- Top of page
- Supporting Information
The simulated lifetime of the patients in the heterogeneous cohort covered on average 25.7 ± 12.9 years. A comparison of the outcomes in both treatment strategies is presented in Table 3. With watchful waiting, the occurrence of conversion from OHT to POAG was 14.6% within the first 5 years and 25.2% after 10 years. Ultimately, 57.0% of the patients not treated for OHT conversed to POAG somewhere during their lifetime and 1.5% went blind. With direct treatment, the occurrence of conversion from OHT to POAG was 7.7% after 5 years, 14.7% after 10 years and 36.5% in the patients’ lifetime. Blindness occurred in 0.4% of the simulated patients. The lifetime use of medication was higher when OHT patients were treated directly, but the incidence of LT and surgery was lower.
Table 3. Average lifetime clinical outcomes of simulated patients in a heterogeneous cohort of ocular hypertension patients.
| ||Watchful waiting||Direct treatment|
|IOP in follow-up (mmHg)||23.6||17.7|
|Occurrence of POAGa||57||37|
|Occurrence of blindnessa||1.5||0.4|
|Average number of medications||0.5||1.3|
|Occurrence of LTa||23||21|
|Occurrence of TEa||15||12|
|Occurrence of ReTEa||4.8||3.2|
|Occurrence of tube implanta||2.9||2.3|
|Occurrence of CEa||28||28|
|End-of-life MD (dB)||−5.5||−2.8|
The health economic outcomes are listed in Table 4. Within a time horizon of 10 years, direct treatment of OHT resulted on average in slight health gains and additional costs at an incremental cost-effectiveness ratio (ICER) of € 33,645. However, over a lifelong horizon, direct treatment resulted on average in 0.27 QALY's gained and cost reductions of € 649 per patient compared to watchful waiting.
Table 4. Health-economic outcomes of simulated patients in a heterogeneous OHT cohort after 10 years and after a lifelong horizon. Average per patient.
| ||Watchful waiting||Direct treatment||Incremental||ICER|
|Costs||€ 2,302||€ 3,415||€ 1,113||€ 35,573|
|Discounted costs||€ 1,891||€ 2,844||€ 957||€ 33,645|
|Discounted QALY's||7.62||7.65||0.03|| |
|Lifetime (mean 26 years)|
|Costs||€ 18,327||€ 14,343||−€ 3,984||Dominant|
|Discounted costs||€ 7,722||€ 7,073||−€ 649||Dominant|
|Discounted QALY's||17.55||17.81||0.27|| |
A breakdown of the incremental costs is provided in Fig. 1. Differences in costs between the two treatment strategies occurred mainly in two cost categories: medication and care. Direct treatment was associated with higher costs for medication, but lower costs for (informal) care. The figure also illustrates how costs further in the future are discounted more heavily. In particular, the relative contribution of costs for care is much larger in the undiscounted incremental costs than in the discounted incremental costs of direct treatment compared to watchful waiting. Figure 2 illustrates the uncertainty surrounding the ICER as a cloud of possible cost-effectiveness outcomes resulting from the PSA. The cost-effectiveness acceptability curve (Fig. 3) showed that at a willingness-to-pay threshold of € 0 per QALY, the probability that direct treatment is cost-effective was 83%. At thresholds of € 10,000 per QALY and higher, this probability had increased to 100%. Likewise, the EVPI decreased from € 96 per patient to € 0 per patient between the thresholds of € 0 and € 10,000 per QALY.
Figure 1. Distribution of the total costs in eight cost categories in both treatment scenarios (grey) and incremental (black) in a heterogeneous ocular hypertension population. The total height of the bars indicates the undiscounted costs; the solid bars indicate the discounted costs and the dotted portion indicates the amount that is discounted away.
Download figure to PowerPoint
Figure 2. Cost-effectiveness plane showing the average incremental cost-effectiveness of direct treatment in all patients compared to watchful waiting in a heterogeneous ocular hypertension population, both in the base case model as in each of the cohort simulations in the probabilistic sensitivity analysis.
Download figure to PowerPoint
Figure 3. Cost-effectiveness acceptability curve (solid black line) of direct treatment compared to watchful waiting in ocular hypertension patients, and the expected value of perfect information (dashed grey line) at increasing acceptability thresholds for the incremental cost-effectiveness ratio.
