Quantitative information, in the form of retinal thickness maps from optical coherence tomography (OCT), has become a critical element in the care of patients with retinal disease, including age-related macular degeneration (AMD). This is evidenced by more than 50 peer-reviewed articles published in the past 2 years describing OCT numerical information in patients receiving antivascular endothelial growth factor (anti-VEGF) treatment for neovascular AMD. In addition, these quantitative data have now become an important element in determining eligibility and making retreatment decisions in clinical trials.1
As further evidence of our increasing reliance on OCT retinal thickness maps, we have confirmed during informal surveys that many surgeons are using only a single printout containing the OCT map and a single OCT B-scan when assessing therapeutic efficacy among their patients.
OCT ACCURACY
The accuracy of these thickness maps, however, depends on an exact identification of the inner and outer retinal boundaries—currently performed automatically by the Stratus OCT (Carl Zeiss Meditec AG, Jena, Germany). When we examined this more closely, however, we found that boundary detection errors are quite frequent. In fact, in a study of 200 consecutive cases with OCT imaging for retinal disease, we found that >90% had some error and 33% had moderate-to-severe errors. It should be noted that the most severe and common errors were often found in patients diagnosed with choroidal neovascularization (CNV).
We also found that even when the Stratus OCT algorithms work as they should, the retina and subretinal fluid are lumped together in the thickness maps. Thus, even in the best-case scenario, the true retinal thickness is not quantified, and pigment epithelial attachments are ignored.
MANUAL ASSESSMENT
Because accurately assessing all of these different components is nearly impossible with current software algorithms, we have resorted to performing manual identification of the retinal boundaries using trained graders at the Doheny Image Reading Center (Los Angeles). The manual grading software (termed OCTOR) is also available online (www.driamd.org). OCTOR offers the ability to perform a detailed subanalysis and quantification of relevant OCT features and correction for Stratus OCT errors. Relevant OCT features include the true retinal thickness, retina cysts, subretinal fluid and tissue, and retinal pigment epithelium (RPE) elevations (Figure 1).
OCTANE (OCT Analysis Engine), a new software algorithm being developed by Dr. Alex Walsh and colleagues at the Doheny Eye Institute may provide an avenue for automated subanalysis and quantification. This algorithm still needs to be validated in prospective clinical trials, however. Accurate automated algorithms will be crucial in the upcoming era of spectral-domain OCT, where large data sets will be impractical for manual grading approaches.
Regardless of whether subanalysis and quantification is done manually (OCTOR) or automatically (OCTANE), the basic concept is the same (Figure 2). The boundary lines are identified on each of the six radial line scans, including the inner and outer retinal boundary, the subretinal tissue boundaries, the inner RPE boundary, and the estimated original location of the RPE (in cases with pigment epithelial detachments [PEDs]). This approach allows the surgeon to generate separate thickness—or better yet volume—maps of the various spaces and compartments on OCT such as the neurosensory retina, subretinal fluid, subretinal tissue, and pigment epithelial attachment. These various compartments can also be combined providing an overall parameter of total exudation in the eye. In addition, in intergrader reproducibility studies, we have shown that this type of grading can be performed in a highly reproducible fashion. The real question, however, is: Is this OCT subanalysis actually useful in evaluating our patients, and in particular studying the response to anti-VEGF therapies?
To answer this question, we performed two retrospective studies.
COMPARATIVE STUDY OF ANTI-VEGF THERAPIES
In the first study, we examined OCT data from consecutive patients with AMD being treated with intravitreal pegaptanib (Macugen; OSI/Eyetech, New York, NY) (n=18 patients), bevacizumab (Avastin; Genentech, South San Francisco) (n=35), or ranibizumab (Lucentis; Genentech) (n=35). Data were collected at baseline and at 3 months and include patient demographic data, automated Stratus OCT thickness maps, and raw OCT exports for OCTOR analysis.
The average age of patients receiving pegaptanib, bevacizumab, and ranibizumab was 79.7, 80.1, and 79.1 years, respectively. Seventeen percent of pegaptanib patients had previous treatments compared with 40% of patients receiving bevacizumab and 66% of patients receiving ranibizumab. Thus, the groups were not well balanced, which is not unexpected in a retrospective study design. At the 3-month follow-up, the mean total retinal volume (normal value approximately 7.00 mm³) decreased from 8.23 mm³ to 7.36 mm³ (P<.05, mean decrease 8.99%) after bevacizumab treatment. The mean total retinal volume decreased from 7.59 mm³ to 7.52 mm³ (not statistically significant or mean decrease of 0.78%) after pegaptanib treatment. Bevacizumab treatment resulted in a mean decrease of 0.14 mm³ for subretinal fluid, 0.10 mm³ for subretinal tissue, and 0.55 mm³ for PEDs. Patients treated with pegaptanib showed a mean decrease of 0.02 mm³ for subretinal fluid but a mean increase of 0.03 mm³ for subretinal tissue and 0.61 mm³ for PED.
