To determine incidence and risk factors for graft failure following penetrating keratoplasty (PK) in eyes with Ahmed valves (AV).
Retrospective, observational cohort study.
Patients who underwent PK after AV implantation (both performed at our institution through 2004) were studied. Intervals to graft failure (defined as either dysfunction [stromal thickening with retention of clarity] or decompensation [central microcystic edema or loss of clarity]) were analyzed using Kaplan-Meier technique. Risk factors for graft failure were analyzed using Cox proportional hazard models.
Included were 77 eyes (77 patients; first procedure 1993). Following PK, 40 eyes (52%) required increased numbers of glaucoma medications; 10 eyes (13%) required additional glaucoma drainage device(s). Graft failure at 1, 2, and 3 years was 42.4% (95% confidence interval: 32.0%-54.6%), 57.1% (45.6%-69.1%), and 59.1% (47.5%-71.2%), respectively. Prior PK (HR 2.38, P = .006) and stromal vessels (HR 2.90, P = .0005) were associated with increased risk of graft failure. Use of glaucoma medications (HR 0.27, P = .009) and evidence of lower intraocular pressures (IOP) during follow-up (excluding hypotony; HR 0.92, P = .010) were associated with reduced risk of graft failure. Endothelial rejection episodes were observed in 13 eyes (17%); however, rejection was not a risk factor for graft failure ( P = .98).
Long-term survival of corneal grafts is poor in eyes with AV. The majority of graft failures are associated with progressive loss of endothelial function, without observed immunologic rejection. Despite the presence of an AV, escalation in glaucoma therapy often follows PK; graft failure may be related to poor IOP control.
Many patients with glaucoma will develop corneal opacities, either because of endothelial failure resulting from multiple intraocular surgeries or because of the underlying causes of glaucoma, such as trauma, herpetic keratouveitis, or congenital anomalies. The glaucoma associated with these conditions is often refractory to medical therapies, requiring placement of a glaucoma drainage device (GDD) for intraocular pressure control. Thus, penetrating keratoplasty (PK) for visual rehabilitation is often performed on eyes that already have, or will eventually have, 1 or more GDD in place.
Glaucoma itself is considered a risk factor for corneal graft failure, and the presence of a GDD is believed to increase that risk further. In a case-controlled study, graft failure was more common in patients with a GDD than among those whose intraocular pressure (IOP) was controlled medically. That study could not confirm that GDD are directly responsible for graft failure (as opposed to being surrogates for some factor related to worse glaucoma), but there is evidence from other studies that the presence of a GDD does lead to endothelial cell loss. Various mechanisms for endothelial cell damage and loss have been postulated; they include mechanical trauma and immunologic damage attributable to alterations of the immunologic milieu of the anterior chamber. Corneal graft failure in the presence of a GDD is likely multifactorial.
The long-term survival of corneal grafts in eyes with GDD is generally believed to be poor, but the reported 1-year incidence of graft failure varies markedly in different studies, ranging from 8% to 51%. Many of these studies have involved heterogeneous patient populations, in terms of the type of GDD and the temporal relationship between placement of the GDD and performance of the PK.
The present study was designed to determine the incidence of corneal graft failure for a relatively homogeneous population of eyes that had undergone placement of a single type of GDD (Ahmed valves [AV]) prior to performance of PK as a separate procedure. We also sought additional factors that increased the risk for graft failure in this population.
We performed a retrospective chart review of patients who underwent PK in an eye that had a previously implanted AV (New World Medical, Inc, Rancho Cucamonga, California, USA). Included were all patients who had undergone both surgeries by a faculty member at the Jules Stein Eye Institute (authors A.J.A., J.C., R.C., A.L.C., G.N.H., S.K.L., B.J.M.) prior to December 2004, which provided at least 6 months of follow-up for each patient; we collected data through May 31, 2005, when the study was approved by the UCLA institutional review board.
