Glaucoma is a leading cause of blindness worldwide, second only to cataracts.1 Once medical management options have been exhausted, the gold standard procedure for the treatment of elevated intraocular pressure (IOP) in the management of glaucoma is trabeculectomy.2 Previous work has suggested that prolonged treatment with antiglaucoma medications increases the risk of future filtration surgery failure.3 Nevertheless, debate ensues as to whether exposure to ocular antihypertensives is indeed a risk factor for surgical failure.4 Furthermore, it is unclear whether the culprit is the active medicinal ingredient, the preservative, or the drop vehicle itself. On the basis of extensive in vitro data and clinical evidence,5–10 the most likely candidate seems to be the preservative included in most ophthalmic solutions: benzalkonium chloride (BAK).
BAK is one of the most frequently used preservatives in antiglaucoma eye drops.11 This quaternary ammonium detergent exerts its antimicrobial effect by causing nonselective cytoplasmic membrane lysis and protein denaturation.11 Preservatives such as BAK are necessary to prevent bacterial contamination of traditionally designed multidose ophthalmic solution containers.11 Unfortunately, BAK is known to have cytotoxic side effects on the cornea and conjunctiva,7 leading to conjunctival metaplasia12 and tear film breakdown.13–14 Prolonged use of preserved ocular antihypertensives seems to lead to subclinical conjunctival inflammation and an associated increased risk of early trabeculectomy failure,3 although a direct effect of BAK exposure on subsequent risk for surgical failure is yet to be shown.
Ocular antihypertensive medications have evolved significantly, since the key study relating preserved drops with surgical failure was reported by Broadway et al.3 Miotics are no longer commonly used in the management of glaucoma due to poor tolerability in patients and an inconvenient dosing schedule.15 Furthermore, alternatives to BAK as a preservative are now widely available and are being used in multidose containers of travoprost, brimonidine, and timolol. Several studies have shown that ophthalmic solutions are as effective with alternative preservatives as they were when preserved with BAK, despite the greater tissue penetration offered by BAK’s detergent activity.14,16–20
This study is intended to retrospectively evaluate the relationship between BAK exposure and subsequent risk of filtration surgery failure in the modern clinical context. The primary variable examined is the number of BAK-preserved drops used daily by the patients, a simple, clinically useful treatment measure.
After approval by the Research Ethics Board of Sunnybrook Health Sciences Centre had been received, a retrospective chart review of glaucoma patients for whom the principle investigator had performed a trabeculectomy or a phacotrabeculectomy between 2004 and 2006 was carried out in June and July 2010. The surgical technique for the trabeculectomy was similar for both the trabeculectomy-only and the phacotrabeculectomy patients. In both cases, a superior fornix-based conjunctival peritomy was created, with hemostasis achieved using gentle bipolar cautery. A solution of 0.2 mg/mL of mitomycin C (MMC) was soaked onto 2 to 3 fragments of weck-cell sponge and these were inserted into the subconjunctival pocket. The duration of application varied between 0 and 120 seconds depending on the surgeon’s judgment of risk factors and conjunctival quality. The MMC was flushed out with 20 mL of sterile saline. A 0.25-mm setting on a diamond knife was used to outline a 4×4 mm rectangular scleral flap, which was lifted with a crescent blade. The anterior chamber was gently decompressed with either a 25 g needle (trabeculectomy) or a 15-degree blade (phacotrabeculectomy). For a phacotrabeculectomy, a keratome was used to enter the anterior chamber under the flap and the cataract removed and intraocular lens inserted. The trabeculectomy filtering ostomy was placed from the scleral spur anteriorly into clear cornea using the 15-degree blade and scissors, followed by a small peripheral iridectomy. The scleral flap was positioned and secured with 4 to 6 10-0 nylon sutures. Balanced salt solution was irrigated through the paracentesis entry port to ensure filtration. The conjunctiva was replaced to the limbus and secured with a combination of running and interrupted 8-0 vicryl sutures. More balanced salt solution was irrigated through the paracentesis to raise a water tight bleb. Tobramycin/dexamethasone ointment and atropine 1% drops (if phakic) were placed on the eye, which was patched and shielded. The postoperative medication routine was standardized: prednisolone acetate 1%, tobramycin 0.3%, atropine 1% (in phakic patients), and dexamethasone ointment at night. The duration of therapy was necessarily individualized according to the patients’ postoperative course.