Download figure to PowerPoint
The outcomes of the comparison of direct treatment versus watchful waiting in subgroups of OHT patients are listed in Table 5. Direct treatment resulted in health gains irrespective of the initial IOP and additional risk. The health gains were larger as the total risk of conversion in the subgroup increased. The health gains came at additional cost in the subgroups with low additional risk of conversion and in the subgroup with neutral additional risk and an initial IOP of 22 mmHg. In the other subgroups, direct treatment resulted in cost savings.
Table 5. Incremental discounted cost-effectiveness outcomes of direct treatment versus watchful waiting in subgroups of OHT patients based on initial IOP and additional risk of conversion (HRother).
| ||Average 5-year risk of conversiona||Incremental costs (€)||Incremental QALY's||ICER (€ per QALY)|
|Low risk (HRother = 0.5)|
|22 mmHg||4||€ 1,259||0.082||€ 15,425|
|24 mmHg||5||€ 851||0.122||€ 6,954|
|26 mmHg||6||€ 624||0.175||€ 3,563|
|28 mmHg||7||€ 1,127||0.221||€ 5,088|
|30 mmHg||8||€ 807||0.303||€ 2,660|
|32 mmHg||10||€ 49||0.403||€ 121|
|Neutral risk (HRother = 1.0)|
|22 mmHg||8||€ 541||0.149||€ 3,629|
|24 mmHg||10||-€ 193||0.214||Dominant|
|26 mmHg||11||-€ 765||0.293||Dominant|
|28 mmHg||13||-€ 1,085||0.374||Dominant|
|30 mmHg||16||-€ 1,788||0.469||Dominant|
|32 mmHg||18||-€ 2,826||0.571||Dominant|
|High risk (HRother = 2.0)|
|22 mmHg||16||-€ 327||0.231||Dominant|
|24 mmHg||18||-€ 1,276||0.300||Dominant|
|26 mmHg||22||-€ 1,995||0.370||Dominant|
|28 mmHg||25||-€ 3,168||0.497||Dominant|
|30 mmHg||29||-€ 4,045||0.583||Dominant|
|32 mmHg||33||-€ 6,046||0.728||Dominant|
- Top of page
- Supporting Information
In this study, we have used a patient-level simulation model of OHT and POAG to simulate the disease progression of patients with OHT, and used its output to estimate the additional health and costs that can be expected from direct pressure-lowering treatment compared to watchful waiting. The modelling approach provides an efficient method to generate new information from available evidence, without the need to conduct clinical studies. Direct treatment turned out to be a dominant strategy over watchful waiting in a heterogeneous population of OHT patients over the lifetime horizon. Over a shorter time horizon, the cost-effectiveness of direct treatment was less favourable, with a discounted ICER of € 30,597. Although this amount may still be acceptable, it is clear that the time horizon plays an important role in the cost-effectiveness of OHT treatment. Pressure-lowering treatment in OHT is a preventative measure involving current and long-term investments to prevent health loss at a much later time period. The time horizon should therefore be long enough to capture future effects, or the ICER will overstate the contribution of short-term investments.
Direct treatment resulted in better health outcomes in all simulated subgroups. The general tendency across the subgroups was that incremental costs decreased as the initial IOP in the subgroup increased. An exception to this tendency was seen in the subgroups with a low additional risk of conversion (table 5), which showed a local ‘peak’ of incremental costs in the 28 mmHg subgroup. This observation can be explained by a shift in the balance between the short-term costs of treatment and long-term savings in low-vision-related care. Up to an initial IOP of 26 mmHg, monotherapy is sufficient to get most patients below the target pressure of 21 mmHg, but higher initial IOPs will mostly require combination therapy. The marginal costs of extra medication cause a sudden increase in overall treatment costs, which is reflected in the total incremental costs. In the subgroups with neutral and high additional risk, a small deviation from the tendency was observed due to the same effect, but it was far less pronounced because the contribution of treatment costs to the overall costs was smaller in these subgroups. Direct treatment was dominant in all subgroups, except for the subgroups with a conversion risk lower than 10% in 5 years. The latter had ICER's in the range of € 100 to € 15,500 per QALY. The implications of these ICER's for decisions regarding the desirability of direct treatment in low-risk subgroups depend on the way ICER's are used to aid decision-making. In the net monetary benefit calculations, we have assumed an acceptability threshold of € 30,000 per QALY based on ranges mentioned in literature and authority reports, although the threshold may be lower (€ 20,000 per QALY) for preventive care [Raad voor Volksgezondheid en Zorg (Council for Public Health & Health Care) (2006); Cleemput et al. 2008; Verweij et al. 2008]. The method of comparing the ICER to an acceptability threshold to gauge the relative value-for-money of the intervention has been criticized though, and if it were employed, it is very likely that the acceptability threshold varies between jurisdictions, between disease severities, and in time (Gafni & Birch 2006; Cleemput et al. 2008). We can therefore only report the value of the ICER of direct treatment in low-risk OHT patients and not speculate on its acceptability.
