J.S.A.G. Schouten Department of Ophthalmology Maastricht University Medical Centre P.O. Box 5800 6202 AZ Maastricht The Netherlands Tel: + 31(0) 43 3875344 Fax + 31(0) 43 3875343 Email: j.schouten@mumc.nl

Abstract.

Purpose: To determine the cost-effectiveness of ocular hypertension (OH) treatment initiated with latanoprost compared to timolol.

Methods: Two strategies for OH therapy are modelled, (1) ‘starting with timolol’ and (2) ‘starting with latanoprost’. Therapy can be maintained or changed dependent on the achieved intraocular pressure (IOP) and side-effects. Adjustments of therapy to reach a target pressure involve monotherapy, combination therapy and laser. Four drugs are used: latanoprost, timolol, brimonidine and dorzolamide. Once the adjustments of therapy are completed, lifelong follow-up with IOP-dependent conversion to glaucoma and progression to blindness are modelled. Direct medical costs are assigned. The IOP-lowering effect of drugs is based on meta-analyses and applied by Monte Carlo simulation to a hypothetical cohort of patients with OH. The characteristics of the cohort, including the initial IOP distribution, are based on data of 1000 patients.

Results: The IOP decreased from 25,4 mm Hg (mean) to 16.7 (±0.017) mm Hg (strategy 1) and to 16.5 (±0.013) mm Hg (strategy 2). Costs per patient within 15 months of therapy were € 367 and € 469, respectively. Lifetime blindness and costs were 0.0334 years and € 3 514 (strategy 1) and 0.0318 years and € 4 397 (strategy 2). Incremental costs per year of vision saved for strategy (2) in comparison with strategy (1) amount to, given the uncertainties in the model, approximately € 537 000.

Conclusion: For saving 1 year of vision, high costs are needed when OH therapy is initiated with latanoprost compared to timolol, when the cost price of latanoprost remains high.

The approach towards treatment of ocular hypertension (OH) and primary open-angle glaucoma (POAG) has changed in recent years. This is partly because of the introduction of new hypotensive agents, but also because in several randomized trials, the benefit of treatment in patients with OH or in patients with early glaucoma was shown (Kass et al. 1989, 2002; Leske et al. 2003). Recent research in various European countries indicates an increased number of treated patients, more aggressive treatment and a change in therapy prescription pattern (De Natale et al. 2004; Azuara Blanco & Burr 2006; Knox et al. 2006; van der Valk et al. 2006). The aim of OH therapy is to prevent conversion to glaucoma, and so prevent visual deterioration and blindness. The rising costs in treatment can be justified if associated with an adequate reduction in visual disability.

Nowadays, hypotensive lipids are becoming the most commonly prescribed glaucoma drugs (Knox et al. 2006). Of these, latanoprost is often used as the drug of first choice, instead of so far used beta-blocker timolol. However, one may question whether it is cost-effective to change the initiation of therapy with timolol to latanoprost in patients with OH. The hypotensive effect of latanoprost is only slightly better than that of timolol (van der Valk et al. 2005), whereas its costs are much higher. In the clinical trials, an evaluation of the effectiveness of a drug is based on its IOP-lowering effect as monotherapy. However, patient management in the clinical practice often involves therapy switches and even combined therapies, to reach a target pressure for each patient. In the present study, two strategies of initiating therapy in patients with ocular hypertension, starting with latanoprost or starting with timolol, are modelled and evaluated for effects and costs. In these strategies, therapy adjustments to reach a target pressure are included. Considering that this is the aim of both strategies, one may expect small difference in the achieved IOP reduction. Which strategy is more cost-effective may depend on whether the achieved difference in IOP will translate, in the long-term, in a reduction of glaucoma development and blindness that will outweigh the cost expenditure for OH medication. This study is performed from a health-care perspective.

