Article Text

Original research
Comparing Ahmed-FP7 to Baerveldt-250 and Baerveldt-350 surgical outcomes: 1-year results from a retrospective cohort study leveraging the electronic health record
  1. Leo L Shen1,
  2. Xinxing Guo1,
  3. Thomas V Johnson1,
  4. David Friedman2,
  5. Michael V Boland2,
  6. Elyse J McGlumphy1
  1. 1Ophthalmology, Johns Hopkins Wilmer Eye Institute, Baltimore, Maryland, USA
  2. 2Ophthalmology, Massachusetts Eye and Ear, Boston, Massachusetts, USA
  1. Correspondence to Dr Elyse J McGlumphy; elysejoelle{at}


Objective To compare outcomes following Ahmed-FP7 (AGI-FP7), Baerveldt-250mm2 (BGI-250), or Baerveldt-350mm2 (BGI-350) implantation.

Methods and analysis Retrospective cohort study comprising 800 eyes from 800 individuals who underwent surgery 1 January 2016–31 December 2020 at a tertiary-care institution. Data were extracted from standardised fields in the electronic health record. Primary outcome was failure (defined as intraocular pressure (IOP) ≤5 mm Hg or >18 mm Hg or reduction <20% at two consecutive visits from month 3 onwards; or visual acuity (VA) loss ≥3 lines; or return to the operating room (OR)). Secondary outcomes were IOP, VA, number of follow-up visits and return to the OR.

Results A total of 523 AGI-FP7, 133 BGI-250 and 144 BGI-350 cases were analysed. The AGI-FP7 group was more likely to be younger and diagnosed with secondary glaucoma, with a higher mean baseline IOP (28.5±12.2 vs 22.0±7.7 mm Hg in BGI-250 and 23.4±9.0 in BGI-350, p<0.001). Cumulative failure rate at month 12 was 30% (AGI-FP7) vs 39% (BGI-250) vs 33% (BGI-350, p=0.159). Mean IOP at month 12 was lower in the BGI-350 group compared with AGI-FP7 (12.4±4.4 vs 14.8±5.6 mm Hg, p=0.003) but not BGI-250 (vs 13.1±4.6, p=0.710). Target IOP was achieved in 71% of AGI-FP7, 66% BGI-250, and 76% BGI-350. VA loss and rates of return to the OR did not differ between groups. Both BGI-250 and BGI-350 had more follow-up visits than AGI-FP7 (p<0.001).

Conclusion These three glaucoma drainage devices performed similarly within 1 year, with no difference in failure rates despite differing baseline patient characteristics.

  • Glaucoma
  • Treatment Surgery
  • Vision

Data availability statement

Data are available on reasonable request. Data are deidentified participant data and are available from LLS (

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

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 Ahmed Baerveldt Comparison and Ahmed versus Baerveldt studies prospectively compared Ahmed-FP7 to Baerveldt-350 implantation, finding the Baerveldt device to carry a lower failure rate but higher risk of hypotony. However, real-world data on the three most-used glaucoma drainage devices, Ahmed-FP7, Baerveldt-250 and Baerveldt-350, in large, diverse populations are lacking.


  • The three devices performed remarkably similarly despite differences in patient characteristics at baseline, suggesting patient selection may be an important component in surgical outcome.


  • Better understanding the outcomes and risks associated with glaucoma drainage device implantation can help inform surgical decision-making as well as patient counselling preoperatively and postoperatively.


Glaucoma is the most prevalent cause of irreversible optic neuropathy worldwide. Lowering intraocular pressure (IOP) is the only proven therapeutic avenue for slowing or halting disease worsening.1–4 Surgical treatment options for glaucoma have expanded dramatically in the recent past, necessitating robust comparative analyses of long-term postoperative outcomes.5–10 The proportion of glaucoma surgeons who implant tube shunts increased from 18% to 51% between 1998 and 2008 among survey respondents in the American Glaucoma Society,11 likely in large part due to data obtained from the Tube versus Trabeculectomy Study.12

