Identification of the Most Cost-effective Position of Vedolizumab Among the Available Biologic Drugs for the Treatment of Ulcerative Colitis

Abstract

Background and Aims

There are limited data on the most cost-effective sequencing of biologics for ulcerative colitis [UC].

Methods

We used Markov modelling to identify the most cost-effective position for vedolizumab among biologics for steroid-dependent UC, with a base-case of a 35-year-old male. We assessed three treatment algorithms, with vedolizumab use: prior to an initial anti-tumour necrosis factor alpha [anti-TNFα] and azathioprine [Algorithm 1]; prior to a second anti-TNF and azathioprine [Algorithm 2]; and prior to colectomy [Algorithm 3]. The initial anti-TNF could be either infliximab or adalimumab. Transition probabilities, costs, and quality-adjusted life-year estimates were derived from published estimates, Medicare, and the Nationwide Inpatient Sample. Primary analyses included 100 trials of 100 000 individuals over 1 year, with a willingness-to-pay threshold of US$100,000. Multiple sensitivity analyses were conducted to assess our findings.

Results

From a population perspective, when both infliximab and adalimumab are available, vedolizumab was preferred as the first biologic if ≥14% of initial anti-TNF use was adalimumab. If infliximab is the primary biologic, vedolizumab use after infliximab [Algorithm 2] and prior to adalimumab was the most cost-effective strategy. All models were sensitive to biologic pricing.

Conclusions

This simulation demonstrated that the most cost-effective strategy in UC depends on the proportion of patients using adalimumab as the initial anti-TNF. If adalimumab was ≥14%, vedolizumab was preferred as the first biologic. When only infliximab was available for first-line therapy, the most cost-effective position of vedolizumab was prior to cycling to adalimumab.

1. Introduction

Monoclonal antibodies targeting specific components of the inflammatory response have revolutionised medical therapy for many disorders, including Crohn’s disease and ulcerative colitis [UC]. The first biologic that was Food and Drugs Administration [FDA]-approved for UC was infliximab in 2005, which is a monoclonal antibody directed against tumour necrosis factor alpha [i.e., the anti-TNFs].¹ Two additional medications within this class were subsequently approved for UC: adalimumab, in 2012, and golimumab, in 2015.^2,3

However, a significant proportion of patients exhibit primary non-response to anti-TNF monotherapy or combination therapy with azathioprine. Of those who do respond, many will lose response by the end of 1 year.^1,4 Vedolizumab is the first FDA-approved biologic agent for UC that uses a gut-selective mechanism of action. Through binding of the alpha-4 beta-7 integrin on the surface of activated lymphocytes, vedolizumab inhibits trafficking of these cells into areas of active inflammation within the alimentary tract.⁵ Clinical trials have demonstrated vedolizumab to be more effective relative to adalimumab in UC,⁶ and potentially safer when compared with the greater systemic immunosuppressive effect of anti-TNFs.^7–9

The clinical advances afforded by the biologic era come with tangible costs. Through the rapid incorporation of biologics into routine clinical practice worldwide, biologic-related expenditures have considerably increased, becoming a significant driver of medical costs.¹⁰ Therefore, their judicious and appropriate use is of particular import to both maximise the potential benefit of these therapies while also minimising unnecessary societal expenditures.

There are limited data to identify the most cost-effective positioning of biologics for patients with moderate to severe UC. Existing studies have some notable limitations. Previous studies have compared vedolizumab with infliximab or adalimumab, while not considering that from a population standpoint, the first anti-TNF used may be heterogeneous. Recent analyses suggest that both infliximab and adalimumab have approximately equal access as first anti-TNF in UC.¹¹ The ratio of infliximab to adalimumab use may be particularly important, given differences in costs and efficacy for these therapies.

Additionally, previous models have compared vedolizumab head-to-head with other agents.¹² Such analyses in UC have not routinely accounted for the likelihood that individuals often require treatment sequencing through multiple agents over time. Similarly, these analyses often assume that use of a biologic of interest is de novo; these studies therefore do not fully account for well-recognised changes in therapeutic efficacy in the setting of previously failed biologic therapies.⁹ Further, some previous studies have focused on assessing only clinical efficacy of medical therapies, without accounting for the costs associated with these medications.¹³

In this study, we sought to identify the most cost-effective position of vedolizumab among the available biologic drugs for the treatment of UC, via simulation modelling. We employed a novel model structure allowing us to assess the impact of the choice of first anti-TNF medication, biologic sequencing, and the position of vedolizumab within multiple treatment algorithms.

2. Materials and Methods

We constructed a Markov simulation model to assess the most cost-effective position for vedolizumab among the currently available biologics used to treat UC. We employed a previously published simulation of a standard treatment algorithm for a 35-year old male with steroid-dependent moderately to severely active UC who was initiating outpatient steroid-sparing therapy.¹³ This top-down treatment algorithm included: [1] combination therapy with an initial anti-TNF and azathioprine; [2] cycling to a second anti-TNF with azathioprine; and [3] colectomy [Figure 1]. In order to estimate the most cost-effective position of vedolizumab with mixed anti-TNF use, in the primary analysis the initial anti-TNF could be either infliximab or adalimumab, and we assumed that 50% of each cohort started with each of these medications.¹¹ We then compared three separate treatment algorithms, with vedolizumab use: Algorithm 1 [A1] first line, prior to combination therapy with an initial anti-TNF; [A2] second line, prior to transitioning to combination therapy with a second anti-TNF; and [A3] third line, prior to colectomy after failing both anti-TNF therapies. We assumed that colectomy was the least preferred treatment option. For all analyses, the baseline time horizon was 1 year and cycle length was 3 months.

Figure 1.

The base model consists of three different positions for vedolizumab [VDZ] in the treatment algorithm for ulcerative colitis, labelled here as A1 through A3. First-line anti-TNF use could consist of either infliximab or adalimumab, with the second anti-TNF consisting of the one not chosen initially. Both were assumed to be in combination with a thiopurine in the primary analyses. AZA, azathioprine; IFX, infliximab; SAE, significant adverse event; VDZ, vedolizumab; TNF, tumour necrosis factor.

