Retrospective evaluation of two-year results with a filtering trabeculotomy in comparison to conventional trabeculectomy by exact matching

Background: To compare two-year results of a filtering trabeculotomy (FTO) to conventional trabeculectomy (TE) in open-angle glaucoma by exact matching. Methods: 110 patients received an FTO and 86 a TE. FTO avoided the need for an iridectomy due to a preserved trabeculo-descemet window anterior to the scleral flap. TE employed a trabecular block excision and iridectomy. Mitomycin C was used in both. FTO and TE were exact matched by baseline intraocular pressure (IOP) and the number of glaucoma medications. Complete and qualified success (IOP ≤18 mmHg and IOP reduction ≥ 30%, with or without medication) were primary endpoints. IOP, visual acuity (BCVA), complications and intervention were secondary endpoints. Results: 44 FTO were exact matched to 44 TE. The IOP baseline in both groups was 22.5±4.7 mmHg on 3±0.9 medications. At 24 months, complete success was reached by 59% in FTO and 66% in TE, and qualified success by 59% in FTO and 71% in TE. In FTO, IOP was reduced to 12.4±4.3 mmHg at 12 months and 13.1±4.1 mmHg at 24 months. In TE, IOP was 11.3±2.2 mmHg at 12 months and 12.0±3.5 mmHg at 24 months. Medications could be reduced at 24 months to 0.6±1.3 in FTO and 0.2±0.5 in TE. There were no significant differences between the two groups in IOP, medications, complications or interventions at any point. Conclusion: Modifying aqueous flow through a limited trabeculotomy in FTO yielded clinical outcomes similar to traditional TE but allowed to avoid an iridectomy.


Introduction
Since the first description of a guarded filtering procedure in enucleated eyes by Grant in 1958 1 , performance of this surgery in patients in 1961 by Sugar 2 , popularization of the term "trabeculectomy" in 1968 by Cairns 3 , and introduction of mitomycin-C as an antifibrotic to make it more effective 4 , trabeculectomy (TE) has remained a primary surgery in the treatment of glaucoma 5 .Over the years, multiple modifications of this surgery have been explored to improve its effectiveness, to make outcomes more predictable and to reduce postoperative complications and need for interventions.These modifications include, among others, variations in size, localization, and thickness of the scleralflap, different suture techniques, variable intra-and postoperative treatment with antifibrotics or a combination of these approaches with one another 6 .
In this study, we combined elements of deep sclerectomy 7,8 and trabeculotomy 9 with TE in an attempt to improve conventional outflow as well as subconjunctival aqueous humor drainage.Encouraged by a pilot study of filtering trabeculotomy (FTO) with a complete success rate of 79% 10 , we hypothesized that FTO had a higher success rate and lower complication rate than TE.We applied advanced statistics, exact matching 11,12 , to enable a highly balanced comparison of our retrospective data with two-year follow-up.

Study design
This retrospective study was approved by the ethics committee of the University of Würzburg, Germany (#2019101601AS).Because of its retrospective nature, informed consent was waived.
All primary open angle glaucoma (POAG) patients at the University Eye Hospital Würzburg who underwent a modified FTO with mitomycin C (MMC) or TE with MMC by a single surgeon (FG) between 2007 and 2014 were analyzed.The indication for surgery was a failure to control intraocular (IOP) despite maximally tolerated medical therapy.Only one eye was included per patient.If two eyes had been operated on, the first eye was chosen to take advantage of a longer history.
Patients were matched by baseline IOP and the number of glaucoma medications.
From patient records, we obtained from 2019 to 2020 the medical history, the best-corrected visual acuity assessed (BCVA [logMAR]), IOP (Goldmann tonometry [mmHg]), topical glaucoma medications (including a glaucoma medication score (GMS) 13 ), as well as postoperative events and complications.These included hypotony with choroidals, hypotony maculopathy, flat or shallow chamber, an IOP relatively high for surgical glaucoma (above 25 mmHg), bleb leakage, a persistently flat bleb indicating absent flow, hyphema, iris incarceration and need for cataract surgery.We computed the complete and qualified success, according to the guidelines set forth by the World Glaucoma Association 14 .In patients who had either TE or FTO in both eyes, the first eye was chosen to be included in the study.Complete success was defined as a postoperative IOP of ≤18 mmHg with a reduction of ≥30% from baseline without glaucoma medications.Qualified success was a postoperative IOP of ≤18 mmHg with a reduction of ≥30% from the baseline, achieved with or without glaucoma medications 14 .Follow-up visits occurred 7 days, 1 month, 3 months, 6 months, 12 months, and 24 months after surgery.

