Febuxostat in the management of gout: a cost-effectiveness analysis

Abstract

Objective:

To determine the cost-effectiveness of febuxostat vs allopurinol for the management of gout.

Methods:

A stochastic microsimulation cost-effectiveness model with a US private-payer perspective and 5-year time horizon was developed. Model flow based on guideline and real-world treatment paradigms incorporated gout flare, serum uric acid (sUA) testing, treatment titration, discontinuation, and adverse events, chronic kidney disease (CKD) incidence and progression, and type 2 diabetes mellitus (T2DM) incidence. Outcomes were estimated for the general gout population and for gout patients with CKD stages 3/4. Modeled treatment interventions were daily oral febuxostat 40–80 mg and allopurinol 100–300 mg. Baseline patient characteristics were taken from epidemiologic studies, efficacy data from randomized controlled trials, adverse event rates from package inserts, and costs from the literature, government sources, and expert opinion. Eight clinically-relevant incremental cost-effectiveness ratios were estimated: per patient reaching target sUA, per flare avoided, per CKD incidence, progression, stages 3/4 progression, and stage 5 progression avoided, per incident T2DM avoided, and per death avoided.

Results:

Five-year incremental cost-effectiveness ratios for the general gout population were $5377 per patient reaching target sUA, $1773 per flare avoided, $221,795 per incident CKD avoided, $29,063 per CKD progression avoided, $36,018 per progression to CKD 3/4 avoided, $71,426 per progression to CKD 5 avoided, $214,277 per incident T2DM avoided, and $217,971 per death avoided. In patients with CKD 3/4, febuxostat dominated allopurinol for all cost-effectiveness outcome measures.

Conclusions:

Febuxostat may be a cost-effective alternative to allopurinol, especially for patients with CKD stages 3 or 4.

Keywords::

• Rheumatology
• Gout
• Economics, pharmaceutical
• Cost-benefit analysis
• Febuxostat
• Allopurinol

Introduction

Gout is a condition in which serum uric acid (sUA) supersaturates in the bloodstream and begins to precipitate as monosodium urate crystal deposits in joints, tendons, and other soft tissue1. This sUA level (6.8 mg/dL or higher) can be asymptomatic, but the shedding of crystals into joint space may cause joint inflammation and severe pain known as gout flares or gouty arthritis2. Severe cases of gout may include the formation of tophi, chalky crystalline deposits that may break through the skin or form on internal organs.

The prevalence of gout in the US is estimated at 2.13–3.9% (3–8.3 million adults) and is increasing3. More than twice as many men as women suffer from the condition, although the gap decreases after menopause. The annual economic burden of gout is over $3000 per patient, with costs to patients in the US totaling tens of billions of dollars annually4. The prevalence of sex-specific hyperuricemia (sUA >7.0 mg/dl for men and >5.7 mg/dl for women) is more than 21% (43.3 million adults), split equally between men and women3.

sUA is primarily a product of cell degradation with ∼14% contributed by the metabolism of purines in the diet5. Under-excretion of sUA takes place in 90% of cases of hyperuricemia6, so behavior modification is frequently insufficient and medical urate-lowering therapies (ULTs) are needed7. These generally fall into two categories, xanthine oxidase inhibitors (XOIs) that reduce the production of sUA, or uricosuric agents that increase the clearance of sUA in the kidneys7.

Hyperuricemia is associated with a number of chronic diseases, including renal (chronic kidney) disease (CKD), cardiovascular disease, hypertension, and metabolic syndrome5. Observational studies have produced conflicting results about the causal role of hyperuricemia in CKD. One study indicated the need for randomized controlled trials (RCTs) to study the relationship between hyperuricemia and CKD onset and progression8. A recent systematic review and meta-analysis summarizing 13 studies of 190,718 participants found hyperuricemia to be an independent predictor of incident CKD (OR = 2.35, 95% CI = 1.59–3.46)9. Furthermore, evidence suggests that control of hyperuricemia impacts CKD. A meta-analysis of seven RCTs of ULTs showed a reduced risk of CKD progression (RR = 0.30, 95% CI = 0.19–0.46)10. An extension of one trial showed a reduced risk of CKD progression for patients continuing on ULTs for up to 7 years11. A decision was made to not include cardiovascular events associated with hyperuricemia into the model. There is evidence regarding a linkage between hyperuricemia and cardiovascular events5, but the quantification of this linkage for inclusion of this into the model was not clear to the authors. Since inclusion of hyperuricemia-related cardiovascular events would be beneficial to FEBUX, given its superior sUA control profile, exclusion is a conservative assumption.