Only 1 eye per patient was studied. If both eyes of a patient met inclusion criteria, the first eye with an AV to undergo PK was selected as the study eye. For eyes that had undergone multiple PK, the first PK to be performed after placement of an AV was considered the index PK. PK performed before the index PK were designated as prior PK for purposes of analysis. Patients who underwent PK and placement of an AV at the same time were included in the study only if they subsequently underwent repeat PK at a later date; in that case, the later PK was considered the index PK and the PK performed at the time of AV placement was considered a prior PK for purposes of analysis. If patients had multiple AV before the index PK, the AV that was placed most recently before the index PK was considered the index AV for purposes of analysis; all others were considered prior AV. The last examination before the index PK was considered baseline.
The following demographic information was collected for each patient at baseline: age, gender, and race/ethnicity. The following ophthalmic information was collected for each patient at the baseline examination: type of glaucoma (open angle, closed angle, uveitic, congenital); indication for PK; past ocular surgeries; number and location (anterior vs posterior chamber) of AV; glaucoma medications; lens status (phakic, aphakic, pseudophakic with anterior chamber intraocular lens [ACIOL], pseudophakic with posterior chamber IOL [PCIOL]); presence of corneal stromal vessels; and presence of peripheral anterior synechiae (PAS). The highest IOP recorded within 1 month of the index PK was considered the baseline IOP, whether or not it was measured at the baseline visit. Indications for PK were recorded as 1 of the following 6 categories: failed prior PK; noninflammatory corneal disease (including aphakic bullous keratopathy [ABK], pseudophakic bullous keratopathy [PBK], endothelial failure in phakic eyes, keratoconus, and noninfectious stromal scarring); active corneal disease (including bacterial keratitis); herpetic corneal disease (herpes simplex virus keratitis, varicella zoster virus keratitis); congenital corneal opacity; other.
The following data were collected from operative reports for the index PK: additional procedures performed at the same time (cataract extraction/IOL placement, secondary IOL placement, anterior vitrectomy, lysis of PAS, tube trimming/repositioning, retinal surgery, other); diameter of host corneal excision site; suture technique (running, interrupted, combination); operative complications.
With regard to follow-up, data were collected until graft decompensation, loss to follow-up, or the end of the data collection period; the following data were collected: highest and lowest recorded IOP; the maximum number of glaucoma medications used at any given time; the presence of postoperative complications; additional ocular surgeries performed; status of the corneal graft (always healthy, ever dysfunctional, ever decompensated); and whether sutures were removed. The interval from index PK was determined for each parameter recorded. Glaucoma procedures following the index PK were recorded as 1 of the following 3 categories: placement of an additional GDD; extraocular procedures (eg, covering an exposed AV); and intraocular procedures without placement of another GDD (eg, flushing of a tube, repositioning of a tube, removal of a GDD).
With regard to glaucoma medications, commercially available combinations were counted as 2 medications in data analyses. An oral carbonic anhydrase inhibitor was counted as 1 additional medication in analyses that considered total number of glaucoma medications used.
Visual acuity was not evaluated as an outcome measure because multiple factors may have affected vision in this patient population, including glaucomatous visual field loss and post-PK corneal astigmatism. The technique with which IOP was measured was not considered in data analyses.
Corneal graft failure was defined as either corneal graft dysfunction or corneal graft decompensation. Dysfunction was defined as the presence of stromal edema or peripheral microcystic edema (MCE), lack of central MCE, and retention of optical clarity as determined by slit-lamp biomicroscopy. Decompensation was defined as the presence of central MCE or loss of optical clarity as determined by slit-lamp biomicroscopy. In the absence of corneal graft failure, corneas were considered to be healthy. Time to the first observed sign of graft failure was defined as the interval from index PK to the examination at which either corneal graft dysfunction or corneal graft decompensation was identified, whichever was observed first.
Hypotony was defined as an IOP <5 mm Hg. Tube touch was defined as direct contact of any intraocular portion of the AV tube with either peripheral host or donor corneal tissue as determined on slit-lamp biomicroscopy.