The three most commonly used glaucoma drainage devices are the Ahmed FP7 (AGI-FP7, New World Medical, Rancho Cucamonga, California, USA), Baerveldt-250mm2 and −350 mm2 glaucoma implants (BGI-250/350, Johnson & Johnson Vision, Irvine, California, USA).13 While the AGI-FP7 contains a valve that restricts directionality and rate of flow, the BGI devices are non-valved, requiring the surgeon to temporarily ligate the tube until scar tissue formation at the plate provides enough outflow resistance to prevent hypotony.7 10 Numerous studies comparing these two implant types and others have been performed, most notable of which are the Ahmed Baerveldt Comparison (ABC)14 15 and Ahmed versus Baerveldt (AVB)16 trials, which compared AGI-FP7 to BGI-350.17 Other retrospective studies comparing various Baerveldt plate sizes,18 19 other implants (eg, Molteno, Ahmed-S2),20–22 and AGI-FP7 vs BGI-250/35023 have been performed, though most have been limited by small sample sizes and homogeneous patient populations that may preclude the generalisation of their results. Surgeon preference and patient indications also play large roles in determining implant types, therefore reinforcing the need to report real-world outcomes outside of controlled trial populations.

Here, we sought to address these problems by developing an automated pipeline for data extraction from the electronic health record (EHR) to compare 1-year outcomes among a large, diverse population of patients undergoing implantation with either the AGI-FP7, BGI-250 or BGI-350 implant.

Materials and methods

Patient and public involvement

Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

EHR data extraction

Current Procedure Terminology codes 66179, 66180, 66 183 were used to extract tube shunt cases performed at the Johns Hopkins Wilmer Eye Institute between 1 January 2016 and 31 December 2020 from the institutional EHR (EpicCare; Epic Systems, Verona, Wisconsin, USA). Preoperative and postoperative clinic encounters were queried between 1 January 2015 and 31 December 2021, to include 12 months of pre- and postoperative data.

Data extracted included patient demographics, surgical laterality, glaucoma diagnosis, implant type, concurrent procedures performed (eg, cataract extraction), prior ophthalmologic procedures, baseline IOP (at visit closest to and prior to tube shunt surgery), target IOP (tIOP), best-documented visual acuity (VA), central corneal thickness, number of follow-up clinic visits, return to operating room (OR) and procedure performed if returned to OR. Glaucoma diagnoses were categorised as (1) open-angle glaucoma (OAG)—including pigmentary and pseudoexfoliation glaucoma, (2) angle-closure glaucoma (ACG) and (3) secondary glaucoma—including glaucoma secondary to uveitis/neovascularisation/steroids. ICD-10 codes (International Classification of Diseases, 10th edition) used to categorise diagnoses were: (A) OAG (H40.1, H40.13, H40.14), (B) ACG (H40.2) and (C) Secondary (H40.3, H40.4, H40.5, H40.6, E13.3592, E11.39, H20.9, H20.13, H40.8, H40.9, H44.4, H18.20). Implant type (Ahmed FP-7 vs Baerveldt-250mm2 vs Baerveldt-350mm2) was determined using text analysis of operative notes. Prior ophthalmic surgical/procedural history going back to 1 January 2015 was extracted from the charts then categorised by procedure using the billing descriptions through text analysis into: (1) corneal transplantation, (2) tube shunt, (3) trabeculectomy, (4) cyclophotocoagulation, (5) cataract extraction, (6) synechiae lysis, (7) intravitreal injection, (8) other retinal (retinal lesion cryotherapy or photocoagulation for retinopathy) or (9) other (eg, eye drug systems, canthoplasty, repair of corneal lacerations. Target IOP values, chosen and documented by the patient’s surgeon, were automatically extracted from the charts when available in standardised templates (N=218) and supplemented with manual chart review (N=342 values added).

Inclusion/exclusion criteria

Patients were excluded based on the following: (1) age ≤18 years; (2) diagnosis of congenital glaucoma or non-glaucoma diagnosis (eg, unspecified corneal oedema); (3) prior history of tube shunt implantation; (4) prior corneal transplantation and (5) tube shunt implant other than Ahmed or Baerveldt (online supplemental figure 1). Only the first eye that underwent tube implantation during the study period was included per individual.

