Cost–Utility and Budget Impact Analysis for Stopping the Inappropriate Use of Proton Pump Inhibitors After Cessation of NSAID or Low-Dose Acetylsalicylic Acid Treatment

Routinely collected dispensing data were extracted from the Dutch Foundation of Pharmaceutical Statistics database. The Dutch Foundation of Pharmaceutical Statistics collects dispensing data from more than 95% of the 1979 Dutch community pharmacies [35]. The database provides detailed information on drugs dispensed including the codes from the Anatomic Therapeutic Chemical code system [36], the prescribed daily dose and the amount dispensed. As a consequence, drug use periods can be calculated by dividing the amount dispensed by the prescribed daily dose. Information on patients’ sex and year of birth were also available. Medication of a specific patient was tracked within a pharmacy over time by a defined unique anonymous code per patient. For the present study, data from Dutch community pharmacies were used, which had provided complete data for the years 2013–2014. With these data, the proportions of inappropriate PPI continuation were calculated for the year 2014 in PPI users who started this protective co-medication in addition to NSAID or LDASA treatment but continued to use a PPI after cessation of their potentially harmful medication. Data on 2013 were necessary to establish that PPIs were started in 2014 and thereby not used in the year preceding 2014. The study was approved by the Dutch Foundation of Pharmaceutical Statistics advisory board.

Patients aged older than 50 years with PPI initiation in 2014 and an NSAID or LDASA treatment period within 10 days before or after PPI dispensing were selected from the database. This selection aimed to select only those PPI users who started using a PPI as a gastroprotective co-medication.

Patient selection included patients from the age categories for which the guidelines recommend gastroprotection [9, 37]. Younger age categories were also selected for a broader examination of PPI co-treatment initiation and cessation. Subjects with inappropriate PPI use were those with a repeat PPI dispensing and without an NSAID or LDASA treatment period within 10 days before or after this dispensing in the year 2014. Treatment periods were calculated from the amount dispensed divided by the prescribed daily dosage. Thus, NSAID or LDASA use could be taken into account from earlier dispenses. Percentages of inappropriate PPI use were calculated relative to patients who started these drugs as a protective co-medication to NSAID or LDASA treatment.

Current guidelines recommend the initiation of PPI co-treatment in patients aged 60 years and older with NSAID or LDASA use who have additional risk factors for GI damage. Additional risk factors are comorbidity such as rheumatoid arthritis, diabetes mellitus, or heart failure, or co-medications such as selective serotonin reuptake inhibitors or oral corticosteroids [9]. Because the guidelines differentiate between the need to prescribe a protective PPI in patients using an NSAID or LDASA for the age categories between 60–69 years, 70–79 years, and for LDASA also users aged 80 years and older, the results were stratified for these age categories as well as for the 50–59 years age category. Descriptive analyses on frequencies were performed with IBM SPSS Statistics 20 and Microsoft Excel 2007.

Costs were compared for the effectiveness in terms of costs and quality of life in separate models for PPI use during treatment with NSAIDs and with LDASA by two strategies each: (1) PPI use discontinued together with the cessation of the NSAID or LDASA treatment, and (2) PPI use continued after cessation of NSAID or LDASA treatment. For this purpose, Markov models with a time horizon of 5 years were created using Microsoft Excel 2007. The analysis were performed from a healthcare perspective. In view of the age categories involved, costs for productivity losses were less relevant and were not included in the models. The general structure of the models is shown in Fig. 1. Each patient started in a health state affected only by a disorder requiring NSAID or LDASA treatment. After 3-month cycles, patients could transfer to one of the following states: hip fracture, CAP, GI bleeding, death or remain in the initial health state.

Population-based incidence rates of hip fractures, CAP, GI bleeding and death were collected from Dutch Statistics. For the PPI discontinuation strategy, incidence rates were transformed into probabilities and used in this strategy [38, 39]. For the continuation strategy, incidence rates were transformed into health-state-specific probabilities by using PPI risk ratios on these events (see Electronic Supplementary Material [ESM]). The different probabilities were constructed in a manner to increase a subject’s risk to develop certain events with higher age.

