(2012), "Surveillance for ocular hypertension: an evidence synthesis and economic evaluation", Clinical Governance: An International Journal, Vol. 17 No. 4. https://doi.org/10.1108/cgij.2012.24817daa.010Download as .RIS
Emerald Group Publishing Limited
Copyright © 2012, Emerald Group Publishing Limited
Surveillance for ocular hypertension: an evidence synthesis and economic evaluation
Article Type: Health technology assessment From: Clinical Governance: An International Journal, Volume 17, Issue 4
J.M. Burr, P. Botello-Pinzon, Y. Takwoingi, R. Hernández, M. Vazquez-Montes, A. Elders, R. Asaoka, K. Banister, J. van der Schoot, C. Fraser, A. King, H. Lemij, R. Sanders, S. Vernon, A. Tuulonen, A. Kotecha, P. Glasziou, D. Garway-Heath, D. Crabb, L. Vale, A. Azuara-Blanco, R. Perera, M. Ryan, J. Deeks and J. Cook
Glaucoma is a chronic progressive optic neuropathy leading to impaired vision and blindness if inadequately treated. Open-angle glaucoma is the most common form. A raised intraocular pressure (IOP) is the only modifiable risk factor. Ocular hypertension (OHT) is defined as IOP>21 mmHg and the absence of clinical signs of glaucoma.
Around one million people in the UK have OHT with most identified during a routine “sight” test; diagnosis is typically confirmed in secondary care. Treatment (daily eye drops) may be indicated to reduce IOP. Surveillance should identify those who would benefit from treatment, and should be affordable and acceptable to patients.
Clinical management establishes that OHT is truly present. Once confirmed, monitoring includes measuring IOP by tonometry and tests to detect glaucoma (visual field by standard automated perimetry (SAP) and evaluation of structural changes in the optic nerve). Outcomes from all three parameters inform whether or not treatment is necessary. Long-term surveillance requires interpretation of serial tests and, for those requiring treatment, the responsiveness of IOP to treatment. In choosing the monitoring frequency, a strategy that separates “true” long-term change (signal) from short-term variation and measurement error “noise” is required.
Guidelines were published by the National Institute for Health and Clinical Excellence (NICE) in 2009, but few data were available to guide best monitoring practice.
To determine effective and efficient monitoring criteria for OHT.
To identify and validate the most relevant tool(s) for predicting risk of developing glaucoma.
To determine optimal monitoring criteria. (Which tests? How often?).
To determine public preferences for a service, taking into account health outcomes and patient experiences.
To undertake an economic evaluation of different surveillance pathways, considering costs, clinical outcomes, quality-adjusted life-years (QALYs) and willingness to pay (WTP).
To determine risk thresholds for initiating surveillance.
To make recommendations for research.
The study comprised three interlinked substudies, described in Figure 1.
A: risk prediction models (objectives 1 and 6)
This involved systematic review of prediction models estimating risk of progression to glaucoma, and external validation of the most robust using individual patient data from four populations: Moorfields Eye Hospital, London, and Rotterdam Eye Hospital, The Netherlands (randomised controlled trials; RCTs), and Dunfermline and Nottingham (UK observational cohorts, hospital- and community-based, respectively). The five-year risk was calculated using the prediction equation and the patients’ observed or imputed value of the predictors. The discriminatory ability of the model was assessed using Harrell’s c-index. Model calibration was assessed using calibration plots and calibration slopes.
B: optimal monitoring criteria (objectives 2 and 6)
This involved evidence synthesis of the measurement agreement between alternative tonometers and the reference standard, Goldmann applanation tonometry (GAT), and statistical modelling of the variability of IOP and visual field indices over time:
A systematic search was undertaken. Meta-analyses of the mean differences and the standard deviation of the differences were undertaken for the agreement between each tonometer and GAT. Summary 95 per cent limits of agreement were generated. Data on study characteristics, recordability, acceptability, practicality and reliability were summarised.
Secondary analysis of data from the placebo arm of the London trial, conducted in the mid-1990s to evaluate medical treatment for OHT, was used to estimate:
average true long-term change of the whole group;
short-term variability – combination of analytic variability and biological fluctuation around a stable average – the “noise”; and
long-term variability – variation in long-term true change between individuals – the “signal”.
The variability was estimated using a direct method and a linear random-effects model. Three imputation methods were employed for missing data. The effect of age and observer was analysed by fitting a separate model for each covariate in the London data set. The models were externally validated using patient-level data from the Rotterdam trial conducted between 1997 and 2008.
