Strategy for the Management of Uncomplicated Retinal Detachments The European Vitreo-Retinal Society Retinal Detachment Study Report 1 Ron A. Adelman, MD, MPH,1 Aaron J. Parnes, MD,1 Didier Ducournau, MD,2 for the European Vitreo-Retinal Society (EVRS) Retinal Detachment Study Group* Objective: To study success and failure in the treatment of uncomplicated rhegmatogenous retinal detachments (RRDs). Design: Nonrandomized, multicenter retrospective study. Participants: One hundred seventy-six surgeons from 48 countries spanning 5 continents provided information on the primary procedures for 7678 cases of RRDs including 4179 patients with uncomplicated RRDs. Methods: Reported data included specific clinical findings, the method of repair, and the outcome after intervention. Main Outcome Measures: Final failure of retinal detachment repair (level 1 failure rate), remaining silicone oil at the study’s conclusion (level 2 failure rate), and need for additional procedures to repair the detachment (level 3 failure rate). Results: Four thousand one hundred seventy-nine uncomplicated cases of RRD were included. Combining phakic, pseudophakic, and aphakic groups, those treated with scleral buckle alone (n ¼ 1341) had a significantly lower final failure rate than those treated with vitrectomy, with or without a supplemental buckle (n ¼ 2723; P ¼ 0.04). In phakic patients, final failure rate was lower in the scleral buckle group compared with those who had vitrectomy, with or without a supplemental buckle (P ¼ 0.028). In pseudophakic patients, the failure rate of the initial procedure was lower in the vitrectomy group compared with the scleral buckle group (P ¼ 310e8). There was no statistically significant difference in failure rate between segmental (n ¼ 721) and encircling (n ¼ 351) buckles (P ¼ 0.5). Those who underwent vitrectomy with a supplemental scleral buckle (n ¼ 488) had an increased failure rate compared with those who underwent vitrectomy alone (n ¼ 2235; P ¼ 0.048). Pneumatic retinopexy was found to be comparable with scleral buckle when a retinal hole was present (P ¼ 0.65), but not in cases with a flap tear (P ¼ 0.034). Conclusions: In the treatment of uncomplicated phakic retinal detachments, repair using scleral buckle may be a good option. There was no significant difference between segmental versus 360-degree buckle. For pseudophakic uncomplicated retinal detachments, the surgeon should balance the risks and benefits of vitrectomy versus scleral buckle and keep in mind that the single-surgery reattachment rate may be higher with vitrectomy. However, if a vitrectomy is to be performed, these data suggest that a supplemental buckle is not helpful. Financial Disclosure(s): The author(s) have no proprietary or commercial interest in any materials discussed in this article. Ophthalmology 2013;120:1804-1808 ª 2013 by the American Academy of Ophthalmology. *Group members listed online (available at http://aaojournal.org).
The optimal treatment of rhegmatogenous retinal detachment (RRD) has been debated for decades. Different surgical procedures are available to the vitreoretinal surgeon, including scleral buckle, pars plana vitrectomy, and pneumatic retinopexy. However, the method of choice sometimes depends on the individual surgeon or institution as opposed to evidence-based outcomes. Retrospective studies in the literature, while demonstrating the overall high anatomic success rate of all methods, do not provide a consensus regarding the best procedure in terms of outcome.1e4 Few
1804
2013 by the American Academy of Ophthalmology Published by Elsevier Inc.
prospective, randomized clinical trials have been performed, and even they report conflicting outcomes and recommendations.5e8 It is obvious that additional data are needed, either prospective or retrospective, to help formulate accurate recommendations regarding optimal treatment selection. The gold standard for clinical studies is the prospective, randomized clinical trial. Overall, the treatment of RRD is very successful with a low failure rate. This necessitates that each treatment group be very large to attain statistical significance. In fact, the number of patients needed to conduct a meaningful
ISSN 0161-6420/13/$ - see front matter http://dx.doi.org/10.1016/j.ophtha.2013.01.070
Adelman et al
Repair of Uncomplicated Retinal Detachments
study may make the investigation very difficult and costly. Given this, other methods of analysis must be considered. The European Vitreo-Retinal Society (EVRS) is an organization of more than 1900 retina specialists from 75 countries organized in 2001. In 2010, a clinical study was announced to the members of EVRS to record the individual treatment successes and failures of the primary procedure for RRDs with proliferative vitreoretinopathy (PVR) ranging from grade 0 (no PVR) to grade C-1 PVR. A total of 176 surgeons from 48 countries provided information on 7678 RRDs with at least 3 months of follow-up. To our knowledge, this is the largest report of the treatment of retinal detachment in the literature. Herein we discuss the results of the treatment of RRDs in those eyes with no or mild PVR (grade 0 or grade A). Our primary focus in this analysis was uncomplicated retinal detachments without choroidal detachment, significant hypotony, or posterior, large, or giant retinal tears.
