Modelling the lifetime economic consequences of glaucoma in France

Abstract

Objective: To estimate the lifetime economic consequences of glaucoma in France.

Methods: A Markov model estimated the average discounted outcome and cost of glaucoma treatment over a patient's lifetime. Clinical states were defined as first- to fourth-line drugs, no treatment, laser therapy, surgery, blindness and death. After each failure (always after the fourth-line drug) patients could receive either laser treatment or surgery followed by no treatment, or a new treatment. A societal perspective was adopted. Sensitivity analyses were performed.

Results: Discounted medical costs were €7,322 for ocular hypertension treatment (OHT) and €8,488 for a glaucoma patient. Social costs of OHT and glaucoma patients exceeded medical costs. First-line use of the most effective drug would reduce medical and social costs. Societal willingness to pay for the vision benefit would equal the medical costs. Treatment initiated with the most effective drug is a cost saving strategy.

Conclusions: Public health decisions in glaucoma treatment should take a broad economic view embracing the lifetime duration of the disease. There is still a place both within and outside the healthcare system for therapeutic innovations with important economic consequences that bring high added value to patients.

Key words::

• economics
• glaucoma
• life-long treatment
• model

Introduction

Primary open angle glaucoma (POAG) is characterised by a progressive alteration of the optic nerve. Glaucoma is a chronic disease that alters the visual field over a period of years and for some patients leads to blindness. Together with cataract and age-related macular degeneration (ARMD), POAG is one of the three leading causes of blindness in developed countries¹.

Several surveys have been conducted to evaluate the prevalence of ocular hypertension (OH) and glaucoma. Many patients with ocular hypertension will develop POAG². Estimates vary between 1% and 5% in people over 40 and increases with age^3–11. An additional risk factor is high intraocular pressure (IOP)^12–14 associated with POAG, caused by aqueous outflow obstruction. Other factors such as a family history of POAG, diabetes mellitus, race, hypertension and hypotension, severe myopia, and a past history of severe ocular trauma have also been described^15–18.

The risk of bilateral blindness 20 years after glaucoma diagnosis was estimated, retrospectively, in Rochester (Minnesota, USA) as 9%¹⁹. Before this fate ensues, however, glaucoma and visual field loss are already associated with disability, diminished reading ability and television viewing, and a higher risk of accidental falls²⁰. It was recently shown²¹ that untreated glaucoma in a developing country increased more rapidly than was previously estimated. The annual progression rate of visual field defects (VFDs) has been estimated^22,23 at 1.5–6.4 %, depending on the scoring method. Apart from visual consequences, an analysis²⁴ based upon the Framingham cohort showed that patients with high intraocular pressure (IOP) values faced a higher risk of death than patients with normal values.

A previous Ocular Hypertension Treatment Study²⁵ determined that topical ocular hypotensive medication delays or prevents the onset of POAG. In the Advanced Glaucoma Intervention Study²⁶, mean visual field defect scores were close to zero for eyes with intraocular pressures below 18 mmHg at every visit during a 6-year follow-up period, whereas eyes with IOP values above 18 mmHg during less than half the period demonstrated visual field defect problems.

Intraocular pressure is measured by tonometry, a simple out-patient assessment performed by ophthalmologists. Control of IOP is a critical aspect of glaucoma treatment. Several strategies have been developed to attain this objective: drugs, laser therapy and surgery. Amongst drugs, several mechanisms of action have been developed: agents inhibiting the production of aqueous humour (beta-blockers and carbonic anhydrase inhibitors – CAIs), agents increasing uveoscleral absorption (prostaglandin analogues) or trabecular absorption (cholinergic agonist), and agents with both mechanisms of action (adrenergic agonists). The choice of glaucoma medication depends on patient characteristics, clinical history, and comorbidities^27,28 and takes into account probable effectiveness and the adverse effect profile²⁹ (including comfort during administration), and the daily frequency of administration.

Economic studies have already been performed in the field of glaucoma and ocular hypertension treatment^30–35. Kobelt et al.³⁶ estimated in the late 1990s treatment costs during the two first years after diagnosis at $2,160 in Sweden and $2,109 in the USA. Treatment costs were positively related to IOP values at diagnosis. Few studies have attempted to model the costs and consequences of glaucoma and OHT treatment³⁷. The model used by Kobelt et al.³⁷ suggests that the more effective IOP control achieved by prostaglandin analogues could delay the use of rescue treatments (laser and surgery). The hypothesis was subsequently confirmed^38,39.

