Comparative Risk of Gout Flares When Initiating or Escalating Various Urate‐Lowering Therapy: A Systematic Review With Network Meta‐Analysis

We systematically examined comparative gout flare risk after initiation or escalation of different urate‐lowering therapies (ULTs), comparative flare risk with and without concomitant flare prophylaxis, adverse event rates associated with flare prophylaxis, and optimal duration of flare prophylaxis.


INTRODUCTION
Gout is the most common form of inflammatory arthritis, with a prevalence ranging from approximately 0.1% to 10% worldwide. 1It is caused by the formation and subsequent deposition of monosodium urate crystals in the joints and soft tissues of susceptible individuals, which form in the presence of elevated serum urate (SU). 2 Through a complex cascade, the interaction between monosodium urate crystals and the immune system results in an acute inflammatory response, known as a "gout flare."Although self-limiting, flares cause immense pain, have considerable impact on quality of life, and can reoccur frequently. 3eliminary findings of this research have been presented as an oral presentation at the Gout, Hyperuricemia and Crystal-Associated Disease Network (G-CAN) 2022 8 th Annual Research Symposium, the Australasian Pharmaceutical Science Association -Australasian Society of Clinical and Experimental Pharmacologists and Toxicologists (APSA-ASCEPT) 2022 Joint Conference, and the Australian Rheumatology Association (ARA) 2023 Annual Scientific Meeting.
Supported by an Australian Government Research Training Program Scholarship.Drs Reeve and Hopkins' work was supported by the National Health and Medical Research Council of Australia (Investigator grants GNT1195460 and GNT2008119, respectively). 1  High or prolonged monosodium urate crystal burden can also result in tophi, chronic gouty arthritis, gouty bone erosion, and permanent disability. 2 The cornerstone of long-term gout management is uratelowering therapy (ULT), which reduces SU concentrations, thus facilitating crystal dissolution. 2However, initiation of ULT may precipitate a transient increase in the number of flares. 2 Although the exact mechanism remains poorly understood, this increase in flares is thought to be caused by a rapid decrease in SU concentrations resulting in crystal mobilization and an enhanced inflammatory response. 4If poorly managed, adherence to ULT may be compromised, with people with gout potentially attributing these flares to ineffectiveness of ULT. 5 Because of the relatively unpredictable nature of flares when introducing ULT, several treatment guidelines recommend that "gout flare prophylaxis" with lowdose colchicine, nonsteroidal anti-inflammatory drugs (NSAIDs), or corticosteroids be considered in all patients starting ULT to minimize flare risk. 6,7 systematic review published in 2014 found no direct evidence to support the prophylactic use of low-dose NSAIDs or corticosteroids when initiating ULT, 8 but a subsequent review from 2016 reported indirect evidence to support NSAID use.9 This review references a post hoc analysis of three large randomized controlled trials (RCTs) of the ULT medications febuxostat and allopurinol.7,9 There is little consensus on the optimal duration of prophylaxis, with guideline recommendations including at least 8 weeks, 10 3 to 6 months, 6 or 6 months from commencing ULT, 11,12 or until the target SU concentration has been attained.13 In recent years, gradual ULT dose-escalation strategies without prophylaxis have demonstrated comparable flare rates to standard-dose ULT with prophylaxis and may therefore mitigate the need for prophylaxis.14 Avoidance of unnecessary therapy is important given the complexity and multimorbidity often encountered in the gout population and the well-known toxicities and potential drug-disease and drug-drug interactions of the available prophylactic medications.15 A systematic examination of the comparative flare rates following the initiation or escalation of ULT, with and without prophylaxis, at various ULT and/or prophylaxis starting doses and dosing schedules may enable more informed decision-making, particularly for those who may be intolerant or have contraindications to prophylactic therapy or for individuals whose only prophylactic option is corticosteroids.We therefore aimed to systematically examine the (1) comparative flare risk after initiation or escalation of different ULTs, (2) comparative flare risk with and without concomitant flare prophylaxis, (3) adverse event (AE) rates associated with flare prophylaxis, and (4) optimal duration of flare prophylaxis.

