Photodynamic therapy for Barrett ’ s esophagus with high-grade dysplasia : A cost-effectiveness analysis

1Department of Medicine, Division of Gastroenterology; 2Department of Clinical Epidemiology and Biostatistics; 3Centre for Evaluation of Medicines (St Joseph’s Hospital), McMaster University, Hamilton, Ontario Correspondence: Dr John K Marshall, Division of Gastroenterology, Room 2F59, McMaster University Medical Centre, 1200 Main Street West, Hamilton, Ontario L8N 3Z5. Telephone 905-521-2100 ext 76782, fax 905-523-6048, e-mail marshllj@mcmaster.ca Received for publication August 23, 2005. Accepted June 22, 2006 D Comay, G Blackhouse, R Goeree, D Armstrong, JK Marshall. Photodynamic therapy for Barrett’s esophagus with high-grade dysplasia: A cost-effectiveness analysis. Can J Gastroenterol 2007;21(4):217-222.

If esophageal biopsies reveal high-grade dysplasia (HGD), the likelihood of synchronous adenocarcinoma or progression to adenocarcinoma is increased (4-7).The standard therapy for patients with BE and HGD is distal esophagectomy (ESO) (8,9).However, ESO is associated with significant perioperative morbidity and mortality (10,11), and can lead to longterm complications, such as dumping syndrome and dysphagia, that impair health-related quality of life (12).
Techniques for endoscopic ablation of dysplastic epithelium have been advocated as alternatives to ESO in patients with BE with HGD.Of these, photodynamic therapy (PDT) is the best studied and the only approach approved by the United States Food and Drug Administration (13).PDT involves intravenous infusion of a photosensitizing agent, such as porfimer sodium (Photofrin, Axcan Pharma Inc, Canada), that accumulates in dysplastic tissue, causing tissue destruction when activated by an endoscopic light source (14).PDT has proven to be efficacious in the eradication of BE with HGD and in the prevention of cancer (15,16).It has the theoretical advantages of reduced morbidity, negligible mortality and lower upfront costs compared with ESO.
An alternate, more conservative approach to BE patients with HGD is intensive, endoscopic SURV with biopsy after thorough investigation, including endoscopic ultrasound, to exclude synchronous adenocarcinoma (17).While this approach risks progression to advanced adenocarcinoma, it can entirely defer or avoid the morbidity and mortality of ESO.
Each of these approaches to managing patients with BE and HGD offers tradeoffs in terms of costs and health benefits.In the absence of rigorous, prospective, comparative trials, decision analysis can estimate the expected outcomes of alternative strategies as well as identify key determinants of efficacy and cost-effectiveness.A decision analytical model was constructed and probabilistic sensitivity analysis was used to assess the incremental cost-effectiveness of ESO, PDT and continued endoscopic SURV in patients with BE and HGD.

General
A Markov model was constructed using Data 3.5 software (TreeAge Software Inc, USA) to simulate the management of a hypothetical cohort of 50-year-old men with BE and HGD.The incremental cost-effectiveness of three alternative strategies was assessed: immediate ESO, PDT and intense endoscopic SURV.Health outcomes were measured in unadjusted life-years (LYs) and quality-adjusted life-years (QALYs).
A simplified version of the Markov model is provided in Figure 1.The model allowed the study cohort to move among health states relevant to the management of BE with HGD using predefined transition probabilities.For each cycle spent in a given health state, patients accrued specific costs and gained QALYs proportionate to the utility weight assigned to that health state.The cycle length was three months, and the model's time horizon was five years.The model considered direct medical costs from the perspective of a Canadian provincial ministry of health.

Model cohort
The model cohort was comprised of men, 50 years of age, with BE and newly diagnosed HGD.The cohort was assumed to be asymptomatic, treatment naive (except for chronic acid suppression) and fit for ESO.Before model entry, the cohort was assumed to have undergone thorough investigation, including endoscopic ultrasound, to exclude synchronous adenocarcinoma.

ESO strategy
In this strategy, the cohort underwent immediate and curative ESO.No further endoscopic SURV was performed.

