Rate of Corneal Grafting Post-Glaucoma Drainage Device Use in Pediatric and Adult Patients.

PRCIS
Kaplan-Meier analysis was used to compare the rate of corneal grafting after glaucoma drainage device placement in pediatric and adult patients. Adults were at an increased risk of receiving a corneal graft after device placement.


PURPOSE
The goal of this study was to compare the rate of corneal graft implantation after glaucoma drainage device placement in pediatric and adult patients.


PATIENTS
Patients receiving a glaucoma drainage device between January 1, 1985 and December 31, 2017 were selected from the medical records. Patients receiving their first device while <18 years of age were considered children for the extent of the study.


METHODS
We compared the rate of receiving a corneal graft after glaucoma drainage device implantation using Kaplan-Meier analysis. Baseline patient characteristics and surgical characteristics were compared using a generalized estimating equation.


RESULTS
Corneal grafting occurred in 8.6% of adults and 4.7% of children from the original cohort.The rate of receiving a corneal graft at 5, 10, and 15 years was 9.4%, 16.8%, 39.4% and 1.6%, 1.6%, 12.5% for adults and children respectively. However, certain characteristics were different between the two groups.


CONCLUSION
Adult patients were more likely to receive a corneal graft after glaucoma drainage device placement. However, pediatric and adult patients differed in the prevalence of preexisting corneal disease, glaucoma type, glaucoma drainage device type used, and types of previous surgeries. Elucidation of the impact of these factors on corneal graft rate requires a larger cohort size.

T he use of glaucoma drainage devices (GDDs) is increasing. 1 However, GDDs are associated with reduced endothelial cell density 2 and corneal decompensation. [2][3][4][5][6][7][8][9][10][11][12] Many studies have looked at the frequency of corneal decompensation after GDDs. [2][3][4][5][6][7][8][9][10][11][12] However, there is limited data that compares the frequency of corneal decompensation after GDDs between adults and children. 5 Mandalos and Sung found no statistically significant difference in the rate of corneal decompensation between children and adults. 5 However, their study was limited in its power because of their sample size. Furthermore, given that the primary purpose of their study was to characterize the outcomes of GDD placement between children and adults, survival analysis was not used to characterize risk of corneal decompensation, which adjusts for variable follow-up time.
The mechanism for corneal decompensation associated with GDD placement is not known. However, biochemical and mechanical mechanisms including promotion of turbulent flow or the introduction of inflammatory or oxidative mediators into the anterior chamber leading to endothelial cell damage may play a role. 13 In addition, corneal edema and subsequent decompensation may occur in children due their active lifestyle which may increase intermittent tubecorneal contact. This can be complicated by impaired detection of early corneal edema because of difficulty examining a child. 4 Furthermore, children have a more elastic sclera which can result in greater changes in tube position. The tube position may also change as the child's eye grows with age. 5 Given these risk factors, we hypothesized that children may be more likely to receive cornea graft implantation after GDD placement. The purpose of this study was to characterize the rate of corneal grafting in children and adults using survival analysis.

Patient Selection
This was a retrospective cohort study at a single institution which included adult and pediatric patients that underwent surgery for GDD implantation between January 1, 1985 and December 31, 2017. Any patient that received a GDD while <18 years old was classified as a pediatric patient for the extent of the study. Patients receiving a corneal graft before GDD placement and those with a keratoprosthesis were excluded. For surgeries completed outside Mayo Clinic with a missing date, the surgical date was estimated using available exam information. One patient with a prior GDD placed at an outside institution was excluded as no reasonable estimate could be made.

Surgery
GDD type used and tube position were decided by the operating surgeon. GDD implants used were the Baerveldt Glaucoma Implant (Advanced Medical Optics, Santa Ana, DOI: 10.1097/IJG.0000000000001755 CA), Ahmed glaucoma valve (AGV) (New World Medical Inc., Rancho Cucamonga, CA), and Schocket tube. All of the GDD surgeries for the pediatric patients used in this study were conducted by staff physicians. A total of 530 of the GDD surgeries in adult patients were conducted by staff physicians and 27 were conducted by outside surgeons.