Surgical methods

Tube shunt implantation was performed by 12 experienced glaucoma surgeons, though technique was not standardised and variations may exist. Ahmed FP-7 shunts were implanted directly without ligation into the eye. Baerveldt implants were ligated intraoperatively with a dissolvable or releasable suture. Use of a ripcord varied among surgeons.

Cumulative failure definition

Failure was defined as meeting any of the following criteria: (1) IOP at two consecutive visits from month 3 onwards >18 mm Hg (or >21 mm Hg in a separate analysis) or reduction <20% or ≤5 mm Hg; or (2) return to OR for failure or complication or both; or (3) VA loss ≥3 lines at any visit from month 3 onwards.

Outcome definitions

Outcome data are reported at postoperative month 12 (POM12). Lost to follow-up (LTFU) was defined as missing both month 6 and 12 data (eg, patients with month 12 data but missing month 6 were still included in analyses). IOP at postoperative day 1, week 1, months 1/3/6/12 were extracted and compared with baseline (data closest to and prior to tube shunt surgery) and target IOP when available. Hypotony was defined as IOP≤5 mm Hg at any visit from month 3 onwards. VA loss was defined as a decline in best-documented VA by ≥3 lines. Extracted VA data were converted from Snellen and ETDRS (Early Treatment Diabetic Retinopathy Study) values to standard line measures manually (eg, 20/20, 20/100). VA was typically obtained in shorter rooms using electronic systems per clinician preference. Return to the OR within 12 months of surgery was categorised by procedure via text processing of billing descriptions (trabeculectomy/tube, glaucoma destructive, glaucoma drainage anterior segment, glaucoma bleb revision, glaucoma other, cataract, vitreoretinal surgery, corneal transplantation, strabismus, other).

For each case that returned to the OR, charts were reviewed and further categorised by indications as: (1) primary failure of tube shunt to control IOP (‘failure’; eg, needling over plate, destructive procedures such as cyclophotocoagulation, and trabeculectomy), (2) complication related to tube shunt surgery but not as a direct result of high/low IOP (‘complication’; eg, shunt revision due to anterior chamber flattening and choroidal detachment, vitreoretinal procedures such as pars plana vitrectomy for endophthalmitis, and anterior vitrectomy for hyphema) or (3) combination of failure and complication (‘failure and complication’; eg, shunt revisions for exposure/migration/hypotony, anterior chamber reformation in the setting of pupillary block, vitreoretinal procedures for high IOP in conjunction with endophthalmitis or malignant glaucoma). Other procedures were excluded (N=21). These included cataract extraction in the absence of high IOP/glaucoma progression (N=10), pars plana vitrectomy for macular hole or rhegmatogenous retinal detachment (N=3), intraocular lens repositioning (N=2), removal of ectropions/tarsorrhaphy (N=2), corneal biopsy and subconjunctival injection for keratitis (N=1), penetrating keratoplasty for failed graft following trauma (N=1), orbitotomy (N=1) and correction of hypertropia (N=1).

Statistical analysis

Statistical analyses were performed using R V.4.2.1 (R Foundation for Statistical Reporting, Vienna, Austria). GraphPad Prism V.9.5.1 (GraphPad Software, San Diego, California, USA) was used for graphical visualisation. Statistical significance was defined as a two-sided p<0.05. Comparisons were made using linear analysis of variance (ANOVA) (equivalent to two-sample t-test when there are only two samples) for continuous variables and χ2 tests or Fisher’s exact (when expected N≤5) for categorical variables. Post hoc pairwise comparisons were performed when a priori comparisons were significant. P values were corrected using the Tukey (for multiple comparisons following ANOVA) or Bonferroni (following χ2/Fisher’s exact tests) methods. For the purposes of regression analysis, missing values were handled by multiple imputation. Specifically, a classification and regression trees method was used for imputation with five iterations to create a complete dataset. Multivariable linear (for number of glaucoma visits and month 12 IOP) and logistic (for presence of return to the OR or VA loss) regressions were performed on each iteration, with subsequent pooling using Rubin’s rule to obtain a final set of estimates and standard errors. Beta coefficients/ORs, 95% CIs and p values were reported. Independent variables included in each model when appropriate were: age, sex, implant type, diagnosis, baseline IOP, return to the OR, baseline VA, VA loss and hypotony.


