The Diabetic Retinopathy Clinical Research (DRCR)
Network Protocol S demonstrated, in a prospective,
randomized, multicenter trial, the non-inferiority of
intravitreal ranibizumab (IVR) compared to panretinal
photocoagulation (PRP) in the treatment of proliferative
diabetic retinopathy (PDR) (1).
The trial established IVR as a viable alternative
treatment for what has been the mainstay, indeed only
non-surgical, treatment for PDR. The primary outcome
measure in the trial was change in visual acuity letter
score, but secondary measures included rate of vitrectomy,
diabetic macular edema (DME), and peripheral visual field
loss, which were lower with IVR. Notably, patients treated
with IVR had more intravitreal injections and doctor visits
during the 2-year study duration. Interpretation of the
PDR results was somewhat confounded by the coexistence
of DME (although the cohorts were stratified for baseline
DME) and its response to or subsequent need for IVR,
not to mention protocol allowances for rescue treatments
for PDR; hence there were significant degrees of overlap
in the initially-assigned treatment cohorts. An embedded
study was assessment of patient-centered outcomes in
each of these two groups, with the inferred purpose of
determining if the non-inferiority of IVR to PRP was
present for these outcomes. The high level summary of
the findings of that study, reported by Beaulieu et al. (2), is
that PRP and IVR also have similar outcomes as measured
by patient-centered instruments, corroborating the DRCR
Protocol S objectively based findings of non-inferiority of
IVR for PDR.
Clinicians have long been cognizant of the importance
to a patient of subjective or functional (“patient-centered”)
outcomes after any treatment. While intuition suggests
these usually parallel objective measures (notably visual
acuity), an active area of investigation has been the
standardization and application of a variety of surveys and
self-assessed questionnaires. These types of evaluations may
corroborate objective measures of outcomes, but might
offer important insights not captured by standard outcome
measures.
Beaulieu et al. (2), report on the results of self-perceived
visual health and function administered to a subset of
patients from the DRCR Protocol S cohort who had
only one eye randomized to treatment with IVR or PRP
(bilaterally affected patients who were assigned to different
treatment groups by the protocol were excluded). The
authors employ three surveys: the National Eye Institute
Visual Function Questionnaire-25 (NEI VFQ-25), the
University of Alabama Low Luminance Questionnaire
(UAB-LLQ), and the Work Productivity and Activity
Impairment Questionnaire (WPAIQ).
Both the NEI VFQ-25 and the UAB-LLQ focus on
health and visual function as well as the psychosocial impact
of visual disability. The NEI VFQ-25 is an abbreviated
form of the long form 51-item NEI-VFQ (3). It is divided
into several sections. Part 1 includes an assessment of general health and vision, anxiety about vision, and ocular
pain. Part 2 focuses on difficulties in accomplishing
activities such as reading, hobbies, mobility, activities of
daily living, and driving. Part 3 assesses for changes in
daily function and habits secondary to visual impairment.
The NEI VFQ-25 also includes vision-related subscales
concerning general health, vision, social function, driving,
role limitations, well-being/distress, and dependency. The
UAB-LLQ is a 32-item questionnaire with six subscales
on driving, extreme lighting, mobility, emotional distress,
general dim lighting, and peripheral vision. A key difference
between the two surveys is that the UAB-LLQ focuses on
vision in low-light conditions and has historically been used
in studies of age-related macular degeneration (AMD).
While in the context of AMD, the UAB-LLQ has shown to
be equivalent to the NEI VFQ-25 (4), the UAB-LLQ has
not been used previously to evaluate vision loss in diabetic
eye disease. In contrast, the NEI VFQ-25 more generally
assesses visual function and has been utilized in prospective
and randomized clinical trials, including in diabetic eye
disease (5). It has been reported as a reliable indicator of
vision related quality of life in diabetic retinopathy (6), and
measureable gains in the NEI VFQ-25 have been associated
with IVR in the treatment of DME (7). The WPAIQ, like
the UAB-LLQ, has not previously been applied to diabetic
eye disease. It is a 6-item questionnaire that assesses for
impacts on work absenteeism, presenteeism (impairment
while working), productivity loss, and activity impairment.
Beaulieu et al. find no statistically significant differences
between the IVR and PRP cohorts in composite NEI VFQ25
scores or in what would be expected to be important
pertinent qualities of peripheral, color, or driving subscores.
Similarly, the authors report no difference in UAB-LLQ
composite scores. Of note, they identify a small but
statistically significant decrease in work productivity at
one year, identified by the WPAIQ, which was no longer
significant at the two year mark.
Overall, patients have favorable outcomes with respect
to those initially driving with both treatments. Eighty-four
percent of patients in the IVR cohort and 89% of patients
in the PRP cohort were still driving by the end of the
study, with only 3% and 4% respectively having stopped
driving because of their vision respectively (NEI VFQ-25).
