Vision Loss and Recovery After Trabeculectomy: Risk and Associated Risk Factors

Objectives To assess the risk of long-term vision loss after trabeculectomy and to determine the course of long-term vision loss vs transient loss and recovery.

Methods The medical records of 301 eyes in 262 patients undergoing trabeculectomy between January 1999 and September 2003 were retrospectively reviewed. Postoperative vision loss was categorized as mild or moderate (decrease in Snellen visual acuity of 3-5 lines) vs severe (decrease of >5 lines). Postoperative vision loss was considered permanent if visual acuity did not have a return of 3 lines within a 6-month follow-up period.

Results Permanent vision loss occurred in 24 of 301 eyes (8.0%): 13 (4.3%) had mild or moderate vision loss, and 11 (3.7%) had severe vision loss. Ten eyes (3.3%) with permanent mild or moderate vision loss and 6 eyes (2.0%) with permanent severe vision loss had no identifiable cause. Significant risk factors for permanent severe unexplained vision loss were preoperative split fixation on visual fields, preoperative number of quadrants with split fixation, and postoperative choroidal effusions with eventual resolution. Transient vision loss occurred in 170 of 301 eyes (56.5%): 79 (26.2%) had mild or moderate vision loss, with a mean time to recovery of 88 days (range, 6-720 days), and 91 (30.2%) had severe vision loss, with a mean time to recovery of 78 days (range, 6-720 days).

Conclusions Transient vision loss after trabeculectomy is common and may take up to 2 years for recovery. The risk of permanent vision loss is less common but significant. Two percent of our study population experienced permanent severe unexplained vision loss (“snuff-out”), and risk factors included preoperative split fixation on visual fields, preoperative number of quadrants with split fixation, and postoperative choroidal effusions with eventual resolution.

Clinical trials^1-4 have shown that the lowering of intraocular pressure (IOP) is associated with reduced progression of glaucomatous visual field (VF) loss. von Graefe^5,6 was the first to report that central vision may be compromised soon after surgery in eyes with chronic glaucoma and contracted VFs. Since then, the recommendation of glaucoma filtration surgery in patients with severely compromised VFs has been debated for fear of loss of central vision. Evidence for this so-called snuff-out phenomenon arose from earlier studies^7-11 that seemed to confirm the original observations by von Graefe. However, other investigators argued that they found no evidence for rapid postoperative progression of VF defects and that the risk of glaucoma surgery was warranted regardless of the severity of preoperative VF loss.^12-16

Of 6 studies^17-22 that evaluated the risk of severe loss of central vision after trabeculectomy, only half reported cases of snuff-out or severe unexplained vision loss, with Aggarwal and Hendeles¹⁷ reporting a rate of 7.69%, Langerhorst et al²⁰ a rate of 2%, and Costa et al²¹ a 0.95% qualified rate of “eyes at risk.” The observed variability in risk is likely because of the widely disparate sample sizes, population demographics, and methods used in these studies. A 2007 study by Law et al²² reported no occurrence of snuff-out, despite a large sample size and an inclusion criterion of “severe preoperative VF defects.”

Because this matter remains controversial, we undertook a large retrospective review using patient medical records at a tertiary academic glaucoma referral center to assess the risk and risk factors associated with long-term unexplained vision loss after trabeculectomy and to determine the course of long-term vision loss vs transient loss and recovery. To our knowledge, this is the first study to track both the occurrence of decreased vision after trabeculectomy and the course of recovery, with a differentiation of transient and permanent vision loss.