MATERIALS AND METHODS
Study design.This study follows the Preferred Reporting Items for Systematic Review and Meta-Analyses statement 16 and the Preferred Reporting Items for Systematic Review and Meta-Analyses for network meta-analysis (NMA) extension statement 17 (Supplementary Tables 1 and 2).The protocol was preregistered in PROSPERO (CRD42020178479).
Search strategy and selection criteria.The Ovid Medline, Embase, Web of Science, and Cochrane Library and Cochrane Central Register of Controlled Trials databases were searched from inception to November 10, 2021.Searches were supplemented with manual searches for published and unpublished studies using reference lists of key articles and international trial registers (the Australian New Zealand Clinical Trials Register, ClinicalTrials.gov,and the European Union Clinical Trials Register).Searches included terms relating to gout, ULT, and flare prophylaxis (including specific ULT and prophylaxis drug names and classes) (Supplementary Exhibit 1).Duplicate records were removed with EndNote version 20 (Clarivate Analytics).Teams of paired investigators (DM, JMT, MT, and NA) independently used Covidence (Veritas Health Innovation) to screen titles and abstracts, review full-text studies, and extract data on trial design, characteristics of the trial population, intervention, and outcomes (Supplementary Methods).Discrepancies were resolved by discussion or through consensus with a senior reviewer (ER).Study authors were contacted when necessary to determine study eligibility or to supplement incomplete, missing, or unclear data.
Eligibility criteria.Articles were included if they were RCTs or prospective controlled (nonrandomized) clinical trials, included adults (aged ≥18 years) with confirmed gout either initiating ULT or escalating subtherapeutic ULT, and reported on the frequency of (1) flares or (2) flare prophylaxis-related AEs.Trials with or without concurrent administration of prophylaxis (colchicine, NSAIDs, corticosteroids, or inteleukin-1 inhibitors [canakinumab or rilonacept]) were included.To best inform clinical practice, we included only approved ULTs (ie, allopurinol, febuxostat, probenecid, benzbromarone, sulfinpyrazone, rasburicase, pegloticase, topiroxostat, and lesinurad or verinurad [with allopurinol or febuxostat]).Articles were excluded if they investigated flare treatment (ie, intentionally recruited participants currently experiencing a flare) rather than prophylaxis, did not report outcomes

SIGNIFICANCE & INNOVATIONS
• The risk of gout flares when introducing uratelowering therapy varies approximately three-fold depending on the treatment strategy used.• Less intensive urate-lowering therapy and the use of flare prophylaxis is associated with fewer gout flares.• The optimal duration and efficacy of nonsteroidal anti-inflammatory drugs and corticosteroids as flare prophylaxis is unclear.
separately for the different ULT arms, were not published in English, or were conference abstracts.A full list of the inclusion/ exclusion criteria and protocol amendments are available in the Supplementary Methods.
Outcomes and statistical analysis.The primary outcomes were flare rates and prophylaxis-related AE frequencies.The secondary outcomes were flare severity, pain scores, flare duration, and health-related quality of life.For statistical analysis, flare rates were measured as the proportion of participants who had one or more flares using data over the longest available time period from time zero (see Supplementary Table 3 for clarifications).Prophylaxis AEs were measured by the total number of participants who withdrew because of AEs, had any AE (total AEs), had serious AEs, or had treatment-related AEs.
We performed a NMA, which combined both direct and indirect sources of evidence across a network of studies to enable the comparison of three or more interventions within a single analysis.To ensure network validity, studies included in the network must, on average, be similar in all key characteristics that may influence the relative effects. 18Separate NMA models were used to evaluate the (1) flare risk following initiation or escalation of different ULT drugs and/or doses (referred to as the "ULT-based flare network"), (2) flare risk following ULT initiation or escalation with different prophylaxis drugs and/or doses ("prophylaxisbased flare network"), and (3) risk of AEs with different prophylaxis drugs and/or doses (separate analyses were conducted for each category of AE, ie, the "prophylaxis-based AE networks").
To be eligible for the NMA of prophylaxis AEs, trials had to report AEs separately for each prophylactic drug and/or dose compared to a different prophylactic drug and/or dose and/or to no prophylaxis/placebo with the same or similar ULT use between groups.Trials that failed to report flares or prophylaxis-related AEs within the aforementioned format were considered part of the review and included narratively.