Although the study finds that the percentage of participants
who changed their driving habits because of their vision did
not differ between the treatment cohorts (NEI VFQ-25),
objective measures that could be relevant to driving, such
as vision of 20/40 or better in at least one eye and binocular
visual acuity better than 20/40, were slightly better in
the IVR cohort. Beaulieu et al. conclude that although
differences were identified in some work productivity and
driving-related outcomes were slightly more favorable
with IVR, most other patient-centered outcomes were
equivalent.
The data from this study also do not show a change in
perception of peripheral field (by NEI VFQ-25 or UABLLQ).
Moreover, there are no treatment cohort differences
in peripheral vision subscales from the patients who had
a history of PRP in the nonrandomized fellow eye. This
finding contrasts to the DRCR Protocol S report which
found mean peripheral visual field sensitivity loss was
−23 dB in the IVR cohort versus −422 dB in the PRP
group (P<0.001). It would have been interesting to know
the average visual field outcomes of the subset of patients
included in this study and whether they reflected those of
the larger cohort. While field loss was clinically measurable
in DRCR Protocol S, it is interesting that it was not
detectable as a subjective detriment by survey response.
This may represent issues with the sensitivity of the
questionnaires’ ability to identify subjective consequences
of peripheral field changes. Alternatively, the differences
identified by perimetry may not be clinically significant to
daily functioning after patient adaptation to a smaller field.
The lack of measurable deficits in perceived visual field is
consistent with the results from a smaller British study that
investigated the effect of PRP on Esterman binocular visual
field score and found only a modest loss of sensitivity at
6 months after treatment (8).
An unexpected finding in this trial is that driving
subscales (from the NEI VFQ-25) are better in eyes treated
with PRP (compared to IVR) if they have preexisting
DME but more favorable with IVR if they do not have
existing DME. This effect also applies to mobility scores.
This finding conflicts with the reports that PRP worsens
DME (9). It is unclear whether this effect represents
potential issues with the reliability of patients’ subjective
responses and requires further investigation in future
studies. Alternatively this might be a consequence of
limitations in performing post hoc analyses, especially in
smaller subsets.
It is also surprising that eyes treated with IVR report less
loss in work productivity (in the WPAIQ questionnaire)
than those treated with PRP. In addition, neither treatment
cohort shows a benefit in patient perceived absenteeism.
In DRCR Protocol S, patients treated with IVR had more
injections and doctor visits over the course of treatment (median 10 injections and 22 visits compared to a median
of 1 injection and 16 visits). This gap occurred even though
48% of patients randomized to PRP also required IVR for
the treatment of DME. It is possible that outside of the
realm of a protocol, in standard clinical practice, patients
treated with PRP would require even fewer visits which
could translate to less absenteeism.
There are previous reports of cost-utility analysis of
laser and anti-VEGF therapy in the treatment of DME (10,11) and PDR (12). Treatment with IVR is significantly
more costly than PRP in terms of gross healthcare dollar
expenditures. Markov models of the cost-utility of IVR
and PRP for the treatment of PDR do not account for the
non-reimbursable costs accrued by patients such as for
transportation to the doctor’s office (13) and missed work
which would be expected to increase the relative difference
in cost given the need for increased visits in patients
treated with IVR long-term. Assuming non-inferior visual
outcomes for the two treatments, the cost per quality
adjusted-life year clearly favors PRP (12). These differences
are more considerable when the cost of treatment is
extrapolated to lifetime therapy, with near a 10-fold
premium in cost per QALY (poorer cost-utility) with IVR
in comparison to PRP. The cost of therapy and physician
visit burden, and non-reimbursed expenses (14-18) involved
in more frequent office visits, may be important factors in
compliance, and should also be considered in addition to
objective and subjective visual outcomes associated with
anti-VEGF therapy or PRP when making shared treatment
decisions with patients.
In summary, Beaulieu et al. (2), report valuable patientcentered
outcomes from the DRCR Protocol S cohort (1).
Providing metrics indicative of patients’ daily function
and personal experiences is important to validate the total
outcome picture of a therapeutic approach as presented in
this context of PDR. While we acknowledge that survey
responses are subjective in nature and must be viewed
critically, it is reassuring that most of the differences
between the IVR and the PRP cohorts were not statistically
significant, or were minimal. It is especially reassuring that
these findings were found in a rigorously performed, wellcontrolled
clinical trial cohort with excellent compliance
and follow-up data. This serves to validate visual acuity
as a comprehensive surrogate measure for outcomes—
something that has historically been presumed, but not
proven, since the findings of Beaulieu et al. parallel the
initial reports on objective visual outcomes reported in
Protocol S.