The full costs of preventative strategies should be taken into account before deciding on their implementation⁴⁰. In the case of chronic diseases, most guidelines recommend that the total duration of a disease should be taken as the benchmark⁴¹. The first step in the present research programme⁴² was to create a Markov model that would simulate the lifetime consequences of glaucoma. The study reported here is based on an earlier medico-economic model and estimated the lifetime economic consequences of glaucoma in France⁴².

Patients and methods

This observational survey was conducted according to French law. The protocol was submitted and approved by the Comité consultatif sur le traitement de l'information en matière de recherche dans le domaine de la santé (ministère de l'enseignement supérieur et de la recherche), the Commission nationale de l'informatique et des libertés, and the Conseil national de l'ordre des médecins. The observational survey was used to determine the probability of disease progression⁴³ and the predictive factor of costs³⁰ used in the model.

Model characteristics

The decision analytic model used here has been described elsewhere in relation to the clinical outcome results⁴² and the validity of the model was also discussed elsewhere⁴⁴. It was constructed with Data Pro release 4.2 software obtained from Tree Age Software Inc. (Williamston, Massachusetts, USA). Patients entered into the model had either OHT or POAG at the time of diagnosis. Patients with secondary glaucoma (i.e. congenital, inflammatory, neovascular and, after cataract surgery, narrow- or closed-angle glaucoma) were not considered in this model. Their average age was 57 years⁴³.

Clinical states were defined by treatment as first- to fourth-line drugs (Figure 1), no treatment, laser therapy and surgery. Tracker variables and specific Markov states were created to count the number of VFDs. The remaining two categorical clinical states were blindness and death. Stochastic variables used in the Monte Carlo 2nd order simulations were age, gender, visual acuity (VA) at diagnosis, number of VFDs at diagnosis, and loss of VA consequent to disease progression. Transition probabilities and costs were based on French observational data.

Time horizon

A cycle length of 1 month was considered appropriate for the model as it made clinical sense (the expiry date of glaucoma bottles has been fixed by the European Medicines Agency⁴⁵ at 28 days for the majority of the glaucoma drugs and patients are therefore advised not to use a bottle longer than this duration) with regard to a patient's glaucoma status and it was appropriate for states that were single timepoint events, such as laser treatment or surgical procedures. Lastly, a 1-month cycle was well suited to economic analyses.

The flow of patients through Markov states was studied in order to simulate the lifetime evolution of disease in a cohort of patients from diagnosis up to 100 years of age, i.e. life expectancy after 100 years was not taken into account.

Model parameters

Estimates for the model parameters were obtained from the literature and a dedicated French epidemiological survey⁴³. Probabilities of blindness as a function of age, due to all causes other than glaucoma, were found in the Handicap Incapacity and Dependency (HID) surveys^46,47. A linear regression by age group estimated incidence rates from prevalence rates and the probability of being blind according to age. Death transition probabilities were derived from general demographic data and life tables published by the Institut National de la Statistique et des Etudes Economiques (INSEE)⁴⁸.

Other probabilities were estimated from data collected by the French glaucoma survey, using survival analysis with an exponential model to estimate monthly probabilities of the following: continuation of the same drug line^30,43, laser treatment^30,43, surgery^30,43, discontinued topical treatment after laser and surgery^30,43 and increased VFDs during maintenance of a drug line⁴³.

Although glaucoma severity and VA are not strongly associated, patients from the French survey⁴³ with a VFD who lost an average of –4dB on their mean defect (as measured by Humphrey perimetry) also lost 0.123 in VA, on average. Utility was derived from Brown functions⁴⁹ and Sharma⁵⁰.

Economic analysis

Each health state was associated with a standard cost representing the Sickness Fund's perspective (reimbursed portion of medical direct costs) added to the payer's perspective (social allowances), amounting to a sort of ‘semi-societal’ perspective. We decided not to undertake a full societal perspective analysis since the additional costs related to blindness, incorporating loss of productivity, are subject to a methodological debate⁵¹.We therefore decided to adopt a conservative approach.

Total costs and all outcomes of each state were determined for one cycle (one month) each and discounted at 3% annually⁵². Medical consumption items included the following: visits to specialists (ophthalmologists), drugs, and procedures associated with each type of treatment. Social costs included transfer to institutions and social allowances for visual impairment.

The French glaucoma survey which took place in 2001 was used to document these resources and costs for France³⁰.