For each medical treatment option, there were several potential outcomes [Figure 1]. During induction with a medical therapy, individuals could achieve clinical remission, clinical response, or not respond. They were also exposed to risks of serious infection and other serious adverse events [SAEs] associated with medical therapy. Individuals who responded or entered remission remained on the same medication, and could then maintain remission or response, lose clinical response, experience a serious infection requiring a pause in therapy, develop a lymphoproliferative complication modelled after non-Hodgkin’s lymphoma [NHL], or develop an SAE requiring cessation of therapy. Simulated individuals who did not respond to therapy, lost response to therapy, or were exposed to an SAE were then transitioned to the next medication in their specific algorithm. All individuals were exposed to age- and sex-specific all-cause mortality in all transition states of the model, as defined by baseline death rates per US census data.¹⁴

If simulated individuals required colectomy, they were exposed to a 1-month decreased quality of life, as well as age-appropriate and peri-operative risks of mortality. The peri-operative course could also be complicated, with further disutility, or uncomplicated.¹⁵ Those recovering from surgery were assumed to have a quality of life consistent with life with a J-pouch.¹⁶

2.1. Model inputs: transition probabilities, quality-adjusted life-years, and costs

Transition probabilities for medication response, remission, and adverse event rates were derived from relevant prospective controlled trials, when available [Table 1]. Systematic reviews and observational studies were used when no controlled trial data existed. Specifically, the SUCCESS-UC study was employed for induction estimates for response and adverse events for azathioprine monotherapy and combination therapy with infliximab, as it includes the most modern estimates for these medications.¹⁷ The ACT trials were used for maintenance therapy rates for infliximab and azathioprine,^1,18 and a meta-analysis was utilized for estimates related to maintenance therapy with azathioprine monotherapy.¹⁹ Estimates related to adalimumab were derived from the ULTRA-2 trial and stratified by previous anti-TNF exposure when applicable.² Vedolizumab estimates were derived from the GEMINI studies, and stratified by previous anti-TNF exposure.⁹ Relevant aspects of all data sources, including studies, medication databases, and QALY estimates, are described in Supplementary Table 6, available as Supplementary data at ECCO-JCC online.¹³

Table 1.

Cost, transition probability, and QALY estimates.

