Previous research from glaucoma clinics in England has suggested most patients get a similar “diet” of VF testing over time.
17 For example, average number of VF examinations over time has been similar for all patients, irrespective of age, severity of VF loss, and rate of MD loss.
17 This “one size fits all” approach to glaucoma follow-up for VF testing is likely suboptimal from a clinical
18 and health economic
19 perspective. Hedgehog Plots provide a potential novel tool for clinicians to visualize all their glaucoma patients simultaneously. Hedgehog Plots could be helpful in selecting patients who should be seen more or less frequently in follow-up, allowing for the prioritization of monitoring resources. In this study, we also present two parameters, RP and LSY. The former is not novel,
5,6 but it is an important concept for clinicians to incorporate into glaucoma management decisions. LSY is novel and we have shown that it offers complementary but different information to RP.
Key to this work and additional to this report is provision of a purpose written interactive application demonstrating the techniques. We hope this application stimulates design of similar software that could be used in clinics. After all, we know software helps provide a way to standardize clinical assessment and has been proven to improve the agreement between clinicians when making decisions about VF progression occurring.
20 Our application is also designed to allow clinicians to analyze data within their own clinics if MD data entry can be completed. Realistically our tool will only be widely adopted if a version is incorporated into clinically used software, like an electronic medical record (EMR) or in ophthalmic data management systems linked to perimeters. At that stage other design features, such as resetting the baseline of a VF series to times of significant changes in therapy in long-term follow-up, could be considered. Moreover, the software ought to be designed such that a clinician can easily exclude unreliable VF examinations or, for example, use published techniques for minimizing the effects of outlier observations.
21
RP can be ranked for all patients in a clinic to help identify worse cases of VF progression without using inferential statistics. This allows for a comparison between individuals and the general clinical population rather than imposing an arbitrary cutoff based on a
P value. Such ranking could be useful in situations where decisions must be made about allocation of VF monitoring resources. For example, more frequent examinations may be allocated to someone with an observed fast rate of progression or a “priority case.” Furthermore, a patient who seems to have rapid progression in their better eye would need more monitoring than a person who shows no progression in either eye. Further refinement can be achieved by using LSY. A specific example of this is given in
Figure 4, where Patient 84 has rapid visual progression in both eyes (ranked the seventh and eighth worst eyes within this clinic). However, this patient was predicted to have a LSY of 0 years. Patient 54 had relatively slowly progressing eyes (ranked 23rd and 46th, respectively) but considerable VF damage at diagnosis (worse that −12 dB); the patient is relatively young. This patient is predicted to have bilateral blindness for 10 years within their expected lifetime. So, this patient might warrant more careful monitoring (more frequent VFs) or intensified treatment. In addition to the applications we have shown here, the Hedgehog Plot could be used as a tool to compare different clinics by using a large data set as a reference clinic.
10 This would allow clinicians to compare results of their own patients directly to those of others, or could be used for auditing purposes. Hedgehog Plots also might be useful for comparing the entire distribution of VF progression in different arms of a clinical trial.
We used an MD worse than −22 dB as a potential definition of visual impairment, but different cutoff values could be used. This threshold was based on the US SSA benchmark, which defines “statutory blindness” as both eyes reaching this cutoff value.
14 Other criteria, linked to ability to perform functional tasks could be used.
22–25 For example, individuals who reach a MD threshold of worse than −14 dB in both eyes are unlikely to satisfy the VF component for adequate vision to be legally fit to drive in the United Kingdom.
26 Of course this limit will vary by country depending on regulations. Further research in this area would be of interest.
In this study, we used MD measurements, which are well known and well understood by clinicians. The MD is only a summary measure of the VF and is a weighted average across the VF, relative to a group of healthy age-matched individuals. However, the individual VF points may be of greater interest, because of the additional information they provide, such as the spatial nature of VF loss. This information is otherwise lost in global parameters. Although we assume binocular VF impairment at −22 dB, the patient still may have some preserved visual function at this threshold. The remaining visual function may be significantly affected by the location of the VF damage.
27–29 For example, more centrally located progression would arguably affect the patient's quality of life more significantly than peripherally located progression. Additionally, if progression took place in matching locations in each eye, this local progression would likely affect the patient's visual function more significantly than where no such binocular defects existed.
8,30
The assumption of a linear rate of progression in predicting VF loss is another limitation of this study. While this may not be representative of true glaucomatous VF deterioration, it is used commonly in clinical practice and has been shown to provide more robust estimates of future measurements than more complex models.
31,32 Furthermore, we assumed that the RP is constant throughout the patient's predicted remaining lifetime. This may underestimate true deterioration. It also does not take into account future amelioration of progression by intensified treatment. For example, a glaucoma patient, possibly nonadherent or nonresponsive to treatment, may have a significant reduction in RP after intensified treatment.
33 In turn, this could affect their course to predicted significant LSY because RP and LSY are inherently linked. In this situation it might be prudent to consider a new baseline VF assessment. Conversely an eye that, for example, suffers ocular comorbidity or rapid vision loss due to cataract, even late in life, may result in a rapidly changing RP and unexpected LSY. Our tool, therefore, is limited because it does not solve these unchanging dilemmas of managing glaucoma. Yet, as a visualization tool, used in conjunction with other patient data, it still may offer the managing clinician useful information that may have remained unseen in a series of VF charts.
It also is important to note that our life expectancy data were determined using UK census data and will not be representative of populations in other countries. Furthermore, the clinic data we use were extracted from clinics across England. These may not be representative of different demographics and different health care systems. This should be taken into account particularly for interclinic comparisons.