Exclusion criteria
The exclusion criteria consisted of patients aged below 18 years, glaucoma types other than open-angle glaucomas, and a history of TE, trabeculotomy or glaucoma drainage device implantation.

Surgical technique
The eye was rotated downward with a traction suture.A 5 mm fornix-based peritomy was made at the anatomic 12 o'clock position, and a sub-tenon pocket was fashioned to accommodate a sponge soaked with MMC at a concentration of 0.2 mg/ml and 100 µl volume for 3 minutes.
In TE, a 3 mm × 4 mm half scleral thickness flap was created.A 0.8 mm × 2 mm sclerotrabecular block was excised to enter the anterior chamber, as described before 10 .A peripheral iridectomy was made.The scleral flap was secured with 10.0 nylon to allow visible percolation of aqueous, and the conjunctiva was closed with an interlocking running suture resulting in a diffusely forming bleb 15 .
In FTO, the scleral flap was sized 4 mm × 4 mm.A smaller, tongue-shaped flap was dissected underneath to unroof and create access ostia to Schlemm's canal.This second flap was excised similar to deep sclerectomy.The canal was probed with a metal trabeculotomy probe (Mackensen, Geuder Inc., Heidelberg) on both sides while the trabeculo-descemet window at the base of the scleral flap was preserved so that no bulk aqueous outflow could occur and no iridectomy was needed.The remaining steps were identical to those in TE.
Postoperative drops consisted of dexamethasone six times for the first week, which was tapered by one drop per week.Ciprofloxacin eye drops were applied four times a day for one week.All patients received 5 mg of 5-FU (0.1 cc volume) daily for a week unless IOP was below 5 mmHg or a Seidel-positive bleb leak or a corneal erosion was present.5-FU was also given once a week for the first month during return visits using the same criteria.

Amendments from Version 1
We have implemented the changes requested by Reviewer 1 and added a power calculation, effect size calculation.We explain how we selected the eye for the study in patients with two surgical eyes.
Categorical variables were described as the frequency with percentage, whereas continuous and discrete variables as mean with standard deviation (SD) or median with range.The chi-squared test or Fisher's exact test was used for the analysis of categorical variables.Continuous variables were compared using Student's t-test or Mann-Whitney U test, whereas discrete variables were compared using Mann-Whitney U test.The distribution of continuous variables was determined by the Shapiro-Wilk test, equality of variances by Levene's test.Assessment of repeated measures for IOP was performed using repeated measures MANOVA and Tukey's test, while visual acuity (logMAR) and medications were examined using the Friedman test and Wilcoxon signed-rank test.The success of treatment was expressed by a Kaplan-Meier curve and compared between treated groups using the log-rank test.The success of treatment at particular a given time point was determined by odds ratio (OR) with respective 95% confidence intervals and p-values.P-values below 0.05 were considered statistically significant.
Our power calculation before commencing the study indicated that we would need 39 eyes per group to detect a difference of at least 20% at a power of 80% and alpha of 0.05 (continuous endpoints, two independent sample study).We then performed a post-hoc analysis for IOP at year 1 and 2 and determined the effect size using Cohen's d (d (.01) = very small, d (.2) = small, d (.5) = medium, d (.8) = large, d (1.2) = very large, and d (2.0) = huge.