Description	Base case definition	Source
Transition probabilities
Infliximab [IFX] + azathioprine [AZA] induction therapy
Probability of remission during IFX + AZA induction	0.3974	17
Probability of clinical response during IFX + AZA induction	0.3718	17
Probability of SAE during IFX + AZA induction	0.0900	17
Probability of serious infection during IFX + AZA induction	0.00	17
IFX + AZA maintenance therapy
Probability of maintaining remission with IFX + AZA up to 30 weeks	0.6977	1,18*
Probability of maintaining clinical response with IFX + AZA up to 30 weeks	0.7534	1,18*
Probability of maintaining remission with IFX + AZA after 30 weeks	0.8167	1,18*
Probability of maintaining clinical response with IFX + AZA after 30 weeks	0.7818	1,18*
Probability of SAE requiring discontinuation of therapy with maintenance therapy with IFX + AZA over 1 year	0	1,18
Probability of serious infection while on maintenance therapy with IFX + AZA over 1 year	0.072	1,18
Adalimumab [ADA] + AZA induction therapy
Probability of remission during ADA induction	0.0918	2
Probability of clinical response during ADA induction	0.2755	2
Probability of SAE during ADA induction	0.0895	2
Probability of serious infection during ADA induction	0.0156	2
ADA + AZA maintenance therapy
Probability of maintaining remission with ADA	0.5122	2
Probability of maintaining clinical response with ADA	0.304	2
Probability of SAE requiring discontinuation of therapy with maintenance therapy with ADA over 1 year	0.0895	2
Probability of serious infection while on maintenance therapy with ADA over 1 year	0.0156	2
Vedolizumab [VDZ] induction therapy
Probability of remission during VDZ induction when used before anti-TNF	0.3968	9§
Probability of clinical response during VDZ induction when used before anti-TNF	0.246	9§
Probability of SAE during VDZ induction when used before anti-TNF	0	9§
Probability of serious infection during VDZ induction before an anti-TNF	0.0027	9§
Probability of remission during VDZ induction when used after an anti-TNF	0.1413	9§
Probability of clinical response during VDZ induction when used after anti-TNF	0.2934	9§
Probability of SAE during VDZ induction when used after anti-TNF	0	9§
Probability of serious infection during VDZ induction after anti-TNF	0.0054	9§
Vedolizumab maintenance therapy
Probability of maintaining remission with VDZ at 1 year, before anti-TNF	0.7101	9§
Probability of maintaining clinical response with VDZ use before anti-TNF	0.5469	9§
Probability of SAE requiring discontinuation of therapy with maintenance therapy with VDZ over 1 year, before anti-TNF	0	9§
Probability of serious infection while on maintenance therapy with VDZ over 1 year, before anti-TNF	0.0202	9§
Probability of maintaining clinical remission with VDZ over 1 year, after anti-TNF	0.6585	9§
Probability of maintaining clinical response with VDZ over 1 year, after anti-TNF	0.5469	9§
Probability of SAE requiring discontinuation of therapy with maintenance therapy with VDZ over 1 year, after anti-TNF	0	9,§
Probability of serious infection while on maintenance therapy with VDZ over 1 year, after anti-TNF	0.0202	9§
Transition probabilities used in monotherapy sensitivity analyses
Immunomodulator [IM]induction therapy
Probability of remission during IM induction	0.2368	17
Probability of clinical response during IM induction	0.26316	17
Probability of SAE during IM induction	0.0886	17
Probability of serious infection during IM induction	0.0127	17
Immunomodulator maintenance therapy
Probability of flare while in remission over 1 year with IM	0.4435	19
Probability of flare with clinical response over 1 year with IM	0.5856	*
Probability of SAE requiring discontinuation of therapy with maintenance therapy with IM over 1 year	0.0792	19
Probability of serious infection while on maintenance therapy with IM over 1 year	0.0127	17
Infliximab monotherapy induction therapy
Probability of remission during IFX monotherapy induction	0.2208	17
Probability of clinical response during IFX monotherapy induction	0.4675	17
Probability of SAE during IFX monotherapy induction	0.0641	17
Probability of serious infection during IFX monotherapy induction	0.0128	17
Infliximab monotherapy maintenance therapy
Probability of maintaining remission with IFX monotherapy up to 30 weeks	0.6977	1,18
Probability of maintaining clinical response with IFX monotherapy up to 30 weeks	0.7534	1,18
Probability of maintaining clinical response with IFX monotherapy after 30 weeks	0.7818	1,18
Probability of maintaining remission with IFX monotherapy after 30 weeks	0.8167	1,18
Probability of SAE requiring discontinuation of therapy with maintenance therapy with IFX monotherapy over 1 year	0.017	1,18
Probability of serious infection while on maintenance therapy with IFX monotherapy over 1 year	0.017	1,18
Surgical mortality and complication rate
Probability of complicated surgical course	0.34	15
Probability of peri-operative mortality	0.001	15
Other mortality rates
Probability of lymphoma-related death for the base-case per cycle	0.0352	20
Probability of death from serious infection	0.001	19
Probability of age-specific all-cause mortality for 35-year-old male with inflammatory bowel disease	0.00040	13,14
Costs
Cost of a single 40-mg dose of ADA	US$2,048.54	21
Cost of a single 100-mg tablet of AZA	US$0.45	21
Cost of a CBC	US$10.39	22
Cost of a CMP	US$14.19	22
Cost of colonoscopy	US$429.20	22
Cost of colectomy	US$32,672.44	23
Cost of faecal calprotectin	US$26.83	22
Cost of hepatitis B core antibody, surface antibody, and surface antigen	US$44.63	22
Cost of average hospitalisation for ulcerative colitis flare	US$10,629.61	23
Cost of infliximab dose at 5 mg/kg for an 80-kg individual	US$4,453.08	21
Cost of infliximab infusion	US$165.06	22
Cost of serious infection and related hospitalisation	US$39,211.93	23
Cost of therapy for lymphoma, per cycle	US$4568.60	24
Cost of outpatient visit	US$112.69	22
Cost of ostomy supplies, per cycle	US$332.48	25
Cost of quantiferon gold	US$114.89	22
Cost of thiopurine methyltransferase phenotype testing	US$23.46	22
Cost of 300-mg dose of vedolizumab	US$5,212.23	21
Cost of vedolizumab infusion	US$136.41	22
Utilities or quality of life estimates
Utility for remission	0.79	26
Utility for UC in clinical response	0.68	*
Utility for UC flare	0.32	26
Utility for serious infection	0.62
Utility for SAE	0.47	26
Utility for non-Hodgkin’s lymphoma per cycle	0.1175	13,27,*
Utility for immediate postoperative course	0.25	13,27,*
Utility for postoperative remission/pouch	0.68	26
Disutility for complicated surgical course	0.1	*
Hazard ratios related to lymphoma
Baseline hazard of lymphoma with azathioprine use	5.28	28
Baseline hazard of lymphoma with anti-TNF use	1	28

Annual rates for maintenance in this table were converted to quarterly rates given 3-month cycle length. CBC, complete blood count; CMP, comprehensive metabolic panel; QALYs, quality-adjusted life-years; SAE, serious adverse event; TNF, tumour necrosis factor; UC; ulcerative colitis.

*Expert opinion used in the derivation of this estimate.

§Based on both published data on vedolizumab and cohorts derived from these data stratified by prior drug exposures.

Probabilities of rare adverse events not appreciated in clinical trials were derived from large observational studies. The probability of developing a lymphoproliferative disorder with azathioprine [AZA] or with an anti-TNF were modelled independently of each other, using baseline age- and gender-specific non-Hodgkin’s lymphoma rates from Surveillance, Epidemiology, and End Results [SEER]²⁰ and drug-specific hazard ratios derived from the Cancers Et Surrisque Associé aux Maladies Inflammatoires Intestinales En France [CESAME] cohort and TREAT registries.²⁸ Based on observational data by Colombel and colleagues, we employed no increased risk for lymphoma with vedolizumab use.⁷ All simulated individuals who developed lymphoma received standard of care chemotherapy for NHL.²⁷ The probabilities of mortality related to serious infection or modulation secondary to inflammatory bowel disease were derived from previous pooled estimates.²⁹

Quality-adjusted life-years [QALYs] were employed for all effectiveness-related rewards in the model. QALYs are estimates of one’s quality of life over a given time period, and are computed via assessing the average preference, or utility, for a specific health state by the duration spent in that state.³⁰ QALY estimates for this model were derived from previously published estimates and our previous simulation model.^13,14,26,31

Cost data for medications were derived using wholesale acquisition costs [WACs] available in IBM Micromedex RED BOOK.²¹ WAC pricing is the list price paid by wholesalers and distributors for drug purchasing, and was selected given their widespread publication and public availability via multiple online data sources. Laboratory, procedural, and infusion-related costs were derived from Medicare²² and were applied dependent on flaring, response, or remission [See Supplementary Methods and Results, available as Supplementary data at ECCO-JCC online]. Average costs for UC-related hospitalisation and colectomy were estimated through the use of 2013 data from the Nationwide Inpatient Sample via isolating relevant admissions and surgical procedures using ICD-9CM codes [see Supplementary Methods and Results].²³ Ostomy-related and lymphoma-related costs were derived from previously published research.^24,25 All costs were adjusted for inflation to 2017 pricing using the All Urban Consumers Consumer Price Index.