Baseline characteristics of the study cohort

Among the 967 tube shunt cases identified during the study period (1 January 2016-31 December 2020), 167 were excluded (online supplemental figure 1). The remaining 800 analysed cases comprised 523 Ahmed FP-7 (AGI-FP7), 133 Baerveldt-250 mm2 (BGI-250), and 144 Baerveldt-350 mm2 (BGI-350, table 1). Patients in the AGI-FP7 group were significantly younger than those in either BGI group (62±15 vs 66±15 years in BGI-250 and 67±14 years in BGI-350, pairwise p=0.027 and 0.007, respectively), with a greater proportion of secondary glaucoma diagnoses (55% vs 16% in BGI-250 and 13% in BGI-350, pairwise p<0.001 for both). Race differed across groups (p=0.015) with a lower proportion of white patients in the BGI-350 group (33%) compared with AGI-FP7 (46%, p=0.012) and BGI-250 (46%, p=0.049). Mean baseline IOP was significantly higher in the AGI-FP7 group when compared against both BGI groups (28.5±12.2 vs 22.0±7.7 mm Hg in BGI-250 and 23.4±9.0 in BGI-350, pairwise p<0.001 for both), with a lower target IOP as well (17.1±3.5 vs 15.0±2.8 mm Hg in BGI-250 and 15.1±4.3 in BGI-350, pairwise p<0.001 for both). More patients in the AGI-FP7 group had a VA of 20/200 or worse compared with the BGI-250 group specifically (28% vs 16%, p=0.015), but not BGI-350 (vs 19%, p=0.111). The two BGI groups differed in lens status (95% phakic in BGI-250 vs 85% in BGI-350, p=0.029), but AGI-FP7 (91%) did not when compared with either BGI group. Lastly, proportions in ethnicity also differed across groups, though no significant post hoc pairwise comparisons were found.

Table 1

Baseline demographic and ocular characteristics of Ahmed-FP7 versus Baerveldt-250mm2 versus Baerveldt-350mm2 implantation groups extracted from the EHR

Loss to follow-up

At month 12, 238 patients (30%) were LTFU (missing both month 6 and 12 data), comprising 180 (76% of LTFU patients) from the AGI-FP7 group and 29 (12%) each from the BGI-250 and BGI-350 groups (p<0.001, online supplemental table 1). Those LTFU were younger than those who were not (62±15 vs 65±15 years, p=0.027), with a smaller proportion of white patients (35% vs 47%, p=0.013), and a greater proportion of Hispanic patients (9% vs 4%, p=0.008), more secondary glaucoma diagnoses (57% vs 34%, p<0.001), and worse baseline VA (36% vs 20% with 20/200 or worse, p<0.001).

Cumulative failure rate

Failure by month 12 occurred in 159 (30%) AGI-FP7, 52 (39%) BGI-250, and 37 (33%) BGI-350 patients and did not differ across groups (p=0.159, table 2). The most common reason for failure was VA loss for all groups, which occurred less frequently in the AGI-FP7 group (15%) when compared with BGI-250 (30%, p<0.001) and BGI-350 (26%, p=0.005). Return to the OR was the second leading reason for failure, but rates did not differ across groups (14% AGI-FP7 vs 14% BGI-250 vs 10% BGI-350, p=0.405). Failure rate due to persistently high IOP >18 mm Hg or reduction <20% did not differ across groups either (9% AGI-FP7 vs 6% BGI-250 vs 4% BGI-350, p=0.077). Hypotony at two consecutive visits was seen in only one patient (0.4%) in the BGI-250 group with 0 in the other two groups (p=0.166).

