Article Text

Download PDFPDF
Changes in corneal endothelial cell density and the cumulative risk of corneal decompensation after Ahmed glaucoma valve implantation
  1. Kyoung Nam Kim1,
  2. Sung Bok Lee1,2,
  3. Yeon Hee Lee1,2,
  4. Jong Joo Lee1,
  5. Hyung Bin Lim1,
  6. Chang-sik Kim1,2
  1. 1Department of Ophthalmology, Chungnam National University Hospital, Daejeon, Korea
  2. 2Department of Ophthalmology, Chungnam National University College of Medicine, Daejeon, Korea
  1. Correspondence to Dr Chang-sik Kim, Department of Ophthalmology, Chungnam National University Hospital, #282 Munhwa-ro, Jung-gu, Daejeon 301-721, Korea; kcs61{at}


Aims To evaluate changes in the corneal endothelial cell density (ECD) and corneal decompensation following Ahmed glaucoma valve (AGV) implantation.

Methods This study was retrospective and observational case series. Patients with refractory glaucoma who underwent AGV implantation and were followed >5 years were consecutively enrolled. We reviewed the medical records, including the results of central corneal specular microscopy. Of the 127 enrolled patients, the annual change in ECD (%) was determined using linear regression for 72 eyes evaluated at least four times using serial specular microscopic examination and compared with 31 control eyes (fellow glaucomatous eyes under medical treatment). The main outcome measures were cumulative risk of corneal decompensation and differences in the ECD loss rates between subjects and controls.

Results The mean follow-up after AGV implantation was 43.1 months. There were no cases of postoperative tube–corneal touch. The cumulative risk of corneal decompensation was 3.3%, 5 years after AGV implantation. There was a more rapid loss of ECD in the 72 subject eyes compared with the 31 controls (−7.0% and −0.1%/year, respectively; p<0.001). However, the rate of loss decreased over time and statistical significance compared with control eyes disappeared after 2 years postoperatively: −10.7% from baseline to 1 year (p<0.01), −7.0% from 1 year to 2 years (p=0.037), −4.2% from 2 years to 3 years (p=0.230) and −2.7% from 3 years to the final follow-up (p=0.111).

Conclusions In case of uncomplicated AGV implantation, the cumulative risk of corneal decompensation was 3.3%, 5 years after the operation. The ECD loss was statistically greater in eyes with AGV than in control eyes without AGV, but the difference was significant only up to 2 years post surgery.

  • Glaucoma
  • Cornea

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

The corneal endothelium maintains corneal clarity by regulating stromal hydration. To perform this function, the corneal endothelium must consist of healthy cells above a given minimum cell density. Corneal endothelial cell loss is induced by various ocular interventions, including cataract surgery,1 trabeculectomy,2 laser iridotomy,3 cyclophotocoagulation,4 pars plana vitrectomy5 and glaucoma implant surgery.6–8 Endothelial cell loss after glaucoma implant surgery is a particularly significant factor, especially with respect to its continuity.6–8

Glaucoma implant surgery and trabeculectomy are the most commonly performed incisional procedures for the management of glaucoma. Although trabeculectomy has traditionally been preferred over glaucoma implant surgery, except in cases of refractory glaucoma carrying a high risk of trabeculectomy failure, recent survey studies have found an increasing preference among glaucoma surgeons for glaucoma implant surgery over trabeculectomy.9 ,10 Glaucoma implant surgery has shown good results, including for postoperative intraocular pressure (IOP) control, and the risk of complications was comparable to that for trabeculectomy.11–13 Furthermore, a tube versus trabeculectomy study recently reported a higher success rate and lower complication and reoperation rates for glaucoma implant surgery compared with trabeculectomy augmented with mitomycin C during a 5-year follow-up in patients with glaucoma who had previously undergone trabeculectomy and/or cataract surgery.13 Nevertheless, corneal endothelial damage (ie, corneal decompensation) was well known as one of the most characteristic long-term complications of the glaucoma implant surgery and almost twice as many patients in the implant group developed corneal decompensation compared with the trabeculectomy group (17 patients, 16%, vs 9 patients, 9%, respectively).13–15

Of the previous four studies on corneal endothelial cell change after glaucoma implant surgery, including two of our own,6–8 ,16 three studies concluded that there was continuous loss in endothelial cell density (ECD) with 1–2 years follow-up investigations.6–8 However, the patient populations were relatively small and the follow-up periods too short for a thorough evaluation of the long-term change of the ECD and risk of the resultant corneal decompensation. Furthermore, to our knowledge, there was no study that evaluated the change of the ECD and resultant corneal decompensation simultaneously following glaucoma implant surgery. Therefore, in the present study, we attempt to evaluate long-term (5 years) changes in the ECD and cumulative risk of corneal decompensation after Ahmed glaucoma valve (AGV) implantation with a larger subject group.