Allopurinol (ALLO) has been the mainstay XOI since its introduction in 1965, comprising over 90% of ULT use in the US12. Despite early indications of efficacy13; however, a low proportion of patients maintain high levels of adherence14 and, even in the managed environment of randomized controlled trials, discontinuation is high15–17. Although dosing up to 800 mg/day is approved by the FDA18, dosing at more than 300 mg/day is uncommon19, at which fewer than half of ALLO patients achieve target sUA (<6.0 mg/dL)20–22.

This study assesses the cost-effectiveness of the use of febuxostat (FEBUX), an XOI with a record of more patients reaching sUA target15–17. The analyses focus on a direct comparison of FEBUX with ALLO, the most commonly prescribed first line ULT for the treatment of gout. We have strived to model common clinical practice rather than guideline recommendations. ALLO and FEBUX patients are followed over a 5-year base case time horizon, with modeled ULT discontinuation and re-initiation. To preserve the direct ALLO vs FEBUX first line comparison, no ULT switching is modeled, although switching does occur in clinical practice. A novel aspect of the model is its consideration of the incidence and progression of CKD associated with hyperuricemia.

Patients and methods

Patient population

In the base case, the patient population was selected to represent the US gout population3^,23–26 with baseline comorbidity prevalences of CKD and T2DM23^,27^,28. Additional analyses were performed for the US gout population over age 65, the US gout population diagnosed with CKD stages 3/4, and the US gout population over age 65 diagnosed with CKD stages 3/4. The CKD stages 3/4 populations were selected as being of interest due to FEBUX’s perceived suitability compared to ALLO for these patients12^,29. The over age 65 populations were selected as being of interest to US payers who cover these elderly patients.

Comparators

The first line uses of the daily oral ULTs ALLO and FEBUX were compared in the model. ALLO alone comprises over 90% of the ULT market in the US12. ALLO was introduced as a treatment in 1965, while FEBUX was introduced in 20091 2. Both ALLO and FEBUX are XOIs, which reduce the production of uric acid. However, ALLO is cleared exclusively by the kidneys, while FEBUX is cleared by both the kidneys and liver18^,30. This may reduce the need for FEBUX dose adjustment for patients with mild-to-moderate CKD12. ULT switching between ALLO and FEBUX was not modeled to focus on the impact of first line use of the respective ULTs.

While both treatments require titration, the initial dose of ALLO of 100 mg/day is unlikely to be sufficient22, and should be titrated by 100 mg/day until sUA target is reached18; for patients with CKD these values are halved18. In contrast, the initial dose of 40 mg/day of FEBUX is frequently efficacious21, and may be titrated once (to 80 mg/day) at 2 weeks30. A complicating factor with ALLO dosing is allopurinol hypersensitivity syndrome (AHS), which manifests as a severe rash that can result in Stevens Johnson Syndrome, toxic epidermal necrolysis and death7. AHS led to a renal clearance-based dosing algorithm31, which generally restricted ALLO dosing to levels ineffective for most patients29.

Model structure

A Java individual-based state transition cost-effectiveness model (CEM) with an Excel user interface was developed. The associated stochastic microsimulation structure allowed for the use of patient characteristic distributions and for the dynamic modeling of patient disease and care characteristics throughout simulation, as opposed to defining an untenable number of Markov states. A 1-month fixed-advance cycle was paired with a variable-length time horizon (5 years in the base case)32 and a year 2013 US payer perspective. Costs were discounted at a 3% annual rate per recommendation for US analyses33. The 5-year base case time horizon was selected to represent a sufficient period for adherent patients to realize the full flare reduction benefit of successful first line XOI treatment, i.e. no switching, as well as to reflect real-world XOI discontinuations and re-initiations.

