Results

Twenty six episodes of post-trabeculectomy endophthalmitis were identified (seven prospectively) of which three were excluded from analysis: two from patients who presented 5 and 14 days respectively following surgery and the other being the second occurrence of post-trabeculectomy endophthalmitis in the same patient. All cases had undergone glaucoma filtration surgery at Moorfields Eye Hospital. The characteristics of the 23 cases and 69 controls (Table 1) were similar with regard to a wide number of variables including age, sex, eye affected, phakic status and the surgical technique employed.

An episode of blebitis, which occurred in six of 23 cases (26%) and two of 69 controls (3%), was statistically significantly associated with post-trabeculectomy endophthalmitis; odds ratio 11.82 (95% CI 2.21–88.31), p = 0.003. The mean time interval between these treated episodes of blebitis and the subsequent development of endophthalmitis was 9 weeks (SD 4.2). Diabetes mellitus occurred more commonly in cases than controls (five of 23 (22%) versus four of 69 (6%)) and was also statistically significantly associated with post-trabeculectomy endophthalmitis; odds ratio 4.51 (95% CI 1.02–20.29), p = 0.04. Antiproliferatives were widely used intraoperatively with mitomycin C being applied most frequently in both cases and controls. The time interval between filtration surgery and development of endophthalmitis was significantly shorter in antiproliferative treated cases (p = 0.001, rank sum test). Use of an antiproliferative agent (20 of 23 cases (87%) versus 46 of 69 controls (67%)) was associated with subsequent endophthalmitis with an odds ratio of 3.33 (95% CI 0.95–15.19), although these results did not reach statistical significance, p = 0.07.

The clinical characteristics of the 23 patients diagnosed with post-trabeculectomy endophthalmitis and the organisms isolated in 10 cases (43%) are summarised in Table 2. The median time interval between glaucoma surgery and endophthalmitis was 15 months (range 1–180) and the mean reduction in acuity following the episode of endophthalmitis was two Snellen lines (SD 2.6). At the latest follow up (median 7 months, range 1–154) 10 of 23 eyes (43%) required medication to control the intraocular pressure and a further three eyes (13%) were either hypotonous or phthisical.

Discussion

Case-control methodology was applied to patients presenting with post-trabeculectomy endophthalmitis to a single hospital over a 6½ year period in order to clarify the risk factors for developing the condition.

The results at the univariate level provide strong evidence that the odds of endophthalmitis are increased approximately 12-fold (OR 11.82, 95% CI 2.21–88.31, p = 0.003) in a patient in whom blebitis has occurred. This result is in keeping with the recognised ability of blebitis to progress to endophthalmitis¹⁵ and it is likely that the changes in bleb morphology that predispose to bleb related infection increase susceptibility for subsequent endophthalmitis. Although the confidence interval associated with the odds ratio is wide, illustrating the imprecise nature of this estimate, even the lower confidence limit lies well above unity. These data also provide evidence that endophthalmitis is more likely in a patient who has diabetes mellitus (OR 4.51, 95% CI 1.02–20.29, p = 0.04); a finding in keeping with diabetes being a proved risk factor for other types of endophthalmitis.16 17

Although the power of this study was limited—a 57% chance of detecting an odds ratio of 5 (assuming α = 0.05, three controls per case and 70% exposure among controls)—these data provide some evidence that the odds of endophthalmitis may be greater in patients in whom antiproliferatives are used (p = 0.07). A patient treated with antiproliferatives is estimated to be three times as likely to develop endophthalmitis as one who is not; however, the 95% confidence interval is just consistent with there being no difference in odds. An adjusted odds ratio was not computed because of the limited data set and because of concerns that blebitis may lie along a postulated causal path between antiproliferative use and endophthalmitis. An association was also found between antiproliferative use and a shorter time interval to the development of endophthalmitis (p = 0.001). This result may, however, partly reflect ascertainment bias as the study included patients developing endophthalmitis within a 6½ year period (patients undergoing trabeculectomy in the era before widespread antiproliferative agent use could have developed endophthalmitis within a year or two of surgery without being ascertained). Other factors were not found to be associated with endophthalmitis.

These results indicate that blebitis, diabetes mellitus and, possibly, antiproliferative use are associated with subsequent endophthalmitis. Migration of bacteria across the filtering bleb is recognised to be the initial step in the development of blebitis3 18 and ordinarily such spread is prevented by the physical properties of the bleb together with the eye's innate immune response. Antiproliferative use alters the thickness, cellularity, and vascularity of the filtering bleb's conjunctiva. The thinner and relatively avascular blebs that result frequently have areas of absent conjunctival epithelium which are demonstrable both microscopically8 and clinically. These areas of denuded epithelium and the underlying full thickness stromal damage7 compromise the bleb's physical barrier function and affect the conjunctival contribution to the ocular innate immune response. One component of this response is a family of highly conserved cysteine rich antimicrobial peptides, called defensins, that combat infection at a wide range of mucosal surfaces.19 20 The constitutive conjunctival expression of certain defensin isoforms (β₁) at a high level21 suggests that they are important components of the conjunctiva's innate defences. β Defensin expression will be altered in unepithelialised areas of the filtering bleb and we hypothesise that the combination of breaches in the bleb's barrier function with altered conjunctival β defensin production may account for part of the observed increased susceptibility to infection of antiproliferative agent treated blebs.

Direct comparison between published reports of post-trabeculectomy endophthalmitis cases is hampered by variation between studies regarding factors including study design, case definition, case mix, antiproliferative agent use (both type, dose and duration of administration), as well as the nature and site of surgery (superior/inferior filtration) and the treatment received on presentation with endophthalmitis.1 9-11 ¹⁵ 22-25 For these reasons we have limited comparison of the clinical features (Table 2) of this group of patients to noting that the visual outcome and rate of filtering bleb survival are broadly comparable with previous reports (and in a study period partially predating the Endophthalmitis Vitrectomy Study26 that 22 of the 23 cases received intravitreal antibiotics). The poor prognosis associated with post-trabeculectomy endophthalmitis is best highlighted by the fact that just three of 23 eyes achieved satisfactory intraocular pressures without treatment and a Snellen acuity greater than 6/24. Overall episodes of post-trabeculectomy endophthalmitis were associated with a high likelihood of trabeculectomy failure (57%) and a mean reduction in acuity of two Snellen lines. This measurement understates the effect of visual loss in individual patients with poor initial acuity as a result of advanced or congenital glaucoma and the number of patients in the latter subgroup (five of 23) may skew the overall visual results reported. The proportion of cases (52%) occurring more than 1 year postoperatively with some occurring several years later indicates that the risk of post-trabeculectomy endophthalmitis is long lasting, perhaps for at least the lifetime of a functioning bleb.6 This is of particular concern in younger glaucoma patients in whom surgery is the main treatment modality. The design of the study precluded calculation of incidence data owing to the likelihood of acertainment bias and the study may have been biased to detecting fewer cases of late endophthalmitis than if all trabeculectomy cases had been followed prospectively. None the less the rise in the number of cases presenting to Moorfields is compatible with reports6 10 11 suggesting that the frequency of post-trabeculectomy endophthalmitis is increasing. Whether the more favourable bleb morphology achieved with larger areas of antiproliferative treatment27 has the potential for reducing the frequency of this complication remains to be determined.