Study participants

Patients who had undergone preoperative specular microscopic examination and AGV implantation for refractory glaucoma >5 years previously were enrolled consecutively. Neovascular glaucoma, secondary glaucoma resulting from uveitis, ocular trauma or surgery, and glaucoma with a wide conjunctival scar from previous ocular surgery were included as refractory glaucoma. Surgery was performed if any of the following criteria were met: IOP >21 mm Hg and progressive optic nerve head damage, a retinal nerve fibre layer defect and/or a visual field defect. All surgeries were performed by a single experienced glaucoma specialist (C-sK) between 1 August 2003 and 30 April 2009 using a standard technique described previously.6 ,7 When inserting the silicone tube into the anterior chamber, a careful limbal puncture using a 23-gauge needle was first made to prevent leakage around the tube.17 Then the silicone tube (2 mm long) was inserted into the anterior chamber in a bevelled-up position, in parallel with, and as close as possible to but not in contact with, the iris plane. In 10 cases in which intraoperative shallowing of the anterior chamber occurred, the silicone tube and 5-0 nylon as an external stent were ligated with 8-0 vicryl to create a temporary occlusion suture. A portion of the 5-0 nylon was exposed through the conjunctiva and was removed within 4 weeks after surgery. In the present study, all 41 patients enrolled in our previous report7 were included to the study group. They underwent surgery between 1 August 2003 and 31 December 2005. In all patients, a 184 mm2 AGV implant model S2 or FP7 (New World Medical, Rancho Cucamonga, California, USA) was positioned in the superotemporal quadrant. The S2 and FP7 models have a silicone drainage tube of identical size (inner diameter, 0.305 mm; outer diameter, 0.635 mm). They also have the same-sized valve plate body, although the S2 model is made of polypropylene and the FP7 is made of silicone. When both eyes fulfilled the inclusion criteria, one eye was selected randomly for inclusion in this study. We excluded patients with congenital glaucoma, preoperative corneal decompensation, corneal endothelial cell disease, including Fuchs dystrophy, posterior polymorphous dystrophy and iridocorneal endothelial syndrome, previous penetrating keratoplasty, and with other corneal epithelial or stromal disorders, including scarring from previous corneal lacerations, which can influence the quality of specular microscopy. In addition, we excluded patients who had been lost to follow-up without any ocular problem within 6 months of surgery.

We retrospectively reviewed the medical records for age at surgery, sex, laterality of the operated eye, past medical history, types and number of previous ocular interventions, IOP, number of glaucoma medications (a fixed-combination drug was considered to be two different agents), best-corrected visual acuity (BCVA), glaucoma type, results of preoperative and postoperative specular microscopic examinations, follow-up months and the occurrence of corneal decompensation. Corneal decompensation was defined as a newly occurring corneal oedema following AGV implantation and persisting for >6 months. Data collection was stopped when any complication requiring surgical treatment developed or an additional IOP-lowering operation was required because of the possible effects on the corneal endothelium.

Specular microscopic examination

Noncontact-type specular microscopy (Noncon Robo SP-8000; Konan Medical, Tokyo, Japan) was performed on both eyes by an experienced examiner both before and periodically after surgery. This instrument automatically captures images of the endothelium once the subject fixates on a target. Then ECD (cells/mm2), percentage of hexagonal cells (an index of pleomorphism) and coefficient of variation in cell area (SD divided by mean cell area, %, an index of polymegathism) were determined semiautomatically; at least 50 contiguous endothelial cells centred on the screen were hand-marked and a computer algorithm calculated the values. This number (≥50) was specified to reduce sampling errors.18 ,19 We analysed the results on the central area of the cornea.