Corneal graft rejection was defined as the sudden onset of corneal edema in the presence of anterior chamber inflammation (aqueous humor cells, keratic precipitates). Increased anterior chamber inflammatory signs without associated corneal edema, in a patient with a history of uveitis, was considered to be a flare-up of uveitis, and not a graft rejection.
A patient was considered to have PAS only if there were adhesions between iris and peripheral cornea that could be visualized directly through the slit-lamp biomicroscope, as gonioscopy was not performed on all patients.
Data Analysis and Statistical Techniques
Postoperative IOP during the first month after the index PK were excluded from analyses, to eliminate the influence of fluctuations in the immediate postoperative period. IOP as a risk factor for graft failure was difficult to study because of fluctuations in measurements over time, differential follow-up, and nonstandardized examination intervals; we therefore used the following factors as surrogates for IOP control: maximum recorded IOP during follow-up, minimum recorded IOP during follow-up, and number of glaucoma medications. Patients with hypotony were excluded from these analyses.
Statistical analysis was performed using statistical software, SAS version 9.1 (Cary, North Carolina, USA). Kaplan-Meier estimates were obtained for time to first observed sign of graft failure and progression interval from graft dysfunction to decompensation. Log-rank tests were performed to assess the difference in Kaplan-Meier curves for patients grouped on the basis of each risk factor. Cox proportional hazards regression models were used to analyze the relationship between time to corneal graft failure and multiple potential risk factors. A P value of <.05 was considered statistically significant. We considered those associations with P values between .05 and .10 to be weak.
We identified 77 patients who met inclusion criteria. All patients had been diagnosed with glaucoma on the basis of uncontrolled IOP, typical visual field changes, typical optic disc changes, or a combination of these factors. All index AV were Model S-2, and all had been placed in the anterior chamber. The first index PK was performed in August 1993. The mean interval from placement of the index AV to the index PK was 14.0 months (range, 1-71 months). Total follow-up for all patients after index PK was 157.9 person-years (PY).
Table 1 lists demographic and ophthalmic characteristics at baseline examination. The mean IOP of 16.5 ± 8.7 mm Hg was a 44% reduction from a mean of 29.6 ± 10.8 mm Hg prior to the index AV. Baseline IOP was below 21 mm Hg in 67 (82%) of the study eyes. Of those with IOP greater than 21 mm Hg, only 2 patients had baseline IOP above 30 mm Hg. One child with Peter anomaly was reported to have an IOP of 60 mm Hg during examination under anesthesia; after lowering of IOP, she underwent PK, anterior vitrectomy, and tube revision during the same anesthesia session. The other patient had an IOP of 42 mm Hg documented within 1 month of the index PK, but repeat IOP measurement immediately before index PK was not available; the patient underwent tube revision at the time of index PK.
|Patients for whom right eye was study eye||50 (65%)|
|Mean (SD)||54.2 years (24.7 years)|
|Median (range)||62 years (0.3–96 years)|
|Female gender||41 (53%)|
|Type of glaucoma|
|Open angle||23 (30%)|
|Closed angle||38 (49%)|
|Prior glaucoma surgeries other than index AV|
|One or more prior GDD||12 (16%)|
|One or more other procedures, no prior GDD||32 (42%)|
|3 or more||2 (2.6%)|
|Mean (SD)||16.5 mm Hg (8.7 mm Hg)|
|Median (range)||15 mm Hg (0–60 mm Hg)|
|Number of IOP medications (n = 76 patients)|
|3 or 4||9 (12%)|
|Pseudophakic with ACIOL||12 (16%)|
|Pseudophakic with PCIOL||31 (40%)|
|PAS present||39 (51%)|
|Corneal stromal vessels present||26 (34%)|
|Indication for index PK|
|Failed prior PK||33 (43%)|
|Pseudophakic bullous keratopathy||23 (29.9%)|
|Aphakic bullous keratopathy||10 (13.0%)|
|Endothelial failure in phakic eyes||2 (2.6%)|
|Corneal scar (traumatic, resolved ulcer)||3 (3.9%)|
|Congenital corneal opacity||5 (6.5%)|
|Inactive corneal scar in a patient with history of HSV-associated keratitis||1 (1.3%)|
a Number of patients/eyes (percentage) unless otherwise specified. One eye per patient was studied. Finding known for all patients unless otherwise specified.