The outcomes of the PSA showed that even if the input parameter values are randomly varied within their uncertainty margins, the outcome of the analysis shows dominance for direct treatment initiation in the majority of cases. This implies that even though there is uncertainty about the exact value of the model's input parameters, this does not result in decision uncertainty. In addition, the EVPI dropped to zero at willingness-to-pay thresholds higher than € 10,000 per QALY, which suggests that there is no value in further research to reduce uncertainty surrounding any of the population parameters in the model. In addition to parameter uncertainty, we have considered the impact of structural uncertainty. An issue of structural uncertainty in our model is the way both eyes of the patient are handled. In the base case model, we have simulated patients rather than individual eyes and simulated that both eyes underwent similar treatment and disease progression. This structural choice involves uncertainty, as not all patients in clinical practice will present with symmetrically affected eyes. To assess the impact of this assumption, we have performed an additional analysis in which we modelled only the worse eye of the patient and assumed that the other eye remained completely unaffected. The lifetime discounted outcomes with watchful waiting in a heterogeneous OHT population were 18.04 QALY's and € 4,580, whereas the strategy with direct treatment resulted in 18.15 QALY's and € 5,830. The ICER of direct treatment was therefore € 11,523 per QALY gained. The outcomes of the base case model (dominance) and this univariate sensitivity analysis represent the two boundaries of the uncertainty spectrum regarding the symmetry of disease progression in both eyes, and all realistic scenarios encountered in clinical practice will fall within these boundaries. Structural uncertainty also played a role in the way costs related to visual impairment were accounted for. The major difference in cost is due to costs of medication use and cost of care. The latter depended on the amount of VF loss. Early treatment leads to less VF loss and thereby savings of the costs involved with this VF loss since less help and care is needed to support a glaucoma patient. Cost for care for glaucoma patients may differ between countries. Since the costs of care differ most between the two treatment strategies, the results of our analyses showed that low-vision-related costs played an important role in the overall cost-effectiveness of treatment, while there is a considerable degree of uncertainty about the size of these costs and how they increase with progressing disease. Previously, authors investigating health economics of OHT and glaucoma treatment have not included such costs in the analysis (Stewart et al. 2008), considered only nursing home costs (Rein et al. 2009) or assumed resource use in this category only in case of blindness (Kymes et al. 2006; Burr et al. 2007; Peeters et al. 2008). In our study, we have assumed a gradual increase in low-vision-related costs with increasing loss of VF, which was based on measurements in our study in 531 patients representing all levels of OHT and POAG severity and MD values ranging from 0 to −32 dB in the better eye (Van Gestel et al. 2010b). The PSA showed that even when the low-vision-related costs were varied between a factor 0 (i.e. no costs) and 2, the dominance of direct treatment was not affected. On the same note, the EVPI analysis indicated that, despite the uncertainty about low-vision-related costs, there is no value in additional research to reduce that uncertainty in the context of the currently investigated treatment decision. Although the EVPI is expressed as a value for an individual patient to make comparisons between studies, it should also be seen in relation to the total population of OHT patients and in fact future OHT patients. This is because savings by perfect information will be gained by changing the management strategy for the total, future population.