Costs related to GI bleeding were taken from De Groot et al. [32] Costs related to CAP were derived from Spoorenberg et al. [40] and costs for a hip fracture were taken from the Dutch Organisation Veiligheid.nl, which collects information on safety problems in the Netherlands [41]. All costs were updated to 2015 prices by using Dutch consumer price indices [42]. A list for the costs and utilities is given in the ESM.

Utilities for the different health states were collected from the literature. The utilities of hip fractures were based on a systematic review of Peasgood et al. [43]. Utilities for GI bleeding were collected from De Groot et al. for NSAID and LDASA use, respectively [32, 33]. Because no utility for CAP was available from the literature, a utility was constructed based upon the comparison of the severity of this health state to the other health states used in this study. With these utilities, quality-adjusted life-years (QALYs) were calculated for each scenario. As recommended by the Dutch guidelines for pharmacoeconomic evaluations, costs were discounted at an annual rate of 4%, whereas utilities were discounted at an annual rate of 1.5% [44].

Parameter uncertainties were taken into account by performing probabilistic sensitivity analyses with Monte Carlo simulations based on 5000 iterations. For each iteration, a random value for each parameter was taken, based on the parameter distribution [38]. These distributions were based upon the 95% confidence intervals of each parameter. Beta distributions were used for utilities and probabilities, gamma distributions for costs, and lognormal distributions for risk ratios [39]. With the outcomes of the Monte Carlo simulations, a scatter plot of the incremental cost-effectiveness ratios and a cost-effectiveness acceptability curve were calculated. Seventy-year-old patients served as base cases, as this was the average age of NSAID and LDASA users for whom the use of a protective PPI medication is recommended. The analyses had a time horizon of 5 years. The model was based on NSAID or LDASA use for 5 years at an adherence rate of 68% [45].

A BIA was performed for the potential cost savings of a successful intervention to prevent inappropriate PPI use according to international guidelines [44, 46]. The budget impact was calculated for the first year after stopping from a healthcare perspective. First, the costs of both strategies for stopping and continuation of PPI co-treatment were calculated with the model. The difference in costs between the two strategies were then multiplied with the number of patients who were dispensed a PPI repeat prescription without concomitant NSAID or LDASA use. For the BIA, age categories were chosen corresponding to the cost–utility analysis because different healthcare resources are used with increasing age. The total sum of the different age categories was considered as the total budget impact for ceasing inappropriate PPI use after 1 year.

In 2014, the numbers of NSAID and LDASA users dispensed a PPI within 10 days of a period of NSAID or LDASA use amounted to 247,460 and 173,852, respectively (Table 1). Overall, 11.0% of the PPI users who started a PPI in combination with an NSAID continued to use a PPI after NSAID cessation. For LDASA users, the percentage of inappropriate PPI continuation was 5.0%. Inappropriate PPI continuation was highest in former NSAID users aged 80 years and older (16.4%) and lowest in patients aged between 50 and 59 years (9.9%). In LDASA users, inappropriate PPI continuation was highest for patients between aged 70–79 years (5.3%) and lowest in the 50–59 years and 60–69 years categories (4.7% for each category).

For the base case, the strategy of discontinuing PPI co-treatment, as compared to the strategy of PPI continuation, resulted in incremental costs of €170 (95% confidence interval − 282 to − 75) indicating savings, and incremental effects gains of 0.003 (95% confidence interval 0.001–0.005). Stopping PPI co-treatment was a cost-saving strategy for all age categories compared with the strategy of continuation after NSAID or LDASA cessation, and as such inappropriate PPI use (Table 2). Consequently the PPI discontinuation strategy ‘dominated’ its alternative in all 5000 simulations (Fig. 2). This was due to both costs saved for the medication and for the treatment of adverse effects (negative incremental costs) and the gain of QALYs because of a lower incidence of adverse effects resulting from continued PPI use (positive incremental QALYs). In older age categories, the difference between costs and effects for the two strategies increased, implicating that the cost effectiveness of interventions to stop inappropriate PPI use increases with a patients’ age.

The results of the BIA are shown in Table 3. The total budget impact of conducting interventions to discontinue inappropriate PPI use after cessation of NSAID or LDASA treatment was almost €1,050,000. Based on the total budget impact of a 100% successful intervention, this equals €29 per patient.