C: health economic evaluation (objectives 3-6)
This involved two components:
A discrete choice experiment (DCE) to investigate the relative importance of attributes of a monitoring service to the public. Attributes and levels were identified by an advisory panel and a focus group. Attributes included a description of health outcomes (“10-year risk of developing glaucoma, severe glaucoma and visual impairment” and “unwanted effects of treatment”), patient experience (“communication/understanding” and “location”) and a cost attribute (price proxy) to provide a composite monetary measure of utility (WTP). Each DCE question involved a choice between two monitoring programmes (differing in the levels of the attributes) and a no-monitoring alternative. A Bayesian experimental design was used to determine choices, using information from a pilot (n=184). Data were collected using a web-based survey (n=814). The conditional logit model was used to analyse aggregate data and subgroup analyses by age.
A discrete event simulation model to assess the relative efficiency of monitoring strategies for those with OHT, estimated by cost-effectiveness, cost-utility and cost-benefit analyses (using monetary values generated from the DCE). Pathways were informed by NICE guidelines, by the literature and in consultation with clinical experts, service users and the DCE. The model was populated with parameter estimates informed by components of earlier objectives and the literature. Sensitivity analyses, for the cost-utility analysis, explored the effect of monitoring or treating higher risk only, reducing the unit cost of the prostaglandin analogue by 50 per cent and reducing NHS costs. A scenario analysis was conducted to determine the effects of varying estimates of adherence to medication, IOP measurement precision, accuracy of glaucoma detection and rate of progression to glaucoma.
A: risk prediction models
Three models were identified, derived using data from two large multicentre RCTs, the Ocular Hypertension Treatment Study (OHTS) and the European Glaucoma Prevention Study (EGPS), evaluating ocular hypotensive medication. The OHTS-EGPS means model, the most robust, estimates the five-year risk based on age, IOP, central corneal thickness (CCT), vertical cup-to-disc (C/D) ratio and pattern standard deviation (PSD); all variables are routinely collected in clinical practice. The model uses the mean values of the right and left eyes of an individual to calculate eye-specific predictors.
The discriminatory ability was good in the four populations tested, with c-indexes between 0.69 and 0.83; however, in calibration analyses, the OHTS-EGPS means model generally overestimated the risk of glaucoma, although for the Rotterdam cohort the calibration slope was close to 1 (1.09, 95 per cent confidence interval 0.72 to 1.46), the ideal value when there is complete agreement between predicted and observed risks.
This OHTS-EGPS model was developed using selected trial populations and may not include all important predictors. Both trial and observational validation cohorts were highly selected and none satisfactorily covered the full spectrum of risk. Furthermore, missing data for the predictors was considerable in all cohorts and definitions of glaucoma were not standardised. Despite these limitations, the model is useful in conjunction with clinical assessment.
B: optimal monitoring criteria
Systematic review and meta-analysis of tonometers
A total of 102 comparative studies assessed the agreement of at least one tonometer with GAT (HAAG-STREIT, Koeniz, Switzerland). Comparators were dynamic contour tonometer, non-contact tonometer (NCT) (Canon USA, Inc., Lake Success, NY, USA; Keeler Ltd, Windsor, UK; NIDEK Co. Ltd, Gamagori, Japan; Reichert Ophthalmic Instruments, Buffalo, NY, USA; Topcon Corporation, Tokyo, Japan), Ocular response analyser® (ORA) (Reichert Inc., Depew, NY, USA), Ocuton S® (EPSa Elektronik & Präzisionsbav, Saalfeld, Germany), Perkins® (Kowa HA-2, Kowa, Japan), rebound tonometer, TonoPen® (Mentor O&O, Inc., Santa Barbara, CA; Reichert, Inc., Depew, NY, USA) and transpalpebral tonometer. Studies were generally poorly reported. The agreement in IOP (95 per cent limits) varied across tonometers, from 0.2 mmHg (−3.8 to 4.3 mmHg) for NCT to 2.7 mmHg (−4.1 to 9.6 mmHg) for Ocuton S. Sizeable inter- and intraobserver variability was observed for all tonometers, including GAT, casting doubt on the validity of GAT as the default standard.