This example illustrates how the fact that some model input is quite uncertain does not invalidate the entire model, and that it is more important to acknowledge uncertainty and assess its impact than negate the informative power of the aggregated evidence. It also demonstrates that the model input with the highest degree of uncertainty is not necessarily the one with the largest impact on the outcome and is therefore not the most likely candidate for future research. In fact, we have conducted analysis of variance with the PSA input and outcomes and found that uncertainty about the relative risk of IOP on conversion had the largest impact (see Appendix S1).
The dominance that we found for direct treatment relative to watchful waiting differs considerably from the $144,780 per QALY that has been reported previously in a US-based study (2004 € 1 ≈ $ 1.25; Kymes et al. 2006; The European Central Bank). The difference is caused by lower incremental costs (-€ 649 versus $ 7,239) and higher incremental QALY's (0.27 versus 0.05) in our study. We compared the methodology of both studies and identified several issues that might explain the differences. First, the setting of the studies affected the estimates for the cost price of medication, cataract surgery and POAG surgery, as cost prices in the United States are generally higher than those reported for European countries (Kobelt & Jönsson 1999; Oostenbrink et al. 2001; Traverso et al. 2005). Second, Kymes et al. (2006) attributed resource use associated with visual impairment only in case of blindness, not in preceding stages. These two factors are probably the main reason why treatment in the US study was associated with incremental costs rather than cost savings. Additionally, four issues may contribute to the differences in incremental effects. First, the estimated utility loss as a result of disease progression was smaller in the US study than in our study, particularly in advanced stages. Second, the horizon was much shorter in the US study. Kymes et al. (2006) do not report the actual duration of follow-up in their study, but considering the total QALY's reported (13.6) and the utility in early and moderate glaucoma (0.97 and 0.89) it is likely to be around 15 years, whereas the horizon was 26 years in our study. As the results of our study have shown, the length of the time horizon has a considerable impact on the ICER of OHT treatment. Third, the risk of conversion in the US study was distributed towards lower values than in our simulated population. The authors reported that 70% of the patients had an annual conversion risk lower than 2%, whereas this was 44% in our simulated population. Since the incremental effects of direct treatment are smaller with decreasing conversion risk (Table 5), a population with more low-risk patients will result in smaller average incremental effects of direct treatment. Finally, the future QALY gains in the US study were more heavily discounted, which reduces the net present value of future health gains (3% versus 1.5%). The combination of all factors may have resulted in the difference in outcomes of our study compared to those reported earlier. These issues do not necessarily concern ‘wrong’ choices in either of the studies but rather reflect the different decision-making contexts targeted by the two studies.
The study of Burr et al. (2012) encompasses several studies. One study was on the selection and validation of a prediction model to predict the risk of conversion to POAG in OHT patients. The second was on the validity and variability of IOP measurements by diverse instruments. The third study was on the public preferences for the type of surveillance or monitoring of OHT patients. The fourth study was on the cost-effectiveness of five monitoring strategies for OHT patients with an IOP > 21 mmHg. These five were the following: (1) treat-all OHT patients with a prostaglandin and refer if IOP lowering is <15% and measurement of IOP annually by an community optometrist; (2) surveillance for ocular hypertension (SOH) in a primary care (community) setting, monitoring is every 2 years and individuals would only be referred to secondary care if IOP was ‘off target’ or conversion to OAG being detected; (3) SOH in a hospital eye service setting; (4) monitoring according to the NICE intensive guidelines; and (5) monitoring according to NICE conservative guidelines. Treatment in monitoring strategies 2–5 was based on a calculated risk of conversion. Treatment intensity was to lower the IOP with 15% and to adjust if IOP was off target or occurrence of POAG had occurred. In the NICE strategies (4 and 5), treatment was limited until a certain age was achieved. Monitoring intervals in the NICE strategies was every 2 months to assess IOP and every 4–24 months for full assessment depending on the risk of conversion and strategy. The monitoring frequency was every 2 years in strategy 2 and 3. In the treat-all strategy 22.8% converted to POAG in 20 years. This was reduced to the lowest value achieved at 20.6%, which is a reduction of 2.2%. Burr et al. concluded that the strategy SOH in a hospital gave fewer POAG and progression, more QALY's and net benefit as compared to the treat-all strategy. However, in terms of cost-effectiveness, the gain of QALY's was not considered worth the costs.