Methods

General design

Two strategies for the initiation of OH therapy are modelled and compared. In a previous study, we have assessed the impact of these strategies on IOP distribution and drug use (article recently accepted in J Glaucoma). In the present study, we have assigned direct medical costs of treatment and calculated the incremental costs and effects according to the standards of an incremental cost-effectiveness analysis (Gold et al. 1996; Drummond et al. 2005).

In strategy (1), therapy is started with timolol and in strategy (2) with latanoprost. First, the adjustments of therapy are modelled. Patients start therapy with a topical agent and return for evaluation. In case of adverse events, or unsatisfactory IOP decrease, therapy is adjusted and a new evaluation follows. Therapy switches are modelled at most four times. Therapies involve different types of monotherapy, combination therapy and laser trabeculoplasty. Four drugs are used, representing four major generic classes: timolol (beta-blockers), latanoprost (hypotensive lipids), brimonidine (α2-adrenergic agonists) and dorzolamide (topical carbonic anhydrase inhibitors). The modelled IOP-lowering effects are specific for each drug and for each drug combination. The IOP-lowering effect of the specific drugs is applied to a hypothetical cohort of patients with OH by using a Monte Carlo simulation. This cohort is characterized by the initial IOP and age distributions.

Once the therapy adjustments are completed, which covers at most 15 months of therapy, the achieved new IOP and drug distributions are determined per strategy. Subsequently, a lifelong follow-up and the disease progression over time are modelled, including an IOP-dependent conversion to glaucoma and progression to blindness. The time horizon of the model is defined by the life expectancy of patients. The parameter estimates are based on the literature. The age and initial IOP distributions are based on the data of 1000 ophthalmic patients. Glaucoma specialists were consulted to decide which medication should be used in case of a therapy switch.

The model consists of two parts. The first part is a Decision tree, which covers the first 15 months of therapy. We have run Monte Carlo simulations to analyse this decision tree. The patient characteristics, such as age and initial IOP, and the treatment effects are sampled from distributions by use of Monte Carlo simulation. The output of the decision tree, containing patient characteristics, new IOP after treatment and the use of medication, are further used in a Markov model, which covers the lifelong follow-up. A simplified impression of these models is given in Fig. 1.

The therapy adjustments (Decision model)

Therapy adjustments are modelled within the structure of a decision tree. In strategy ‘start with timolol’, patients start therapy with timolol unless contraindicated, in which case latanoprost is used. In strategy ‘start with latanoprost’, all patients start therapy with latanoprost. A switch from latanoprost to timolol and vice versa is possible. In case of a therapy switch first latanoprost and timolol are used, brimonidine and dorzolamide afterwards. Laser treatment follows if there are no more treatment options through medication. At all places in the model where timolol is applied, a proportion of patients with contraindication to this drug receive another medication.

Every 3 months after the prescription of medication, an evaluation is modelled to take place. For each simulated patient, a new IOP, achieved by the prescribed drug, and an occurrence of side-effects are modelled probabilistically. Criteria for treatment adjustments are as follows. (i) If serious side-effects occur, the drug is replaced with another one, regardless of the achieved IOP level. (ii) In the absence of side-effects, the treatment assignment depends on the achieved IOP level as follows: (a) If the achieved IOP is less than or equal to 21 mm Hg, medication is not changed. (b) If the achieved IOP is higher than 21 mm Hg and the IOP reduction is more than 20% of the original IOP level, another drug is added. (c) If the achieved IOP is higher than 21 mm Hg and the IOP reduction is <20% of the original IOP level, the currently used medication is substituted by another one. Laser treatment is assigned in the same way as a new drug, as an addition to the medication or as a replacement of it. Occurrence of side-effects is independent of age or IOP.

The lifelong follow-up (Markov model)

The lifelong follow-up is implemented within the structure of a Markov model. In a Markov model, a set of independent states is defined. Patients switch from one state to another at a regular interval (cycle) according to the transition probabilities. Health states were defined for all possible medication assignments following the therapy adjustments. Patients are allocated in these health states after the first 15 months of therapy. Subsequently during the follow-up, these patients can be reallocated in the following health states: ‘death’, ‘glaucoma’ and ‘blindness’ (see Fig. 1). The transition probabilities include age and sex-dependent survival probabilities, an IOP-dependent (per mm Hg) glaucoma incidence rate and a yearly probability of becoming blind from glaucoma. Transitions take place each 6 months. This cycle length reflects the common time between consecutive outpatient visits.