Model flow within the CEM was based on treatment of gout with ULTs and assumed a mixture of guideline-based and real-world treatment paradigms. Patients entering the model were randomly assigned baseline values for a number of characteristics based on distributions in the US including age and gender, sUA band (≥8–< 9, 9–<10, or ≥10 mg/dL), CKD status (stage 0/1, 2, 3/4, or 5), type 2 diabetes mellitus (T2DM) status, and the presence of tophi. Characteristics of gout care included the presence or absence of recommended testing for sUA level and titration, willingness to titrate ALLO above 300 mg/day, and the prescription of prophylaxis to reduce flare incidence upon initiation of ULT. All patients entering the model had been diagnosed with gout and uncontrolled hyperuricemia, and initiated treatment on either ALLO or FEBUX, but were not necessarily treatment naive.

Figure 1 depicts the model flow, reflecting treatment and disease progression and treatment response stochastically governed by a set of probabilities. At each time advance, patient sUA level was estimated based on patient baseline characteristics and their ULT treatment and dosage. T2DM and CKD statuses were updated based on baseline characteristics and sUA level. Treatment for flare was provided with colchicine or non-steroidal anti-inflammatory drugs (NSAIDs). The combination of sUA testing and ULT titration was assigned as a patient care characteristic at baseline. If sUA testing was performed, treatment was continued without change, titrated, or terminated based on the results. If sUA testing was not performed, patients remained on the given ULT treatment and dosage. Treatment discontinuation could result from an adverse event or, reflecting real-world discontinuation rates from RCTs, could also take place per patient decision. ULT treatment was re-initiated after a flare if previously discontinued per patient decision. Treatment switching was not allowed in the modeled analyses. Patients could die at natural or excess mortality rates depending on age, sex, and the statuses of CKD, T2DM, and sUA.

Figure 1. Model flow. CKD, chronic kidney disease; N, no; sUA, serum uric acid; T2DM, type 2 diabetes mellitus; Tx, treatment; Y, yes.

As simulated patients followed the model flow, tallies of costs and events were recorded. These included the costs of ULT treatment, flares, T2DM comorbidity, CKD comorbidity, adverse events, and gout care. Events that were recorded included reaching sUA target, flares, CKD incidence, CKD progression, progression to stage 3/4 CKD, progression to stage 5 CKD, T2DM incidence, and deaths.

For the analyses, 10,000 patients were simulated in each treatment arm. Simulation results were stored for each patient, and averaged to produce the study results for each arm. The quantity of 10,000 simulated patients was sufficient to sample the heterogeneity of the patient characteristics and to produce stable results. Half-cycle corrections were not implemented due to the short 1-month fixed-advance model cycle.

Outcomes

The model estimated eight cost-effectiveness measures relevant to gout clinical outcomes. These included costs per patient reaching sUA target, per flare avoided, per incident case of CKD avoided, per CKD progression avoided, per progression to CKD stages 3/4 avoided, per progression to CKD stage 5 avoided, per incident case of T2DM avoided, and per death avoided. The corresponding clinical outcomes were also estimated. In addition to mean total cost per patient, the disaggregated mean costs for ULT, flares, T2DM comorbidity, CKD comorbidity, adverse events, and gout care were also estimated. Outcomes of particular interest to the authors were two cost-effectiveness outcomes: cost per patient reaching sUA target and cost per progression to CKD stage 5 avoided, and their clinical counterparts.

Input probabilities and distributions

The distributions of baseline characteristics of gout patients such as age, gender, T2DM prevalence, presence of tophi, and CKD status were derived from census and epidemiological sources. The distribution of sUA levels was sourced from a previously performed analysis of anonymized electronic medical records. Annual flare rates by sUA levels were sourced from the literature34. Background rates and sUA-related rates and relative risks of CKD and T2DM onset and CKD progression were derived from epidemiological studies35–38. Values and citations are presented in Table 1. Natural mortality was based on a 2009 US life table39. The probabilities of reaching target sUA (<6 mg/dL) were derived from recent RCTs15–17 (Table 1). The probabilities were derived by weighting the percentages of FEBUX and ALLO patients reaching target from the RCTs.

Table 1. Gout population, comorbidities, and treatment success probabilities.