Results
A total of 196 patients were included.The unmatched demographic data of FTO and TE had significant differences in preoperative IOP (p=0.017),glaucoma medication score (p=0.001), and pseudophakia (p=0.012).In total 88, eyes (44 in each group) could be matched as exact pairs eliminating key differences in IOP and medications.The IOP at baseline was 22.6±4.7 mmHg in FTO and 22.6±4.7 in TE while on 3.0±0.9medications in both (Table 1).There were no significant differences between FTO and TE in gender, age, best-corrected visual acuity, type of glaucoma, or surgical side.In FTO, 13 eyes were pseudophakic  compared to 3 in TE (p=0.006).Two patients in FTO had a pars plana vitrectomy and one a retinal cryopexy.In TE, there were no prior ocular surgeries other than phacoemulsification.
There were no statistically significant inter-group differences in complete or qualified success at any time (p=0.403for complete success; p=0.204 for qualified success at 24 months; Figure 1 and Figure 2).The complete success rate in FTO ranged from 79% at 6 to 78% at 12 months, and 59% at 24 months.In TE, it was 81%, 85%, and 66%, respectively.Similarly, IOPs of FTO and TE were not significantly different at any time (12 months: p=0.983, 24 months: p=1.000, Figure 3).At one year, the IOP had declined to 12.4±4.3mmHg in FTO and 11.3±2.2mmHg in TE with medications (qualified success).At two years, IOP remained at 13.1±4.1 mmHg in FTO and 12.0±3.5 mmHg in TE (qualified success), respectively.The posthoc analysis showed that we could have detected an IOP difference of 16.1% at one year and an IOP difference of 17.6% at two years with a power above 80%.Cohen's d was 0.32 at one year and 0.29 at two years, respectively, in both cases classifying as small.The postoperative visual acuity was not significantly different in FTO and TE at any time (p=0.894after 12 months; p=0.443 after 24 months; Figure 4 and Figure 5).
There was a reduction in glaucoma medication from 3±0.9 to 0.6±1.3 in FTO and from 3±0.9 to 0.2±0.5 in TE after 24 months.There was no significant difference between the two groups in glaucoma medications at 24 months with 19% of patients in FTO and 12% of patients in TE using glaucoma drops (p=0.471, Figure 6 and Figure 7).
The number of postoperative interventions was the same in the two groups (p=0.087).Bleb needling, conjunctival suture, and scleral flap suture were the most common interventions.There was no statistically significant difference between both groups in early or late interventions (Table 3), except 5-FU and laser suture lysis, which was performed more often in FTO.

Discussion
TE with MMC remains a primary surgery in the management of advanced glaucoma 5 despite its potential for serious complications that include choroidal effusions, maculopathy, blebitis, endophthalmitis, and suprachoroidal hemorrhage 17 .Numerous modifications have been explored over the years to reduce the rate of these.They include a smaller scleral flap 18 , limbus-versus fornix-based conjunctival closure 19 , releasable flap sutures 20,21 , a combination of trabeculectomy with deep sclerectomy 22 ,  different concentrations and exposure times of MMC 23 or sutureless tunnel trabeculectomy without iridectomy 24 .The iridectomy, which is part of traditional trabeculectomy with a trabecular block excision, can cause hyphema, inflammation, posterior synechiae, iridodialysis, and cataracts 25,26 .FTO addresses some of these issues by creating a more spread-out intake of aqueous humor, thereby reducing iris aspiration and avoiding the need for an iridectomy.The trabeculotomy and sclerectomy 8,27,28 , that are part of the FTO, were meant to remove some of the post-trabecular outflow resistance [29][30][31] .
We used matching, a nonparametric method of controlling the confounding influence of pretreatment variables in observational data 32 .Before matching, FTO and TE had significant differences.We have previously used coarsened exact matching, propensity score matching, and, more recently, exact matching.Coarsened exact matching applies multiple imputations to fill in missing data to not distort any relationships contained in the data while enabling the inclusion of all observed data from moderately uneven groups.Such was the case when we compared patients with a primary IOP indication to patients with a mixed indication for both cataract removal and IOP reduction [33][34][35] .Propensity score matching is helpful to compare even more divergent groups, for instance, patients undergoing tube shunt surgery with patients undergoing trabectome surgery 36,37 .By contrast, exact matching is well suited to compare similar pathological conditions and similar treatments, for instance, phaco-iStent or phaco-trabectome 38,39 .A downside of exact matching is that a certain number of datasets must be excluded from the analysis because the algorithm accepts only identical primary criteria  matches.Overall, our data loss was acceptable because of the high similarity that already existed at baseline.We were able to retain a large number of eyes (about 45%) as identical pairs of preoperative IOP and medication to focus on the preeminent questions of success in IOP and medication reduction.We found that FTO was as successful as TE with a similar reduction of IOP and medications.Both had a similar intervention and complication rate, notwithstanding numerical hypotony within the first six weeks after surgery.We observed a remarkably low rate of hyphema compared to ab interno trabeculectomy 40 that occurs when the