Optimal frequency of monitoring intraocular pressure and tests to detect glaucoma
Statistical modelling was performed on ocular measures from the London placebo group (n=153; mean IOP 24.4 mmHg; standard deviation (SD) 3.5 mmHg, 14 4-monthly visits). Validation was performed using the Rotterdam placebo data (n=132, mean IOP 25.7 mmHg; SD 2.5 mmHg; 21 visits biannually).
A linear random-effects model, using the last value carried forward to impute missing data, was the best fit to IOP data. The average change in IOP over time for the whole group was <1 mmHg in three years, although a ≥5 mmHg change occurred in 25 per cent of participants. For most individuals any true change in underlying IOP (“signal”) was smaller than the estimated “noise”. Observed changes, using a single measure of IOP at each time point, of ≤3 mmHg can be explained as “noise”. Assuming independence of repeated measures, the mean of two baseline IOP readings increased the signal-to-noise ratio such that true change in IOP of 2 mmHg could be detected at two years. With three baseline measures averaged, true change could be detectable between one and two years.
For lower baseline IOP (<26 mmHg) the model suggested that a true change in IOP would be unlikely within three years. The model may have underestimated the small proportion of individuals with a large change in IOP; this was adjusted for in the economic model. Mean deviation (MD) data, a visual field index measured by SAP, were available only in the London data set. MD fluctuated, increasing and decreasing, with minimal signal detected over four years.
Because of limited patient data on sequential measures of visual fields, the determination of optimal monitoring frequency was based on IOP variability.
C: health economic evaluation
Discrete choice experiment
There was a general public preference for monitoring of individuals with OHT. Individuals were willing to pay £28 per year for a service, everything else being equal. Coefficients representing each attribute, other than hospital location, were significant predictors of preferences at the 1 per cent level. Specifically, marginal valuations of the risk of glaucoma and sight loss over ten years were statistically significant in the expected direction but small. Side effects of treatment reduced the value of any service, with more disbenefits as side effects increased. Good communication with the health professional and understanding of the testing process were important predictors of the value of alternative services. Preferences varied according to age, with those aged >50 years being less concerned with the risk of sight loss (compared with those aged <50 years), but more concerned about treatment side effects and the importance of good communication and understanding of the process.
Economic modelling evaluation
Five pathways were compared. Two were based on NICE guidelines with monitoring interval depending on initial risk stratification: “NICE intensive” (four-monthly to annual monitoring) and “NICE conservative” (six-monthly to biennial monitoring) with treatment according to baseline risk stratification by age, IOP and CCT; two further pathways, differing in location (“surveillance for ocular hypertension (SOH) hospital” and “SOH primary care”), included monitoring biennially with treatment initiated for a ≥6 per cent five-year glaucoma risk. The pathways included repeated IOP measurements, within four months, following treatment initiation or change. A “treat all” pathway involved treatment if IOP was >21 mmHg, measurement of IOP annually in community optometry and referral to secondary care if treatment response was inadequate (<15 per cent IOP reduction).
“Treat all” was the least costly pathway and “NICE intensive” the most costly pathway. The “SOH hospital” pathway reduced the number of cases of conversion to glaucoma compared with the “treat all” pathway and provided more QALYs but the incremental cost per extra QALY (incremental cost-effectiveness ratio; ICER) was considerably more than £30,000. The “NICE intensive” pathway also avoided conversion to glaucoma, but NICE-based pathways were dominated (more costly and less effective) by the “SOH hospital” pathway. In the cost-benefit analysis, compared with “no monitoring” the “SOH hospital” pathway was the only pathway to show net benefit.
Results were sensitive to the risk threshold for initiating treatment, NHS costs and treatment adherence. If treatment was initiated when the five-year risk of developing glaucoma was >10 per cent (e.g. a 60-year-old with an IOP of 27 mmHg, CCT of 560 μm, vertical cup-to-disc (VCD) ratio of 0.4 and a PSD of 1.4 dB has a 10.3 per cent risk in at least one eye), an “SOH hospital” pathway was less costly and more effective than a “treat all” pathway. The SOH pathways had ICERs of <£30,000 compared with the “treat all” pathway when service cost for repeat IOP measurement, in response to treatment change, was <£60. Differences in treatment adherence between the “treat all” and SOH pathways of approximately 40 per cent or higher led to the SOH pathways having ICERs of ≤£30,000. NICE-based pathways were more costly and either were dominated or had ICERs well above £30,000 per QALY.