Outcome Measures
Baseline characteristics of patients including sex, race, age, prior surgeries, preexisting corneal disease, and glaucoma type were recorded. Rate of corneal grafting after GDD placement was the outcome of interest.

Statistical Analysis
Categorical variables are presented with the frequency and the percentage. Continuous variables are summarized using the mean and standard deviation, unless noted otherwise. Baseline characteristics were compared between the groups using generalized estimating equation models. These models were used to attempt to account for the potential correlation between eyes from the same patient. The rate of corneal grafting was estimated using Kaplan-Meier method. Potential risk factors of corneal grafting were assessed using Cox proportional hazards models. Sandwich estimators were used to account for the multiple eyes in the analysis for some of the patients. All analysis was completed using SAS version 9.4 (Cary, NC).

RESULTS
A total of 621 eyes (64 pediatric and 557 adult) were included from 481 adult and 44 pediatric patients. The median age (standard deviation) was 68 (17) years for adults and 8 (6) years for children. The median follow-up time (standard deviation) for adult and pediatric patients was 34 (52) months and 94 (61) months, respectively. A majority of adult (89%) and pediatric (88%) patients were white. The proportion of female and male patients between the 2 groups was similar and evenly distributed (P = 0.40, Table 1). However, glaucoma type differed between the 2 groups. Open-angle glaucoma (P = 0.05) was more common in adults, while congenital glaucoma (P < 0.001) and uveitic glaucoma (P < 0.001) were more common in children. In this cohort, the most common diagnosis for adults was open-angle glaucoma, while uveitic glaucoma was the most common diagnosis in pediatric patients ( Table 2). A greater proportion of adults (22%) had preexisting corneal disease ( Table 3, P = 0.05). Furthermore, adults had a greater number of laser surgeries for glaucoma (ie, selective laser trabeculoplasty, argon laser trabeculoplasty, laser peripheral iridotomy) (mean number 0.51 vs. 0.03 in adults and children, respectively, Table 4, P < 0.001).
There was a significant difference in GDD type used in pediatric and adult patients. Baerveldt 350 (P < 0.001) and Ahmed FP7 (P = 0.01) implants were more common in adults, while the Baerveldt 250 (P < 0.001) and Ahmed S2 (P < 0.001) implants were more common in pediatric patients ( Table 5). The superior temporal position was preferred for plate placement for both children and adults ( Table 5). The tube was most commonly placed in the anterior chamber and no difference was noted in the proportion of pediatric (86%) and adult patients (86%) with anterior tube placement (data not shown, P = 0.99).
Corneal grafting occurred in 48 eyes (8.6% of the original cohort) in the adult group and 3 eyes (4.7% of the original cohort) in the pediatric group. The risk of corneal grafting after GDD implantation was greater in adult patients ( Fig. 1, P = 0.006). The rate of receiving a corneal graft at 5, 10, and 15 years was 9.4%, 16.8%, 39.4% and   1.6%, 1.6%, 12.5% for adults and children, respectively ( Table 6). The median time to graft was 36 months in adults and 163 months in pediatric patients. When comparing GDD tube location in the anterior chamber versus pars planar, sulcus, or unknown placement, a statistically significant difference is observed in the rate of corneal grafting between adult and pediatric patients with placement in the anterior chamber ( Fig. 2A, P = 0.02) but not between the 2 groups when the tube is placed outside the anterior chamber ( Fig. 2B, P = 0.73).