PDT strategy
In this strategy, patients underwent PDT with porfimer sodium (Photofrin) as described by Overholt (18).The number of PDT sessions required to achieve ablation and the rate of esophageal stricture requiring dilation were estimated from the literature (15,16,19).Patients in whom eradication of HGD was confirmed underwent intense SURV strategy until HGD recurred, cancer developed or the model terminated.Patients who were found to have persistent or recurrent HGD after PDT and those who subsequently progressed to adenocarcinoma underwent ESO as per the ESO strategy.

Endoscopic SURV strategy
In this strategy, the cohort underwent SURV endoscopy with biopsy every three months as per the Seattle protocol (3).Patients who progressed to adenocarcinoma underwent ESO.All others continue to undergo endoscopic SURV indefinitely.

Probability assumptions
For each transition probability, a literature review was conducted to estimate its base-case value and a plausible range for use in sensitivity analysis (Table 1).To estimate the efficacy of PDT, a MEDLINE search was performed in November 2004 using the terms 'Barrett esophagus' and 'photochemotherapy', with results limited to porfimer sodium (Photofrin).Previously published reviews and proceedings from key conferences were hand-searched for pertinent abstracts and references.Fixed event rates derived from the literature at various time points were fitted to an exponential curve to determine three-month cycle probabilities (20).All costs and outcomes were discounted at an annual rate of 3% per Canadian guidelines (21).

Mortality
Survival with BE and HGD and early esophageal adenocarcinoma was estimated from the National Cancer Institute Surveillance Epidemiology and End Results (SEER) database (22).The survival of model cohort subjects with BE and HGD was assumed to equal the SEER program survival for men with histologically confirmed, in situ cancer.The survival of model cohort subjects who developed early adenocarcinoma was assumed to be equal to the weighted mix of localized and in situ esophageal cancer in the SEER program.
Patients in the model who survived ESO for either BE with HGD or cancer assumed the age-specific mortality of the Canadian male population (23).All patients who survived to the model's five-year time horizon were assigned the predicted life expectancy of a 55-year-old Canadian man (23.38 years) as estimated from Canadian life tables (23).

Health utility weights
The model estimated health outcomes using both unadjusted LYs and QALYs.For the latter, death was assigned a utility weight of zero and the post-ESO state was assigned a utility weight of 0.80, as estimated by Provenzale et al (24).All other health states were assigned a utility value of one (Table 1).The transient disutility of undergoing endoscopy (including PDT) was not considered.

Cost inputs
Direct medical costs were considered from the perspective of a provincial ministry of health (Table 1).Canadian data sources were used wherever possible.American costs were converted to 2003 Canadian costs using prevailing Bank of Canada exchange rates (US$1=CDN$1.34).Standard resource profiles were developed for each health state cycle using expert opinion and consensus.Costs for physician services were obtained from the Ontario Ministry of Health and Long-Term Care Schedule of Benefits (April 2002) (25).Costs for surgical admissions and hospital-based endoscopic procedures were estimated by adapting the London Health Sciences Centre Case Cost Program (LHSCCP) administrative database, which is part of the Ontario Case Costing Initiative.The LHSCCP fully allocates hospital global budget expenditures among standardized functional cost centres such as nursing, clinical laboratory, diagnostic imaging and pharmacy in accordance with Guidelines for Management Information Systems in Canadian Health Care Facilities (26).The LHSCCP assigns each resource or service a variable cost for material supplies, plus a workload-indexed proportion of fixed or overhead costs, such as labour, equipment and capital infrastructure (including depreciation).Total costs were estimated by multiplying resource units by the total unit cost (ie, fixed plus variable).
Resource profiles for performing PDT were adapted from Overholt et al (15).Costs per session of PDT included the facility costs for endoscopy; physician reimbursement fee for endoscopy; and the per patient costs of administering PDT, including 2 mg/kg doses of the sensitizing drug (Photofrin), disposable light-emitting fibre and centring balloon, and laser-light generator.Table 1 shows cost parameters used in the model.Market prices for Photofrin, disposables and laser-light generator were obtained from the manufacturer on October 2003.Cost per use of the light source was amortized assuming a five-year life expectancy with 1000 procedures.Each patient required an average of 1.32 PDT sessions based on available trial data (15).To assess treatment response to PDT, each patient had a follow-up endoscopy at three months, which incurred the costs of the endoscopy and physician reimbursement for the procedure, esophageal biopsy and a subsequent clinic visit.Although Nd:YAG laser therapy and endoscopic ultrasound were used by Overholt et al (15), these procedures were not included in the model, given their limited availability in Canada at the time of study conception.Rates of complications of PDT (eg, esophageal stricture requiring dilation) were estimated from the literature (Table 1).Each event was assigned a standard resource profile for management.Costs incurred from these complications included both treatment costs and physician reimbursement fees for the treatment and clinic follow-up.
Costs for ESO included both facility costs and physician reimbursement for the consultation, surgical procedure and follow-up visits.A routine post-ESO endoscopy was not included in the model.Direct costs from postoperative complications would have been captured by the LHSCCP (see above).Similarly, SURV arm costs included physician reimbursement for the consultation, endoscopy procedure and follow-up visits, and the direct cost per endoscopy.No complication rate was assumed for nontherapeutic endoscopy.