Cost assignment

The costs of outpatient visits, medication and laser therapy related to the first 15 months of treatment are added up for each simulated patient. The costs related to treatment during the follow-up are assigned to the health states of the Markov model. These costs reflect the treatment requirements related to 6 months of treatment of a particular state. For patients with OH, these long-term follow-up costs are calculated by the model, based on the medication used after the period of 15 months. The calculated costs depend on the cost prices of the specific type of medication. For patients with POAG, the costs represent the average costs of a patient with glaucoma in the Netherlands.

Model input

In Tables 1 and 2, an overview is given of the parameters used in the model. The hypothetical cohort consists of patients with OH (without glaucomatous changes of the optic nerve head or visual field loss) of at least 40 years old. Mean age is 64.5 years. Initial IOP within this population includes values from 22 to 35 mm Hg. Mean initial IOP value is 25.4 mm Hg. The age and IOP distribution for the population have been determined from the charts of 1000 patients visiting a nonreferral general ophthalmic practice in Maastricht (Medisch Centrum Maastricht Annadal) (van der Horst et al. 2003). Sex has been distributed equally. The survival probabilities are taken from general population data provided by the Dutch Central Bureau for Statistics (Centraal Bureau voor de Statistiek (CBS) (2003).

Table 1. Characteristics of the patient cohort simulated in the model.

Age categories per 10 years

Age distribution Proportion

40–49 years

0.13

50–59 years

0.22

60–69 years

0.27

70–79 years

0.24

80+ years

0.14

IOP (mm Hg)

Initial IOP distribution Proportion

22

0.17

23

0.16

24

0.15

25

0.13

26

0.11

27

0.08

28

0.05

29

0.04

30

0.03

31

0.02

32

0.02

33

0.02

34

0.01

35

0.01

Table 2. Parameters used in the simulation model.

Therapy intervention

Transition probabilities for the Decision tree

IOP reduction (%) (Mean ± SD)

The sources of all estimates are listed in the article, in the section “Methods”, subsection “Model input”.

Costs of medication and health care are based on several sources (Oostenbrink et al. 2001; Commissie Farmaceutische hulp (CFH) van het College voor zorgverzekeringen (Health Care Insurance Board) (2004); Peeters et al. 2008). The prices of medical drugs represent the prices in the Netherlands (incl. value added tax (VAT)). The unit prices related to outpatient visits and laser trabeculoplasty are as determined at the University Hospital Maastricht. This was done according to the micro-costing method (Gold et al. 1996). The average treatment costs of a patient with POAG were taken from a previous study of the authors (Peeters et al. 2008). All costs used in the model are in Euro, 2003 year’s values. The glaucoma costs were originally estimated in 2001 year’s value and were indexed with the retail price index. The costs are in the model discounted with 4% per year. The frequency of the health care use is modelled in accordance with the specialist’s opinion and the recommendations of the American Academy of Ophthalmology (American Academy of Ophthalmology 2000).

The parameters for the model were based on data from systematic reviews when available. For the choice of studies other than systematic reviews, the following issues were taken into account: date of publication, study design, type and size of a patient population, length of follow-up and outcome, e.g. IOP lowering, incidence of POAG, progression. When available, always several appropriate studies were selected. The IOP-lowering effects of the specific drugs are based on a systematic review (van der Valk et al. 2005). This review analyses trials with IOP level as a primary outcome and gives exact figures by estimating peak and trough IOP reductions of every included drug separately. In the practice, an IOP level achieved by a specific drug is individual and varies between patients. This results in a certain distribution of the achieved IOP levels, once a drug is administered to a large number of patients. Therefore, in the model, the IOP-lowering effects of the specific drugs are present in the form of a distribution. The distributions for the model were obtained from the studies collected for the earlier mentioned meta-analysis, by calculations in which the number of patients, change of IOP in terms of percentage with respect to baseline and its standard deviation were taken into account. The effects of extra IOP lowering in case of an ‘add on’ are based also on a systematic review (Webers et al. 2007).