Statistical analysis

A Kaplan–Meier survival analysis was performed to define the frequency of corneal decompensation after AGV implantation in a total of 127 enrolled patients. In a subgroup analysis, the rates of the annual corneal ECD change (%) were determined using linear regression (the slope of each regression equation represented the rate of ECD change) for the 72 patients included in this evaluation of the 127 enrolled patients, who had been evaluated a minimum of four times using serial specular microscopic examination (of which the first, preoperative, and final, postoperative, measurements were a minimum of 2 years apart). This rate of ECD change was compared with the control group, which consisted of 31 fellow glaucomatous eyes treated using glaucoma medications rather than glaucoma implant surgery. All statistical analyses were performed using SPSS V.18.0 (SPSS, Chicago, Illinois, USA). The demographics were compared between the patients with and without corneal decompensation using the Mann–Whitney test or Fisher's exact test. In the subgroup analysis, Student's t test, χ2 test or Fisher's exact test was applied to compare the demographics and rate of annual ECD change (%) between the patient and control groups. Univariate regression analysis was used to find the variables associated with the annual ECD change rate (%). A value of p<0.05 was taken to indicate statistical significance.


In total, 127 eyes of 127 patients who had undergone AGV implantation >5 years previously were enrolled consecutively (mean age, 54.0±13.1 years; 87 males, 40 females). The mean follow-up was 43.1±20.5 months (range 6–60); with 67 eyes (53%) followed up 60 months. There were no definite complications affecting corneal endothelium such as tube–corneal touch and collapse of the anterior chamber. Table 1 provides the demographics of the enrolled patient.

Table 1

Demographics of patients with glaucoma with Ahmed glaucoma valve implantation

Corneal decompensation occurred in three patients during the follow-up period. Kaplan–Meier survival analysis of the corneal decompensation indicated a survival rate of 98.1% at 2 years (n=97) and 96.7% at 5 years (n=67) (figure 1). Demographics of the three patients with corneal decompensation are described in supplementary table 1 .

Figure 1

Kaplan–Meier survival curve showing survival for corneal decompensation. Among the total of 127 enrolled patients, the survival rate was 98.1% at 24 months (n=97) and 96.7% at 60 months (n=67).

In the subgroup analysis to determine the ECD loss rate, 40 eyes (55.6%) of the 72 AGV-implantation patients and 17 eyes (54.8%) of the 31 controls without AGV were followed up for 60 months. The mean durations from the preoperative baseline to the final postoperative specular microscopic examination for the subjects and the controls were 45.3 months (range 24–60) and 46.4 months (range 24–60), respectively, while the mean numbers of specular microscopic examinations were 6.0 (range 4–10) and 6.1 (range 4–10), respectively. The characteristics of the 72 subjects and 31 controls are summarised in table 2. There was a more rapid progressive loss of ECD in the patients with AGV implantation compared with the controls (−7.0±5.2 and −0.1±2.4%/year, respectively; p<0.001). With the inevitable exceptions such as the IOP (37.1±12.7 vs 15.0±3.8; p<0.001), the number of glaucoma medications (3.6±0.7 vs 1.9±1.1; p<0.001) and preoperative BCVA (0.6±0.5 logarithm of the minimum angle of resolution (logMAR) vs 0.1±0.5 logMAR; p<0.001), there were no significant differences between the two groups (all p>0.1).

Table 2

Demographics of the subject group with Ahmed glaucoma valve implantation and the control group (fellow eyes receiving glaucoma medications without Ahmed glaucoma valve implantation)

Of the 31 controls, two eyes received AGV implants due to visual field progression despite the IOP of the control being under 22 mm Hg: one eye at 26 months and the other eye at 38 months after enrolment. Cataract surgery was performed in four cases due to a progressive decrease in visual acuity. Data collection in these six patients was stopped before these surgical procedures. The difference in IOPs between subjects and controls at 6 months after AGV implantation was statistically significant, but was not clinically significant at the corneal endothelium (19.0 vs 15.7 mm Hg). Other postoperative time points did not show a significant difference in IOPs between the two groups (all p>0.05, table 3).