FTO (n, (%)) TE (n, (%)) P-Value
No. of interventions IOP is at or below episcleral venous pressure allowing blood to reflux into the anterior chamber.This could indicate reduced patency of collector channels in advanced glaucoma that qualifies for filtering surgery, which has been observed ex vivo 41 .
Trabeculotomy ab externo has been applied to adult POAG before but, compared to TE, was noted to have a lower success rate of 70% at one year, presumably due to a reapproximation or regeneration of the disrupted trabecular meshwork 42 .A study by Chihara et al. was in agreement with this, finding that a modified trabeculotomy ab externo lowers IOP to an average near 16 mmHg in a safe fashion 43 , by not as much as TEs.Ogawa et al. compared a nonpenetrating trabeculectomy with or without trabeculotomy 44 using a technique that was very similar to the one applied in our study, except without MMC.Despite the absence of this antifibrotic, the authors achieved a two-year IOP of 13 mmHg, not unlike our patients.
Our two-year TE results match Kirwan et al.'s multicenter study of TE with an iridectomy in 428 eyes well 45 .These eyes achieved an IOP of 12.4±4 mmHg and 80% did not have to use glaucoma medications anymore, slightly more than in our study population.The authors performed needling in 17% of patients, not unlike our rates, and numerical hypotony during the first six months occurred in 7.2% of patients, which is in a similar range as our 9% in FTO and 2% in TE.Leakage was observed in 14% compared to 9% in our FTO and 2% in our TE.However, Kirwan et al.'s cataract surgery rate was 31%, much higher than our 1% in both FTO and TE, which might represent a difference in practice pattern or simply easier access to this elective procedure, which is done as an outpatient surgery in the UK.
It is interesting to note that TEs in the Tube Versus Trabeculectomy (TVT) study were performed without an iridectomy 46 .In that study, IOP reduced from 25.6±5.3mm Hg to 12.7±5.8mm Hg at one year while the number of glaucoma medications declined from 3.0±1.2 to 0.5±0.9 47, which is also relatively similar to our results although these investigators included eyes with prior intraocular surgery, including glaucoma procedures.An early complication rate of 37% was observed by these authors, primarily consisting of a shallow or flat anterior chamber in 20% and choroidal effusions in 10% 6 .In another study of TE without an iridectomy by Jea et al., IOP reduced from 26.3±10.9mmHg to 10.2±4.1 mmHg at two years 48 .The number of glaucoma medications decreased from 2.2±1.6 to 0.5±1.0.A complete success occurred in 76.6% at one year and 66.2% at two years, matching ours.
In all of these studies, TE was performed as an outpatient patient procedure, a practice that emerged in the late 1980s 49,50 and became a standard for most types of eye surgeries and countries [51][52][53][54] .Even before the now-common use of antifibrotics, there was no significant difference in success or complication rates between inpatients and outpatients 50 , an observation that has been confirmed with antifibrotics as well 53 .In the country of our study, TE and FTO are only reimbursable as inpatient procedures.It has been argued that meticulous micromanagement after TE for several days may be associated with better long-term outcomes of TE.However, this hypothesis is challenged by the present study and by the findings of others 49,50,53,55 .One could argue that the considerably lower cataract surgery rate in our data compared to Kirwan et al. 45 might indicate a higher threshold for cataract surgery.These would have typically also been done as inpatient procedures to better handle post-cataract surgery bleb care.
Our study was limited by its retrospective nature and nonrandomized design.In patients with surgery in both eyes, we had selected the first eye undergoing glaucoma surgery for our analysis to take advantage of a longer history with more data.This selection might have favored the eye with more severe glaucoma.However, as this occurred in both groups, a similar bias will have occurred.Certainly, randomized controlled trials are a more sophisticated tool to reduce bias when trying to detect differences.However, given that exact matching already allows for a highly balanced comparison of retrospective data, such an effort might be difficult to justify.The indication for postoperative interventions and length of hospitalization was at the discretion of the treating physicians and was not standardized for both groups.Despite reducing confounding through the exact matching of IOP and medications, other confounding factors might have contributed to small differences in early postoperative patient management, as reflected by the fact that FTO patients received more 5-FU injections.Although these patients had a higher rate of numerical hypotony they were hospitalized for a slightly shorter time and experienced results that were not significantly different.
In conclusion, our results are largely in line with other FTO and TE outpatient studies.Combining elements from both yields reasonable two-year rates of surgical success, postoperative complications, and interventions while avoiding an iridectomy.
The choice of the exact matching criteria.Obviously, extending the matching criteria reduces the number of available matches.But besides the criteria chosen the authors may discuss whether the inclusion of criteria directly linked to failure, such as the presence of drug related ocular surface inflammation, could have led to a further reduction of a significant confounding factor.More precisely, if a grading system for preoperative ocular surface inflammation existed, this could have been included as a matching criterion.However, given the retrospective nature of this study, such a grading might not have been available.
Discussion.The results suggest that both interventions may be used interchangeably.Could the authors provide arguments which speak in favour or against one or the other technique?-maybe with respect to certain details, such as: Lens status.Should FTO be considered rather in phakic eyes as the risk for cataract development may be lower?But why was the rate of pseudophakic patients significantly higher in FTO compared to TE? 1.
Surgery time and/or need for general anaesthesia (TE faster?therefore better for local anaesthesia?). 2.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound?