Using inverse-variance random effects NMA models, risk ratios (RRs) with 95% confidence intervals (CIs) of flares and AEs were estimated for each pairwise comparison.When a trial had an arm with zero events, it was included via 0.5 event continuity correction. 19Trials reporting zero events in all arms were excluded from the respective NMA.Heterogeneity in random effect distributions were assessed via t 2 and I 2 statistics. 20Equal heterogeneity was assumed across all comparisons, and correlations induced by multi-arm trials were considered.Inconsistencies between direct and indirect comparisons were evaluated both statistically and visually, including via the design-by-treatment test and backcalculation method for pairwise comparisons. 21The transitivity assumption for indirect comparisons was evaluated via inspection of the distribution of key characteristics (ie, potential effect modifiers) across trials. 18Network diagrams were used to graphically depict the structure of network interventions.NMA estimates (based on direct and indirect evidence) were presented via forest plots and league tables displaying the relative effects between interventions with associated uncertainty.RR < 1 corresponded to beneficial effects for the first intervention relative to the comparator.Direct evidence-only estimates were also presented within league tables.P-scores (range 0 to 1) were used to rank interventions by taking the precision of frequentist point estimates into account. 22Interventions with larger P-score were considered more beneficial.[Correction was added on 27 March 2024, after first online publication: 'P values' has been corrected to "P-score" in this version.]Comparison-adjusted funnel plots and the Egger's test was used to assess publication bias.For outcomes in which NMA was not possible, findings were presented descriptively.Analyses were performed using the "netmeta" package (2.1-0) in R (version 4.1.2).
Risk-of-bias assessment and data availability statement.Included studies were critically appraised by two independent reviewers (DM and JMT) using the revised Cochrane risk-of-bias tool for randomized trials, 23 with discrepancies resolved by consensus.Risk of bias was assessed separately for the co-primary outcomes of flares and AEs related to prophylaxis.The statistical analysis plan and dataset for this systematic review and NMA can be available from the corresponding author on reasonable request.No individual level data are included in this manuscript.All data are aggregated data from clinical trials.
Risk-of-bias assessment.Quality assessment indicated that the overall risk of bias was low for 4 (14%) of 28 and 2 (18%) of 11 eligible publications presenting flare and prophylaxis-related AE data, respectively (Supplementary Figures 1-4).This was driven by 61% and 45% of publications being determined to have high risk or some concerns toward bias because of missing outcome data, and 64% and 73% of publications were determined as having a high risk or some concerns toward bias in the selection of the reported results (flare-and prophylaxis-related AE data, respectively).
NMA and ULT-based flare network.Figure 2 shows the network of eligible comparison for the co-primary outcomes (see Supplementary Table 8, Supplementary Exhibits 5-9, and Supplementary Figures 5-25 for full NMA results).Thirteen trials involving 7,816 participants were included within two disconnected ULT-based flare networks [25][26][27][28]33,34,36,41,42,44,46,48,51 (Figure 2A), all of which were ULT initiation trials. Flare results are summariz per trial in Supplementary Table 3, pairwise comparisons are presented in Figure 3, and forest plots of direct comparisons are presented in Figure 4A and B. For subnetwork 1, compared to placebo plus any prophylaxis, the RR (95% CI) of flares ranged from 1.08 (0.87-1.33) for febuxostat 40 mg plus any prophylaxis to 2.65 (1.58-4.45)for febuxostat 80 mg plus lesinurad 400 mg plus any prophylaxis (Figure 4A).Flare rates for febuxostat 40 mg plus any prophylaxis, allopurinol approximately 300 mg plus any   (A and B) and flare prophylaxis-related adverse events (C).Each intervention