DISCUSSION
In our study, adults were more likely to have a corneal graft after GDD implantation. Mandalos and Sung conducted a study that compared the frequency of corneal decompensation between children and adults directly. Their study did not show a difference in the frequency of corneal decompensation between children and adults, but was limited in power. 5 Most previous studies calculate the raw frequency of corneal decompensation to be between 2.2% and 11.9% for children 3-5 and 2.1% to 16% in adults. 5,7,[10][11][12] However, this method does not control for loss to follow-up. The frequency of corneal grafting in our cohort was similar to some of these studies (8.6%). However, since this method is impacted by follow-up, a longer follow-up time could result in detecting a greater number of events. Survival analysis allows for correcting for loss to follow-up to reduce the impact of differences in follow-up times.
Using Kaplan-Meier analysis, adults were found to be more likely to have a corneal graft following GDD placement. A previous study had a 3.3% 5-year cumulative risk of corneal decompensation. 2 This is lower than the 9.4% 5-year cumulative risk of corneal grafting in our study. However, our study contains patients that were referred after GDD placement. It is possible that this difference could be because of a referral bias with more complicated cases being at a greater risk of corneal decompensation. When adults are stratified based on where they received their surgery, no difference is noted in the probability of receiving a corneal graft (P = 0.13, data not shown). However, it is possible that the higher rate of corneal grafting could be because of a selection bias where more complicated patients are remaining within the health system for follow-up and those less likely to receive a graft are lost to follow-up.
Given the presence of a foreign body in the anterior camber, it is believed that anterior chamber placement of the GDD tube is likely to increase the likelihood of corneal decompensation and subsequent need for a corneal graft compared with pars planar or sulcus placement of the GDD    tube. 14 In fact, stratification by GDD tube location showed that GDD tube location acts as an effect modifier. When the tube is placed in the anterior chamber a difference is observed in pediatric and adult patients but not for tube location outside the anterior chamber (Fig. 2).This may be because of mechanical and biochemical changes observed in the aqueous humor of patients with GDD tubes placed in the anterior chamber. 15 Despite our original hypothesis, it appears that adults have a higher rate of corneal grafting. Various factors related to endothelial cell density, including glaucoma and age, could explain this increased rate. For example, the magnitude of intraocular pressure and duration of elevated intraocular pressure have been associated with reduced endothelial cell density. 16 Likewise, a continual reduction of endothelial cell density of 0.6% per year is observed above the age of 18. 16 In conjunction with the higher prevalence of preexisting corneal disease in adults (Table 3), it is reasonable to assume that on average adults in this cohort are starting with a lower baseline endothelial cell density given their age and longer period of time living with elevated intraocular pressure. However, when controlling for preexisting corneal disease there is still a significant difference in the rate of corneal grafting between children and adults (P = 0.01, data not shown). However, assessing baseline endothelial cell density was not considered standard of care; thus, there is limited data available for comparison between the 2 groups. Further studies recording baseline endothelial cell density between the 2 patient groups are necessary to determine if endothelial cell density before surgery carries another prognostic benefit. There was also a difference in GDD type used between pediatric and adult patients. Adults were more likely to receive a Baerveldt 350 or an Ahmed FP7 implant and pediatric patients were more likely to receive an Ahmed S2 or a Baerveldt 250 implant (Table 4). While this preference makes sense given the smaller eye size in children, 5 it is important to consider the GDD type used as a potential confounding factor. In addition to being a confounding factor between the groups, GDD type may impact outcomes in pediatric patients. 5 As such, we attempted to stratify pediatric patients by GDD type. However, there were not enough events (corneal grafting) in each group to allow for statistical analysis to stratify by GDD type in the pediatric group. When adults were stratified by the 3 most common GDD types no difference was observed in the rate of corneal decompensation (Supplemental Fig. 1, Supplemental Digital Content 1, http://links.lww.com/IJG/A508, P = 0.51). This suggests that GDD type does not appear to impact graft rate within adults. However, this does not preclude the impact that GDD type can have on the difference in the rate of corneal decompensation between pediatric and adult patients. In addition, the groups differed on the number of previous surgeries (Table 4) and glaucoma type ( Table 2). The impact that these factors have on the probability of corneal grafting cannot be ruled out.
Our study was limited by the number of events in each group; thus, determining the impact the variables discussed above had on the probability of grafting was not possible. While we cannot rule out the impact that these variables have on graft probability, the strength of our study was in utilizing survival analysis to compare the rate of corneal grafting between pediatric and adult patients. From these results, it appears that adults are at a higher risk of receiving a corneal graft after GDD implantation in comparison to children when the GDD tube is placed in the anterior chamber regardless of history of corneal disease. Given that children have a longer predicted life span, characterizing the rate of corneal grafting after GDD, one of the most common complications of GDD implantation, can hold important clinical and prognostic utility when treating this patient population.