Analysis

The following outcomes were determined per strategy: (i) mean IOP after 15 months, (ii) mean costs after 15 months, (iii) blindness within the lifetime follow-up and (iv) lifetime costs. Both costs and effects were discounted by 4%. In a subgroup analysis, years of blindness and the lifetime costs were determined for the 40, 50, 60 and 70 years old. In a one-way sensitivity analysis, contraindication for timolol, side-effects of medication, cost of glaucoma therapy and the cost price of latanoprost have been varied. A two-way sensitivity analysis was carried out for the IOP-lowering effect of timolol and latanoprost.

Results

Outcomes after 15 months of treatment: IOP and expected costs

The initial IOP decreased with approximately 34% from 25.4 mm Hg (sampled mean) to 16.7 (±0.017) mm Hg in strategy ‘start with timolol’ and to 16.5 (±0.013) mm Hg in strategy ‘start with latanoprost’. For patients with an initial IOP of 30 mm Hg, the IOP decrease was about 69%, in both strategies to about 17.5 mm Hg. The calculated costs per patient 15 months after initiation of therapy were € 367 in strategy ‘start with timolol’ and € 469 in strategy ‘start with latanoprost’. For patients with an initial IOP of 30 mm Hg, these costs were € 441 and € 496, respectively.

Outcomes within a lifetime follow-up: blindness caused by glaucoma and expected costs

In Table 3, the years of blindness caused by glaucoma and the expected lifetime costs per patient are given. The computed time with blindness per patient is 0.0923 (33.7 days) years in strategy ‘start with timolol’ and 0.0870 (31.8 days) years in strategy ‘start with latanoprost’ (without discount). This corresponds to approximately 1 month of blindness per patient, over a period of 18.7 years, which is the computed mean life expectancy in the cohort. The difference between the strategies is 2 days of blindness, a difference of about 6%, in favour of strategy ‘start with latanoprost’. For reference, a ‘no treatment’ simulation (0% IOP reduction through medication and laser) showed 0.492 expected years of blindness per patient, which corresponds to approximately 6 months.

Table 3. Results of the cost-effectiveness analysis of two different strategies for initiating treatment of ocular hypertension.

Strategy 1 ‘start with timolol’

Strategy 2 ‘start with latanoprost’

The values represent the expected lifetime costs and effects for a treated patient with OH (an average computed from 5 Monte Carlo simulations of size 20 000 each).

IOP = intraocular pressure.

* Standard deviation for the average, computed from 5 Monte Carlo simulations of size 20 000 each.

† Mean expected time spent in blindness per person within 18.7 years of life expectancy.

^{‡} The incremental cost-effectiveness ratio shows extra costs per year of vision saved when starting therapy with latanoprost compared to starting with timolol.

4% discount (costs and effects)

Initial IOP (sampled from distribution)

Costs (€)

3 514 ± 13.670*

4 397 ± 15.014

Years of blindness^{†}

0.0334 ± 0.0004

0.0318 ± 0.0004

Incremental C/E^{‡}

–

536 852

Initial IOP (fixed = 25 mm Hg)

Costs (€)

3 396 ± 4.056

4 304 ± 4.378

Years of blindness

0.0361 ± 0.0004

0.0344 ± 0.0003

Incremental C/E

–

547 276

Initial IOP (fixed = 30 mm Hg)

Costs (€)

4 215 ± 6.394

4 583 ± 10.708

Years of blindness

0.04301 ± 0.0002

0.04296 ± 0.0004

Incremental C/E

–

7 068 037

0% Discount (costs and effects)

Initial IOP (sampled from distribution)

Costs (€)