is represented by a colored circle.The size of the circles is proportional to the number of randomly assigned participants within each network.Randomized comparisons between the interventions are shown by the lines and numbers between the circles.The width of the lines is proportional to the number of trials within each network or subnetwork.In (A) the ULT-based flare network, nodes represent different ULT drug, dose (either fixed or up-titrated), and flare prophylaxis combinations (prophylactic options were aggregated).In (B) the prophylaxis-based flare network and the (C) prophylaxis-based adverse event networks, nodes represent different prophylactic drug and/or dose combinations (ie, different ULT options were aggregated).Because of variability across trials, all colchicine doses were aggregated.Fourteen trials were ineligible for inclusion in (A) the ULT-based flare network because 8 did not report data in a suitable format for analysis, 29,30,35,40,43,45,47,50 4 had no common comparator arm to connect to the network, 24,32,37,39 1 did not report the ULT dose or dosing schedule, 49 and 1 was excluded because of violations of the transitivity assumption.38 Twenty-two trials were ineligible for inclusion in (B) the prophylaxis-based flare network because 13 did not report flares separately for each prophylactic drug used in the trial, [24][25][26][27][28]30,[33][34][35][36][37][38]40 6 had no comparator group, 29,32,39,[41][42][43] and 3 did not report data in a suitable format for analysis.45,47,50 Seventeen trials were ineligible for inclusion in (C) the prophylaxis-based adverse event networks because 9 did not report on flare prophylaxis-related adverse events, [28][29][30]33,34,37,38,41,42 5 reported limited data that were not suitable for analysis, 24,27,35,40,45 and 3 did not use flare prophylaxis.32,39,43 ULT, urate-lowering therapy.Color figure can be viewed in the online issue, which is available at http:// onlinelibrary.wiley.com/doi/10.1002/acr.25309/abstract.

GOUT FLARES WHEN INTRODUCING OR ESCALATING URATE-LOWERING THERAPY
prophylaxis and febuxostat up-titration to 40 mg were not statistically different to placebo plus any prophylaxis (Figure 4A).For subnetwork 2, compared to allopurinol uptitration without any prophylaxis, the RR (95% CI) of flares was 0.27 (0.12-0.58) for febuxostat up-titration plus any prophylaxis, 0.40 (0.31-0.51) for allopurinol up-titration plus any prophylaxis, and 1.43 (0.06-33.88) for benzbromarone uptitration plus any prophylaxis (Figure 4B).Prophylaxis-based flare network.Five trials involving 758 participants were included in the prophylaxis-based flare network (Figure 2B), all of which were RCTs of prophylaxis. 44,46,48,49,51All but one study 51 included a placebo arm.A forest plot of direct comparisons and a league table of pairwise comparisons are presented in Figures 5 and 6, respectively.ULT plus prophylaxis had a lower RR of flares compared to ULT without any prophylaxis, with the RR (95% CI) ranging from 0.35 (0.25-0.50) for rilonacept 160 mg to 0.43 (0.31-0.60) for rilonacept 80 mg and 0.50 (0.35-0.72) for colchicine (Figure 5).There was no significant difference in the RR of flares between the different prophylactic treatments.
Prophylaxis-based AE networks.Seven trials 44,[46][47][48][49][50][51] and one post hoc analysis 7 of three trials 25,26,36 were included within the prophylaxis-based networks (Figure 2C).AE results are summarized per trial in Supplementary Table 5, and pairwise comparisons are presented in Figure 6.Compared to ULT without prophylaxis, the respective NMA did not identify any statistically significant differences in withdrawals caused by AEs, total AEs, or serious AEs with prophylaxis, and no statistically significant difference was observed between the different prophylactic drugs.ULT with rilonacept 160 mg (RR 2.99 [95% CI 1.40-6.38])or 80 mg (RR 2.29 [95% CI 1.19-4.43])but not colchicine (RR 1.69 [95% CI 0.61-4.69])was associated with a higher risk of treatment-related AEs compared to ULT without prophylaxis.There was no statistically significant difference in treatmentrelated AEs between these prophylactic options.