Once the chart had been identified as appropriate for inclusion in the analysis, the number and type of ophthalmic drops used preoperatively by the patients were recorded, as were relevant demographics. The criteria used to evaluate surgical outcomes included IOP levels, need for postoperative ocular antihypertensives, further laser treatments to the operated eye, 5-fluorouracil (5-FU) needling to the bleb, and subsequent incisional surgery in cases of bleb failure. Possible confounding variables collected were age, sex, race, glaucoma type diagnosis, time since glaucoma was first diagnosed, laterality of involved eye, type of surgery performed, ocular surgical and laser treatment history, and length of MMC application during surgery. Patients were examined for outcome criteria over a minimum postoperative follow-up period of 2 years.
A total of 203 patients were identified who had filtration surgery between 2004 and 2006 with a minimum 2-year postoperative follow-up period. Exclusion criteria included recent (up to 1 y) preoperative use of miotics or dipivalyl epinephrine (based on the finding that the addition of these classes to β-blockers increases the risk of surgical failure),3 use of ocular steroids in the immediate preoperative period, and previous trabeculectomy on the same eye. Of the 203 original patients, 75 were excluded based on the above exclusion criteria, leaving 128 patients in the final sample.
The primary outcome measure used in this study stratified patients into complete surgical success, qualified success, or surgical failure. Complete success was defined as a final recorded IOP<18 or a minimum 20% IOP decrease from preoperative baseline, no need for postoperative ocular antihypertensives, no 5-FU needling revisions to the bleb later than 6 weeks after surgery, no postoperative laser treatments to the operated eye (except for YAG capsulotomy or argon laser suture lysis), and no further incisional surgery for control of IOP. Qualified success was defined as a violation of any of the above success criteria, apart from further incisional surgery or transscleral YAG laser treatment, which comprised the surgical failure category.
Because of the variability in postoperative follow-up periods, the data were assessed using survival analysis techniques. The censoring variable was surgical success, and the time to event variable measured the period from surgery until surgical failure, qualified success, or final recorded follow-up. The main model of interest examined the number of drops containing BAK instilled into the eye per day as the independent variable. Kaplan-Meier survival curves were compared using log rank (Mantel-Cox) testing to identify significant divergence between survival distributions at different levels of BAK exposure. When represented graphically, BAK exposure level groups that contained <5 patients were binned with contiguous groups. To examine possible confounders’ contribution to the main effects model, a reduced (parsimonious) Cox proportional hazards regression model and an expanded model were analyzed. The reduced model variables included age, sex, race (white or other), ocular surgical history (previous phacoemulsification and intraocular lens implantation or none), and duration of MMC exposure during surgery. A second, expanded, model was used to examine further variables of interest, which omitted sex and ocular surgical history in favor of a more descriptive racial profile (African, South East Asian, East Asian, or white), use of argon laser suture lysis, laser trabeculoplasty history, surgical procedure type, and glaucoma diagnosis (neovascular or uveitic vs. other). Time since glaucoma was first diagnosed could not be included in either model because the data were generated from patient recollection. The time period data were therefore unreliable and often recorded as a range in years.
To follow convention, type 1 error was set at 0.05. Data analysis was performed using SAS for the a priori power analysis and Cox regressions, PASS for the post hoc power analysis, and SPSS for all other tests including Kaplan-Meier survival analyses.
Medications used by patients are referred to by their trade name, and by their generic name, to convey the preservative formulation associated with a particular brand. The brand names, the preservative concentration, and the suppliers of the medications used in this study are listed in Table 1.
Median patient age was 72 years (SD 11.7), ranging from 23 to 90. Most patients were male (52.3%), and slightly more surgeries were performed on the left eye (58.6%). The average length of time with a glaucoma diagnosis was 8.2 years (SD 5.5), ranging from 4 months to 34.8 years. Mean postoperative follow-up time was 4.3 years (SD 1.0), ranging from 2.0 to 6.3 years. None of the patients had a recorded history of conjunctival disturbing surgery, and only 6.3% had undergone a previous phacoemulsification and intraocular lens implantation. There was no significant difference between the complete success and partial success groups in preoperative IOP levels (20.5 vs. 21.6 mm Hg, respectively, P=0.3953, t test). Patients received between 1 and 8 BAK-containing drops per day, with a median of 3. The aforementioned BAK exposure corresponded to the use of between 1 and 5 medications to control IOP levels daily, again with a median of 3. The number of medications used preoperatively did not significantly correlate with preoperative IOP levels (Pearson r=0.122, R2=0.015, P=0.169). See Table 2 for a summary of the patients’ racial profile, glaucoma diagnosis, surgery type, and cause of partial success (for the subpopulation in which complete surgical success was not maintained).