Table 2

Cumulative failure rate at month 12 with IOP>18 mm Hg cut-off and reasons for failure comparing Ahmed-FP7 versus Baerveldt-250mm2 versus Baerveldt-350mm2 implantation using EHR data

Increasing the IOP cut-off to 21 mm Hg decreased the cumulative failure rate in the two groups disproportionately, revealing a difference in failure rates across groups (26% AGI-FP7 vs 38% BGI-250 vs 31% BGI-350, p=0.025), though no post hoc pairwise comparisons were significant (online supplemental table 2). Failure rates due to high IOP declined, with no statistical difference across groups (3% AGI-FP7 vs 1% BGI-250 vs 1% BGI-350, p=0.077).

IOP outcomes

Mean IOP values at baseline, postoperative day 1 (POD1), week 1 (POW1), and months (POM) 1/3/6/12 are shown in table 3 and graphically in figure 1A. Despite their higher baseline IOP, AGI-FP7 eyes had significantly lower IOP at POD1 and POW1 when compared with both BGI groups (POD1: 11.2±8.0 vs 16.6±10.4 mm Hg in BGI-250 and 17.5±9,3 in BGI-350, p<0.001; POW1: 12.1±6.2 vs 17.9±9.4 mm Hg and 18.4±10.0, p<0.001). From month 3 onwards, mean IOP was higher in AGI-FP7 compared with both BGI groups. At month 12, mean IOP in the AGI-FP7 group was 14.8±5.6 compared with 13.1±4.6 mm Hg in BGI-250 (p=0.074) and 12.4±4.4 in BGI-350 (p=0.003, table 3 and figure 1B). The absolute change from baseline at month 12 was greater in the AGI-FP7 group compared with BGI-250 (14.3±11.7 vs 9.0±9.0, p=0.005) but not BGI-350 (11.1+9.6, p=0.100). Percentage change from baseline did not differ across groups (40±35 in AGI-FP7 vs 33±37 in BGI-250 and 35±58 in BGI-350, p=0.473). Hypotony (IOP≤5 mm Hg at any visit from month 3 onwards) rates were higher in the BGI-250 group compared with AGI-FP7 (8% vs 2%, p=0.041) but not BGI-350 (4%, p=0.913). IOP outcomes did not differ between BGI-250 and BGI-350 across the board.

Figure 1

Intraocular pressure (IOP) outcomes after Ahmed-FP7, Baerveldt-250 mm2 or −350 mm2 implantation. (A) Mean IOP during the first year of follow-up (day 1, week 1, months 1/3/6/12). Shaded regions represent 95% CI. (B) Violin plot of IOP distribution between groups at month 12. Median—solid line; Q1/Q3—dashed lines. Multiple comparisons with Tukey’s test comparing means. (C) Proportion of eyes meeting IOP cut-offs (≤18, ≤15, ≤12 mm Hg or ≥30% reduction from baseline) for AGI-FP7 versus BGI-250 versus BGI-350 over time. Percentages are shown above their corresponding bars. (D) Postoperative IOP in relation to target IOP (≤5 mm Hg, >5 mm Hg and ≤target IOP, ≤target IOP+3, ≤target IOP+7, >target IOP+7) over time. Sample size is depicted above each bar. POM, postoperative month.

Table 3

Intraocular pressure (IOP) and change from baseline in the 12-month postoperative period comparing Ahmed-FP7 versus Baerveldt-250mm2 versus Baerveldt-350mm2 implantation using EHR data

Proportions of each group reaching IOP cut-offs of ≤18, ≤15, or 12 mm Hg, or ≥30% reduction from baseline at months 6 and 12 are shown in figure 1C. In all three instances with flat cut-offs, a greater proportion of BGI eyes met the cut-off than AGI-FP7. However, a similar proportion met a reduction of ≥30% at month 12 in all groups (69% AGI-FP7 vs 55% BGI-250 vs 69% BGI-350). Postoperative IOP was categorised in relation to baseline targets, with 49% of AGI-FP7, 43% of BGI-250 and 36% of BGI-350 eyes falling into the desired >5 mm Hg and ≤ target IOP group (figure 1D). For AGI-FP7, BGI-250 and BGI-350, respectively, percentages meeting the target were: 52% vs 51% vs 57% at month 3, 68% vs 70% vs 75% at month 6 and 71% vs 66% vs 76% at month 12.