2.2. Statistical analyses

Analyses were conducted using TreeAge Pro 2018 [TreeAge Software, Inc., Williamstown, MA]. Expected costs and QALYs were calculated for all algorithms at the end of 1 year. The incremental cost-effectiveness ratios, or ICER, were calculated using mean cost and QALY estimates derived from First Order Monte Carlo Simulation [FOMCS] of 100 000 subjects over 100 iterations of the model. A baseline willingness-to-pay threshold of US$100,000 was used, and analyses were repeated considering US$150,000 and US$50,000 thresholds when appropriate.

2.3. Sensitivity analyses

We conducted a number of structural and input-related sensitivity analyses to assess assumptions made in the construction of our model. One-way sensitivity analyses were performed using microsimulation varying all transition probabilities by 25%, and QALY estimates by 15%. To assess the accuracy of WAC pricing used in our model, we also performed one-way sensitivity analyses for these values, varying estimates by +/-15%.

In order to assess the relationship between VDZ positioning and use of either infliximab or adalimumab, we performed additional deterministic sensitivity analyses with: [1] 100% infliximab; and [2] 100% adalimumab use as first anti-TNF, each in combination with AZA. These analyses allow us to assess situations where only one of these agents may be available as first-line therapy. We then conducted a one-way sensitivity analysis altering the ratio of infliximab to adalimumab use in our simulated population of 100 000 individuals, to determine if there was a threshold of mixed use that would exceed our willingness-to-pay threshold or alter the preferred strategy.

We repeated our primary analyses using probabilistic, or second-order Monte Carlo, methods to most fully capture the impact of uncertainty inherent in incorporated estimates.³² To conduct these analyses, all model inputs were converted to relevant distributions,³³ the parameters for which were derived directly from clinical trials and, in the case of rare outcomes, relevant observational research. QALY-related inputs employed normal distributions. Costs were modelled using gamma distributions. Transition probabilities were modelled using Dirichlet distributions. Probabilistic analyses were conducted using cohorts of 50 000 simulated individuals.

We also performed analyses to assess how the introduction of biosimilar infliximab and adalimumab may influence our model. For all biosimilar anti-TNFs, identical efficacy rates between biosimilar and bio-originator compounds were assumed. Using the base model with 50% utilization of adalimumab and 50% utilization of infliximab in a population, first-order Monte Carlo simulations of 50 iterations of 100 000 individuals, with infliximab costs ranging from 100% of WAC pricing [the base value in our model] to 50% of WAC pricing, were performed to calculate mean costs, QALYs, and ICERs. We selected the estimate of 50% reduction in pricing based on recently published estimates of the impact of biosimilar pricing in European markets.³⁴ The threshold at which the preferred strategy would change, when comparing vedolizumab as first-line biologic therapy with last-line biologic therapy, was then identified, when considering a WTP threshold of US$100,000. Whereas only biosimilar infliximab is currently available in the USA, due to the possible availability of biosimilar adalimumab after 2023, these analyses were repeated considering both biosimilars to be available with similar ratios of biosimilar to bio-originator use for both drugs, ranging from 100% bio-originator costs for both to 50% cost reduction for both.

We also conducted several additional structural sensitivity analyses. We included an iteration of the model where azathioprine could be used as first-line therapy, resulting in four potential algorithms of care. We also assessed the impact of dose escalation on the results of our 100% infliximab model. For this analysis, all individuals who flared while receiving a medication first underwent an attempt at dose escalation before transitioning to the next assigned drug in their algorithm. For infliximab, the dose was escalated to 10 mg/kg every 8 weeks, and for adalimumab, the dose was escalated to 80 mg every other week. Response and remission rates for anti-TNF dose optimisation were assumed to be similar between anti-TNFs.³⁵ Vedolizumab-related rates of response and remission were derived from a recent analysis by Loftus and colleagues.³⁶ Last, we explored time horizons longer than 1 year, repeating our analyses for infliximab, adalimumab, and mixed combination therapy at 3, 5, and 7 years. These analyses were conducted using probabilistic transition probabilities and reward estimates.

3. Results

In our base model assessing 50% utilization of infliximab or adalimumab in combination with AZA, Algorithm 1 [vedolizumab use as first-line therapy] was the preferred strategy, strongly dominating Algorithms 2 and 3 at the end of 1 year [Table 2]. Algorithm 1 yielded the best quality of life at 1 year (Algorithm 1 expected value [EV]: 0.607 QALYS, Algorithm 2 EV: 0.573, Algorithm 3 EV: 0.585). Algorithm 1 was also the least expensive option [Algorithm 1: US$64,109.53, Algorithm 2: US$67,755.44, Algorithm 3: US$67,335.77].

Table 2.

Quality-adjusted life years, costs, and incremental cost-effectiveness ratios when considering a population using 50% infliximab and 50% adalimumab as first anti-TNF.

Strategy	Mean cost	Mean incremental cost	Mean effectiveness [in QALYs]	Mean Incremental effectiveness	Mean ICER
Algorithm 1: VDZ->first anti-TNF + AZA- >second anti-TNF + AZA->surgery	US$64,109.53	-	0.607	-	-
Algorithm 2: First anti-TNF + AZA-> VDZ-> second anti-TNF + AZA->surgery	US$67,755.44	US$3,645.91	0.573	-0.034	*
Algorithm 3: first anti-TNF + AZA->second anti- TNF + AZA->VDZ->surgery	US$67,335.77	US$3,226.24	0.585	-0.022	*

Results of deterministic analysis examining different infliximab to adalimumab ratios as initial anti-TNF selected.

ICER, incremental cost-effectiveness ratio; AZA, azathioprine; QALY: quality-adjusted life-year; VDZ, vedolizumab; TNF, tumour necrosis factor.

*Negative ICER not reported as per modelling convention.