Duration of flare prophylaxis.Four trials reported flare rates during and immediately after cessation of prophylaxis. 24,26,48,51Schumacher et al 48 reported an increase in flares in the four weeks after rilonacept was ceased at week 16; however, this was comparable with placebo cessation.Yamanaka et al 51 randomized participants to three arms, one of which included fixed-dose febuxostat (40 mg) with colchicine for 12 weeks followed by fixed-dose febuxostat alone.Although the statistical significance was not reported, this arm reported a greater mean number of flares per participant in weeks 13 to 24 compared with the first 12 weeks in which colchicine was administered (1.53 vs 1.33 flares, respectively).Although the Febuxostat versus Allopurinol Controlled Trial (FACT) 26 reported an increased rate of flares in all arms in the 4 weeks after cessation of prophylaxis at week eight, no significant increase in flares was observed in a lesinurad escalation trial in which prophylaxis was ceased at month five. 24iminishing flare rates over a 12-month study period were reported in the other lesinurad trials 27,35 in which prophylaxis was also ceased at month five.Two trials 25,41 that continued prophylaxis throughout reported relatively high flare rates during the first 8 weeks.Karimzadeh et al 45 was the only trial to investigate the optimal duration of prophylaxis, with participants randomized to either 3 to 6, 7 to 9, or 10 to 12 months of colchicine while initiating allopurinol.Flares were lowest with 10 to 12 months of prophylaxis, although the authors concluded that 7 to 9 months was as efficacious but more cost effective.This trial was at high risk of bias across multiple domains.
Other results.There were insufficient data to perform the prespecified subgroup analyses of flare risk categorized by the definition of a flare, participants who achieved the SU target of <0.36 mmol/L, participants with tophaceous gout, and ULT treatment-naive versus treatment-established participants.Secondary outcomes and detailed sub-group analyses information are presented in Supplementary Tables 6 and 7, and Supplementary Exhibits 3 and 4.
Large heterogeneity was only detected in the prophylaxisbased total treatment-related AEs NMA; however, this was not statistically significant (I 2 = 61.8%[95% CI 0.0%-87.2%]).There was no clear evidence of global or local inconsistency (Supplementary Exhibits 7 and 8).Treatment rankings based on the P-score for each NMA are presented in Supplementary Exhibit 9. [Correction was added on 27 March 2024, after first online publication: 'P values' has been corrected to "P-score" in this version.]We found no evidence of publication bias (Supplementary Figures 19-25).

DISCUSSION
This systematic review represents the most comprehensive data synthesis regarding the relative risk of gout flares when initiating or escalating ULT.It is the first NMA investigating the relative risk of gout flares with different ULT drugs, ULT dosing schedules and prophylaxis, and AEs with prophylaxis.Fourteen trials were included in the NMA of flares with 26 direct and 31 indirect comparisons, and the results of 10 trials were included in the NMA of AEs related to prophylaxis with 17 direct and 11 indirect comparisons.The data were too limited to make definite conclusions about optimal ULT drug and/or dosing strategies, but higher fixed doses of febuxostat and febuxostat in combination with lesinurad were generally associated with greater flare risk.Prophylaxis with either colchicine or rilonacept significantly reduced the incidence of flares, and the safety of prophylaxis was generally comparable to placebo or no prophylaxis.There was insufficient evidence regarding gradual up-titration strategies, NSAIDs and corticosteroids as prophylaxis, and the optimal duration of prophylaxis.
A dose-response relationship has been established between ULT and reductions in SU concentrations. 52,53However, high ULT starting doses have greater flare risks.In clinical practice, ULT dosing has shifted away from commencing at fixed standard/high-dose regimens (eg, allopurinol 300 mg daily) toward approaches that start low and up-titrate gradually until the target SU is reached. 6,11,12This shift in dosing was implemented to reduce the risk of the rare but potentially lifethreatening allopurinol hypersensitivity syndrome. 54It was then subsequently observed that this dosing strategy may reduce the risk of gout flares at ULT initiation. 54Only one gradual versus fixed ULT dosing trial was included within the NMAs. 51Although this trial demonstrated fewer flares with a step-wise dose increase, it should not be inferred that all gradual up-titrations cause fewer flares, as febuxostat was up-titrated to a low dose (40 mg daily) rather than to a specific SU target as is generally recommended. 12Because they could not be connected to the network, we were unable to conclude whether flare risks differ for individuals re-initiating ULT or escalating subtherapeutic ULT compared to ULT-naive individuals.With a high proportion of the gout population either underdosed 55 or poorly adherent to ULT, 56 a greater understanding of flare risk during treatment escalation or reinitiation will inform prophylaxis recommendations in these populations.