There was considerable variation in the medications prescribed to patients preoperatively for IOP control. A list of medication combinations that occur in the charts of at least 3 patients can be found in Table 3. When calculating the number of medications used for IOP control, combined formulations (such as dorzolamide/timolol or brimonidine/timolol) were treated as 2 distinct medications but were only counted once toward the total number of BAK-containing drops used per day. The latter drop count also included preserved medications other than ocular antihypertensives (such as ketorolac). Furthermore, in the single patient on an oral medication (acetazolamide) for the control of IOP, this was included in the total medication count.
Complete surgical success was achieved in 47.7% of patients. Only 4 patients suffered a complete surgical failure requiring a repeat trabeculectomy or a transscleral YAG laser treatment; however, in all 4 cases, the surgical failure was preceded by a qualified surgical success as a consequence of the need for postoperative ocular antihypertensive medication or 5-FU treatment. Time to qualified success was therefore used for the survival analysis rather than the time to complete surgical failure.
On the basis of Kaplan-Meier survival analysis with log rank testing for significant survival variation, time to surgical failure in patients receiving higher preoperative daily doses of BAK was significantly shorter than in patients who had less BAK exposure (P=0.008). A graphical representation of the survival curves is included in Figure 1. BAK exposure level groups that contained <5 patients were binned with contiguous groups for clarity. Log rank testing on the grouped exposure levels was still significant at P=0.036.
The reduced Cox proportional hazards regression model did not identify any significant interaction between the tested potential confounding variables (age, sex, previous phacoemulsification, nonwhite race, and duration of intraoperative MMC exposure) and survival time. The number of BAK-preserved drops used daily was still significant as a variable within the model (P=0.0425), at a hazard ratio (HR) of 1.21; however, the overall model did not reach statistical significance (P=0.408). In contrast, the expanded Cox model did identify a significant interaction between survival time and a diagnosis of uveitic or neovascular glaucoma that resulted in a greater chance of surgical failure with an HR of 4.06 (P=0.0238). The HR for BAK exposure was unchanged at 1.21 (P=0.0319), but again the overall Cox model was not significant (P=0.241). When the cumulative daily concentration of BAK was examined, rather than the number of BAK-preserved drops used per day, the effect on surgical survival was still observed (P<0.001) (Fig. 2).
A secondary proportional hazards analysis was executed to examine a potential effect of the number of different medications used to control IOP. The number of medications variable replaced the number of BAK-containing drops variable in both the reduced and expanded Cox models for this analysis. The number of medications used did not have a significant effect on survival as a variable in the models (P=0.887, 0.611, respectively). A diagnosis of uveitic or neovascular glaucoma was still found to be significant (P=0.038) with an HR=3.68. The Kaplan Meier survival analysis for the effect of the number of medications used on surgical survival is shown in Figure 3.
The results of this study reveal that a dose response curve exists with respect to the amount of preoperative BAK exposure. Increased amount of preserved drops used per day increases the risk for early qualified failure. Indeed, for each additional drop containing BAK, the risk of early failure increases by a factor of 1.21 (assuming proportional hazards and a linear association between the number of daily drops). The number of different medications used to control IOP was not found to significantly affect survival time, and with an HR of 1.01, any effect masked by potential type II error is negligible. The impact of BAK exposure is therefore independent of medication amount in terms of risk for early surgical failure. It also follows that disease severity did not play a significant role in early failure risk in this data set, as more severe disease would likely require more aggressive treatment with multiple medication classes, which, as previously described, did not significantly interact with early failure risk.
There are many potential risk factors that can adversely impact the outcome of filtration surgery. The typical variables that are expected to have such an effect include young age, black race, secondary types of glaucoma, and a history of conjunctival disturbing surgery, or laser trabeculoplasty.2,21–22 None of the patients in the study had a recorded history of relevant surgery. Of all the other aforementioned factors, only uveitic and neovascular glaucoma were found to significantly impact surgical outcome. It is likely that due to the relatively small number of patients that fell into the traditional high-risk groups, the study was underpowered to detect their effect on the main model. Nevertheless, the HRs of all potential risk factors were >1, conforming to the expectations from the literature. Interestingly, black race had the same HR as exposure to BAK-preserved drops (1.21), although not at a significant level.
The univariate Kaplan-Meier curves generally reflect the trend of increased risk of failure with increased BAK dosage. The only group that seems not to conform is that of patients receiving 4 BAK-containing drops per day. However, this group was smaller than the others, containing only 9 patients. The small size makes this group inherently more sensitive to outliers, which in this case led to the appearance of greatly reduced survival because 7 of the 9 patients had not enjoyed complete surgical success. This group was not transformed into a binned variable, because it is both large enough for log rank testing, and because there is no clinical guide in choosing whether it should be combined with the group receiving 3 BAK-preserved drops per day or the group receiving 5. The other small groups, however, were at both ends of the BAK exposure spectrum and were therefore binned into the low exposure group (1 or 2 daily BAK-containing drops) and the high exposure group (receiving 6 or more such drops).