Base-case and incremental cost-effectiveness analysis
Expected lifetime costs, LYs and QALYs were first calculated for each strategy using base-case parameter estimates.General principles of assessing cost-effectiveness among multiple strategies were then applied (27).First it was determined whether any treatment dominated, having both higher costs and lower health benefits by a single strategy (strict dominance) or a linear combination of two strategies (extended dominance).Any dominated strategies were eliminated from further analysis.Among nondominated strategies, incremental cost-effectiveness ratios (ICERs) were calculated as the ratio of the difference in cost to the difference in outcome between two alternatives.Beginning with the least costly treatment, alternatives were compared with the next most costly strategy when calculating ICERs.This process produced an 'efficiency frontier' of increasingly more effective but more costly treatment strategies.
PDT for Barrett's with high-grade dysplasia Can J Gastroenterol Vol 21 No 4 April 2007 219

Probabilistic sensitivity analysis
To assess the impact of the joint uncertainty of model variables on cost-effectiveness, a probabilistic sensitivity analysis was conducted (28).Here, transition probabilities and cost estimates used in the model were assigned unique distributions (Table 1).Cost parameters were assumed to be normally distributed and transition probabilities were assigned either beta or log-normal distributions.Distribution parameters and 95% CIs for each variable are shown in Table 1.With each model simulation, a value for each parameter was chosen by random selection with a resultant estimate of expected costs and outcomes.Results from 10,000 Monte Carlo simulations were aggregated to define distributions of the expected costs and health outcomes of each strategy (29).

Acceptability curves
The results of the probabilistic sensitivity analysis were also depicted using cost-effectiveness acceptability curves (30).This format is particularly informative when a payer's willingnessto-pay (WTP) threshold is uncertain.WTP thresholds are the maximum ceiling costs per unit of benefit, below which a payer deems an intervention to be cost-effective (31).Specifically, acceptability curves use Monte Carlo simulation results to estimate the proportion of times that each strategy is the most costeffective alternative at each WTP threshold.

Base-case analysis
The base-case model results are presented in

Incremental cost-effectiveness analysis
ICERs and cost-effectiveness efficiency frontiers for unadjusted LY and QALY are shown in Table 2 and Figure 2.For the unadjusted LY, all three strategies formed the frontier.The ICER of PDT relative to SURV was $814/LY and that of ESO relative to PDT was $3,397/LY.For QALYs, the efficiency frontier comprised only the SURV and PDT strategies, because ESO was dominated by PDT (Figure 3).The ICER of PDT relative to SURV was $879/QALY.

Probabilistic sensitivity analysis
A scatterplot of the incremental cost and QALY pairs for PDT and ESO relative to SURV (reference), generated from the 10,000 Monte Carlo simulations, is shown in Figure 4.The cost-effectiveness acceptability curve for an unadjusted LY (cost per LY) is shown in Figure 5. ESO had the highest probability of being the most cost-effective strategy for WTP thresholds above $3,500/LY.This probability reached 0.99 at a ceiling ratio of $25,000/LY.The cost-effectiveness acceptability curve for a QALY (cost per QALY) is presented in Figure 6.
Here, PDT had the highest probability of being the most costeffective strategy for WTP thresholds above $1,000/QALY.The probability that PDT is the most cost-effective strategy reached 0.99 at a ceiling ratio of $25,000/QALY.