5 456 ± 23.929

6 794 ± 27.790

Years of blindness

0.0923 ± 0.0012

0.0870 ± 0.0010

Incremental C/E

–

253 011

Initial IOP (fixed = 25 mm Hg)

Costs (€)

5 269 ± 18.760

6 628 ± 17.188

Years of blindness

0.0977 ± 0.0011

0.0924 ± 0.0011

Incremental C/E

–

254 150

Initial IOP (fixed = 30 mm Hg)

Costs (€)

6 556 ± 18.142

7 099 ± 13.767

Years of blindness

0.1175 ± 0.0008

0.1173 ± 0.0012

Incremental C/E

–

2 009 703

The lifetime costs include the costs of the first 15 months of therapy as well. The costs are lower for strategy ‘start with timolol’ than for strategy ‘start with latanoprost’, namely € 5 456 against € 6 794 per patient (€ 3 514 against € 4 397 when discounted). For patients with an initial IOP of 30 mm Hg, the difference in costs is smaller, namely € 6 556 against € 7 099 (€ 4 215 against € 4 583 when discounted). The outcomes of the subgroup analysis are given in Table 4. The difference in costs and effects between the strategies becomes smaller with an increasing age.

Table 4. Results of the subgroup and sensitivity analyses.

4% discount (costs and effects)

Initial IOP = 25 mm Hg

Initial IOP = 30 mm Hg

Strategy 1 ‘Start with timolol’

Strategy 2 ‘Start with latanoprost’

Strategy 1 ‘Start with timolol’

Strategy 2 ‘Start with latanoprost’

The values represent the expected lifetime costs and effects for a treated ocular hypertension (OH) patient (an average computed from 5 Monte Carlo simulations of size 20 000 each).

IOP = intraocular pressure.

^{*} Standard deviation for the average, computed from 5 Monte Carlo simulations of size 20 000 each.

^{†} Mean expected time spent in blindness per person within 18.7 years of life expectancy.

^{‡} The incremental cost-effectiveness ratio shows extra costs per year of vision saved when starting therapy with latanoprost compared to starting with timolol.

^{§} Higher costs, less effect in strategy ‘start with latanoprost’.

^{¶} Lower costs, more effect in strategy ‘start with latanoprost’.

Subgroup analysis

Age: 40 years

Costs (€)

5 414 ± 1.364*

6 797 ± 2.222

6 712 ± 7.813

7 261 ± 7.620

Years of blindness^{†}

0.1218 ± 0.0006

0.1156 ± 0.0007

0.1453 ± 0.0004

0.1447 ± 0.0003

Incremental C/E^{‡}

–

222 632

–

912 363

Age: 50 years

Costs (€)

4 716 ± 2.157

5 942 ± 4.177

5 866 ± 9.311

6 356 ± 14.177

Years of blindness

0.0749 ± 0.0005

0.0712 ± 0.0004

0.0892 ± 0.0002

0.0893 ± 0.0004

Incremental C/E

–

334 809

–

−8 786 078^{§}

Age: 60 years

Costs (€)

3 848 ± 5.543

4 874 ± 4.547

4 804 ± 9.425

5 219 ± 5.903

Years of blindness

0.0380 ± 0.0002

0.0361 ± 0.0003

0.0454 ± 0.0003

0.0452 ± 0.0003

Incremental C/E

–

552 352

–

3 320 995

Age: 70 years

Costs (€)

2 885 ± 4.400

3 668 ± 1.688

3 607 ± 5.437

3 929 ± 1.861

Years of blindness

0.0145 ± 0.0001

0.0138 ± 0.0001

0.0175 ± 0.0001

0.0175 ± 0.0000

Incremental C/E

–

1 057 025

–

18 182 688

Sensitivity analysis

Cost glaucoma therapy = 350 €/year

Costs (€)

3 347 ± 8.779

4 253 ± 7.471

4165 ± 5.290

4530 ± 9.849

Years of blindness

0.0361 ± 0.0002

0.0343 ± 0.0003

0.0428 ± 0.0006

0.0427 ± 0.0005

Incremental C/E

–

506 138

–

3 453 151

Cost glaucoma therapy = 600 €/year

Costs (€)