Gout population	% of total	% female				Source(s)
Age (years)
<35	4.90%	24.53%				[3,23,24]
35–44	10.80%	11.51%
45–54	17.90%	12.93%
55–64	23.50%	16.66%
65–74	21.10%	28.15%
≥75	21.80%	35.53%
	≥8–< 9	9–<10	≥10
Initial sUA level (mg/dl)	42.43%	30.84%	26.73%			[25]
Annual flare rate (by sUA)	0.69	0.93	0.93			[34]
% with initial tophi	22.00%					[26]
Comorbidities	CKD 0/1	CKD2	CKD3/4	CKD 5	T2DM
% of gout population
Age (years)
<35	97.10%	2.90%	0.00%	0.00%	1.90%	[23,27,28]
35–44	53.75%	37.84%	8.41%	0.00%	5.97%
45–54	45.95%	47.55%	5.73%	0.77%	14.97%
55–64	32.95%	52.49%	14.32%	0.24%	22.63%
65–74	6.79%	64.69%	26.54%	1.98%	30.70%
≥75	0.49%	42.58%	55.93%	1.00%	24.60%
CKD (by age)	20–39	40–49	50–59	60–69	≥70
Annual CKD 0/1→2	6.00%	1.50%	3.00%	14.00%	17.50%	[27,28]
Monthly CKD 2→3/4	0.75% (all ages)					[35]
Monthly CKD 3/4→5	0.37% (all ages)					[35]
T2DM (by age)	20–44	45–64	≥65
Annual onset	0.42%	1.32%	1.28%			[36]
Relative risks by sUA	≥8–< 9	9–<10	≥10
CKD onset	1.43	1.43	1.43			[37]
CKD progression	1.98	1.98	1.98			[35]
T2DM onset	1.40	1.60	1.70			[38]
Treatment success
Probability sUA <6 mg/dl	≥8–< 9	9–<10	≥10
ALLO, ≤300 mg/day	0.54	0.4047	0.2678			[15–17]
FEBUX, 40 mg/day	0.6007	0.4711	0.2651			[15–17]
FEBUX, 80 mg/day	0.8228	0.7241	0.5575			[15–17]

Treatment success probabilities applied at initiation/re-initiation for a single 1-month model cycle.

ALLO, allopurinol; CKD, chronic kidney disease; FEBUX, febuxostat; sUA, serum uric acid; T2DM, type 2 diabetes mellitus.

Table 2 presents provider treatment pattern inputs used in the analysis. The percentage of providers who prescribe prophylaxis with ULT treatment was set to 52%44, with half using colchicine and half using NSAIDs55. The percentage who test sUA and potentially titrate ALLO (up to 300 mg/day) was set to 60.5%44. The percentage who test sUA and potentially titrate FEBUX from 40 mg/day to 80 mg/day was set to 80%, based on the assumption that there is a higher propensity to test and titrate the non-generic FEBUX therapy, which has a more efficacious higher dose available at the same price as the lower dose. An additional analysis assumption is that no patients will be titrated to ALLO dosing greater than 300 mg/day to approximate actual clinical practice as opposed to guideline recommendations. Other published FEBUX vs ALLO cost-effectiveness models made the same assumption regarding ALLO dosing greater than 300 mg/day32^,56^,57.

Table 2. Other probabilities, proportions, risk ratios, time periods, and costs.

Sequencing	Value	Source
Monthly elevated flare probability
ALLO	0.058	[17]
FEBUX	0.081	[17]
Months of elevated flare risk and duration of prophylaxis
ALLO	6	[40]
FEBUX	6	[40]
Months until zero flare risk for sUA <6 mg/dl	36	[26]
Annual probability of patient decision discontinuation
ALLO	0.289	[32]
FEBUX	0.122	[32]
Prophylaxis flare RR
Colchicine	0.2769	[15–17]
NSAID	0.2769	[15–17]
Flare RR due to tophi	1.46	[41]
Percentage of ALLO patients with rash	0.02	[20]
Percentage of ALLO rash patients with AHS	0.1925	[20,42,43]
Percentage of AHS fatal	0.21	[31]
Percentage of providers that
Test sUA and titrate ALLO	60.5%	[44]
Test sUA and titrate FEBUX	80%	Analysis assumption
Dose ALLO >300 mg	0%	Analysis assumption
Prescribe prophylaxis	52%	[44]
Mortality HR due to comorbidity
T2DM	1.0	[45]
CKD 1/2	1.0	[46]
CKD 3/4	1.8	[46]
CKD 5	5.9	[46]
Costs
ULT, $ per month
ALLO	$5	[47]
FEBUX	$155	[47]
Prophylaxis
Colchicine	$330	[48]
NSAID	$30	[48]
Gout care, $ per episode
Physician visit	$114.41	[49,50]
sUA lab test	$30	[51]
Gout flare
Physician visit	$240	[52]
AHS treatment	$10,412	[43]
Comorbidities, $ per year
CKD stages 3/4	$7105	[53]
CKD stage 5, age <65	$149,040	[54]
CKD stage 5, age ≥65	$84,384	[54]

AHS, allopurinol hypersensitivity syndrome; ALLO, allopurinol; CKD, chronic kidney disease; FEBUX, febuxostat; HR, hazard ratio; NSAIDs, non-steroidal anti-inflammatory drugs; RR, relative risk; sUA, serum uric acid; T2DM, type 2 diabetes mellitus.