Approximately half of study eyes had PAS or histories of angle closure or both. The majority (44 patients [57%]) had undergone prior glaucoma surgeries other than the index AV, and a majority (49 of 76 patients [64%]) were being treated with at least 1 glaucoma medication, despite the presence of 1 or more GDD. At the time of the index PK, the majority (44 patients [57%]) were phakic or had a PCIOL, providing a barrier to vitreous humor in the anterior chamber. The majority (68 patients [88%]) of index PK was performed for endothelial failure, either of the host cornea or a prior graft; no patient had active inflammatory disease.
Table 2 shows surgical factors related to the index PK. The majority (48 patients [62%]) had 1 or more additional surgical procedures at the time of the index PK, including intraocular manipulation of the AV tube (21 patients [27%]). In only 4 patients (5.2%) was the diameter of excised host tissue larger than 8 mm.
|Additional surgery performed with index PK|
|Cataract extraction/IOL implantation||5 (6.5%)|
|Secondary IOL implantation||11 (14%)|
|Anterior vitrectomy||14 (18%)|
|Lysis of PAS||24 (31%)|
|GDD tube trim or reposition||21 (27%)|
|Retinal surgery b||3 (3.9%)|
|Median diameter of excised host tissue (range)||7.5 mm (5.5–9.0 mm)|
|Interrupted only||62 (81%)|
|Combined interrupted/running||15 (19%)|
a Number of patients/eyes (percentage) unless otherwise specified. One eye per patient was studied. Finding known for all patients.
b Includes intraocular silicone oil renoval, pars plana vitrectomy with silicone oil placement, drainage of choroidal hemorrhage (1 eye each).
We observed signs of graft failure (either dysfunction or decompensation or both) of the index graft in 46 patients (60%). In 9 grafts, the first sign of graft failure was decompensation (without observation of a prior state of graft dysfunction). Graft decompensation was eventually observed in a total of 34 eyes. The incidence of graft failure (based on first signs of either graft dysfunction or decompensation) was 29/100 PY. The incidence of graft decompensation was 22/100 PY. The median time to the first observed signs of failure was 1.4 years (range, 0.03-6.0 years), and the median time to graft decompensation was 3.6 years (range, 0.2-6.0 years; Figure 1 ). Of the 35 grafts whose first observed sign of graft failure was graft dysfunction, the median interval from dysfunction to decompensation, based on Kaplan-Meier analysis, was 0.5 years (range, 0.02-3.3 years, Figure 2 ). Kaplan-Meier survival analysis showed the probability of a graft failure at 1, 2, and 3 years to be 42.4% (95% confidence interval [CI]: 32.0%-54.6%), 57.1% (95% CI: 45.6%-69.1%), and 59.1% (95% CI: 47.5%-71.2%), respectively ( Figure 1 ). Graft failure continued to occur after 3 years but the number of at-risk patients was small (9 eyes at 4 years; 4 eyes at 5 years).
Table 3 describes ophthalmic factors during follow-up; a substantial number of study eyes required glaucoma medications or additional glaucoma surgery during follow-up. There was an increase in the number of study eyes that required glaucoma medications during follow-up and an increase in the number of glaucoma medications used per eye. Overall, when compared to baseline, 40 eyes (52%) required a greater number of topical glaucoma medications, 2 eyes (3%) required fewer medications, and 35 eyes (45%) required the same number of medications during follow-up. Glaucoma surgical procedures were performed on 18 eyes (23%); 6 of these eyes required multiple such procedures. IOP above 21 mm Hg was recorded at least once during follow-up in 56 study eyes (73%).