Table 3

Comparison of the IOP and the number of glaucoma medications between the AGV implantation group and the control group (fellow eyes that received glaucoma medications without AGV implantation)

Figure 2 shows the comparison of the ECD loss rates according to each follow-up time between the subject and control groups. The annual loss rate decreased with time: −10.7% from preoperatively to 1 year, −7.0% from 1 year to 2 years, −4.2% from 2 years to 3 years and −2.7% from 3 years to the final follow-up (mean of 45.3 months) and statistical significance compared with control eye disappeared after 2 years postoperatively (p<0.001, 0.037, 0.230 and 0.111, respectively).

Figure 2

Corneal endothelial cell densities with respect to the time interval following Ahmed glaucoma valve (AGV) implantation. The solid line and dashed lines represent the endothelial cell density in 72 patients with an AGV implant and 31 controls (fellow glaucomatous eyes without AGV implantation but receiving glaucoma medications), respectively. The annual rates of the endothelial cell density change, regression coefficient (B), for the subject group decreased gradually: −10.7% from preoperatively to 1 year (p<0.001, Mann–Whitney test); −7.0% from year 1 to the second year (p=0.037); −4.2% from the second year to the third year (p=0.230); and −2.7% from 3 years to final follow-up (p=0.111; mean follow-up, 45.3±20.6 months). SE bars are given at 1, 2 and 3 years.

A univariate regression analysis showed that the annual ECD loss rate (%) was significantly associated with the presence of an AGV implant (B=−6.895; p<0.001). In contrast, age, sex, laterality of the eye, diabetes mellitus, hypertension, presence of uveitis, the number and type of previous ocular interventions, IOP, the number of glaucoma medications, BCVA, glaucoma type, preoperative ECD, hexagonality of the endothelial cells and the coefficient of variation of the cell area of the corneal endothelial cells were not associated with the rate of ECD loss (table 4).

Table 4

Results of univariate regression analyses of the clinical variables associated with the rate of endothelial cell loss

We used the S2 and FP7 AGV implant models in this study. The S2 model was used in 77 (60.6%) patients, and the FP7 model was used in the other 50 (39.4%) patients. Two of the three patients with corneal decompensation used the S2; the other patient used the FP7. The S2 and FP7 models were used in 37 (51.4%) and 35 (48.6%) patients, respectively, in the subgroup analysis of 72 patients. We compared the rate of ECD loss in patients who used the S2 and FP7, but no significant difference was observed between the two groups at any time interval; from preoperatively to 1 year, from 1 year to 2 years, from 2 years to 3 years or from 3 years to the final follow-up (all p>0.100, Mann–Whitney test).


According to previous reports, glaucoma implant surgery is uniquely associated with continuous damage to the corneal endothelium6–8 ,14 ,20 contrary to other intraocular surgeries or laser therapy, including cataract surgery, trabeculectomy, laser iridotomy, cyclophotocoagulation and pars plana vitrectomy; such damage is thought to occur only during the treatment.1–8 To our knowledge, this is the first long-term study, >2 years postoperatively, on change of the ECD and resultant corneal decompensation simultaneously following glaucoma implant surgery. In the present study, an AGV implantation is associated with continuous ECD loss from preoperative baseline to 2 years postoperatively in patients with glaucoma. This result is comparable with previous studies. On the other hand, endothelial cell loss after 2 years of follow-up was not significant compared with controls.

Recently, there have been numerous reports of good surgical outcomes for AGV implantation;12 ,17 ,21–23 and not only has the frequency of glaucoma implant surgery increased significantly, its indications have also been broadened.11 ,24 However, corneal decompensation is a known leading cause of vision decrease and is one of the most important and frequent post-AGV-implantation complications.20 ,21 ,25 Reports of the frequency of corneal decompensation after glaucoma implant surgery with long-term (≥2 years) follow-up range from 5% to 27%.16 ,21 ,26 In the present study, corneal decompensation occurred in only three patients (2.4%), and its cumulative probability according to the Kaplan–Meier analysis was 3.3% at 5 years. We considered that the low incidence of corneal decompensation in the present study relative to those in other investigations might reflect subject characteristics. For example, Topouzis et al21 reported the development of corneal decompensation (including graft failure) in 16 eyes (27%) after AGV surgery, and among the 60 enrolled patients, 16 patients had had previous penetrating keratoplasty, among whom, complete graft failure occurred in nine eyes. In contrast, in our study, we excluded the patients with previous penetrating keratoplasty because, other than glaucoma implant surgery, penetrating keratoplasty is a well-known cause of continuous ECD loss.27 ,28 Another factor differentiating our study from others is that none of the cases in whom tube–corneal contact developed after AGV implantation. Tube–corneal contact is an important risk factor for endothelial damage.20 ,21 In particular, the relevant previous studies did not include preoperative baseline corneal endothelium values. Therefore, direct comparison of our study with the others is inappropriate.