Table 2 reports other non-cost values used in the model. These include treatment-related sUA mobilization elevated flare probabilities17^,40, duration and benefits of prophylaxis15–17^,40, ULT discontinuation rates32, time until zero flare risk with sUA <6 mg/dl26, elevated flare risks due to tophi41, ALLO-related AHS probabilities20^,31^,42^,43, and T2DM and CKD mortality hazard ratios45^,46. The presence of an ALLO-related rash was modeled as an adverse event that resulted in immediate treatment discontinuation. ULT re-initiation only occurs after a patient decision discontinuation and subsequent flare.

Input costs

All costs used in the model, along with source citations, are shown in Table 2. These values were taken from a number of sources50^,51^,53, included retrospective database analyses47^,52, an economic analysis of colchicine48, Medicare reimbursement rates49, and an annual report on the costs of end stage renal disease (ESRD), as represented by renal dialysis costs54. Source values have been inflated as necessary to 2013 equivalents using the medical care component of the US Consumer Price Index58. Pharmacy acquisition costs for ALLO and FEBUX are from a retrospective analysis of MarketScan Commercial and Medicare administrative claims databases for patients initiating urate-lowering therapies. The costs reflect the average US acquisition costs47. All costs represent commercial payer costs, with the exception of CKD stage 5 (ESRD) costs, which are differentiated for patients ages 65 years and older to reflect Medicare costs54. No mortality costs were included in the analysis. The only ULT-related adverse event was ALLO-related AHS43.

Assumptions

All pharmacoeconomic models are dependent upon assumptions. For the modeled ULTs, the following assumptions were implemented: no price difference between 40 mg and 80 mg FEBUX or between any doses of ALLO; all doses of ALLO are equally efficacious; if a serious ALLO adverse event occurs, it does so immediately upon treatment initiation; and patients that reach target on an elevated dose of ALLO remain at target if their dose is lowered due to CKD progression. For the impact of ULT: a treatment has failed if a patient has not reached sUA target at maximum dose in a single 1-month cycle32^,59; sUA remains at baseline if a patient continues on a failed ULT32; sUA returns to baseline in a subsequent 1-month model cycle post-successful ULT discontinuation32; and elevated flare risk due to sUA mobilization begins at treatment initiation. Regarding the administration of ULTs: one cycle is the minimum length of treatment; persistent patients are fully adherent; no downward ULT titration; if a discontinued patient re-initiates treatment, they re-start on the same ULT and dose; prophylaxis (if utilized) begins at treatment initiation and re-initiation; a patient will continue on unsuccessful ULT if sUA not tested (until patient decision discontinuation), and sUA testing is discontinued once a patient reaches target. The following model structure assumptions were applied: a 1-month time advance is sufficiently granular; a simplified model of diabetes and renal failure was sufficient; and tophi presence was a permanent patient characteristic.

Sensitivity analysis

Sensitivity analysis is carried out for two purposes: to identify the inputs that most influence the results of a model and to test alternative inputs when there is uncertainty in the base case values. Fifteen sensitivity analyses were carried out for this model. The base case, upper, and lower boundary values may be seen in Table 3.

Table 3. One-way sensitivity analysis inputs.