working sheet was sent to each contributor, masking the name of the other contributors, so that cleaning accuracy could be agreed on.
Success Criteria Because of the high percentage of success during the study, it was decided that the failure data were the most important parameter to evaluate. The failure rate was presented in 3 categories. The level 1 failure rate was the true failure rate declared by the surgeon and represented eyes with detached retina judged to be inoperable by the conclusion of the study. The level 2 failure rate was the percentage of eyes that had not been declared as a level 1 failure, but silicone oil was remaining in the eye at the conclusion of the study. In this level, we do not know if the result was a success or a failure. A level 2 failure group might have had a successful outcome after the removal of the silicone oil. The level 3 failure rate was the percentage of eyes that had been declared as a success but had a recurrence of the detachment or a complication after the initial procedure and required an additional surgery.
Statistical Analysis
Methods The EVRS Retinal Detachment Study was a nonrandomized, retrospective, multicenter study in which the goal was to analyze surgical strategy in RRD repair (grades 0, A, B, and C-1 PVR), focusing on the influence of initial clinical details and the primary surgical procedure on final anatomic results. This analysis was based on the choice of primary procedure and not on subsequent procedures performed, if indeed the initial procedure failed to repair the detachment. A request was made in 2010 for all members of EVRS who were interested in reporting the RRDs that they operated on from April 2010 through April 2011. A portal was created on the EVRS website that contained the reporting questionnaires to be filled out for each patient treated. By the cutoff of July 2011, the study organizers received complete data on 7678 RRDs operated on by 176 contributors from 48 countries, with follow-up ranging from 3 months to 1 year. The results were analyzed independently of the investigators by the French National Institute of Statistics and Economic Studies. Because this study was performed in 48 countries on 5 continents, the regulations and institutional review board requirements were different in different countries. Thus, each participant was responsible for following the rules and regulations of his or her own country and institution. The EVRS committees also approved the design and ethical aspects of the study. Because this was a nonrandomized study, this study carried a risk: even if the input was anonymous and data on all of the surgeries of each surgeon were requested, a small number of surgeons might have selected the cases that they wanted to contribute, therefore affecting the quality of the results. To face this problem, the Institute of Statistics, which analyzed the study, made 2 decisions. First, one should not present the results of a technique as an individual result, but always in comparison with at least 1 of the other techniques, so that the possibility of bias could have a comparative effect in both groups. Second, a high number of surgeons are required for each technique so that the effect of selection error will be negligible.
The National Institute of Statistics and Economic Studies first performed univariate and bivariate analyses for the entire database to have a graphical representation of the results. This was the first step taken to identify the variables that were linked to failure rate. Multivariate analysis then was performed. A step-by-step logistic regression was performed for the entire database on clinical findings and surgery parameters for identification of the variables that were linked independently to the failure rate. A logistic regression was performed in cases where vitrectomy was performed for determination of vitrectomy machine parameters and their association with failure rate. These data were reported independently and were combined to formulate a strategy based on the results. Statistical significance was defined as a 2-tailed P value of less than 0.05.
Results The details regarding the treatment and outcome of the primary procedure for 7678 cases of RRD was reported by 176 retinal surgeons from 48 countries. Of the 7678 cases reported, 2349 were grade 0 (30.6%), 2829 were grade A (36.8%), 1390 were grade B (18.1%), and 1110 were grade C-1 (14.5%). Baseline demographic data, including lens status and initial procedure performed, are displayed in Table 1.