Kalinina Ayuso et al29 reported a correlation between longer-duration uveitis and progressive ECD loss. In contrast, in the present investigation, uveitis displayed no association with the rate of ECD loss, and we thus suggest that its effect may have been obscured by the study's retrospective design, according to which data on the severity, number of recurrences and duration of uveitis were unavailable (only its history in our patients was considered).

Although many theories have been proposed, the exact mechanism causing corneal endothelial damage after glaucoma implant surgery remains unclear. To date, the proposed mechanisms of corneal endothelial damage are as follows: the jet flow around the tube end caused by the pulse, retrograde flow from the encapsulated reservoir to the anterior chamber, inflammation in the anterior chamber, intermittent tube–corneal contact and foreign-body reaction to the tube material.13 ,16 ,21 Although we cannot completely exclude the possible presence of the intermittent tube–corneal touch not discovered during office hours, there was no case with tube–corneal contact developed in our study. Therefore, we speculate that, in this situation, damage to the corneal endothelium after AGV surgery is prominent in the short term, within couple of years, then the rate of ECD loss is settled down, and the risk of the corneal decompensation caused by long-term progressive ECD loss after AGV implantation might not be as high as it is known.

This study had several limitations. First, although 53% of the enrolled patients were followed up for 60 months, a significant number were lost to follow-up. However, we consider that, to some extent, this dropout is inevitable in such long-term follow-up studies.13 Second, although we used the same fixation target and same way of marking of the central corneal endothelial cell at each examination, it is possible that we may not have examined the identical area of the cornea at every visit. Third, in evaluating the ECD loss rate and analysing its relevant clinical factors, we used the fellow eye of each subject as a control. This may not represent an ideal control group because there were differences in preoperative IOPs and numbers of antiglaucoma medications, which might have affected the corneal endothelium. However, all of the other demographic factors were well matched between the two groups except for the BCVA (table 2), and the follow-up values for IOPs and numbers of glaucoma medications were relatively well matched (table 3). Lastly, this study was retrospective in design; we had data only for the central cornea, which is the most important area clinically as it directly affects visual acuity, and not for other areas. Based on the results of our previous study, the ECD at the superotemporal cornea tube insertion site was expected to have more severe damage than the central area.6 ,7 The central cornea was probably less affected by intraoperative tube insertion and/or inserted tube status as regards the length of tube inserted and the distance from corneal endothelium than the superotemporal cornea. However, we expected the central ECD to reflect the superotemporal ECD to some degree because corneal endothelial cells are unable to reproduce but compensate for damage by sliding, rearrangement and enlargement. In other words, central ECD loss might be caused not only by central damage itself but also by compensation for the superotemporal area, through sliding, rearrangement and enlargement of corneal endothelial cells.30

In conclusion, we found that after AGV surgery, in patients with glaucoma without direct mechanical complication such as tube–corneal touch, the ECD loss from baseline value was statistically significant only until 2 years postoperatively and cumulative risk of corneal decompensation for 5 years is 3.3%.


View Abstract


  • Contributors Involved in design of study: KNK and C-sK. Data collection: HBL, SBL, JJL and YHL. Analysis and interpretation of data: KNK, HBL, SBL, YHL, and JJL. Writing of article: KNK. Critical revision of article: C-sK and SBL. Final approval of article: KNK, HBL, SBL, YHL, JJL and C-sK.

  • Competing interests None declared.

  • Ethics approval Institutional Review Board of Chungnam National University Hospital and conducted in accordance with all Declaration of Helsinki specifications.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles

  • At a glance
    Keith Barton Jost Jonas James Chodosh