	One-way sensitivity analyses			Comments
	Base	Upper	Lower
Test-titrate ALLO/FEBUX	60.5–80%	100–100%	60.5–60.5%
Prophylaxis use	52%	100%	26%	100% use and 50% of base case
Model time horizon	60	120	24	Months
Untreated flare rate	Base	150%	50%	Base is sUA-dependent
CKD onset RR	1.43	2.15 (150%)	1	to CKD Stage 2
CKD progression RR	1.98	2.97 (150%)	1	to CKD Stage 3/4 & 5
FEBUX vs. ALLO increased benefit	Base	150%	50%	Increase/decrease FEBUX RR advantage over ALLO
ULT initiation flare probability	Base	150%	50%	Base is ULT-dependent
Flare-related ULT reinitiation rate	100%	100%	50%
Patient decision ULT discontinuation	Base	150%	50%	Base is ULT-dependent
Prophylaxis flare RR	27.69%	41.54%	13.85%	±50% of base
Annual cost discount rate	3%	5%	0%
FEBUX cost	$155	$203.96	$155	Upper cost is WAC
Flare cost	$240.40	$480.80	$120.20	200% to 50% of base
CKD costs	Base	125%	50%	Base costs are CKD stage-dependent

ALLO, allopurinol; CKD, chronic kidney disease; FEBUX, febuxostat; RR, relative risk; ULT, urate-lowering therapy;

WAC, Wholesale Average Cost.

The first two tests concerned aspects of care. In the base case, patients who were allocated FEBUX were more likely to be titrated than those who received ALLO. The sensitivity analysis tested titration for all patients and for the same percentage of patients, regardless of treatment. The second sensitivity analysis tested use of prophylaxis for all patients and for 50% as many as in the base case.

The third sensitivity analysis tested the model time horizon. In the base case the time horizon was 60 months (5 years), a fairly short time period for chronic conditions like CKD or T2DM to develop or progress. The input was tested at 120 months (10 years) and 24 months (2 years).

Several inputs were tested at 50% higher and lower than base case. Exceptions to this included the following: (1) annual discount rates were tested at recommended alternative rates for the US, (2) CKD onset and progression relative risks were tested at 50% greater and at no increased risk over background rates, (3) the cost of FEBUX was tested at Wholesale Average Cost (WAC) price, (4) flare costs were tested at 50% lower than the median cost and at the mean cost taken from a database analysis, and (5) CKD costs were tested at 25% higher and 50% lower than base case values.

Sub-group analysis

The population in the base case was representative of the total US gout population. Sub-group analyses were performed for the US gout population over age 65, the US gout population diagnosed with CKD stages 3/4, and the US gout population over age 65 diagnosed with CKD stages 3/4.

Ethics

No patient records/information was utilized in this study, thus there was no need to procure approval from an ethics committee or institutional review board. For the same reason, there was no need to anonymize or de-identify patient records/information prior to analysis. No clinical records were utilized in this study, thus there was no need to obtain written informed consent from participants. For the same reason, there was no need to anonymize or de-identify patient records/information prior to analysis.

Results

The cost-effectiveness results for the base case population and time horizon are shown in Table 4. The total cost per patient was $1264 higher for FEBUX than for ALLO. The incremental cost-effectiveness ratios (ICERs) for FEBUX over ALLO ranged from $1773 per flare avoided to $221,795 per incident CKD case avoided. The ICER for patient reaching target sUA was $5377. ICERs which appear most affirmative for use of FEBUX were those related to patients reaching sUA target, flares avoided, and CKD progression avoided, all reporting ICERs less than $75,000. ICERs greater than $200,000 were reported for CKD incidence, T2DM incidence, and deaths avoided.

Table 4. Cost-effectiveness results.

Cost-effectiveness outcome	Events per 1000 patients			ICER
	ALLO	FEBUX	Incremental
Cost effectiveness for US gout population
Total cost per patient	$34,126	$35,390	$1264
Patients reaching target sUA	433	668	235	$5377
CKD incident cases	85	79	−6	$221,795
CKD progression	323	279	−44	$29,063
Progression to CKD 3/4	263	228	−35	$36,018
Progression to CKD 5	104	87	−18	$71,426
T2DM incident cases	57	51	−6	$214,277
Flares	3687	2974	−713	$1773
Deaths	177	171	−6	$217,971
Cost effectiveness for US gout population with CKD 3/4
Total cost per patient	$96,906	$93,509	−$3397
Patients reaching target sUA	433	668	235	Dominant
CKD incident cases	0	0	0	NA
CKD progression	282	243	−39	Dominant
Progression to CKD 3/4	0	0	0	NA
Progression to CKD 5	282	243	−39	Dominant
T2DM incident cases	56	50	−6	Dominant
Flares	3598	2912	−686	Dominant
Deaths	229	222	−7	Dominant

ALLO, allopurinol; CKD, chronic kidney disease; FEBUX, febuxostat; ICER, incremental cost-effectiveness ratio; sUA, serum uric acid; T2DM, type 2 diabetes mellitus.