Evaluation of the Variables Linked to Failure Rate An initial univariate analysis identifying major independent explanatory variables of the failure rate found that the presence of choroidal detachment and significant hypotony were associated with higher failure rates independent of other factors (Table 2, available at http:// aaojournal.org). Subsequent analysis of other variables was carried out in eyes without either choroidal detachment or significant hypotony. The relationship between PVR and failure rates was as expected, with increasing level of PVR corresponding to increasing Table 1. Baseline Demographic Patient Data
Surgery Reports Surgeons independently chose the surgical technique according to the clinical situation. For each case of operated RRD, they had to describe 25 items: 6 on the RRD clinical findings, 14 on the surgical procedure details, 1 on perioperative complications, and 4 on the postoperative results. After having cleaned the database, the global
Initial Procedure Performed Lens Status
Vitrectomy Alone
Vitrectomy With Scleral Buckle
Scleral Buckle Alone
Pneumatic Retinopexy
Phakic Pseudophakic
1159 1076
261 227
1103 238
92 23
1805
Ophthalmology Volume 120, Number 9, September 2013 failure rate. A comparison of grade 0 and grade A PVR is shown in Table 3 (available at http://aaojournal.org). No significant difference was found, and grade 0 and grade A PVR were grouped together. Therefore, in this article, we analyzed further the 5178 cases that were grade 0 or grade A PVR (67.4%). A bivariate analysis then was performed to find additional variables linked to failure rate. The impact of size of the largest retinal break on failure rate was investigated. Small and medium breaks were defined as less than 1 clock hour in size. Breaks between 1 and 3 clock hours were considered to be large, whereas those more than 3 clock hours were giant tears. In those with grade A PVR, when a small or medium break was present, the level 2 failure rate was 1.8%, but when a large or giant tear was present, the level 2 failure rate was 6.7%. This was a statistically significant difference (P ¼ 10e10), and the size of the retinal break was determined to be a major independent explanatory variable of the failure rate. Lens status and total number of tears were not linked to the failure rate. Next, a multivariate analysis was performed that included all of the clinical findings to discover the variables definitely linked to each failure rate (Table 4). Considering that the number of detached quadrants was not independent of time, it was not kept as criteria for analysis. Given this analysis, those eyes with choroidal detachment, significant hypotony, grade B or C-1 PVR, and a large or giant retinal tear were excluded. In addition, eyes with posterior breaks, including eyes with a macular hole, were excluded because a vitrectomy in those cases is mandatory. This left us with a pool of 4179 primary procedures for uncomplicated RRDs.
Evaluation of the Treatment of Uncomplicated Retinal Detachments The first analysis of this pool was to compare failure rates of those who were treated with vitrectomy, with or without a supplemental buckle, and those who were treated with scleral buckle alone. The true failure rate was higher in the group who underwent vitrectomy with or without a supplemental buckle, and this was statistically significant (P ¼ 0.04; Table 5, available at http://aaojournal.org). When comparing those who underwent vitrectomy alone (without buckle) with those who underwent a scleral buckle procedure alone, the level 1 failure rate was not statistically significantly different (P ¼ Table 4. Variables Independently Linked to the Rate of Failure
b Coefficient P Value Odds Ratio Variables independently linked to level 1 failure Grade C-1 PVR 4 Detached quadrants Choroidal detachment or hypotony Variables independently linked to level 2 failure (the rate of remaining silicone oil) Pseudophakia Grade B PVR Grade C-1 PVR 3 Detached quadrants 4 Detached quadrants Small break Choroidal detachment or hypotony PVR ¼ proliferative vitreoretinopathy.