In concordance with previous systematic reviews, 8,9,15,57 we found that colchicine and rilonacept (an interleukin-1 inhibitor) were beneficial in reducing flares when initiating ULT.Additionally, our study was the first to directly compare colchicine and rilonacept and identify that they have similar effects with regards to flare suppression.Incomplete reporting of flares stratified by the prophylactic therapy received impeded the ability to assess NSAIDs and corticosteroids.Despite these therapies being recommended in guidelines, we highlight a significant gap in evidence that could be filled either by RCTs or a re-analysis of individual patient data from previous RCTs of ULT that included nonrandomized NSAIDs and/or corticosteroids.Furthermore, there is emerging evidence of an association between a recent gout flare and increased risk of myocardial infarction and stroke, 58 making it essential to determine the most effective and safe regimen to prevent gout flares.
We found no increase in the risk of total or serious AEs or withdrawals because of AEs with the addition of prophylaxis or between different prophylactic drugs, but there were significant limitations in the available data.Very few studies analyzed AEs associated with prophylaxis separately from ULT AEs, and most were post hoc analyses, which may introduce bias.Furthermore, methods to capture and report AEs were generally inconsistent, 59 and many trials implemented safety measures that may not occur in routine practice, such as co-prescribing proton pump inhibitors.People with gout commonly experience comorbidities 55 and are therefore at a greater risk of AEs from prophylactic medications, but people with significant comorbidities are generally excluded from clinical trials. 60Avoiding unnecessary or prolonged prophylactic drug exposure is important given the risks of prophylaxis drugs combined with the additional medication costs, monitoring, and pill burden.
The optimal duration of prophylaxis remains contentious and guidelines provide conflicting recommendations. 61This is unsurprising given that only one study in our review, which was judged to have a high risk of bias, investigated different durations of prophylactic treatment. 45Indirect inferences from other trials were again limited by trial reporting, wherein few trials reported flare rates for both the period during and immediately after prophylactic cessation.3][64][65][66] Additionally, established risk factors for flares, such as the presence of tophi, require quantifiable measurement by dual-energy computed tomography 67 or ultrasonography 68 to accurately assess the risk, and access to these in practice may be limited.
Our literature search was comprehensive and included unpublished data.Although we cannot rule out the possibility that some unpublished data were not identified, our comparisonadjusted funnel plots were not suggestive of small study effects or publication bias.In comparison to previous systematic reviews, 8,15 we allowed for the inclusion of RCTs with nonrandomized prophylaxis.Another strength was our NMA design, which facilitated analysis according to ULT and/or prophylaxis by drug and/or dose.
Our study had some limitations.The risk of bias for both primary outcomes was assessed as high for most publications.For flare data, this was primarily the result of unbalanced withdrawals caused by flares.This is unlikely to affect our results because those who withdrew because of flares would, for most studies, still be captured within our primary flare outcome.The data and comparisons within our NMA were also heavily limited by trial reporting. 69Some trials did not report data in a suitable format for analysis, whereas others reported data over selected timeframes.Many of our networks had single-standing nodes that were connected to only one other node, which increased the number of network estimates based solely on indirect evidence.There was also a disproportionality high number of trials with fixed ULT dosing compared to the newer, gradual up-titrated dosing.Lastly, because of our strict inclusion/exclusion criteria, not all available therapies were included within our review, primarily because of inapplicable study designs or trials with no suitable comparator.Additionally, some drugs included have since been discontinued (eg, lesinurad) or are not licensed for use in gout (eg, rilonacept).Nevertheless, the range of medications in our review reflects contemporary therapies for gout. 70n conclusion, high fixed doses of ULT were associated with a greater risk of flares.However, because most treatment guidelines have shifted to a gradual ULT up-titration approach, coupled with the reductions in flare risk noted with prophylaxis, there should be a greater emphasis on the need for future studies to compare gradual up-titration approaches against the prophylactic therapies currently used in clinical practice (ie, colchicine, NSAIDs, and/or corticosteroids) and to define the optimal duration of prophylaxis.Emphasis should be placed on ensuring more consistent and comprehensive trial reporting of flares as well as prespecification and thorough reporting of prophylaxis-related AEs in future clinical trials.