When considering a population where 100% of individuals use infliximab as the first-line anti-TNF, vedolizumab as first-line biologic therapy prior to combination therapy [Algorithm 1] yielded the greatest quality of life benefit when compared with other strategies at the end of 1 year [Table 3] (Algorithm 1 expected value [EV]: 0.6171 QALYS, Algorithm 2 EV: 0.6123, Algorithm 3 EV: 0.6094). With regards to annual costs, Algorithm 2, or vedolizumab as second-line therapy, was the least expensive strategy [Algorithm 1: US$61,768.68, Algorithm 2: US$60,339.71, Algorithm 3: US$61,985.68]. Based on these results, the most cost-effective option at the end of 1 year was Algorithm 2 [vedolizumab use prior to adalimumab]; the ICER for Algorithm 1 was US$300,696.75, exceeding our willingness-to pay threshold of US$100,000. Algorithm 3, or vedolizumab use as third-line therapy after both infliximab combination therapy and adalimumab combination therapy, was strongly dominated [ICER not reported]. We appreciated similar results in our probabilistic analyses as well, with Algorithm 2 preferred, and an ICER of US$286,506.05 for Algorithm 1 [see Supplementary Table 1, available as Supplementary data at ECCO-JCC online]. Over 100 000 iterations of the model, Algorithm 2 was preferred 44.5% of the time, with Algorithm 1 preferred in 32% of iterations and Algorithm 3 in 23.6%.

Table 3.

Quality-adjusted life-years, costs, and incremental cost-effectiveness ratios when considering 100% infliximab use as first anti-TNF.

Strategy	Mean cost	Mean incremental cost	Mean effectiveness [in QALYs]	Mean incremental effectiveness	Mean ICER
Algorithm 2: first anti-TNF + AZA->VDZ- >second anti-TNF + AZA->surgery	US$60,339.71	US$0.00	0.612	-
Algorithm 1: VDZ->first anti-TNF + AZA- >second anti-TNF + AZA->surgery	US$61,768.67	US$1,428.97	0.617	0.005	US$ 300696.7
Algorithm 3: first anti-TNF + AZA->second anti- TNF + AZA->VDZ->surgery	US$61,985.68	US$217.01	0.609	-0.008	*

Results of deterministic analysis of primary model, demonstrating that vedolizumab [VDZ] use after combination therapy with infliximab and azathioprine [AZA] but before cycling to adalimumab and AZA is the most cost-effective strategy. ICER, incremental cost-effectiveness ratio; AZA, azathioprine; QALY, quality-adjusted life-year; VDZ, vedolizumab; TNF, tumour necrosis factor.

*Negative incremental cost-effectiveness ratio [ICER] not reported as per modelling convention.

When considering a population where 100% of individuals use combination therapy with adalimumab as the first-line anti-TNF, vedolizumab as first-line biologic therapy [Algorithm 1] yielded the best quality of life [Algorithm 1 EV: 0.5963 QALYs, Algorithm 2 EV: 0.5330 QALYs, Algorithm 3 EV: 0.5607 QALYs], and was also the least expensive option at the end of 1 year [Algorithm 1: US$66447.42, Algorithm 2: US$75178.07, Algorithm 3: US$72678.76] [Table 4]. Algorithm 1 strongly dominated both Algorithms 2 and 3 [ICERs not reported]. In our probabilistic analyses, we appreciated similar results, with Algorithm 1 strongly dominating the other options [see Supplementary Table 2, available as Supplementary data at ECCO-JCC online]; this algorithm was preferred in 97.8% of model iterations.

Table 4.

Results when considering adalimumab as first-line therapy.

Strategy	Mean cost	Mean incremental cost	Mean effectiveness [in QALYs]	Mean incremental effectiveness	Mean ICER
A1: VDZ->ADA + AZA->IFX + AZA->surgery	US$66,447.43	US$0.00	0.596	-	-
A3: ADA + AZA->IFX + AZA->VDZ->surgery	US$72,678.76	US$6,231.33	0.561	-0.036	*
A2: ADA + AZA-> VDZ-> IFX + AZA-> surgery	US$75,178.07	US$8,730.64	0.533	-0.063	*

Results of deterministic analysis model considering adalimumab [ADA] as first-line combination therapy, demonstrating that vedolizumab [VDZ] use as first-line therapy strongly dominates other strategies.

AZA, azathioprine, IFX, infliximab; QALY, quality-adjusted life-year.

*Negative incremental cost-effectiveness ratio [ICER] not reported as per modelling convention.

We then assessed at which threshold of mixed adalimumab and infliximab use Algorithm 2 becomes the preferred strategy. In this simulation, if 14% or more individuals used adalimumab as their first-line anti TNF, vedolizumab use after an initial anti-TNF but before cycling to a second anti-TNF [Algorithm 2] strongly dominated other strategies [Table 5]. If more than 6% but less than 14% of individuals in a population used adalimumab as their first-line anti-TNF, vedolizumab was preferred when considering a WTP threshold of US$100,000. However, if fewer than 6% of individuals utilized adalimumab as first-line therapy, then vedolizumab use prior to cycling to a second anti-TNF was the preferred strategy. [Figure 2].

Table 5.

Quality-adjusted life-years, costs, and ICERs at various levels of mixed adalimumab and infliximab use as first anti-TNF.