In the cost-utility analysis, surveillance was not compared with a “no monitoring” alternative as this was not an acceptable option given current NHS policy. A “treat all” pathway was included based on emerging findings from the literature. The acceptability to users and health-care professionals of a “treat all” pathway was not explored. The modelling took a 20-year time horizon, which may be insufficient to capture longer-term benefits. Sensitivity analyses conducted may not fully capture the uncertainty surrounding parameter estimates. Although patient views were consulted when developing the DCE, the results were based on public preferences, which may differ from those of patients.
Implications for health care
The best available prediction model (OHTS-EGPS means model) estimates the five-year risk of glaucoma based on age and the ocular predictors IOP, CCT, VCD ratio and PSD. An IOP measurement algorithm using the average of repeat measurements at one visit reduces noise. Our findings support the clinical importance of establishing a true baseline IOP prior to initiating monitoring or treatment. IOP measurement using the NCT or hand-held applanation tonometer appears to give the closest agreement with GAT with >75 per cent of measurements within 3 mmHg. However, findings suggest that GAT may not be the most appropriate reference standard. The same type of tonometer should be used to compare IOP measurements in an individual.
Our findings, based on a small sample, suggest biennial IOP monitoring for untreated or stable treated OHT. The optimal frequency of clinical testing (perimetry or optic nerve evaluation) to detect glaucoma remains uncertain. The economic evaluation suggests no clear benefit in intensive monitoring to detect glaucoma; any service reconfigurations should consider patient experiences, ensuring adequate time to explain the purpose of monitoring and avoid treatment side effects. If the NHS costs for repeat visits to monitor IOP response to treatment are minimised, biennial hospital-based monitoring appears optimal. The economic model may not have fully captured data uncertainties or the opportunity cost of resource use. The feasibility of community care pathways should be explored.
Recommendations for research
A prospective cohort study including a representative sample of newly diagnosed OHT to update the risk prediction model and evaluate the optimal interval of serial glaucoma tests; standardisation of a measure of perimetry or optic nerve analysis with consensus on glaucoma conversion criteria; a comparison of alternative tonometers; costs and patient preferences for surveillance and treatment; and an updated economic model.
Further development of tonometers to meet the needs of patients and the NHS.
Funding for this study was provided by the Health Technology Assessment programme of the National Institute for Health Research.
© Crown copyright
J.M. Burr, P. Botello-Pinzon, A. Elders, K. Banister, C. Fraser, A. Azuara-Blanco and J. Cook are all based at the Health Services Research Unit, University of Aberdeen, Aberdeen, UK.
Y. Takwoingi and J. Deeks are based at the Public Health, Epidemiology and Biostatistics Unit, University of Birmingham, Birmingham, UK.
R. Hernández and L. Vale are based at the Health Services Research Unit and Health Economics Research Unit, University of Aberdeen, Aberdeen, UK.
M. Vazquez-Montes, P. Glasziou and R. Perera are all based at the Department of Primary Health Care, University of Oxford, Oxford, UK.
R. Asaoka and A. Kotecha are based at the Glaucoma Research Unit, Moorfields Eye Hospital NHS Foundation Trust, London, UK and the Department of Optometry and Visual Science, City University, London, UK.
J. van der Schoot and H. Lemij are based at Glaucoma Services, Rotterdam Eye Hospital, Rotterdam, The Netherlands.
A. King and S. Vernon are based at the Queen’s Medical Centre, University Hospital, Nottingham, UK.
R. Sanders is based at the Queen Margaret Hospital, Dunfermline, UK.
A. Tuulonen is based at the Eye Centre, Tampere University Hospital, Tampere, Finland.
D. Garway-Heath is based at the Glaucoma Research Unit, Moorfields Eye Hospital NHS Foundation Trust, London, UK.
D. Crabb is based at the Department of Optometry and Visual Science, City University, London, UK.
M. Ryan is based at the Health Economics Research Unit, University of Aberdeen, Aberdeen, UK.
Burr, J.M., Botello-Pinzon, P., Takwoingi, Y., Hernández, R., Vazquez-Montes, M., Elders, A., Asaoka, R., Banister, K., van der Schoot, J., Fraser, C., King, A., Lemij, H., Sanders, R., Vernon, S., Tuulonen, A., Kotecha, A., Glasziou, P., Garway-Heath, D., Crabb, D., Vale, L., Azuara-Blanco, A., Perera, R., Ryan, M., Deeks, J. and Cook, J. (2012), “Surveillance for ocular hypertension: an evidence synthesis and economic evaluation”, Health Technology Assessment, Vol. 16 No. 29