Direct costs:

The cost of generic drug timolol was used for all other generic drugs, which was appropriate in France as the generic glaucoma market was <1% of turnover. For the sensitivity analyses, it was arbitrarily fixed (conservative approach) that the costs of a neuroprotective agent and associated medical resources would be similar to those of the current market case-mix.

Monthly institutionalisation costs were estimated from data collected in 1999 by the French Ministry of Health (Direction Générale de l'Action Sociale)⁵³. Total French institutional costs, including hostel (57%) and medical costs (47%), were estimated at €2,500/month (average).

Social allowances:

These allowances were paid mostly by public institutions such as Sickness Funds, Family Allowance Funds, local communities and charities. Data to quantify these allowances were extracted from the HID surveys performed in France by INSEE⁵⁴. The databases identified three groups: blind, low vision (LV) and control (all others). The net costs attributable to visual disorders were estimated by subtracting the costs of individuals without visual disorders from those of blind and low vision persons. A linear regression was performed to adjust social allowances for gender, age and number of handicaps^55,56. The probability of living in an institution when aged more than 55 years was 6.13% for blind people, 5.91% for LV persons, and 1.14% for the control group. An individual had 5.4 times more chance of living in an institution if blind, and 5.2 times if LV⁵⁷.

Institutionalised blind people received a mean extra amount of €112.64 per month, as compared to €484.33 for blind and control persons living at home. Amounts for LV subjects were €19.22 and €201.52 per month, respectively.

In the absence of a threshold fixed by the French authorities, we chose arbitrarily €50,000/QALY as the monetary value of maintaining vision, an empirical value supported by a previous decision.

Sensitivity analysis:

The following variables were included in the sensitivity analysis: patients diagnosed with OHT or open-angle glaucoma at diagnosis, treatment line duration (lines 1 and 2) increased by 50%, treatment started with a glaucoma drug not amongst the most efficacious. Lastly, the effect of adding a drug that would directly diminish the likelihood of developing a new VFD, without any IOP change, was estimated. This will be described as a ‘neuroprotective’ effect in this paper.

Results

The model predicted a life expectancy of almost 22.5 years, with death at an average age of 81 years. Clinical outcomes have already been reported⁴². Table 1 describes medical unit costs.

Discounted costs (Table 2) were €7,322 for OHT patients and €8,488 for glaucoma patients. Drugs represented the major cost, followed by surgery, examinations, visits and laser treatment. Discounted social costs were much lower for OHT than glaucoma patients (€7,276 vs. €23,103). The social costs of OHT patients were similar to the medical costs of OHT or glaucoma patients. Institutionalisation costs represented one-third of social costs for OHT patients and one-half for glaucoma patients. Glaucoma patients lost 5.9 months of full quality of life compared with OHT patients. Based on a threshold value of €50,000/QALY, the discounted social value-for-money of early diagnosis (i.e. at the time of OHT) was estimated at about €56,600.

A drug that would increase the duration of first-line treatment by 50% would decrease the discounted medical cost by €189 for an OHT patient and by €241 for a glaucoma patient. Savings would be mainly on drugs and surgery. Fewer patients would be institutionalised and social allowances would be limited to visual impairment. Social savings amounted to about 10% of medical costs for a glaucoma patient. The monetary value associated with maintained vision almost equalled the total medical costs of OHT and exceeded those of glaucoma. A drug that would increase the duration of second-line treatment by 50% would still produce savings, but to a lesser degree. Savings would be similar for OHT and glaucoma patients. Nonetheless, the value for money from preserving vision still remained important, i.e. €2,083 for an OHT patient and €2,500 for a glaucoma patient.

By contrast, treatment strategies (surgery, laser therapy, drug monotherapy, or drug combinations) beginning with a first-line drug giving less IOP control than a second-line treatment strategy, would occasion serious negative economic consequences. It would increase total discounted medical costs (OHT €160 and glaucoma €68) by increasing drug and surgery costs. Institutionalisation costs and social allowances would also increase, especially for glaucoma patients. Lastly, the social discounted value relating to maintained vision would become negative. Further to the additional resources used by the medical and social sectors, the cost to a patient in terms of lost QALYs (economically stated as ‘willingness to pay to avoid the loss’) was equivalent to €5,417 (72.4% of all medical costs) for an OHT patient and €2,083 (24.4% of all medical costs) for a glaucoma patient.