Participants were patients who underwent trabeculectomy with mitomycin C at the Doheny Eye Institute, Keck School of Medicine, University of Southern California, Los Angeles, an academic tertiary care center, between January 1999 and September 2003. The Institutional Review Board at the University of Southern California, Los Angeles, approved the study protocol, and all study procedures conformed to tenets of the Health Insurance Portability and Accountability Act and the Declaration of Helsinki for research involving human participants. All participants had provided written consent before trabeculectomy. Inclusion criteria were a minimum 6-month follow-up period and diagnosis of primary open-angle glaucoma, low-tension glaucoma, exfoliation glaucoma, pigmentary glaucoma, or chronic angle-closure glaucoma. Exclusion criteria were aphakia, prior intraoperative complications, other concurrent surgical procedure, or a diagnosis of neovascular glaucoma, pediatric glaucoma, or uveitic glaucoma. Preoperative characteristics were determined, including age, sex, race/ethnicity, diagnosis, lens status, prior filtration surgery, IOP, Snellen visual acuity (VA), cup-disc ratio, Humphrey VF (HVF) mean deviation, presence of split fixation based on HVF measurements, and number of quadrants with split fixation. Split fixation was defined as a sensitivity of less than 10 dB in any of the cardinal quadrants (supratemporal, infratemporal, supranasal, and infranasal) on HVF examination. Also obtained at intervals of 1 day, 1 week, and 1, 3, 6, 12, and 24 months were postoperative data, including VA, IOP, complications, choroidal effusions, flat or shallow anterior chamber, hypotonia (IOP ≤5 mm Hg), and postoperative procedures.

Data were examined for documentation of postoperative vision loss, which was categorized as mild or moderate vs severe. Postoperative mild or moderate vision loss was defined as a decrease in Snellen VA of 3 to 5 lines compared with preoperative measures. Postoperative severe vision loss was defined as a decrease in Snellen VA of more than 5 lines or as semiquantitative categories of low vision (eg, counting fingers, hand motion, light perception, and no light perception). Occurrences of permanent vision loss were then determined. Postoperative vision loss was considered permanent if Snellen VA did not have a return of at least 3 lines within a 6-month follow-up period. Preoperative, intraoperative, and postoperative variables were then examined among patients with permanent vision loss vs patients with transient or no vision loss. Postoperative vision loss was considered transient if a loss of 3 or more lines of Snellen VA had a return of at least 3 lines during the postoperative follow-up period. The following specific variables were examined: cup-disc ratio (<0.9 vs ≥0.9), preoperative IOP (<21 vs ≥21 mm Hg), HVF mean deviation (<−12 vs ≥−12 dB), and IOP at all postoperative intervals (≤5 [hypotonia]) vs >5 mm Hg). Determined for each postoperative time point was the occurrence of mild or moderate vs severe vision loss, as well as the mean (range) number of days until recovery of vision loss.

Statistical analysis was performed using commercially available software (SAS, version 9.2, for Windows; SAS Institute, Inc, Cary, North Carolina). Descriptive statistics were used to compare baseline demographics of the postoperative groups with permanent vision loss vs other (transient or no vision loss). Independent-samples t test was used to evaluate statistical significance of continuous variables for which there was a single measurement. To examine the significance of the association (contingency) between the 2 postoperative groups, the Fisher exact test was used for categorical data that resulted from binary classification. χ² Test was used to evaluate statistical significance of categorical variables for which there were large samples. Statistical significance was defined as P < .05.

Our study had 29 patients in whom both eyes were included. For these patients, we first analyzed the data using one eye at random and then pooled these data on finding that the results were not statistically significantly different.

Three hundred one eyes in 262 patients were included in the study, and all eyes underwent trabeculectomy with mitomycin C without reported intraoperative complications. Table 1 gives baseline demographics of the study cohort. Table 2 lists postoperative variables in the outcome categories of unexplained permanent vision loss vs other (including explained, transient, and no vision loss).

The Figure shows the categorical occurrences of vision loss. Twenty-four eyes (8.0%) had permanent vision loss, of which 13 (4.3%) had mild or moderate vision loss and 11 (3.7%) had severe vision loss. Ten eyes (3.3%) with permanent mild or moderate vision loss and 6 eyes (2.0%) with permanent severe vision loss had no readily identifiable cause on examination. A total of 170 eyes (56.5%) experienced transient vision loss, of which 79 eyes (26.2%) had mild or moderate vision loss and 91 eyes (30.2%) had severe vision loss, with a median time to recovery of 30 days (range, 6-720 days). A total of 107 eyes (35.5%) experienced no vision loss of more than 2 lines of Snellen VA throughout the 6-month follow-up period.