1806
1.48 0.89 0.69
10e5 510e5 0.005
4.4 2.4 2.0
0.38 0.54 1.09 0.56 0.79 e0.47 0.70
0.001 0.0005 <10e5 0.0002 <10e5 0.0002 <10e5
1.5 1.7 3.0 1.8 2.2 0.6 2.0
0.134; Table 6, available at http://aaojournal.org). This can be explained by the following fact: when we compared the group who underwent vitrectomy alone with those who underwent vitrectomy with a buckle placed, it was found that a supplemental buckle did not help and was actually associated with an increased level 1 failure rate (P ¼ 0.048; Table 7, available at http://aaojournal.org). Some variability among those undergoing a scleral buckle procedure did exist. Either a segmental or encircling implant was used based on surgeon preference. It is important to point out that the failure rates of segmental versus encircling scleral buckle were not statistically significantly different, and therefore, these were grouped together (P ¼ 0.5; Table 8). The vitrectomy group was broken down further by the type of pump in the machine. Those eyes that underwent a vitrectomy using a vacuum control mechanism, or venturi pump, had statistically significantly higher level 1 and level 2 failure rates than those not undergoing vitrectomy (P ¼ 0.045; Table 9, available at http://aaojournal.org). Phakic and pseudophakic eyes with retinal detachment were investigated separately. When vitrectomy, with or without a supplemental buckle, was compared with scleral buckle alone in phakic eyes, just as before, the level 1 failure rate was significantly higher in the group who underwent a vitrectomy (P ¼ 0.028; Table 10). When the same comparison was made in pseudophakic eyes, no significant difference in the level 1 failure rate was observed. However, the number of eyes with remaining silicone oil was higher in the vitrectomy group (P ¼ 0.01). In both phakic and pseudophakic eyes, level 3 failure rates were significantly higher in cases where scleral buckle alone was used (P ¼ 0.029 and P ¼ 310e8, respectively). The use of pneumatic retinopexy then was evaluated in the setting of uncomplicated detachment with either a small tear or an equatorial hole. When compared with failure rates of scleral buckle without vitrectomy, pneumatic retinopexy had equivalent results when an equatorial hole was the culprit (Table 11, available at http:// aaojournal.org). However, when the detachment was associated with other types of breaks, pneumatic retinopexy had higher level 1 and 2 failure rates (Table 12, available at http://aaojournal.org).
Discussion The question remains for any particular RRD, what is the best treatment option: pneumatic retinopexy, scleral buckle, or vitrectomy? It is difficult to answer because there is no largescale clinical trial that has shown statistically significant differences among these treatment options. Randomized, masked clinical trials are the gold standard of clinical studies. The Scleral Buckling Versus Primary Vitrectomy in Rhegmatogenous Retinal Detachment study tried to answer the question of buckling versus vitrectomy, analyzing 681 retinal detachments.8 However, there was no conclusive recommendation. One problem with such a study is that one may interpret no statistical difference as being no Table 8. Failure Rates of Segmental Versus Encircling Scleral Buckle
Level 1 failure Level 2 failure Level 3 failure
Segmental Buckle (n [ 721)
360 Buckle (n [ 351)
P Value
1.0% 2.8% 21.8%
0.6% 0% 17.4%
0.5 0.3 0.093
Adelman et al
Repair of Uncomplicated Retinal Detachments
Table 10. Failure Rates of Vitrectomy, With or Without a Supplemental Buckle, Versus Scleral Buckle Alone in Phakic and Pseudophakic Eyes
Phakic eyes Level 1 failure Level 2 failure Level 3 failure Pseudophakic eyes Level 1 failure Level 2 failure Level 3 failure
Vitrectomy With or Without Buckle
Scleral Buckle Alone
P Value
1.3% 2.6% 11.4%
0.5% 0.4% 14.3%
0.028 10e5 0.029
1.1% 3.9% 10.5%
0.9% 0.5% 23.4%
0.81 0.01 310e8
difference at all. This leaves a surgeon to choose a strategy without much evidence. Some clinical trials report negative results because of inadequate power and insufficient sample size.9,10 In retinal detachment repair, the failure rate generally is low. Thus, a large number of cases are needed to detect a difference. To obtain a statistical difference (with a P value less than 0.05) between 2 techniques presenting, respectively, 1% and 2% failure rates, a minimum of 2400 cases is needed. Such a large number of subjects makes a prospective study prohibitively difficult and costly. We conducted a multicenter study recording treatment outcomes in uncomplicated RRDs. The failure rate of different treatment methods was evaluated. When comparing all of the patients who had undergone vitrectomy, with or without a supplemental scleral buckle, with those who underwent a scleral buckle procedure alone, our study suggests that applying a scleral buckle is better than performing a vitrectomy. Overall, the level 1 failure rate was significantly lower when a scleral buckle alone was used. This held true when phakic eyes were analyzed separately, but in pseudophakic eyes, treatment with vitrectomy did not lead to a higher level 1 failure rate. There is evidence in the literature implying that uncomplicated pseudophakic detachments are treated better with vitrectomy versus scleral buckle alone.11 However, in pseudophakic eyes in the present study, the rate of remaining silicone oil was significantly higher in those receiving vitrectomy. Level 3 failure rates, or cases that needed an additional procedure to repair the detachment, were significantly higher when scleral buckle alone was the primary procedure in both phakic and pseudophakic eyes. In phakic eyes, this can be forgiven considering that treatment with primary scleral buckle resulted in a lower final failure rate. In pseudophakic eyes, this presents a dilemma where eyes receiving scleral buckle alone more often needed a second procedure to repair the detachment and eyes treated with vitrectomy more often needed a second procedure to remove silicone oil. The importance of single-surgery reattachment in this context must be emphasized. Regardless of lens status, eyes undergoing vitrectomy had a higher single-surgery reattachment rate than those treated with scleral buckle alone. Considering that visual acuity data are not available, it is difficult to know if either successful primary repair or successful final repair is most important to functional outcome. It is entirely possible