There are more filtration surgery failure risk factors that can act as potential confounders of the multivariate Cox model than the model could accommodate at once, because the convention is to limit model variables to an order of magnitude less than the number of “events” in the data set. This study found 67 patients with a qualified surgical success (67 “events”), thereby capping the model at 7 variables. For that reason, we included both a reduced and an expanded model with a different complement of variables in each, as explained in the methods section. The primary model of interest was set to be the more parsimonious one, and so the expanded model is treated as a secondary analysis to support the main model. Consequently, any results from the expanded model should be interpreted with caution, as it is more susceptible to spurious associations than the parsimonious main model.
BAK concentration in the medications used in this study ranged from 0.005% to 0.02%. To account for this variability, the number of BAK-preserved drops administered daily was weighted for the BAK concentration in each drop. The new weighted variable was then tested in a similar manner to that used for the main model. Unsurprisingly, the univariate model found that the exact BAK dose administered was also significantly related to early surgical failure (P<0.001). Nevertheless, it is more clinically efficient to try to control the total number of BAK-preserved drops a patient may receive daily than to calculate weighted BAK dosages.
Alternatives to BAK as a preservative are actively being developed. BAK alternatives in current use include single-use, nonpreserved formulations, barriers to microbiological contamination through a valve mechanism (COMOD) or antimicrobial filters (ABAK), and less toxic preservatives.10 Three new, less toxic preservatives currently available are Purite, a stabilized oxychloro complex, Sofzia, a buffer system containing boric acid, propylene glycol, sorbitol, and zinc chloride, and Polyquad (polyquaternium-1), a polycationic polymer previously used in personal care products and for contact lens care.10 Unit dose medications that are completely free of all preservative also are available, such as Cosopt Preservative Free.
Given the current availability of both non-BAK using and preservative free ocular antihypertensive agents, future work should prospectively evaluate surgical outcomes in patients receiving preserved versus BAK-free versus completely nonpreserved medical regimens preoperatively. Such a study would offer very strong evidence for the impact of BAK on surgical success, without the biases inherent in retrospective, smaller studies. We recognize that this study is limited by its retrospective nature. It would have been very interesting to perform our analysis against the lifetime cumulative exposure to BAK, but obtaining reliable information to enable such an analysis was impossible. Another limitation of our study relates to standardization of postoperative management. The decision to intervene in the postoperative period, and in what manner, depended on clinical judgment taken at the time; no formal criterion could be applied during chart review. Nevertheless, this study offers the most conclusive evidence to date for the detrimental effect of BAK on surgical outcomes in the context of the modern antiglaucoma pharmacopeia. Failure of filtration surgery most commonly results from subconjunctival fibrosis after fibroblast proliferation, collagen synthesis, and glycosaminoglycan deposition.23 Fibroblasts from the perifistula zone contribute to this effect, and previous work demonstrates a significant contribution to this inflammatory cascade from topical medical treatment.3,12 This study identifies BAK specifically as the most likely etiologic agent.
Acknowledging the data generated by the Collaborative Initial Glaucoma Treatment Study, there is no impetus to change the current paradigm of treating to maximal tolerated topical drops before surgery24; however, ophthalmologists managing glaucoma patients may wish to use ocular antihypertensives that are preserved with the less cytotoxic alternatives to BAK. In cases where patients cannot be weaned off of BAK before surgery, consideration could be given to adding an anti-inflammatory or steroid drop 1 month before the operation.25–27
The authors thank Marko Katic (Institute for Clinical Evaluative Sciences) for his kind assistance with statistical analysis.
1. Quigley HA, Broman AT.The number of people with glaucoma worldwide in 2010 and 2020.Br J Ophthalmol.2006;90:262–267.
2. Broadway DC, Chang LP.Trabeculectomy, risk factors for failure and the preoperative state of the conjunctiva.J Glaucoma.2001;10:237–249.
3. Broadway DC, Grierson I, O’Brian C, et al..Adverse effects of topical antiglaucoma medication: II. The outcome of filtration surgery.Arch Ophthalmol.1994;112:1446–1454.
4. Flach AJ.Does medical treatment influence the success of trabeculectomy?Trans Am Ophthalmol Soc.2004;102:219–224.
5. Baudouin C.Detrimental effect of preservatives in eye drops: implications for the treatment of glaucoma.Acta Ophthalmol.2008;86:716–726.