VA, follow-up and return to the OR outcomes

A total of 52 (18%) patients experienced VA loss ≥3 lines at month 12, with no difference between AGI-FP7, BGI-250 and BGI-350 groups (16% vs 20% vs 22%, p=0.536, table 4). AGI-FP7 patients, however, had fewer glaucoma clinic visits by month 12 than BGI-250 (6.7±5.7 vs 9.3±6.9, p<0.001) and BGI-350 (9.4±6.4, p<0.001). A total of 72 (14%) AGI-FP7 eyes, 19 (14%) BGI-250, and 14 (10%) BGI-350 patients returned to the OR within 12 months of implantation due to failure/complication/both (p=0.700). Rates of procedures performed on return to the OR did not differ between groups except for trabeculectomy/repeat tube shunt, which occurred more frequently in the BGI-250 group compared with AGI-FP7 (6 (32% of BGI-250 eyes that returned) vs 5 (7% of AGI-FP7), p=0.035, online supplemental table 3).

Table 4

Month 12 outcomes for visual acuity loss ≥3 lines, number of follow-up visits and return to the or comparing Ahmed-FP7 versus Baerveldt-250mm2 versus Baerveldt-350mm2 implantation using EHR data

Factors associated with postoperative outcomes

To examine factors associated with postoperative outcomes including return to the OR, number of postoperative clinic visits, VA loss and IOP at month 12, we performed linear/logistic regressions using imputed datasets. With respect to return to the OR, only hypotony at any visit from month 3 onwards was associated with a greater odds of return (OR 7.26 (95% CI 3.23 to 16.32), p<0.001; online supplemental table 4). Positive predictors of postoperative visits included BGI-250 (β 1.77 (95% CI 0.58 to 2.95), p=0.004) and BGI-350 implants (β 2.12 (95% CI 1.00 to 3.25), p<0.001), higher baseline IOP (β 0.06 (95% CI 0.02 to 0.10), p=0.002) and return to the OR (β 4.55 (95% CI 3.33 to 5.76), p<0.001, online supplemental table 5). Negative predictors of postoperative visits included a diagnosis of secondary glaucoma (β −1.68 (95% CI –2.74 to –0.62), p=0.002) and baseline VA between 20/40 and 20/200 (β −1.21 (95% CI –2.18 to –0.25), p=0.014) or worse than 20/200 (β −2.57 (95% CI –3.66 to –1.48), p<0.001).With respect to VA loss at month 12, only a baseline VA between 20/40 and 20/200 was associated with an increased odds ratio (1.77 (95% CI 1.10 to 2.84), p=0.020, online supplemental table 6). Month 12 IOP was negatively predicted by BGI-350 implant type (β −0.97 (95% CI –1.92 to –0.02), p=0.048) and hypotony at any visit from month 3 onwards (β −3.58 (95% CI –5.70 to –1.47), p=0.003, online supplemental table 7).


Using an automated pipeline for EHR data extraction, we retrospectively examined 800 eyes that underwent AGI-FP7, BGI-250 or BGI-350 implantation, finding remarkably similar failure rates (30%, 39% and 33%, respectively, p=0.159) and performance at 1-year despite differences in patient characteristics at baseline.

While other studies have examined outcomes following tube shunt implantation, some do not differentiate between BGI-250 and BGI-350,24–26 or compare AGI-FP7 to BGI-350 only,27–29 or in one instance (Kilgore et al23) do compare the AGI-FP7/BGI-250/BGI-350 implants separately but are limited by population homogeneity and inclusion of multiple eyes per patient without statistical correction, which may confound effects. These studies also vary in success/failure definitions. Our study, therefore, supplements this literature with data from a large, diverse population comparing the three most-used implants and a pipeline for data extraction and analysis that can be used in the future to continually provide surgeons with information on their outcomes.