to have a case of a successfully attached retina after initial failure that sustained a loss of function. With this information, it is reasonable to conclude that in phakic eyes, scleral buckle alone can be considered. However, in pseudophakic eyes, there is no clear winner and the surgeon should consider higher risk of level 3 failure with scleral buckle versus higher risk of level 2 failure with vitrectomy. Combining all uncomplicated RRD patients, when vitrectomy alone was compared with scleral buckle alone, there was a higher rate of remaining silicone oil after vitrectomy and no difference in the level 1 failure rate. The argument could be made that the increased risk of subjecting the patient to a second procedure to remove silicone oil makes vitrectomy less attractive. Interestingly, in these cases of uncomplicated detachments when a vitrectomy was performed, the cases treated with an additional scleral buckle demonstrated a higher failure rate than when a buckle was not placed. It is not entirely clear why patients fared worse when a buckle was added. It is possible that even those cases singled out as uncomplicated by this investigation could have been determined by the surgeon to be more likely to fail, and thus a combined vitrectomy and buckle procedure was performed. We attempted to minimize this sort of error by separating out cases with more extensive PVR. Similar results were seen for noncomplex pseudophakic retinal detachments where the addition of a buckle was not found to improve anatomic success rates but were associated with higher complication rates, including macular pucker, macular edema, and glaucoma.11e13 Pneumatic retinopexy was found to be equivalent to scleral buckle in cases where the retinal detachment was induced by atrophic holes. However, in cases where a flap tear is present, scleral buckle has a lower failure rate than pneumatic retinopexy. Given this information, pneumatic retinopexy may be considered in cases of retinal detachment with grade 0 or grade A PVR and atrophic holes. There are recognized inherent weaknesses in this sort of study. Using failure as a primary outcome measure is not optimal; however, it does serve as a method to compare treatment groups. Certainly using visual acuity as the primary outcome measure for success would have been preferred, but interpretation of nonstandardized acuities would be difficult. Another limitation is the short follow-up period. It is likely that level 2 failure rates would fall if the patients were followed up for longer periods. Also, as with any evaluation that involves self-reporting, the validity of the data relies on those reporting it. This may lead to selection bias and may affect the quality of the results. This problem can be minimized if a high number of surgeons are participating, as in this study, so that many surgeons will be involved in each technique and the effects of selection may be canceled or reduced. Randomization can solve all these issues; however, the study would be very expensive and lengthy when more than 5000 cases are to be included. Unfortunately, such a large-scale randomized clinical trial has not been performed for retinal detachment repair. Although pitfalls exist, this type of study has strengths. Considering that such a large number of physicians from 48 different countries spanning 5 continents participated in the study and a wide variety of techniques were performed
1807
Ophthalmology Volume 120, Number 9, September 2013 based on surgeon preference, the outcomes reported here are representative of practicing retinal surgeons globally. Given that the collection of these data was not limited to a few centers, or even to a few countries, the results are applicable to retinal surgeons worldwide. In summary, this large-scale study by 176 surgeons demonstrated that in cases of uncomplicated phakic retinal detachments, treatment with scleral buckle is a good option. Segmental buckle and 360-degree buckle were not statistically significantly different. For uncomplicated pseudophakic retinal detachments, the surgeon should balance the risks of vitrectomy versus scleral buckle and keep in mind that singlesurgery reattachment rates may be higher with vitrectomy. Also, the addition of a scleral buckle in cases where a vitrectomy was performed did not seem to be efficacious. Largescale, prospective, randomized clinical trials are needed to determine optimal treatment for retinal detachments. Acknowledgments. Anja Leppich, the EVRS administrative director, led the data collection. Marianne Borzic, Vice Director of the National Institute of Statistics and Economic Studies Pays de la Loire survey analysis department, led the statistical analysis of this study. Zofia Michalewska, MD, PhD, and Jack Sipperley, MD, contributed to the writing of EVRS retinal detachment manuscripts.