|Length of follow-up b|
|Mean (SD)||2.1 years (1.8 years)|
|Median (range)||1.7 years (0.1–8.6 years)|
|IOP c (n = 75)|
|Mean (SD)||24.4 mm Hg (8.6 mm Hg)|
|Median (range)||23 mm Hg (4–52 mm Hg)|
|Mean (SD)||9.8 mm Hg (4.7 mm Hg)|
|Median (range)||9 mm Hg (1–30 mm Hg)|
|Maximum number of glaucoma medications at any point (n=75)|
|3 or 4||31 (42%)|
|Additional glaucoma surgeries|
|Exchange/revision of existing GDD||6 (7.8%)|
|Removal of existing GDD||3 (3.9%)|
|Placement of an additional GDD||10 (13%)|
|Conjunctival revision||5 (6.5%)|
|Additional nonglaucoma surgeries|
|Cataract extraction/IOL implantation||3 (3.9%)|
|Other d||15 (19%)|
a Findings for all 77 patients (77 eyes), unless otherwise indicated (number of patients (percentage), unless otherwise indicated).
b Length of follow-up was defined as the interval from index PK to graft decompensation or most recent examination, whichever occurred first.
c Excludes IOP during first month after index PK.
d Includes Nd:YAG laser capulotomy, pars plana vitrectomy, drainage of choroidal hemorrhage, repair of traumatic wound dehiscence.
Table 4 identifies complications following the index PK. Endothelial graft rejection was observed in 13 eyes; 2 patients were observed to have 2 discrete episodes of rejection each. Of the 13 eyes, 10 developed graft dysfunction within 0 to 4.5 years after the rejection episode, and 5 were observed to progress to graft decompensation. A total of 27 glaucoma-related complications occurred in 12 eyes (16%); only 4 of these eyes had a single glaucoma-related complication (AV exposure in 3 eyes; tube-cornea touch in 1 eye).
|Factor||Number of Eyes (%)|
|Yes a||12 (16%)|
|Tube touch b||5 (6.5%)|
|Hypotony c||9 (12%)|
|Impatent tube||3 (3.9%)|
|Choroidal detachment||3 (3.9%)|
|Exposed tube||7 (9.1%)|
|Infectious endophthalmitis||1 (1.3%)|
|Persistent epithelial defect||3 (3.9%)|
|Bacterial keratitis||3 (3.9%)|
|Herpetic keratitis||1 (1.3%)|
|Endothelial rejection||13 (17%)|
|Traumatic wound dehiscence||1 (1.3%)|
|Band keratopathy||2 (2.6%)|
|Wound leak||2 (2.6%)|
|Uveitis flare-up||1 (1.3%)|
|Retinal detachment||3 (3.9%)|
a Some study eyes experienced more than 1 glaucoma-related complication.
b Tube touch involved host corneal tissue only in 2 eyes.
Table 5 displays potential risk factors for failure of the index graft, based on univariate Cox proportional hazards regression models. The following factors at baseline were associated with a significantly increased risk of graft failure: prior failed PK (relative risk [RR] 2.38, 95% CI 1.28-4.43, P = .006) and the presence of corneal stromal vessels (RR 2.90, 95% CI 1.60-5.27, P = .0005). There were weak associations with the following factors at baseline: higher IOP (as a continuous variable; RR 1.03, 95% CI 1.00-1.06, P = .073, per mm Hg) and angle-closure glaucoma as the indication for index AV placement (RR 1.91, 95% CI 0.89-4.13, P = .099). Lysis of PAS during surgery was also weakly associated with an increased risk of graft failure (RR 1.80, 95% CI 0.98-3.33, P = .060), although the presence of PAS alone was not a significant risk factor (RR 1.55, 95% CI 0.85-2.81, P = .153). The following postoperative factors were associated with a significantly decreased risk of graft failure: use of glaucoma medications (HR 0.27, 95% CI 0.10-0.72, P = .009), a lower minimum IOP (HR 0.92, 95% CI 0.86-0.98, P = .01, per mm Hg), and suture removal (HR 0.32, 95% CI 0.16-0.65, P = .001). There was a weak association between decreased risk of graft failure and older age at baseline (RR 0.91, 95% CI 0.81-1.02, P = .095, per 10 years).