Comparing our results to the 1-year pooled ABC trial data,14 which used an IOP>21 mm Hg cut-off for failure, we find higher failure rates (26% in our study vs 16% in ABC for AGI-FP7; and 29% vs 14% in BGI-350). However, the AVB trial30 used an IOP>18 mm Hg cut-off and reported failure rates higher than ours in the case of AGI-FP7 (43% in AVB vs 30% in our study) but not BGI-350 (28% vs 33%). The reasons for failure also differed, with VA loss being the largest contributor in our study, whereas high IOP was the most common reason for failure in both the ABC14 and AVB30 trials. This variability largely stems from differences in success/failure definitions. Our study defined VA loss as ≥3 lines lost, while the trials used the more conservative loss of light-perception as their criterion. However, when looking at the trials’ secondary VA outcomes, we find similar rates of VA loss in our study. The higher rates of VA loss in the BGI groups compared with the AGI-FP7 group may be due to the worse baseline VA of the latter, potentially creating a ‘floor’ effect for VA loss, while BGI patients had more initial VA to lose.

Another study that contextualises our work is Kilgore et al’s23 report on outcomes following AGI-FP7, BGI-250 and BGI-350 implantation. There, they also find the three implants to have similar failure rates, but contrary to our study, do not find differences in postoperative IOP between groups. They report month 12 mean IOP of 12.6, 11.6 and 13.3 mm Hg (p=0.327) in AGI-FP7, BGI-250 and BGI-350, respectively, compared with 14.8, 13.1 and 12.4 mm Hg (p=0.002) for the same groups in our study, perhaps reflecting a difference in patient selection/populations at baseline. Notably, their study population is 95% white compared with 43% in this study.

The ABC/AVB trials did not look at the proportion of patients meeting their target IOP, which is an important indicator of success. In this study, 71% of AGI-FP7, 66% of BGI-250 and 76% of BGI-350 eyes were in their target range by month 12. Neither the trials nor other retrospective studies examining glaucoma drainage devices, to our knowledge, have looked at the number of postoperative glaucoma clinic visits. We reasoned that this information may be useful to surgeons when counselling their patients regarding what to expect postoperatively. Indeed, we found both BGI groups to have significantly more postoperative visits, which may be due to the non-valved nature, and therefore, delayed IOP lowering effects secondary to suture dissolution of the Baerveldt implants requiring closer follow-up,31 32 or a difference in patient populations between groups at baseline.

With regard to those differences at baseline and possible confounders, AGI-FP7 patients were far more likely to be LTFU. This, in conjunction with the characteristics of those LTFU (younger age, more non-white race, more Hispanic, worse VA and more secondary glaucoma diagnoses) may reflect underlying social determinants of health and correspondingly more medically complicated patients who face barriers in accessing care.

The differences in age, race, ethnicity, diagnosis and baseline ocular characteristics between groups in our study may additionally reflect surgeon preferences and differences in indications between patients. For example, valved AGI shunts may be deliberately selected for patients who need rapid IOP lowering.14 17 31 Indeed, our data are consistent with this indication, as mean preoperative IOP was higher in the AGI-FP7 group but fell rapidly in the immediate postoperative period.

While rates of return to the OR within 12 months did not differ between groups, the proportion of procedures did. BGI-250 patients were more likely to undergo a trabeculectomy/repeat tube shunt or revision than both the AGI-FP7 and BGI-350 groups. We speculate that these findings may be related to the preoperative characteristics of these patients which led their physician to place the BGI-250, such as stage of glaucoma and IOP target or other characteristics, but we cannot exclude plate size as a possible contributor.

With regard to plate size, BGI-250 and BGI-350 performed remarkably similarly in this study. Their baseline patient characteristics, failure rates and secondary outcomes largely did not differ. However, only the BGI-350 group had a lower mean IOP when compared with AGI-FP7 (12.4±4.4 vs 14.8±5.6, p=0.003), while the BGI-250 found no difference (13.1±4.6, p=0.074). The mean IOP at month 12 did not differ between the two BGI groups, which is consistent with previous data.33 Rate of hypotony also only differed between BGI-250 and AGI-FP7 (8% vs 2%, p=0.041), but not in any of the other pairwise comparisons, which may be a due to the non-valved nature of the BGI implant, though it is unclear why the same effect is not true for the BGI-350 group. Ultimately, our study is limited by its retrospective nature with differences between groups that could not be controlled for.