References 1. Afrashi F, Erakgun T, Akkin C, et al. Conventional buckling surgery or primary vitrectomy with silicone oil tamponade in rhegmatogenous retinal detachment with multiple breaks. Graefes Arch Clin Exp Ophthalmol 2004;242:295–300. 2. Miki D, Hida T, Hotta K, et al. Comparison of scleral buckling and vitrectomy for retinal detachment resulting from flap tears in superior quadrants. Jpn J Ophthalmol 2001;45:187–91. 3. Oshima Y, Yamanishi S, Sawa M, et al. Two-year follow-up study comparing primary vitrectomy with scleral buckling for macula-off rhegmatogenous retinal detachment. Jpn J Ophthalmol 2000;44:538–49.
4. Wolfensberger TJ. Foveal reattachment after macula-off retinal detachment occurs faster after vitrectomy than after buckle surgery. Ophthalmology 2004;111:1340–3. 5. Ahmadieh H, Moradian S, Faghihi H, et al, Pseudophakic and Aphakic Retinal Detachment (PARD) Study Group. Anatomic and visual outcomes of scleral buckling versus primary vitrectomy in pseudophakic and aphakic retinal detachment: six-month follow-up results of a single operationdreport no. 1. Ophthalmology 2005;112:1421–9. 6. Sharma YR, Karunanithi S, Azad RV, et al. Functional and anatomic outcome of scleral buckling versus primary vitrectomy in pseudophakic retinal detachment. Acta Ophthalmol Scand 2005;83:293–7. 7. Brazitikos PD, Androudi S, Christen WG, Stangos NT. Primary pars plana vitrectomy versus scleral buckle surgery for the treatment of pseudophakic retinal detachment: a randomized clinical trial. Retina 2005;25:957–64. 8. Heimann H, Bartz-Schmidt KU, Bornfeld N, et al, Scleral Buckling versus Primary Vitrectomy in Rhegmatogenous Retinal Detachment Study Group. Scleral buckling versus primary vitrectomy in rhegmatogenous retinal detachment: a prospective randomized multicenter clinical study. Ophthalmology 2007;114:2142–54. 9. Freiman JA, Chalmers TC, Smith H Jr, Kuebler RR. The importance of beta, the type II error and sample size in the design and interpretation of the randomized control trial. Survey of 71 “negative” trials. N Engl J Med 1978;299: 690–4. 10. Williams HC, Seed P. Inadequate size of ‘negative’ clinical trials in dermatology. Br J Dermatol 1993;128:317–26. 11. Arya AV, Emerson JW, Engelbert M, et al. Surgical management of pseudophakic retinal detachments: a metaanalysis. Ophthalmology 2006;113:1724–33. 12. Stangos AN, Petropoulos IK, Brozou CG, et al. Pars-plana vitrectomy alone vs vitrectomy with scleral buckling for primary rhegmatogenous pseudophakic retinal detachment. Am J Ophthalmol 2004;138:952–8. 13. Weichel ED, Martidis A, Fineman MS, et al. Pars plana vitrectomy versus combined pars plana vitrectomy-scleral buckle for primary repair of pseudophakic retinal detachment. Ophthalmology 2006;113:2033–40.
Footnotes and Financial Disclosures Originally received: August 18, 2012. Final revision: January 28, 2013. Accepted: January 30, 2013. Available online: April 16, 2013.
Presented at: European Vitreo-Retinal Society Annual Meeting, October 2011, Valletta, Malta. Manuscript no. 2012-1275.
1
Department of Ophthalmology & Visual Science, Yale University School of Medicine, New Haven, Connecticut. 2
EVRS, Nantes, France.
*A full listing of the European Vitreo-Retinal Society Retinal Detachment Study Group is available at http://aaojournal.org.
1808
Financial Disclosure(s): The author(s) have no proprietary or commercial interest in any materials discussed in this article. Correspondence: Ron A. Adelman, MD, MPH, Department of Ophthalmology & Visual Science, Yale University School of Medicine, 40 Temple Street, Suite 3D, New Haven, CT 06510. E-mail:
[email protected].