Sensitivity analysis

Sensitivity analysis was performed for 15 inputs on each of two outcomes: cost per patient reaching target sUA and cost per flare avoided. For cost per patient reaching target sUA, the most sensitive input is the relative risk of CKD progression, driven by the high cost of renal dialysis. The second-most sensitive input is the time horizon, but in this case it is due to the slow progression of CKD associated with uncontrolled sUA level. For both outcomes the sensitivity of inputs are roughly in the same order. The tornado diagrams presented in Figure 2 illustrate the variability of the ICER associated with the cost per patient reaching target sUA outcome.

Figure 2. Sensitivity analysis tornado diagram of cost per patient achieving target sUA ICER. The analysis is of a gout population including all ages and CKD stati over a 5-year time horizon. Baseline ICER = $1264. ALLO, allopurinol; CKD, chronic kidney disease; Disc., discontinuation; FEBUX, febuxostat; ICER, incremental cost-effectiveness ratio; Prob., probability; RR, relative risk; Reinit., re-initiation; sUA, serum uric acid; ULT, urate-lowering therapy.

For cost per flare avoided, the time horizon was by far the most sensitive input. This is consistent with the medical knowledge of the treatment and condition. Upon initiation of treatment, FEBUX rapidly lowers the sUA level, which often increases the frequency of flares. In the longer term, however, success in meeting sUA target can translate into a very low likelihood of having a flare. In this case it ultimately drives the ICER for FEBUX to cost-savings. The second-most sensitive input was the relative risk of CKD progression. This is driven by the cost of renal dialysis associated with end stage renal disease. The results are considerably less sensitive to the rest of the analyzed inputs. These results are presented in the Online Supplementary Appendix.

Sub-group analysis

Table 4 presents cost-effectiveness estimates for the US gout population with stage 3/4 CKD. FEBUX is dominant (less costly and more effective) vs ALLO for all cost-effectiveness outcomes estimated. Analyses were also performed for the US gout population under 65 years of age and the under 65 population with stage 3/4 CKD. The results can be seen in the Online Supplementary Appendix. The analysis for the population under 65 is much the same as the base case analysis. One significant difference is that the ICER for death avoided rises to $662,243. The analysis for the under-65 stage 3/4 CKD population is very similar to the all ages stage 3/4 CKD population analysis, with FEBUX dominating ALLO on all outcomes. Patients with CKD stage 3/4 advance more quickly as a group to the costly CKD stage 5 than do the all gout patients cohort. Utilization of the more efficacious FEBUX treatment rather than ALLO slows progress to CKD stage 5, resulting in large cost savings.

Discussion

Our cost-effectiveness analysis of FEBUX in gout employed a model of diagnosed gout patients from a US payer perspective over a 5-year time horizon. The 5-year horizon reflects the first line FEBUX vs ALLO comparison, with no ULT switching allowed. A total of eight cost-effectiveness outcomes were evaluated in the base case. The analyses found FEBUX vs ALLO to have cost per event ICERs less than $10,000 for patients reaching target sUA and flares avoided. CKD progression ICERs were higher, but less than $75,000 in all cases. ICERs for CKD incidence, T2DM incidence, and death events avoided were substantially higher. FEBUX was dominant (less expensive and more effective) compared to ALLO for the six cost-effectiveness outcomes analyzed in the gout population with stages 3/4 CKD (CKD incidence and CKD 3/4 progression outcomes not applicable). Given the high CKD prevalence in patients with gout, and the high potential costs associated with ESRD as a result of CKD progression, an argument can be made for the use of FEBUX in all gout patients, and especially those with pre-existing CKD.

We believe that our US-focused model comparing first-line use of FEBUX and ALLO, with its focus on real-world treatment patterns and on medical events, is a valuable addition to the literature. Other cost-effectiveness models have been published comparing FEBUX to ALLO. One model is a European-focused cost-effectiveness model with a cost per QALY ICER perspective32. This publication partially informed our work. Additional published US-focused studies only presented cost-effectiveness results as 5-year costs per treatment success, defined as sUA level <6 mg/dl56^,57. One study did allow for dosing of ALLO greater than 300 mg/day using a simple efficacy algorithm. It implemented a cost per QALY ICER perspective, with no differentiation on impacts of improved sUA control aside from that on QALYs, flare rates, and average costs60.