Strategy	Mean cost	Mean incremental cost	Mean effectiveness [in QALYs]	Mean incremental effectiveness	Mean ICER
85% infliximab, 15% adalimumab
A1: VDZ->first Anti-TNF + AZA->second anti-TNF + AZA- >surgery	US$62,470.96	US$0.00	0.614	-	-
A3: first anti-TNF + AZA->second anti-TNF + AZA->VDZ- >surgery	US$62,559.29	US$88.33	0.600	-0.013	*
A2: first anti-TNF + AZA->VDZ->second anti-TNF + AZA- >surgery	US$63,588.94	US$1,117.98	0.602	-0.012	*
90% infliximab, 10% adalimumab
A2: first anti-TNF + AZA->VDZ->second anti-TNF + AZA- >surgery	US$61,819.53	US$0.00	0.604	-	-
A1: VDZ->first anti-TNF + AZA->second anti-TNF + AZA- >surgery	US$62,236.19	US$416.66	0.615	0.011	US$39,346.65
A3: first anti-TNF + AZA->second anti-TNF + AZA->VDZ- >surgery	US$63,054.42	US$818.24	0.605	-0.010	*
95% infliximab, 5% adalimumab
A2: first anti-TNF + AZA->VDZ->second anti-TNF + AZA- >surgery	US$61,079.58	US$0.00	0.608	0.000	US$0.00
A1: VDZ->first anti-TNF + AZA->second anti-TNF + AZA- >surgery	US$62,002.11	US$922.53	0.616	0.008	US$120,481.13
A3: first anti-TNF + AZA->second anti-TNF + AZA->VDZ- >surgery	US$62,518.21	US$516.09	0.607	-0.009	*

Results of deterministic analysis examining different infliximab to adalimumab ratios as initial anti-TNF selected.

AZA, azathioprine; VDZ, vedolizumab; QALY, quality-adjusted life-year; TNF, tumour necrosis factor.*Negative incremental cost-effectiveness ratio [ICER] not reported as per modelling convention.

Figure 2.

Sensitivity analysis examining the relationship between the percentage of adalimumab versus infliximab use among a population and the incremental cost effectiveness ratio [ICER] comparing using vedolizumab as first-line therapy with vedolizumab after an initial anti-TNF combined with azathioprine. With greater percentages of infliximab as first anti-TNF drug, the ICER of first-line vedolizumab use increases, exceeding a willingness-to-pay-threshold of US$100,000 if 94% or greater of first anti-TNF use is infliximab. TNF, tumour necrosis factor.

3.1. Sensitivity analyses

In one-way sensitivity analyses conducted using microsimulations, our model was particularly sensitive to costs of both infliximab and vedolizumab. If the cost of infliximab increased above US$4613 [a 3.6% increase], the ICER for vedolizumab as first line decreased below the willingness-to-pay threshold of US$100,000. If the cost of infliximab exceeded US$4715, a 5.9% increase, Algorithm 1 became the preferred strategy [Figure 3]. Similarly, if the cost of vedolizumab decreased to below US$5065 [a 2.8% decrease], the ICER for Algorithm 1 decreased below US$100,000; and if the price decreased below US$4970 [a 4.6% decrease], Algorithm 1 became the preferred strategy [Figure 4]. The model was also sensitive to the rate of clinical response with vedolizumab when used prior to an anti-TNF. If this value increased above 29.0% [an absolute increase of 4.4% and relative increase of 17.9%], the ICER for Algorithm 1 decreased below US$100,000. If this value increased above 37.1% [a 12.5% absolute increase and 50.8% relative increase], Algorithm 1 was preferred. The model was not sensitive to other transition probabilities or costs.

Figure 3.

One-way sensitivity analysis examining the impact of the cost of each dose of infliximab, at 5mg/kg, on the incremental cost-effectiveness ratio [ICER]. Thresholds at which the a priori willingness-to-pay threshold and preferred strategy are crossed are highlighted by vertical bars. Abbreviations: ICER: incremental cost-effectiveness ratio, IFX: infliximab, IM: immunomodulator, VDZ: vedolizumab, WTP: willingness-to-pay.

Figure 4.

One-way sensitivity analysis examining the impact of the cost of each dose of vedolizumab on the incremental cost-effectiveness ratio [ICER]. Thresholds at which the a priori willingness-to-pay threshold and preferred strategy are crossed are highlighted by vertical bars. ICER, incremental cost-effectiveness ratio; IFX, infliximab; IM, immunomodulator; VDZ, vedolizumab; WTP, willingness-to-pay.

3.2. Biosimilar costs and model outputs

When considering the use of biosimilar infliximab in our base model with 50% utilization of infliximab and 50% utilization of adalimumab as first-line anti-TNF, the ICER for first-line vedolizumab use in comparison with last-line use increased linearly with decreasing infliximab costs [Figure 5]. The WTP threshold of US$100,000, comparing first-line vedolizumab use with last-line therapy, was exceeded if infliximab pricing was reduced by 29% or more. A similar linear relationship was appreciated between increasing use of both biosimilar infliximab and biosimilar adalimumab, which more strongly affected the ICER. When considering both biosimilar agents, vedolizumab as last-line therapy became preferred, with a price reduction exceeding 16% with a WTP threshold of US$100,000.

Figure 5.

The relationship between percent cost reduction in anti-TNFs due to biosimilar use and incremental cost effectiveness ratio [ICER] was assessed in the base model with 50% utilization of each. For infliximab biosimilar use and both infliximab and adalimumab biosimilar use, a relatively linear relationship existed between percent anti-TNF cost reduction and the ICER for using vedolizumab as first-line therapy as compared with last-line therapy before surgery, with reductions in anti-TNF pricing increasing the ICER value. TNF, tumour necrosis factor.

3.3. Structural sensitivity analyses

A model with azathioprine monotherapy as first-line therapy, followed by combination therapy with infliximab, combination therapy with adalimumab, and lastly surgery, results in four different points of vedolizumab insertion [see Supplementary Figure 1, available as Supplementary data at ECCO-JCC online]. In this model, vedolizumab as first-line therapy yielded the best quality of life at 1 year [see Supplementary Table 3, available as Supplementary data at ECCO-JCC online]. However, given the low cost of azathioprine in comparison with biologic therapies, the incremental cost-effectiveness exceeded the willingness-to-pay threshold for this strategy, and initial thiopurine use was the most cost-effective strategy.