As summarized in Table 2, significant risk factors for permanent severe unexplained vision loss were preoperative split fixation on VFs and postoperative choroidal effusions with eventual resolution. Preoperative split fixation on VFs yielded an odds ratio of 6.23 (95% confidence interval, 1.36-28.53; P = .01) compared with the absence of preoperative split fixation. Postoperative choroidal effusions with eventual resolution yielded an odds ratio of 4.76 (95% confidence interval, 1.68-13.52; P = .006). The preoperative number of quadrants with split fixation was also significantly associated with permanent unexplained vision loss (mild, moderate, or severe vision loss vs other; P = .02).

No other variables examined were predictive of long-term vision loss. These included age, sex, race/ethnicity, diagnosis, lens status, preoperative IOP, cup-disc ratio, HVF mean deviation, prior filtration surgery, and hypotonia.

During the postoperative courses among 10 eyes with permanent unexplained mild or moderate vision loss, 2 patients underwent anterior wound revision and 5 patients underwent laser suture lysis of the scleral flap. Among 6 eyes with permanent severe unexplained vision loss, the postoperative courses were unremarkable for additional intervention or ocular events, except for 3 patients who underwent laser suture lysis of the scleral flap and 1 patient who underwent cataract extraction. One patient who underwent laser suture lysis at the 1-week postoperative visit developed a choroidal effusion but had complete resolution of the choroidal effusion on subsequent visits. These additional postoperative interventions had no identifiable adverse contribution to vision loss in either patient group.

Our findings suggest that snuff-out (or severe long-term unexplained vision loss), although rare, occurs after trabeculectomy with mitomycin C treatment and indicate that transient vision loss with recovery may be significantly more common than previously reported. To our knowledge, only 6 published studies^17-22 have examined the risk of snuff-out after trabeculectomy. However, these data are difficult to compare because concomitant procedures were included in some studies,^19-21 various VF defect requirements were used, and severe vision loss was inconsistently defined across studies, as summarized in Table 3.

Of these, 3 studies^18,19,22 reported no occurrences of snuff-out, with the authors concluding that sudden vision loss is an exceedingly rare complication of trabeculectomy in advanced glaucoma. In a retrospective review with a follow-up period of 2 months, Martinez et al¹⁸ showed that postoperative VA results in 52 of 54 eyes were the same as their preoperative baselines; the other 2 patients experienced only a 1-line decrease in Snellen VA. In a prospective study by Topouzis et al,¹⁹ 9 of 12 patients had both eyes included in the study, yielding a total of 21 eyes, a small number to detect a phenomenon that has long been regarded as rare at best.

In 2007, Law et al²² published a retrospective analysis that included strictly defined VF defect requirements for inclusion in the study (on HVF examination, a sensitivity of ≤5 dB in >85% of test points or >75% of test points that include 3 of 4 central test points). The authors reported no occurrences of severe unexplained vision loss but identified 7 cases of severe vision loss attributable to hypotonia maculopathy (3 patients), uncontrolled IOP (2 patients), cataract increase (1 patient), and severe inflammation (1 patient). However, because of the strict inclusion of patients with only specifically defined severe preoperative glaucomatous VF defects, the observers may have inadvertently excluded other patients with mildly involved fixation who may have experienced permanent vision loss. Increased rates of surgical complications and higher preoperative IOPs were the only measures found to be statistically significantly correlated with loss of central vision.

The findings of 3 studies^17,20,21 showed a small but measurable risk of severe unexplained vision loss after trabeculectomy. In a prospective study of 26 patients with primary open-angle glaucoma, Aggarwal and Hendeles¹⁷ reported 1 case of severe unexplained vision loss and 2 cases of severe vision loss attributed to cystoid macular edema and persistent hypotonia among 9 of 26 eyes (34.6%) that experienced reduction in VFs. In a large-scale retrospective study without formal preoperative VF requirements, Costa et al²¹ reported that 4 of 508 eyes experienced central vision loss without clear cause; however, application of the data is made difficult because of the use of a control group composed of 85 eyes that were randomly selected among 466 eyes that had not experienced sudden vision loss. On the basis of their findings, Langerhorst et al²⁰ came to the same broad conclusion as our group in that most patients having glaucoma with central islands had a transient loss of visual function after surgery; however, there is uncertainty as to whether their single case of snuff-out occurred secondary to trabeculectomy or concomitant cataract surgery.