6. Martone G, Frezzotti P, Tosi GM, et al..An in vivo confocal microscopy analysis of effects of topical antiglaucoma therapy with preservative on corneal innervations and morphology.Am J Ophthalmol.2009;147:725–735.
7. Baudouin C.The ocular surface in glaucoma.Cornea.2009;28:S14–S19.
8. Baudouin C, Riancho L, Warnet J, et al..In vitro studies of antiglaucomatous prostaglandin analogues: travaprost with and without benzalkonium chloride and preserved latanoprost.Invest Ophthalmol Vis Sci.2007;48:4123–4128.
9. Ammar DA, Noecker RJ, Kahook MY.Effects of benzalkonium chloride-preserved, polyquad-preserved, and sofZia-preserved topical glaucoma medications on human ocular epithelial cells.Adv Ther.2010;27:837–845.
10. Baudouin C, Labbe A, Liang H, et al..Preservatives in eyedrops: the good, the bad and the ugly.Prog Retin Eye Res.2010;29:312–334.
11. Noecker R.Effects of common ophthalmic preservatives on ocular health.Adv Ther.2001;18:205–215.
12. Broadway DC, Grierson I, O’Brian C, et al..Adverse effects of topical antiglaucoma medication: II. The conjunctival cell profile.Arch Ophthalmol.1994;112:1437–1445.
13. Burstein NL.Corneal cytotoxicity of topically applied drugs, vehicles, and preservatives.Surv Ophthalmol.1980;25:15–30.
14. Ishibashi T, Yokoi N, Kinoshita S.Comparison of the short-term effects on the human corneal surface of topical timilol maleate with and without benzalkonium chloride.J Glaucoma.2003;12:486–490.
15. Schuman JS.Antiglaucoma medications: a review of safety and tolerability issues related to their use.Clin Ther.2000;22:167–208.
16. Yamazaki S, Nanno M, Kimura T, et al..Effects of switching to SofZia-preserved travaprost in patients who presented with superficial punctuate keratopathy while under treatment with latanoprost.Jpn J Ophthalmol.2010;54:7–14.
17. Lewis RA, Katz GJ, Weiss MJ, et al..Travoprost 0.004% with and without benzalkonium chloride: a comparison of safety and efficacy.J Glaucoma.2007;16:98–103.
18. Easty DL, Nemeth-Wasmer G, Vounatsos JP, et al..Comparison of a nonpreserved 0.1% T-Gel eye gel (single dose unit) with a preserved 0.1% T-Gel eye gel (multidose) in ocular hypertension and glaucomatous patients.Br J Ophthalmol.2006;90:574–578.
19. Hamacher T, Airaksinen J, Saarela V, et al..Efficacy and safety levels of preserved and preservative-free tafluprost are equivalent in patients with glaucoma or ocular hypertension: results from a pharmacodynamics study.Acta Ophthalmol.2008;242:S14–S19.
20. Uusitalo H, Kaarniranta K, Ropo A.Pharmacokinetics, efficacy and safety profiles of preserved and preservative-free tafluprost in healthy volunteers.Acta Ophthalmol.2008;242:S7–S13.
21. Mietz H, Raschka B, Krieglstein GK.Risk factors for failures of trabeculectomies performed without antimetabolites.Br J Ophthalmol.1999;83:814–821.
22. Dally LG, Ederer F, Gaasterland DE, et al..The advanced glaucoma intervention study (AGIS): 11. Risk factors for failure of trabeculectomy and argon laser trabeculoplasty.Am J Ophthalmol.2002;134:481–498.
23. Skuta GL, Parrish RK II.Wound healing in glaucoma filtering surgery.Surv Ophthalmol.1987;32:149–170.
24. Lichter PR, Musch DC, Gillespie BW, et al..Interim clinical outcomes in the collaborative initial glaucoma treatment study comparing initial treatment randomized to medications or surgery.Ophthalmology.2001;108:1943–1953.
25. Baudouin C, Nordmann JP, Denis P, et al..Efficacy of indomethacin 0.1% and fluorometholone 0.1% on conjunctival inflammation following chronic application of antiglaucomatous drugs.Graefes Arch Clin Exp Ophthalmol.2002;240:929–935.
26. Broadway DC, Grierson I, Sturmer J, et al..Reversal of topical antiglaucoma medication effects on the conjunctiva.Arch Ophthalmol.1996;114:262–267.
27. Breusegem C, Spielberg L, Van Ginderdeuren R, et al..Preoperative nonsteroidal anti-inflammatory drug or steroid and outcomes after trabeculectomy: a randomized controlled trial.Ophthalmology.2010;117:1324–1330.
© 2013 by Lippincott Williams & Wilkins.