but the overlapping components of GAS and GMIM suggest that it might. A parallel effort to identify a multivariable outcome measure for acute gout might be worthwhile. Moreover, a direct comparison of the performance of GAS and GMIM might be informative.

There is an important limitation with respect to the GMIM as it is currently formulated. There is no adjustment for baseline variables with scores so low that the patient cannot achieve a specified (20%, 50%, or 70%) improvement regardless of the efficacy of treatment. This same problem applies to the American College of Rheumatology criteria for assessment of treatments for rheumatoid arthritis[39,40] and has been used as a justification for composite measures that assess absolute disease activity vs change from baseline.[40] This limitation might be overcome by adjusting response criteria to include reduction of a given endpoint to 0 when baseline scores are sufficiently low. We addressed this issue by making the response criterion for a response improvement, 3 of the 5 GMIM parameters in addition to achieving sU < 6 mg/dL and a decrease in flares. Finally, there is a general concern about composite criteria and whether they might overestimate responses has been articulated[41,42] For this reason, GMIM was structured to have the 2 most frequently required components, decrease in sU < 6.0 mg/dL and decrease in flares to be mandatory components.