3 446 ± 10.226

4 340 ± 8.563

4303 ± 2.503

4671 ± 8.284

Years of blindness

0.0353 ± 0.0003

0.0336 ± 0.0002

0.0429 ± 0.0002

0.0429 ± 0.0002

Incremental C/E

–

528 171

–

5 226 625

Cost latanoprost 0.005% = 10 €/month

Costs (€)

2 964 ± 9.027

3 326 ± 11.926

3 666 ± 6.358

3 756 ± 6.301

Years of blindness

0.0360 ± 0.0003

0.0339 ± 0.0004

0.0425 ± 0.0001

0.0425 ± 0.0001

Incremental C/E

–

174 394

–

−2 271 388^{§}

Cost latanoprost 0.005% = 6 €/month

Costs (€)

2 751 ± 5.945

2 828 ± 5.806

3 368 ± 3.385

3 307 ± 4.260

Years of blindness

0.0359 ± 0.0003

0.0341 ± 0.0003

0.0428 ± 0.0003

0.0426 ± 0.0002

incremental C/E

–

44 095

−

−324 031^{¶}

IOP reduction: timolol 2% less and latanoprost 2% more

Costs (€)

3 467 ± 7.314

4 344 ± 13.292

4 324 ± 12.942

4 626 ± 9.037

Years of blindness

0.0360 ± 0.0003

0.0313 ± 0.0002

0.0423 ± 0.0003

0.0414 ± 0.0002

Incremental C/E

186 068

−

317 103

Sensitivity analyses

The outcomes of the sensitivity analyses are partially shown in Table 4. The variation of contraindication for timolol and side-effects of medication within the given ranges (see Table 2) had hardly any influence on the outcomes (data not shown). Lowering the costs of glaucoma therapy from € 450/year to € 350/year, or increasing it to € 600/year, gives only a slight decrease or increase of the lifetime costs in both strategies, about 0.8–1.5%. Analyses with a reduced price of latanoprost, from € 18.7/month to € 10/month and € 6/month, show a marked reduction and vanishing of the cost difference between the strategies. With a reduced price of latanoprost, for patients with an initial IOP of 30 mm Hg, in strategy ‘start with latanoprost’, the lifetime costs are lower than in strategy ‘start with timolol’. The IOP-lowering effect of timolol and latanoprost varied within the range of a 95% confidence interval around the given value, in which the lower and upper bounds for both drugs represent about 2% more or less IOP reduction. A two-way sensitivity analysis with 2% more IOP reduction through latanoprost and 2% less IOP reduction through timolol shows better blindness prevention for a cohort with an initial IOP 25 mm Hg in strategy ‘start with latanoprost’ and with an initial IOP 30 mm Hg in both strategies.

Cost-effectiveness

Strategy ‘start with timolol’ is cheaper than strategy ‘start with latanoprost’ but yields more years of blindness. The given incremental cost-effectiveness ratios represent extra costs per year of vision saved when initiating therapy with latanoprost compared to initiating with timolol. If therapy is initiated with latanoprost, the computed extra costs per year of vision saved in comparison to initiation with timolol amount to € 537 000. For patients with an initial IOP of 30 mm Hg, the extra costs per year of vision saved become much higher, namely € 7 068 000.

The subgroup analyses per 10-years age categories show that extra costs per year of vision saved are increasing with age, from € 222 632 for the 40-years-old to € 1 057 025 for the 70-years-old. When the price of latanoprost is reduced, the cost-effectiveness ratio becomes smaller and drops even to € 44 095, if the price is € 6/month. For patients with an IOP of 30 mm Hg strategy ‘start with latanoprost’ then becomes cost saving.