A new chemical entity that would reduce the probability of a new VFD by 25%, for a given level of IOP control, could be assumed to provide a ‘neuroprotective effect’. The micro-economic consequences of such a compound would be dramatic. Medical costs would not change very much, but social costs would fall sharply (–€1,176 for OHT patients and –€2,503 for glaucoma patients). More importantly, the discounted value to the patient in terms of QALYs gained (economically stated as ‘willingness to pay for the benefit’) would be €26,250 for an OHT patient and €35,000 for a glaucoma patient.

Discussion

A Markov model was developed to estimate health outcomes and costs associated with the long-term follow-up of glaucoma patients in France. Mean, discounted cumulative levels of utility (QALY) and both medical and social costs of glaucoma were calculated. The analysis, therefore, permitted a comparison between investments (medical costs) and outcomes (institutionalisation, visual impairment social allowances and societal ‘willingness to pay’ to maintain life-long vision), since both dimensions were expressed in monetary terms.

Certain conclusions can be drawn from this model. The discounted, average medical cost of a patient treated for OHT or glaucoma does not rank amongst the most expensive of chronic diseases. Nonetheless, social costs equalled medical costs for OHT patients and were three- to four-times higher for glaucoma patients, demonstrating that the social consequences of visual impairment are serious and lie mainly outside the healthcare sector: the IOP lowering effect and its vision preserving consequences are the major economic factors in glaucoma treatment. A drug providing the best available IOP control is an effective alternative and should be prescribed first-line, provided its incremental cost-effectiveness ratio is acceptable. As a direct result, drug costs being equal, medical and social costs would fall moderately, but the societal ‘willingness to pay’ for the outcome benefit would counter-balance the medical costs. A neuroprotective drug effect would impact dramatically upon the economic consequences of glaucoma, mostly affecting the social costs of blindness.

We estimated that surgery and laser use would decrease when using a more efficacious drug⁴². The economic consequences appeared to be weak. This is explained by the effect of discounting since rescue treatment like surgery and laser are used years after drugs.

The results of the authors' model should be compared with information available in existing published studies. To the authors' knowledge, no analysis has been carried out on the long-term outcome and costs of glaucoma. However, a previous the outcome article by the same group validated the main assumptions adopted here against external sources⁴². Reduced use of surgery and the associated savings has already have been reported, based upon patient data³⁹.

Social costs were highly underestimated. Assistance time, time spent caring for visual impairment, loss of revenue, medical devices, house modifications, and home moves were not evaluated by the model. The omitted costs can represent up to 90% of social costs⁵¹. Total medical costs were also underestimated as the costs of depression^58,59, anxiety, falls and hip fracture⁶⁰, death^61–65 related to visual impairment, and other co-morbidities related to visual impairment are unavailable.

The finding that savings, made by increasing the duration of second-line treatment, were similar for OHT and glaucoma patients stems from the greater risk of new VFDs with second-line treatment. Thus, IOP control during first-line treatment is more important than during subsequent drug treatment lines. This supports the policy of prescribing drugs that achieve the greatest control IOP as first-line treatment.

The neuroprotective effect did not impact upon medical costs, which might appear surprising. However, in our model, a neuroprotective drug was defined as ‘a new chemical entity that would reduce the probability of developing a new VFD by 25%, for the same level of IOP control’. Any lack of IOP control would incur increased costs. As, according to our definition, the IOP control provided by a hypothetical new chemical entity was to be equivalent, no major effect on medical costs could be expected.

A threshold of €50,000/QALY was used to define a cost-effective alternative treatment aimed at preserving vision. This sum was derived from an empirical decision made by health authorities when allocating resources for innovative medical products. In the field of eye diseases, the French authorities accepted verteporfin, without limiting its reimbursement, when a published cost/QALY value was much higher than our threshold^66,67, although the latter citations do not in fact report the French cost-effectiveness ratio. Similar thresholds (US$50,000/QALY) were used for ranibizumab in the USA⁶⁸ and for ARMD screening in the UK (£30,000/QALY)⁶⁹.

Public health decisions should take into account the natural time horizon of diseases, e.g., life-long for OHT and glaucoma patients. As it is not possible to organise data collection over such a long period, a model must be used. The present model allowed us to estimate the costs and consequences of treating patients with the therapeutic arsenal existing in 2005, and to assess the economic consequences of anticipated new treatments. The limitations of models are well-known (external validity, stable environment needed, association functions between variables to enable predictions, etc.), and readers should keep these in mind when interpreting the results. Lastly, longitudinal data are required to refine the model, with modifications to incorporate new therapeutic practice.