When we examined the various OCT subcomponents of the CNV disease process, we discovered that pegaptanib-treated patients demonstrated an enlargement of PEDs in the short term, versus the bevacizumab patients who showed a reduction in all of the parameters. Most importantly, these differences in therapies were either not apparent or less apparent when evaluating the Stratus OCT data alone.
WHAT ABOUT RANIBIZUMAB?
In ranibizumab-treated patients, the mean total retinal volume decreased from 7.39 mm³ to 7.00 mm³ after treatment (P<.05, mean decrease 5.48%). Ranibizumab-treated patients also had a mean decrease of 0.32 mm³ for subretinal fluid, 0.04 mm³ for subretinal tissue, and 0.79 mm³ for PED (Figure 3). The magnitude of the reduction with ranibizumab appeared to be less than that of bevacizumab, however, it is critical to keep in mind that this was a small retrospective study, and the groups were not balanced at baseline.
In particular, the baseline OCT findings were quite different, with considerably more exudation present in the bevacizumab group compared with the other groups. Thus, the ranibizumab and pegaptanib groups, which had less exudation at baseline, also had less room for improvement. Still, it is reassuring to see that ranibizumab-treated patients show a reduction in all of the subcomponents that is similar to bevacizumab-treated patients. And again, these differences were not apparent when looking at the automated Stratus OCT data alone.
LONGITUDINAL STUDY
In addition to looking at the differences between the various pharmacological therapies, we wanted to see if we could learn something about the longitudinal effect of treatment on different components of the disease.
In a larger longer retrospective study of 110 consecutive patients undergoing ranibizumab therapy for neovascular AMD, we examined data from OCT subanalysis performed at baseline, 1 week, 1, 3, 6, and 9 months after initiation of therapy. The mean change in volume compared with baseline for each of the different components was used as the outcome measure.
In an OCT subanalysis of neurosensory retinal volume, we found a rapid reduction in retinal volume to a nadir by 1 month following ranibizumab therapy with a gradual "escape," or steady increase in retinal volume despite continued ranibizumab treatment (Figure 4). The retinal volume, however, still remained lower than at baseline. At this point, we are not sure of the cause of this apparent escape phenomenon and will probably not know for certain until it is addressed in future prospective dosing trials. Until then, we can speculate that it may have been caused by cystic retinal degeneration, tachyphylaxis to therapy, or an artifact of the as-needed dosing strategy used in treating these patients.
Analysis of subretinal fluid also showed a rapid decline to a nadir by month 1, but unlike the neurosensory retinal volume, this reduction appeared to be sustained. Subretinal tissue and PEDs demonstrated a more gradual and progressive reduction, and these parameters still appeared to be declining at the final month-9 time point in this study (Figure 5).
CONCLUSION
Both ranibizumab and bevacizumab were associated with a short-term reduction in volume in all of the various subspaces on OCT. Pegaptanib-treated eyes also showed a reduction in most subspaces, but also showed a short-term increase in PED volume (Figures 6 and 7). The effects of treatment were less apparent and inconsistent by automated Stratus OCT analysis.
Although no statistically significant differences between ranibizumab and bevacizumab therapy were observed, bevacizumab was associated with significantly greater reduction in exudation compared with pegaptanib. We need to be cautious of any comparison, however, due to the limitations of the retrospective study design. Nonetheless, the observations from this study suggest that clinicians should not rely on the Stratus OCT numerical data in assessing the response of their AMD patients to anti-VEGF therapy, as significant treatment effects may be completely missed. In addition, these studies suggest that quantitative OCT subanalysis can be very useful in studies of neovascular AMD. Hopefully, this type of analysis will be incorporated into future trials to more precisely define the anatomic affects of neovascular AMD therapies.
SriniVas R. Sadda, MD, is Associate Professor of Ophthalmology and Director of the Doheny Imaging Unit at the Doheny Retina Institute, Los Angeles. Dr. Sadda states that he has no financial interest in the products or companies mentioned. He may be reached at SSadda@doheny.org.
_1773249222.png?auto=compress,format&w=75)