Discussion

The results of the study show that initiating therapy in patients with OH with latanoprost brings along extremely high incremental costs, € 537 000 extra for saving 1 year of vision when compared to initiating therapy with timolol. If treatment with latanoprost only and timolol only would be compared, there would undoubtedly be a significant difference in treatment effects between those two drugs. However, a comparison of clinically based strategies, where different first therapy choice is followed by similar steps for therapy adjustments to achieve a target pressure, shows minimal difference in the achieved IOP and blindness through glaucoma between the two analysed strategies. The cost difference, on the other hand, is large. As the IOP reduction is similar in both strategies, with a consequence that only slight differences occur in glaucoma development and blindness, the corresponding glaucoma costs will be comparable in both strategies as well. The different OH therapy costs are thus responsible for the main cost difference, as confirmed by the results of the sensitivity analysis.

Many patients with OH can reach their target pressure with timolol as well as with latanoprost. Patients whose target pressure cannot be easily reached and controlled with monotherapy need more therapy adjustments including combination therapy or laser treatment. This is, however, the case in both strategies, and the costs for these patients will thus be comparable. Long-term use of monotherapy with latanoprost is a largest contributor to the higher cost expenditure in the strategy ‘start with latanoprost’. When therapy is at once started with latanoprost, the less expensive alternative, namely to start with timolol, is not considered. The latter strategy, however, has a sufficient effect for patients who reach the target pressure with both drugs. This is the underlying reason why the strategy ‘start with latanoprost’ is not preferable from the cost-effectiveness point of view, in comparison to the strategy ‘start with timolol’. The modelled simulations show that even for young patients with OH, who have a longest life expectancy and thus most time to develop glaucoma and blindness, it is not cost-effective to initiate therapy with latanoprost.

A simulation model was used to analyse the two treatment strategies. Modelling techniques are increasingly applied in the field of ophthalmology for a comparison of different treatment alternatives (Tuulonen et al. 2009). To avoid modelling or programming errors, a process known as “debugging” the tree was applied (Detsky et al. 1997). As we were able to incorporate data of recent, adequately performed, randomized clinical trials, we believe that the results of our modelled strategies validly represent the current practice. By applying a Monte Carlo simulation, which is run 100.000 times, a reliable estimate of the clinical result of a strategy given the underlying variation is obtained (see the calculated standard deviations given in Tables 3 and 4). Extensive sensitivity analyses were performed to establish the impact of a variation of the uncertain values in the model on the outcomes and the study conclusions. It showed that these conclusions were robust. Only a reduction of the cost of latanoprost leads to a favourable cost-effectiveness ratio for the strategy ‘start with latanoprost’.

The model showed approximately 34% IOP reduction through medication and laser. In the Ocular Hypertension Treatment Study (OHTS) (Kass et al. 2002), the design of which corresponds well with the modelled therapy approach, this was 22.5 ± 9.9%. However, the target pressure criterion differs, this being 24 mm Hg in the OHTS study and 21 mm Hg in the model. Laser treatment was not applied in the OHTS study participants. Furthermore, in the OHTS study, the average IOP across follow-up visits is given, whereas the model gives a mean IOP value after the therapy adjustments. Lastly, the treatment choice was limited, when the OHTS study participants started their therapy. The yearly estimate for blindness might be slightly overestimated in the model, because no difference is made between beginning and advanced glaucoma. The age and IOP distributions of the population are verified to be consistent with the literature (Koch 1989; Schappert 1995; Ellwein & Urato 2002). IOP values above 35 mm Hg were not addressed in the model, because such values are very likely to be already associated with glaucoma. The IOP reduction through specific medication is based on the best available data, namely on meta-analyses of randomized controlled trials (van der Valk et al. 2005; Webers et al. 2007). Recently, another systematic review has been published including medical interventions for POAG and OH (Vass et al. 2007). However, as this study does not show concrete drug-related IOP levels, we could not use its outcomes for our model.

