Our time-to-event analysis estimated that in this cohort around one in five patients’ VFs deteriorated such that the patients failed a surrogate measure for the UK legal fitness to drive. Previously, the rate of VF loss in POAG with low IOP has been shown to be very similar to that of POAG at approximately 3% loss per year (equivalent to a loss of 0.7 dB mean deviation per year).
30 Moreover, it has been suggested that an average 60-year-old patient in the United States who developed VF loss would probably not go blind in either eye in his or her lifetime.
31 A slow rate of VF loss has been found in patients with treated POAG, with 19% of eyes becoming legally blind after 22 years of follow-up.
32 In a population survey of untreated POAG in St. Lucia, West Indies, the probability of blindness at 10 years was only 16%.
33 Recent randomized controlled trials have established evidence of the benefit of treatment on VF progression in ocular-hypertensive patients and patients with early OAG and have also indicated that conversion and progression rates are generally modest.
34 35 Most patients in our analysis did not progress sufficiently to reach the endpoint of failing the driver’s license criteria during follow-up, thus echoing these previous findings about OAG’s generally being a slowly progressing disease. However, a significant proportion did fail the criteria during follow-up, in line with previous retrospective reports. We therefore suggest this criterion, or milestone to blindness, to be a meaningful endpoint on which clinical management decisions could be based. Of interest, the time-to-event analysis suggests there was a subgroup of those with rapid progression to disability (
Fig. 2indicates that approximately half of the patients who failed the criteria did so in the first 2 years of follow-up), making the practical problem to be solved that of how to determine rate of loss in such a short follow-up period. Still, it is imperative to establish the risk of progression to functional disability in patients who may undergo unnecessary treatment and to determine those patients who are unlikely to progress such that it impinges on their quality of life. The framework of IVF and using rates of progression provides a mechanism to determine this risk.
The main purpose of this study was to identify which monocular and binocular measures of VF loss best predict which patients in a normal-tension glaucoma clinic are likely to meet the DVLA criteria for VF loss, and thus potentially lose their driver’s licenses. IVF sensitivity indices provided a better fit to our dataset than monocular measurements, and the IVF over the central 20° field in particular provided a good fit to the observed data. In terms of predicting which patients will lose their driver’s licenses in the future, it is important to know both the mean IVF value in the central 20° field at baseline (how severe the binocular defect is at the start of follow-up) and the rate of change in the IVF field at 2 years (how well a patient does over time). This combination appears to be better than simply monitoring the worse eye or the most rapidly declining eye in isolation. It has also been shown that both baseline binocular VF loss and VF loss at 2 years after baseline are significantly associated with driving cessation in older adults.
36 Patients who come into a clinic with poor IVF and/or patients who rapidly lose IVF sensitivity over time can be flagged for possible intensified intervention or closer monitoring, to prevent their failing the DVLA VF criteria. It is important to bear in mind that we are predicting a surrogate measure of DVLA failure and as such, we are not able to confirm whether patients did in fact lose their drivers’ licenses.
In our best-fitting model shown in
Table 3 , the variable for glaucoma surgery is statistically significant, so that patients who underwent glaucoma surgery were approximately three times as likely to fail the DVLA test during follow-up than were patients who did not undergo surgery. This finding is almost certainly due to selection bias (i.e., patients with demonstrable disease progression or high risk of further progression are identified early and tend to be referred for surgery).
The value of a model may be further improved by applying it to an independent sample of patients and assessing its performance in predicting failure the VF criteria to prevent loss of driver’s license. In the absence of an independent dataset, we used bootstrap techniques to validate our findings. The large-sample approximations for the Cox regression coefficients were broadly confirmed by applying the same model to the bootstrap samples. The superiority of the central 20° IVF baseline sensitivity and 2-year change values as predictors of DVLA failure was supported in 93% of the bootstrap samples
(Table 5) .
Our modeling necessitated excluding patients who failed the surrogate DVLA criteria within a short period from the start of follow-up (2 years), because one of our model variables was the rate of change at 2 years. These patients showed rapid rates of loss of VF sensitivity. The patients remaining in our sample had vision that was deteriorating less rapidly. This may explain why the rate of loss parameter is only just equivalent to the initial loss at presentation as a risk factor for failure to meet the surrogate DVLA criteria
(Tables 3 4) . In practice the former is probably more important, but the clinical conundrum is estimating rapid rate of loss reliably over a shorter period than say 2 years with a clinically realistic number of fields. Statistical estimation of the rate of loss must be done on a sufficient number of points in time. Perhaps frequency of testing could be increased when monitoring patients with newly diagnosed glaucoma,
37 but this issue is beyond the discussion of the results from this study. It should also be noted that the patients in our dataset are likely to be highly motivated, having attended the clinic over a long period, and have received excellent care and so are likely to have more stable glaucoma.
Historically, there has been a great deal of uncertainty about how best to treat POAG with normal IOP (or normal tension glaucoma), since there was little evidence to show that lowering IOP prevents continued VF loss. It has since been shown that IOP-lowering treatments can slow VF loss or progression of optic disc damage
38 ; however, the research literature remains mixed: a Cochrane review concluded that there is “not enough evidence about the effects of surgery or medications on vision loss from normal tension glaucoma.”
39 Given the weak research evidence for effectiveness of treatment in these patients, and the risks of side effects of treatment, many patients with POAG and low IOP are left untreated and instead are monitored frequently for signs of VF damage. We believe that the methods outlined in this study could greatly assist the clinician with the latter.
The IVF has a simple appeal in terms of functional VF status, since it is a measure of binocular sensitivity and thus represents a patient’s functional field of view. Furthermore, IVF sensitivity is easily calculated directly from output of monocular sensitivities: It is the highest of the left and right monocular sensitivity values. The IVF has been shown to provide accurate predictions of binocular VF sensitivity in patients with glaucoma, so it can be considered a suitable measure when evaluating patients’ disability in terms of activities of daily living and quality of life. Of course, the IVF cannot provide an estimate of field loss beyond the central areas tested with either 24-2 or 30-2 HVFA, but evidence that peripheral constriction occurs in isolation from central defects in glaucoma is vague.
40 Our retrospective examination of a large group of patients with bilateral glaucoma indicated that peripheral constriction rarely occurred (Viswanathan AC et al.
IOVS 1998;39:ARVO Abstract 2270).
The relationship between VF loss and driving accidents has not yet been established with any certainty, due in part to the poor methodology of studies investigating this relationship, although the better the methodology, the stronger the link between VF defects and crashes.
41 It has been shown that involvement in crashes for older drivers is significantly predicted by binocular VF loss, but not by monocular field loss.
42 Other factors contributing to the inconclusive association between VF loss and driving accidents include the tendency for drivers to compensate for poor vision and change their behavior
43 and findings that the determinants of accidents are multifactorial, including cognitive and perceptual skills as well as visual function.
44 However, standards for VF loss, along with visual acuity, remain the criteria by which the vision component for legal fitness to drive in the United Kingdom is assessed.
In summary, losing one’s driver’s license is a significant and adverse milestone on the way to functional blindness in progressive glaucoma; 20% of patients in our dataset failed the DVLA VF criteria at some point during follow-up. The development of the IVF should make it possible for clinicians (and others) to start integrating binocular considerations into the diagnosis and management of patients in a user-friendly and clinically valid way. The status of the binocular IVF at presentation and the rate of binocular IVF sensitivity loss may help identify patients who could benefit from intervention or intensified treatment, thus avoiding this milestone to blindness. Furthermore, the rate of IVF loss may provide a way of helping identify those who do not need treatment.