In comparing the study of Burr et al. with ours, it is important to notice that Burr et al. compare strategies in which at least a group, those with the highest risk or all patients are being treated. The difference is further in the setting and the criteria for follow-up. Moreover, their treatments were not based on target pressures to be achieved but based on achieving a reduction of at least 15%. In our studies, two extremes are compared. ‘Treat all’ or postpone treatment in all OHT patients untilled POAG has occurred. Moreover, we set a target pressure that implies that high IOP values need to be reduced to below 21, if that could be achieved with eye drops or laser treatment. Altogether, our two strategies are two extremes in effectiveness. Moreover, our utilities were based on the HUI-3 that is a better utility scale for ophthalmology because it encompasses visual impairment in its questionnaire. It could be that this gives larger differences between stages of OHT/POAG. The study of Burr et al. gives a refinement of the monitoring strategy when at least a (large) subgroup is treated.
Another difference from the study of Burr et al. is that they included a measure for compliance. We did not take that into account since we assumed that this was an implicit part of the outcome of treatments in other studies.
The results are findings that show the best management strategy from a cost-effectiveness point of view. Patient management is, however, influenced by other factors that can be identified by ophthalmologists and patients who may have their own preferences. The results are based in average findings, but still not all patients are at risk, leaving room for individualized care.
The average benefits of direct treatment initiation in the heterogeneous population were considerable, and the best chances of optimal health outcomes in an individual patient are therefore with treatment rather than watchful waiting. The subgroup analyses further indicated that it is not necessary to consider IOP or additional risk of conversion in this treatment decision. However, there is a considerable likelihood in OHT patients that, in hindsight, treatment was not necessary. Indeed, 43% of the simulated patients in the watchful waiting strategy did not convert to POAG during their entire lifetime. The problem is that one cannot tell in advance who these patients are going to be, even if one calculates the risk of conversion with high evidence-based risk calculators. The best chance of optimal health outcomes in OHT patients is with direct treatment initiation. The implication of this finding is that the attitude towards treatment initiation in OHT could change from ‘do not treat, unless the risk of glaucoma is too high’, which is basically up to the ophthalmologist's judgment, towards ‘treat, unless the burden of treatment is too high’, which is much more up to the patient's judgment. Moreover, these issues should be discussed with the patient. The results of our study may aid the ophthalmologist and patient in deciding on the best treatment.
Two strategies were compared. One in which treatment was started directly and IOP lowered in a stepwise fashion with the first target pressure of 21 and 18 and 15 when POAG occurred and progression occurred respectively. In the other strategy the same stepwise approach was postponed and started until after the period of no treatment conversion to POAG had occurred. In essence prevention is compared to ‘cure’ when the disease has occurred. To have a fair comparison to prove whether prevention is better, the same treatment strategies were compared. However, from a clinical point of view, other comparisons are possible. One could question, for example, whether early treatment with a stepwise approach is better as compared to direct intensive treatment and direct lowering of the IOP below 15 when glaucoma has occurred. The latter gives less VF loss and lower costs as compared to a stepwise lowering approach (van Gestel, 2012). Another strategy could be to look at the age of discovering OHT. In an older person of 85, this may not be very cost-effective. Others could be to include generic preparations in the model. It has been shown that generic preparations may cause adverse events different from brand preparations (Takada et al. 2012). Moreover, generic preparations could give lower IOP values (Narayanaswamy 2007). Although the generic preparations could be cheaper, it could lead to more burden for the patient and ophthalmologist and costs in future. It would be worth to calculate its effect on costs and effects in our model. In addition to this, the results can be different in other countries when there are no costs of care for visually impaired or blind glaucoma patients.In conclusion, we found that direct pressure lowering treatment is a dominant strategy compared to watchful waiting in a heterogeneous population of OHT patients, and that the efficiency of direct treatment increases with increasing initial IOP and the presence of additional risk factors for conversion. The implementation and consequences of the results should be discussed with ophthalmologists and individual patients.
In conclusion, we found that direct pressure lowering treatment is a dominant strategy compared to watchful waiting in a heterogeneous population of ocular hypertension patients, and that the efficiency of direct treatment increases with increasing initial IOP and the presence of additional risk factors for conversion. The implementation and consequences of the results should be discussed with ophthalmologists and individual patients