Are sufficient details of methods and analysis provided to allow replication by others? Yes
If applicable, is the statistical analysis and its interpretation appropriate?Yes Are all the source data underlying the results available to ensure full reproducibility?Partly

Are the conclusions drawn adequately supported by the results? Yes
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Glaucoma I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.A post-hoc power analysis could further improve this excellent paper to prove that the sample size of 44 eyes per group was sufficient to find a difference or no difference between the two study arms

Is the study design appropriate and is the work technically sound? Yes
Are sufficient details of methods and analysis provided to allow replication by others?difference was found."We have now deleted this statement and inserted the text as described below.Our power study before commencing the study indicated that we would need 39 eyes per group to detect a difference of at least 20%.Our posthoc calculation showed that we had a power to detect a difference of 16% at 1 year and 17.6% at 2 years.We added to Methods: "Our power calculation before commencing the study indicated that we would need 39 eyes per group to detect a difference of at least 20% at a power of 80% and alpha of 0.05 (continuous endpoints, two independent sample study).We then performed a post-hoc analysis for IOP at year 1 and 2 and determined the effect size using Cohen's d (d (.01) = very small, d (.2) = small, d (.5) = medium, d (.8) = large, d (1.2) = very large, and d (2.0) = huge [1])." We added to Results: "The posthoc analysis showed that we could have detected an IOP difference of 16.1% at one year and an IOP difference of 17.6% at two years with a power above 80%.Cohen's d was 0.32 at one year and 0.29 at two years, respectively, in both cases classifying as small."

Figure 2 .
Figure 2. Kaplan-Meier curve for qualified success in FTO and TE.

Figure 5 .
Figure 5. Scatter plots of visual acuity of both groups.

Figure 7 .
Figure 7. Scatter plots of IOP of both groups.Few patients required medications postoperatively.

Reviewer Report 18
November 2020 https://doi.org/10.5256/f1000research.29560.r74186© 2020 Bussel I.This is an open access peer review report distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.Igor I. BusselGavin Herbert Eye Institute, University of California, Irvine School of Medicine, Irvine, CA, USA Most glaucoma surgery clinical trials are often non-randomized, usually limited by a short duration of follow-up (<1 year), lack hypothesis, and appropriate comparison to a gold-standard.The investigators of this trial managed to overcome lack of randomization by using novel exact matching methodology.This allowed for a balanced comparison of retrospective data.FTO was compared to the gold standard, TE.It is unclear if the exact matching process created a sample bias of cases with higher success.The authors demonstrated that FTO and TE yield similar twoyear rates of surgical success, postoperative complications, and interventions.

Is the work clearly and accurately presented and does it cite the current literature? Yes Is the study design appropriate and is the work technically sound? Yes Are sufficient details of methods and analysis provided to allow replication by others? Yes If applicable, is the statistical analysis and its interpretation appropriate? Yes Are all the source data underlying the results available to ensure full reproducibility? No Are the conclusions drawn adequately supported by the results? Yes Competing Interests:
No competing interests were disclosed.

have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
Reviewer Report 30 October 2020 https://doi.org/10.5256/f1000research.29560.r73193 Töteberg-Harms: Are all the source data underlying the results available to ensure full reproducibility?-No Authors: We have made all underlying data available at Open Science Framework: Retrospective evaluation of 2-year results with a filtering trabeculotomy in comparison to conventional trabeculectomy by exact matching, https://doi.org/10.17605/OSF.IO/KDYF356.Data are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).