|Risk Factors||First Sign of Graft Failure a||Progression b|
|Relative Risk c||95% CI||P Value||Relative Risk c||95% CI||P Value|
|Gender (male vs female)||1.00||0.56, 1.82||.989||0.88||0.45, 1.74||.709|
|Age at index PK (per 10 years)||0.91||0.81, 1.02||.095||0.89||0.77, 1.02||.096|
|Baseline ophthalmic factors|
|Closed angle||1.91||0.89, 4.13||.099||0.87||0.36, 2.08||.752|
|Uveitic||1.50||0.46, 4.92||.500||0.31||0.06, 1.54||.153|
|Congenital||1.91||0.71, 5.17||.201||0.59||0.18, 1.91||.376|
|Indication for index PK|
|Noninflammatory corneal disease d||reference||reference|
|Failed prior PK||2.38||1.28, 4.43||.006||1.07||0.51, 2.22||.861|
|All other e||0.47||0.11, 2.07||.319||0.87||0.19, 4.02||.858|
|Interval from index AV to index PK (per month)||1.00||0.98, 1.02||.868||0.97||0.93, 1.01||.159|
|Prior glaucoma surgery (other than index AV)|
|One or more prior GDD||0.94||0.40, 2.21||.883||1.48||0.57, 3.82||.422|
|One or more other procedures, no prior GDD||0.73||0.39, 1.39||.343||0.64||0.40, 1.39||.263|
|IOP prior to index AV (per mm Hg)||1.00||0.97, 1.03||.907||0.99||0.96, 1.02||.621|
|IOP at baseline (per mm Hg)||1.03||1.00, 1.06||.073||1.02||0.98, 1.06||.249|
|Number of glaucoma medications at baseline|
|1||0.73||0.34, 1.57||.421||0.80||0.33, 1.94||.624|
|2 or more||0.77||0.39, 1.52||.444||0.50||0.22, 1.11||.088|
|Lens status at baseline|
|Aphakic||0.80||0.33, 1.97||.629||0.84||0.32, 2.19||.719|
|Pseudophakic with ACIOL||1.02||0.37, 2.84||.963||0.71||0.20, 2.44||.581|
|Pseudophakic with PCIOL||0.82||0.35, 1.89||.638||0.67||0.26, 1.73||.412|
|PAS at baseline||1.55||0.85, 2.81||.153||1.30||0.64, 2.61||.469|
|Corneal stromal vessels at baseline||2.90||1.60, 5.27||.0005||2.41||1.20, 4.82||.013|
|Diameter of excised host tissue (per mm)||0.73||0.42, 1.28||.271||0.68||0.43, 1.08||.105|
|Suture technique (combined vs interrupted only)||1.03||0.51, 2.09||.930||0.65||0.28, 1.50||.311|
|Lysis of PAS with index PK (yes vs no)||1.80||0.98, 3.33||.060||1.27||0.63, 2.54||.507|
|GDD tube trim or reposition with index PK (yes vs no)||0.92||0.47, 1.78||.797||0.92||0.42, 2.03||.831|
|Follow-up ophthalmic factors|
|Suture removal (yes vs no)||0.32||0.16, 0.65||.001||0.87||0.41, 1.82||.706|
|Post-op glaucoma medications|
|1 or more||0.27||0.10, 0.72||.009||0.27||0.09, 0.84||.023|
|Maximum IOP f (per mm Hg)||0.96||0.90, 1.02||.147||0.96||0.92, 1.02||.162|
|Minimum IOP f (per mm Hg)||0.92||0.86, 0.98||.010||0.91||0.86, 0.98||.006|
|Endothelial rejection episode (yes vs none)||0.99||0.46, 2.14||.978||0.62||0.24, 1.60||.319|
|Tube touch (yes vs no)||0.31||0.04, 2.32||.255||1.75||0.61, 5.04||.298|
|Any ocular surgery g (yes vs none)||0.67||0.32, 1.40||.291||0.51||0.23, 1.11||.090|
|Additional GDD (yes vs none)||0.70||0.25, 1.96||.504||0.22||0.05, 0.91||.037|