Our review had no formal VF requirements and included only patients undergoing trabeculectomy with mitomycin C treatment. Among 301 eyes in 262 patients, 6 of 11 eyes had permanent severe unexplained vision loss. This 2.0% proportion is consistent with the summary given in Table 3, in which the proportion of permanent severe unexplained vision loss ranges from 0% to 3.8%.

Permanent vision loss, whether mild, moderate, or severe, occurred in 8.0% of eyes herein, which is significant and should be included in patient education about the risks and benefits of trabeculectomy. Among 8 eyes with permanent explained vision loss, there were 4 cases of hypotonia maculopathy, 2 cases of corneal opacification, 1 case of severe inflammation, and 1 case of retinal hemorrhage.

In the Tube Versus Trabeculectomy Study, Gedde et al²⁴ found that, after trabeculectomy with mitomycin C treatment, transient choroidal effusions significantly increased the risk of vision loss, which we also found on univariate analysis in our study. However, unlike Gedde et al, we chose not to classify postoperative choroidal effusions as an explanation of vision loss because we analyzed vision loss only after the effusions had completely resolved and because only 7 of 47 patients with transient postoperative choroidal effusions experienced permanent vision loss. However, we recognize that there is a statistically significant risk of permanent vision loss associated with choroidal effusions.

Because of our long follow-up period of up to 24 months, we were able to track recovery of VA. Of 170 eyes (56.5%) having transient vision loss, 79 (26.2%) with mild or moderate vision loss had a mean time to recovery of 88 days (range, 6-720 days), and 91 (30.2%) with severe vision loss had a mean time to recovery of 78 days (range, 6-720 days). Whereas we excluded eyes that underwent concomitant procedures to avoid potentially confounding risk factors, we included eyes with VF defects that did and did not involve fixation so that we might observe the full spectrum of postoperative vision loss.

We assigned precise definitions in our study to allow clinically relevant comparisons between transient vs permanent vision loss and between mild or moderate vs severe vision loss. Vision loss was considered mild or moderate when Snellen VA decreased by 3 to 5 lines and was considered severe when Snellen VA decreased by more than 5 lines. We considered vision loss permanent if there was no return of at least 3 lines within a 6-month follow-up period. Because only 3 of 301 eyes had a preoperative VA of 20/25 or better, our sample was similar to that in the study by Law et al.²²

Our study is limited by its retrospective design and by the lack of regular postoperative HVF examinations. The decision to pursue trabeculectomy was made by the treating physician on the basis of overall patient status. In addition, because snuff-out is so rare, it was necessary to include both eyes from some patients to gain proper statistical power for this retrospective review. However, there was no statistically significant difference in the results when only one eye of these patients was randomly selected for the analysis. There may have also been a bias toward underestimating the incidence of snuff-out if surgeons had a preconception that this may occur in patients with advanced VF loss and consequently did not operate as frequently on these patients.

In conclusion, our findings suggest that unexplained permanent vision loss (or snuff-out) occurs after trabeculectomy with mitomycin C treatment. Risk factors for long-term vision loss are preoperative split fixation on VFs, preoperative number of quadrants with split fixation, and postoperative choroidal effusions with eventual resolution. Transient vision loss is common and may take up to 2 years for recovery.

Correspondence: Brian A. Francis, MD, MS, Doheny Eye Institute, Department of Ophthalmology, Keck School of Medicine, University of Southern California, Los Angeles, 1450 San Pablo St, Ste 4804, Los Angeles, CA 90033 (bfrancis@usc.edu).

Submitted for Publication: July, 2, 2010; final revision received November 91, 2010; accepted December 15, 2010.

Funding/Support: This study was supported in part by grant EY03040 for basic research from the National Eye Institute.

Online-Only Material: This article is featured in the Archives Journal Club. Go here to download teaching PowerPoint slides.