Our model is one of the few cost-effectiveness models that have been created for gout and the first which incorporated CKD in its evaluation. The reason for this omission in earlier models has likely been the state of science concerning hyperuricemia and CKD, which may yet be unsettled. The co-morbidity of gout and CKD is frequent and well established3^,61. However, Feig8, describing observational studies of the matter, reported over one-third of the studies found sUA with no causal relationship. RCTs that have investigated the issue have been positive, but small62. Those investigating the matter using FEBUX have been observational retrospective analyses of FEBUX arms of RCTs63.

As seen in the sub-group analyses, the acceptance of hyperuricemia as a causal factor for CKD incidence and progression places a premium on reaching target sUA, especially for patients with stage 3/4 CKD. The higher probability for patients taking FEBUX to reach target sUA in modern RCTs makes a strong argument for its use, especially in the CKD stage 3/4 population. Excess mortality associated with CKD stages 3/4 and stage 5 is modeled and, thus, the CKD-related mortality is reduced with FEBUX vs ALLO.

Based on the sensitivity analyses, it can be argued that a longer time horizon would better allow the differences in the ULTs to fully manifest themselves. A longer time horizon would benefit FEBUX with its superior sUA control by allowing more time for flare frequency to dissipate and for CKD to progress further in the less-controlled ALLO arm.

Limitations

Our model has a number of limitations. First, efficacy results for ALLO from modern RCTs are available only for 300 mg/day dosing. Studies have been performed on the efficacy of ALLO dosing greater than 300 mg/day64–66, but these were small studies and not FEBUX or ALLO registration studies. Additionally, a gout management treatment pattern study indicated that only 2.9% of ALLO patients received doing greater than 300 mg/day19. We have, therefore, assumed equal efficacy for all doses of ALLO.

The analysis did not include cost per quality-adjusted life year (QALY) or per life year (LY) gained outcomes. The US focus of the analysis, where use of QALYs is not common, in conjunction with the low, short-term impact of flares (the most common manifestation of gout) on quality-of-life, factored into this decision. Instead we concentrated on clinical outcomes to which physicians and patients can easily relate.

The use of a 1-month fixed advance in the model may be a limitation, particularly as guidelines recommend titration of ALLO and FEBUX on weekly and biweekly frequencies, respectively. Also, in light of the importance of CKD to the analysis, a more sophisticated sub-model of CKD may have been desirable.

The costs of ALLO and FEBUX were based on a pharmacy claims study which reported costs 24% lower than WAC prices47^,67. Despite the lower apparent adherence, no adjustments were made to treatment efficacies.

Conclusions

From an overall cost perspective, use of FEBUX in the treatment of gout can be justified given its higher efficacy relative to ALLO 300 mg/day in achieving sUA target levels of <6.0 mg/dl, the under-utilization of higher ALLO daily dosing, and inconsistent rates of sUA testing and titration upon ULT initiation. Compared to patients treated with ALLO, over a 5-year time horizon more patients achieve sUA target of <6 mg/dl and fewer patients experience flares and CKD progression. Smaller, yet positive, impacts of FEBUX vs ALLO were shown in reduced CKD incidence, T2DM incidence, and deaths. Use of FEBUX as a treatment option for patients with gout should be considered by the clinician. This appears particularly to be the case in the gout population with stage 3/4 CKD. Further studies should include larger RCTs that definitively demonstrate the effectiveness of ULTs in preventing CKD progression.

Transparency

Declaration of funding

Takeda Pharmaceuticals International, Inc., the maker of febuxostat, funded the research that informed the cost-effectiveness model on which this manuscript is based.

Declaration of financial/other relationships

LJS is an employee and equity owner of Medical Decision Modeling Inc., which provides for-fee research services for Takeda Pharmaceuticals International, Inc. JCG is an employee of Medical Decision Modeling Inc., which provides for-fee research services for Takeda Pharmaceuticals International, Inc. GM is a former employee of Takeda Pharmaceuticals USA, Inc. AS is an employee of Takeda Pharmaceuticals International, Inc. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

The authors would like to thank Ronald C. Wielage of Medical Decision Modeling Inc. for his work on this study. His assistance was funded with support from Takeda Pharmaceuticals International, Inc., the maker of febuxostat. His contributions included data identification, input on model development, and proofreading of the manuscript.