When considering infliximab monotherapy as first-line therapy, followed by combination therapy with infliximab, combination therapy with adalimumab, and lastly surgery, we again considered four potential treatment algorithms incorporating vedolizumab use [see Supplementary Figure 2, available as Supplementary data at ECCO-JCC online]. In this model, vedolizumab use as first-line therapy yielded the best quality of life over 1 year [see Supplementary Table 3, available as Supplementary data at ECCO-JCC online]. However, due to the increased costs related to this strategy relative to use after combined infliximab and azathioprine, the ICER for this strategy greatly exceeded our willingness-to-pay threshold, with an ICER of US$1,084,080.

We also explored the impact of longer time horizons in models considering combined infliximab and azathioprine as first line, adalimumab and azathioprine as first line, and 50% infliximab and 50% adalimumab [Supplementary Table 4, available as Supplementary data at ECCO-JCC online]. When considering infliximab combination therapy as first line, vedolizumab after combination therapy but before transition to adalimumab combination therapy remained the most cost-effective strategy. Similar results were seen with 50% infliximab and adalimumab utilization. However, when considering adalimumab as first-line therapy, the time horizon affected our findings: At 3 years, vedolizumab as first-line therapy remained preferred. However, when considering time horizons of 5 and 7 years, adalimumab as first-line therapy became preferred, with vedolizumab as first-line therapy exceeding the predetermined WTP threshold of US$100,000.

We also assessed the impact of biologic dose escalation in our primary model at 1, 3, 5, and 7 years [see Supplementary Table 5, available as Supplementary data at ECCO-JCC online]. At all time horizons, vedolizumab after initial combination therapy with infliximab but before transition to a second anti-TNF was preferred.

4. Discussion

Identifying the most cost-effective position for biologics is crucial for maximising the patient’s quality of life and simultaneously controlling the burgeoning costs of medical care for inflammatory bowel disease. Although previous cost-effectiveness analyses have compared vedolizumab head-to-head with other medications, none of the prior studies have accounted for mixed utilization of either infliximab or adalimumab as first-line therapy in ulcerative colitis, which is common in real-world scenarios. Head-to head comparisons also do not uniformly account for medication sequencing. Given the proportion of patients that experience primary non-response and subsequent loss of response rates with biologics in both UC and Crohn’s disease, a large proportion of patients are sequentially exposed to multiple biologic drugs.

The research presented here is unique in several ways. Whereas previous studies have assessed the cost-effectiveness of medication sequencing in Crohn’s disease in European health care systems,^37,38 we have focused on sequencing in ulcerative colitis. Further, earlier CD- and UC-based sequencing models have focused on European or Asian markets, whereas our focus is on US-based population management.³⁹ Importantly, we were also able to simulate a cohort of individuals with access to either infliximab or adalimumab as their first-line anti-TNF. Building upon our prior simulation assessing the effectiveness of different UC treatment algorithms, coupled with the best available clinical and cost-related estimates, our model predicts that the most cost-effective position for vedolizumab is before cycling to a second anti-TNF when considering infliximab as first-line therapy.¹³ In contrast, when considering adalimumab as first-line anti-TNF therapy, vedolizumab was preferred as the first biologic when considering a 1-year time horizon. However, in most clinical settings, some patients are treated first with infliximab and others with adalimumab. Using our novel model structure, we were able to determine that vedolizumab as the first biologic therapy was the dominant strategy if 14% or more of simulated individuals utilized adalimumab as their first anti-TNF, and remained cost-effective if more than 6% of patients utilized adalimumab as their first anti-TNF. When fewer than 6% of patients used adalimumab as the first anti-TNF drug, vedolizumab use between the first and second anti-TNF drugs was the most cost-effective strategy.

These findings have important implications, given a rapidly shifting payer landscape with regards to biologics. First, infliximab remains the most cost-effective first-line therapy for moderate to severe ulcerative colitis, likely secondary to its well-demonstrated efficacy and similar costs to other therapies. Although there may be more serious adverse events with infliximab combination therapy than with vedolizumab, these were relatively rare over a 1-year time frame and thus did not substantively affect costs in these analyses. Another important finding is vedolizumab’s position in relation to adalimumab. We consistently appreciated that vedolizumab was more cost-effective than adalimumab. Further, modelling for either infliximab or adalimumab as first-line combination therapy demonstrated that, in a given market, the overall utilization of adalimumab as first line only needed to exceed 14% for vedolizumab to become preferred as first-line therapy.

This study has several important strengths. We apply novel modelling methods to attempt to assess the cost-effectiveness of sequential medications in UC, as opposed to comparing specific medications head-to-head in isolation. As more than 50% of patients with moderate to severe steroid-dependent UC will have to cycle between therapies, particularly when considering longer time horizons, similar analyses should be requisite for approval of and determination of payer policies related to new agents as they are developed. We were also able to assess several different sequences of anti-TNFs by alternating the position of infliximab and adalimumab in the treatment algorithm in sensitivity analyses. These data are of particular import when considering a payer’s decision to select a specific anti-TNF as their preferred therapy in that class. These analyses also resulted in one of our key findings regarding the preferred vedolizumab position when considering adalimumab as first-line therapy. Further, our simulation model allowed us to, for the first time, assess the impact of mixed initial use of infliximab or adalimumab. For payers with reduced restrictions on selection of an initial anti-TNF, our findings suggest that early use of vedolizumab should be considered in those regions where 14% or more of individuals consider adalimumab their preferred first anti-TNF.

Our results used the most up-to-date transition probabilities and US cost data. However, another strength of our work is the extent of our sensitivity analyses in both mixed utilization and with adalimumab and infliximab models. Our findings highlight the relative differences in utilization levels and costs required for each of the agents in the model to significantly impact on the preferred strategy, which can be used to provide guidance for other health care systems. Our model construct can also be applied with limited modifications to incorporate pricing from other markets, if necessary. Further, we performed analyses examining the impact of biosimilar use: although not yet a major component of US pricing, these results highlight the future cost reductions required for infliximab and adalimumab to affect the preferred strategy, further expanding the generalisability of our results.