Among other limitations of our study is the reliance on EHR for automated data extraction. Data not entered in specific fields could not be reliably extracted, resulting in more missing values than if charts were manually reviewed. Prior surgical history was also limited to data extracted during the study period, leaving open the possibility that patients could have had prior procedures that were not accounted for in our study. Moreover, surgeon preference and surgical variations such as ripcord and ligature versus ligature alone, method for venting prior to ligature opening and size of ligature suture, among others, all present additional confounders. Our study is also limited by its retrospective cohort nature, with important differences in patient characteristics at baseline. Lack of statistical significance may also not reflect a lack of clinical significance. Furthermore, indication bias is certainly present here as patient characteristics between groups differed substantially. We also cannot be sure if other factors associated with outcome played a role in surgical decision-making, which is a limitation of all retrospective studies. Defining the baseline visit as the one closest to and prior to the procedure may also present important confounders when certain patients require more emergent surgery than others, such as in the case of acutely elevated IOP. True baselines would have to be determined by averaging historical data for each patient. Moreover, data for social determinants of health, disease course/severity, comorbidities, and surgical complications were not recorded and may present possible confounders. Hypotony was also limited to an IOP cut-off rather than clinical diagnosis, though this is consistent with the ABC/AVB trials and other retrospective studies.14 23 30 Glaucoma medication use was also not assessed.

Taken together, while indications and baseline characteristics differed between those receiving AGI-FP7, BGI-250, or BGI-350 implants, these findings reflect real-world practice patterns and demonstrate remarkably similar outcomes at 1 year after surgery. Our largely automated data extraction and analysis using the EHR will also facilitate future comparisons between other surgeries/devices and reanalyses of cohorts with longer-term outcome data.

Data availability statement

Data are available on reasonable request. Data are deidentified participant data and are available from LLS (

Ethics statements

Patient consent for publication

Ethics approval

The Johns Hopkins University School of Medicine Institutional Review Board approved the study protocol (IRB00147410). The need for written informed consent was waived due to the retrospective nature of the analysis. This study adheres to the tenets of the Declaration of Helsinki and to the US Health Insurance Portability and Accountability Act of 1996.


Supplementary material


  • Twitter @LeoLShen, @elysejoelle

  • Contributors All authors attest that they meet the current ICMJE criteria for authorship. Conceptualisation; LLS, XG and EJM. Data curation; LLS, XG and EJM. Formal analysis; LLS and XG. Funding acquisition; EJM and TVJ. Investigation; LLS, XG and EJM. Methodology; LLS, XG and EJM. Project administration; LLS and EJM. Supervision; EJM. Visualisation; LLS. Roles/writing—original draft; LLS. Writing—review and editing; LLS, XG, TVJ, DF, MVB and EJM. EJM was the guarantor of the study and accepts full responsibility for the work and conduct of the study, had access to the data, and controlled the decision to publish.

  • Funding MAPS award from the American Glaucoma Society awarded toEJM. The Shelley and Alan Holt Professorship in Ophthalmology, Research to Prevent Blindness Career Development Award, and National Eye Institute (K08-EY031801) grant awarded to TVJ. LLS and XG have no financial disclosures. TVJ is a consultant for Abbvie and receives additional grants from the National Eye Institute (R13EY034018, R21EY034332), BrightFocus Foundation (G2002005S), Alcon, Perfuse Therapeutics, iCare USA and InjectSense. DF is a consultant for the National Outcomes Research Center, has received payment/honoraria for lectures from Thea pharmaceuticals, and Bausch and Lomb, and receives additional grants from Genentech and the National Institutes of Health. MVB is a consultant for Carl Zeiss Meditec, Allergan, Janssen, and Topcon Healthcare. EJM is a consultant for Gore and InjectSense.

  • Disclaimer These funding organisations had no role in the design or conduct of this research.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

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

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.