Figure 1 .
Figure 1.Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram.*Considered a flare prophylaxis-randomized trial rather than a ULT-randomized trial.ANZCTR, Australian New Zealand Clinical Trials Register; CENTRAL, Cochrane Library and Cochrane Central Register of Controlled Trials; ULT, urate-lowering therapy.

Figure 2 .
Figure2.Network meta-analysis diagram for gout flares (A and B) and flare prophylaxis-related adverse events (C).Each intervention is represented by a colored circle.The size of the circles is proportional to the number of randomly assigned participants within each network.Randomized comparisons between the interventions are shown by the lines and numbers between the circles.The width of the lines is proportional to the number of trials within each network or subnetwork.In (A) the ULT-based flare network, nodes represent different ULT drug, dose (either fixed or up-titrated), and flare prophylaxis combinations (prophylactic options were aggregated).In (B) the prophylaxis-based flare network and the (C) prophylaxis-based adverse event networks, nodes represent different prophylactic drug and/or dose combinations (ie, different ULT options were aggregated).Because of variability across trials, all colchicine doses were aggregated.Fourteen trials were ineligible for inclusion in (A) the ULT-based flare network because 8 did not report data in a suitable format for analysis,29,30,35,40,43,45,47,50 4 had no common comparator arm to connect to the network,24,32,37,39 1 did not report the ULT dose or dosing schedule,49 and 1 was excluded because of violations of the transitivity assumption.38 Twenty-two trials were ineligible for inclusion in (B) the prophylaxis-based flare network because 13 did not report flares separately for each prophylactic drug used in the trial,[24][25][26][27][28]30,[33][34][35][36][37][38]40 6 had no comparator group,29,32,39,[41][42][43] and 3 did not report data in a suitable format for analysis.45,47,50 Seventeen tials were ineligible for inclusion in (C) the prophylaxis-based adverse event networks because 9 did not report on flare prophylaxis-related adverse events,[28][29][30]33,34,37,38,41,42 5 reported limited data that were not suitable for analysis,24,27,35,40,45 and 3 did not use flare prophylaxis.32,39,43ULT, urate-lowering therapy. Color figure cn be viewed in the online issue, which is available at http:// onlinelibrary.wiley.com/doi/10.1002/acr.25309/abstract.

Figure 3 .
Figure 3. League table: ULT-based flare network.Network meta-analysis results are presented in the lower left diagonal and direct evidence-only estimates in the upper right diagonal.The risk ratios, with 95% confidence intervals, represent the relative difference in the proportions for one or more gout flares.A risk ratio <1 favors the column-defining intervention.Interventions from ULT-based flare subnetwork 1 are shown in blue, and interventions from ULT-based flare subnetwork 2 are shown in gray.The results with significant difference are bold.A, allopurinol; A + P, allopurinol 300 mg + prophylaxis; B, benzbromarone; F, febuxostat; F up-titration, febuxostat 10 mg up-titrated to 40 mg; L, lesinurad; NA, not applicable; P, prophylaxis.Color figure can be viewed in the online issue, which is available at http://onlinelibrary.wiley.com/doi/10.1002/acr.25309/abstract.

Figure 6 .
Figure 6.League table: prophylaxis-based flare network and prophylaxis-based adverse-event networks.Network meta-analysis results are presented in the lower-left diagonal and direct evidence only estimates in the upper-right diagonal.The risk ratios, with 95% confidence intervals represent the relative difference in the proportions for (1) one or more gout flares (shown in green), (2) withdrawals because of AEs (shown in white), (3) total AEs (shown in light blue), (4) serious AEs (shown in medium blue), and (5) total treatment-related AEs (shown in dark blue).A risk ratio below 1 favors the column-defining intervention.The results with significant difference are bold.Colchicine doses varied between 0.5 mg and 0.6 mg once or twice daily.Naproxen was dosed at 250 mg twice a day.AE, adverse event, NA, not applicable; ULT, urate-lowering therapy.Color figure can be viewed in the online issue, which is available at http://onlinelibrary.wiley.com/doi/10.1002/acr.25309/abstract.