Conclusion

The discounted medical costs of an OHT patient were about €7,300 and increased to about €8,500 for a glaucoma patient. The social costs of OHT and glaucoma patients exceeded the medical costs. Use of the most effective first-line drug would reduce medical costs. Value for money in terms of vision benefit, from the patient's standpoint, equalled discounted medical costs. An agent with a direct neuroprotective effect would attract very high societal ‘willingness to pay’.

Figure 1. Lifelong model description.

Table 1. Costs of health states (€ per month).

Health state	Cost
Line 1 (€)	25.60
Line 2 (€)	38.98
Line 3/Line 4 (€)	50.31
No treatment (€)	16.48
Laser (€)	148.18
Surgery (€)	2102.73
Allowance for blindness (€)	329
Allowance for low vision (€)	52
Institutionalisation rate for blindness (%)	6.13
Institutionalisation rate for low vision (%)	5.91
Monthly cost of institutionalisation (€)	2,500

Line 1: Treatment started for newly diagnosed patients. Line 2: Treatment initiated after first-line treatment failure. Line 3: Treatment initiated after second-line treatment failure. Line 4: Treatment initiated after third-line treatment failure. Treatment line description is provided in Denis et al.³⁰. Lines 1–4: ‘No treatment’ costs were estimated from Denis et al.³⁰, allowances for blindness and low vision, and the probability of institutionalisation were obtained from the HID survey^55–57. The monthly cost of institutionalisation is a tariff published by the Direction Générale de l'Action Sociale⁵³.

Table 2. Lifetime micro-economics of glaucoma treatment in France in euros.

Treatment costs (€)	Base case scenario		Line 1 increase 50%		Line 2 increase 50%		Switch 1st line to 2nd line		Neuro-protective effect
	OHT	Glaucoma	OHT	Glaucoma	OHT	Glaucoma	OHT	Glaucoma	OHT	Glaucoma
Drugs	2,722	3,307	2,623	3,186	2,722	3,315	2,851	3,349	2,649	3,165
Visits	1,225	1,517	1,217	1,502	1,220	1,509	1,227	1,519	1,200	1,491
Examinations	1,627	1,847	1,614	1,831	1,623	1,846	1,638	1,851	1,611	1,828
Laser	72	72	67	68	69	69	72	73	70	76
Surgery	1,676	1,744	1,611	1,658	1,612	1,682	1,694	1,764	1,679	1,789
Total medical	7,322	8,488	7,132	8,247	7,245	8,420	7,482	8,555	7,208	8,439
	Ref	Ref	–189	–241	–76	–67	160	68	–113	–139
Institutionalization	2,909	11,294	2,770	10,962	2,858	11,177	3,003	12,458	2,258	11,088
Disability allowances	4,367	11,809	4,262	11,325	4,338	11,650	4,439	11,840	3,842	9,511
Total social	7,276	23,103	7,031	22,288	7,197	22,827	7,442	24,298	6,100	20,599
	Ref	Ref	–245	–815	–79	–276	165	1,196	–1,176	–2,503
Grand total	14,598	31,590	14,164	30,534	14,442	31,247	14,924	32,853	13,308	28,948
	Ref	Ref	–434	–1,056	–156	–343	326	1,263	–1,290	–2,642
QALM	172.4	158.8	174.3	161.4	172.9	159.4	171.1	158.3	178.7	167.2
Gain in QALM	Ref	Ref	1.9	2.6	0.5	0.6	–1.3	–0.5	6.3	8.4
Value (ICER = €50,000/QALY)*	Ref	Ref	7,917	10,833	2,083	2,500	–5,417	–2,083	26,250	35,000

*Value (ICER = €50,000/QALY) = Gain in QALM * (50,000/12).

Discount rate: 3%. OHT, ocular hypertension; ICER, incremental cost-effectiveness ratio; QALM, quality adjusted life months. Gain/QALY included only the vision-related benefit, i.e. grand total balance was not included in the calculation. A negative figure reported in total medical, total social or grand total is associated to savings. A positive figure reported in value is associated with positive clinical outcome.

Acknowledgement

The survey was supported by an unrestricted grant from Alcon France SA, Rueil-Malmaison, France, was conducted according to local laws and was contracted to Cemka, Bourg-la-Reine, France.