However, there are several limitations to consider regarding this research. First, we based our findings on US wholesale acquisition costs and Medicaid data. It is possible that contracting with specific centres may alter pharmacological, laboratory, and procedural reimbursements. This is likely true for many biologic medications in the USA, including those in our model. Unfortunately, these contractual rates are not published or publicly available. To assess how these prices influenced our model, we performed one-way sensitivity analyses on all costs. Our model was sensitive to the costs of both infliximab and vedolizumab, highlighting that specific contracting with payers may directly influence the most cost-effective option. These costs may be further influenced by biosimilar agents as they are approved and adopted. There is currently limited utilization of these medications in the USA.⁴⁰ However, we provide further examples of specific scenarios in which biosimilar infliximab and adalimumab pricing may influence our model as noted previously.

Our model used efficacy estimates derived from randomised controlled trials to estimate transition probabilities. As discussed with our previous model, there are several potential concerns regarding this approach, including variance in patients enrolled in these studies over time and lack of head-to-head data.¹³ Further, it is possible that the real-world effectiveness of these therapies may differ from these data. This gap between efficacy and effectiveness has been well described in multiple medical disorders. Similarly, infliximab was the first FDA-approved biologic therapy for ulcerative colitis, and there are limited data available to estimate how infliximab or adalimumab perform when used after biologics with alternative mechanisms of action. Importantly, although sensitive to medication costs, our simulation results were not sensitive to reasonable variation in the efficacy of modelled treatments. However, as real-world effectiveness data are accrued, particularly with regards to newer therapies and infliximab use after failure of other agents, these analyses should be repeated.

We also chose to model combination therapy utilizing azathioprine/6-mercaptopurine [6-MP], and did not model the use of methotrexate. This decision was supported by several factors. First, the most robust and modern data regarding infliximab monotherapy and combination therapy induction rates can be derived from the SUCCESS UC trial.¹⁷ Second, maintenance estimates from studies such as ACT primarily focused on thiopurine use. Further, recent data from the MERIT-UC trial have demonstrated that methotrexate is not an effective therapeutic option in ulcerative colitis.⁴¹ The benefit of its use in combination therapy is still uncertain but, based on these factors, we chose not to explore this treatment. We also did not simulate tofacitinib, which was recently FDA-approved for UC, as this medication is most often being used as last-line therapy due to increased infectious and possible thromboembolic risks.^42,43 Were we to include this medication, it would be used as last-line therapy before surgery, and therefore would minimally affect the overall results. In fact, in our primary mixed model with 50% utilization of infliximab or adalimumab as first-line therapy, only 5.1%, 8.1%, and 8.0% of patients would have transitioned to tofacitinib before colectomy at 1 year in Algorithms 1 through 3, respectively.

We also used clinical response and remission in our model as opposed to mucosal healing. There are some data that suggest that ‘deep remission’ with mucosal healing may lead to more durable remission rates. However, there are limited data at this time to inform estimates of medication optimisation or changes in those with clinical response but continued disease activity. Further, it is uncertain what impact, if any, mucosal healing in and of itself would have on quality of life in comparison with clinical remission. As most current guidelines are focused on obtaining clinical remission, and there are ongoing debates regarding mucosal healing as an appropriate endpoint for all patients, we selected clinical response and remission as the more realistic endpoint. As more robust data on mucosal healing become available, one could consider repeating these analyses using this clinical endpoint.

Last, we also estimated the most cost-effective therapy using progressively longer time horizons. As demonstrated in our Supplementary Results, time horizons beyond 5 years demonstrated results that are not consistent with our primary findings. Interpretation of these results requires significant caution, given the limited data available on the long-term durability of all therapies included beyond 1 year.

In summary, we used state-of-the-art simulation methods to assess the most cost-effective sequencing of biologic medications for UC. This simulation demonstrated that when both infliximab and adalimumab are available as first-line therapy, vedolizumab use was most cost-effective when 14% or more of patients selected adalimumab as first-line therapy. When all patients received infliximab as the first anti-TNF drug, vedolizumab use was most cost-effectively positioned after infliximab in combination with azathioprine failure but before cycling to adalimumab. Additionally, if adalimumab was used prior to infliximab, vedolizumab was preferred as first-line biologic therapy. Further research to assess the distribution of first-line anti-TNF drug use for UC in the USA and long-term outcomes with biologic therapies will help to answer remaining areas of uncertainty.

Funding

FIS and JDL received support via a research grant provided by Takeda Pharmaceuticals USA, Inc.

Conflict of Interest

FIS reports grants from Takeda Pharmaceuticals USA and Janssen Pharmaceuticals, and personal fees from Evidera. Inc., PRIME Incorporated, Janssen Pharmaceuticals, and Merck Pharmaceuticals. JDL reports personal fees from Shire, Janssen Pharmaceuticals, AbbVie, Immune Pharmaceuticals, AstraZenecca, Amgen, MedImmune, Merck, Nestle Health Science, Takeda Pharmaceuticals North America, Pfizer, Lilly, Gilead, Samsung Bioepis, Bristol-Myers Squibb, and Johnson and Johnson, and research funding from Takeda Pharmaceuticals North America and Nestle Health Science, and non-financial research support from AbbVie. RM reports consulting fees from Roche, and funding from the NIH/National Cancer Institute [K23-CA187185]. ML reports employment at Takeda Pharmaceuticals USA, Inc., Deerfield, IL. KL reports employment at Takeda Pharmaceuticals USA, Inc., Deerfield, IL. The remaining authors report no conflict of interest.

Author Contributions

FIS planned the study, interpreted the data, constructed the simulation model, performed model-related analyses, and drafted and edited the manuscript. He is the guarantor. JDL edited the research proposal, interpreted the data, and edited the manuscript. YS assisted in model construction and edited the manuscript. BF, MG, RV, RM, KL, and ML analysed the results, and edited the manuscript.