There have been some studies published, addressing the cost-effectiveness of various treatments with glaucoma drugs (Bernard et al. 2003; Noecker & Walt 2006). However, no sensible comparison of our results with these studies can be made, because they include treatment of glaucoma. The drug persistency in glaucoma treatment is different than in ocular hypertension. Moreover, these studies use IOP as an outcome measure. In one epidemiological study, by using a questionnaire in 1 513 patients, a drug use in ocular hypertension in Belgium was established (Walckiers & Sartor 1996). This was done in 1992, and 96% of patients were then using a topical beta-blocker. In 5 years, 77% of patients used timolol alone. In the model, about 66% of patients maintain the originally prescribed timolol after 15 months of treatment. Another study in France found that 82.6% of the patients with OH used monotherapy, mostly a beta-blocker (Denis et al. 2004). In the model, approximately 90% of patients use monotherapy medication. In the same study, the annual treatment costs for patients with OH were € 275, € 376 and € 476 for patients with 0, 1 and 2 treatment changes. In the model, the mean costs calculated for 12 months of treatment were approximately € 295 for strategy ‘start with timolol’ and € 375 for strategy ‘start with latanoprost’. The proportion of patients with no treatment change in the model was 68% in strategy ‘start with timolol’ and 77.1% in strategy ‘start with latanoprost’. The average treatment costs of a patient with POAG used in the model match with the outcomes of other glaucoma cost studies (Kobelt et al. 1998; Kobelt & Jonsson 1999; Iskedjian et al. 2003; Traverso et al. 2005; Lee et al. 2006; Lindblom et al. 2006).

The analysis was performed from a health-care perspective. The large difference in long-term costs between the strategies is mainly because of the cost of medication for OH treatment. Even if the analysis would have been done from a societal perspective, we expect the conclusions to be the same, as the process of care, e.g. number of outpatient visits and clinical outcomes are so similar. The costs because of blindness in connection with the usage of disability facilities in the Netherlands could not be retrieved at that moment. Such costs depend on the care and support given and might vary per country, and even within one country. In one study, which took the societal perspective, the cost of blindness because of macular degeneration was 10 134 (2159–26 847) Euro (converted from English pounds; 2003 year’s value) in the first year and slightly less thereafter (Meads & Hyde 2003). It is highly unlikely that the costs per year of glaucoma blindness are more than the incremental costs of 537 000 Euro. In the model, a 4% discount rate is used, while internationally a 3% rate is recommended (Gold et al. 1996). This was done because of the guidelines in the Netherlands, however, with a 3% discount rate, no substantial changes occurred.

In the present study, cost-effectiveness is based on clinical outcome in the form of years of vision saved. It is not a cost-utility analysis. The latter would imply to take into account patient-perceived benefits and harms. However, side-effects have been taken into account because therapy is changed in the model in occurrence of an adverse event. This implies that, because the adverse events are short-lived, patients are bothered by these adverse events for only a short period of time, at most 3 months between the outpatient visits. Moreover, in a recent study in 3 333 patients using glaucoma drugs, we have reported that there is no difference in chance of changing glaucoma drugs because of adverse events or a difference in patient satisfaction with these drugs. The number and frequency of self-reported ocular adverse events were statistically significantly different between the drugs, but this was because of more reported adverse events in the small proportion of patients (0.7 percent) using three or more drugs (Beckers et al. 2008).

The model approaches clinical practice closely, the recent estimates of treatment efficacy are used, and the results are in general consistent with existing reports. The difference in outcome is not of clinical relevance, not even when this is defined as 10 days over a life time period of more than 18 years. This implies that we can assume that there is no difference in outcome and that the difference in cost determines the best strategy from a cost-effectiveness approach.

In conclusion, initiating therapy in patients with OH with latanoprost brings along extremely high incremental costs to prevent blindness when compared to initiating therapy with timolol. Given the current cost price of latanoprost, saving 1 year of vision would incur € 537 000 of expenditures. These incremental costs are even higher for patients with a high initial IOP and are increasing with age.

Acknowledgements

Financial support has been provided by the Dutch Health Care Insurance Council, Diemen, the Netherlands. No author, nor the financer, has any commercial (proprietary or financial) interest in any drugs, devices or techniques discussed in this article.