DISCUSSION
This cost-effectiveness analysis uses state-of-the-art modelling techniques, including probabilistic sensitivity analysis, to demonstrate that PDT is a cost-effective alternative to endoscopic SURV and immediate ESO in male patients with BE and HGD.In particular, PDT is favoured when survival is adjusted to reflect reduced quality of life after ESO.Several assumptions made in the design of this model should be acknowledged explicitly and may influence the interpretation of results.First, the model did not allow spontaneous regression of HGD in the setting of BE because there   remain few published data on the rate and significance of this phenomenon.Second, rates of progression from HGD to cancer were estimated from observational studies.While diagnostic error is captured implicitly in these data, false-positive error was not modelled explicitly, and we assumed diagnoses of HGD and adenocarcinoma to be accurate.Third, we assumed that all members of the cohort were free of cancer at baseline and remained eligible for surgery throughout the model, neither of which may be clinically realistic.Fourth, patients who underwent ESO for HGD or cancer were deemed cured with no need for further endoscopic SURV, which may not be standard practice at some centres.ICERs offer meaningful comparisons of costs incurred to achieve additional units of effectiveness among competing strategies, and are informative for decision-makers who must allocate constrained resources (32).In such cases, a management strategy may be adopted if its ICER falls below a predefined WTP threshold.No such threshold is accepted universally, but values between $50,000/LY and $100,000/LY have been proposed and are broadly applied (33).Because all ICERs in our base-case analyses are well below these thresholds, our model suggests that ESO is the most cost-effective strategy for unadjusted LYs and PDT is the most cost-effective strategy for QALYs in patients with BE and HGD.
Fixed estimates of ICER may be misleading, because they do not communicate the inherent uncertainty with which input parameters are estimated.Monte Carlo techniques capture and measure overall uncertainty, which is presented graphically as cost-effectiveness acceptability curves (30).If the exact acceptable WTP threshold is not known, decisionmakers may refer to the acceptability curve to identify WTP ranges for which a strategy is most likely to be cost-effective.For the unadjusted LY, our results suggest that PDT is the strategy most likely to be cost-effective for WTP thresholds between $100/LY and $3500/LY, whereas ESO is the most costeffective at higher WTP thresholds.For QALYs, PDT is the strategy most likely to be cost-effective for all WTP thresholds above $1000/QALY.
Other models have been used to assess the cost-effectiveness of PDT for BE with HGD (34)(35)(36).Although our results confirm the conclusions of those analyses in many respects, two important differences warrant mention.First, our study is the only study to assume a non-American perspective.Although Canada and the United States share similar health care standards (37,38), unit costs and practice patterns often vary, and costeffectiveness cannot be assumed to be transferable.Second, our study is the only analysis of PDT that uses state-of-the-art probabilistic sensitivity analysis to assess simultaneously the joint uncertainty of all model parameters, providing more comprehensive information to decision-makers.

SUMMARY
PDT is a cost-effective alternative to ESO and continued endoscopic SURV for the management of patients with BE and HGD.Assuming reasonable WTP, PDT is the strategy most likely to be cost-effective for gains in QALYs.However, ultimately, management must still be individualized to the patient, considering his or her preferences and comorbidities, as well as local expertise.
ACKNOWLEDGEMENTS: Unrestricted financial support was provided by Axcan Pharma Inc (Canada).

TABLE 1
Selected base-case parameter values ranges for probabilistic analysis *Axcan Pharma Inc, Canada.HGD High-grade dysplasia; LHSCCP London Health Sciences Case Cost Program; PDT Photodynamic therapy

Table 2
. Endoscopic SURV incurred the lowest costs but also yielded the shortest survival, both in unadjusted LYs and QALYs.ESO incurred the highest costs and provided the longest unadjusted LYs but intermediate QALYs.PDT incurred intermediate costs but provided the most QALYs.

TABLE 2
Results of base-case analysis