Methods

The medical records of all patients with open globe injury admitted to 15 tertiary referral hospitals in China between 1 January 2001 and 31 December 2005 were retrospectively reviewed. Each case was recorded in a standardised preformulated data sheet and in the computerised database of eye injury. This study included 4968 eyes of 4865 patients with open globe injuries admitted to the 15 tertiary referral hospitals of China over 5 years. A total of 173 eyes (one bilateral rupture of a male eye) enucleated or eviscerated within 24 h after trauma were excluded.

In this study, classification and definition of ocular trauma were based on the Birmingham Eye Trauma Terminology (BETT). Open globe injury indicates a full-thickness wound of the eyeball. A rupture wound is produced by an inside-out mechanism caused by a blunt object, whereas a laceration wound (including penetrating, intraocular foreign body (IOFB) and perforation) occurs at the impact site by an outside-in mechanism caused by a sharp object. Self-sealing wounds indicate that the wounds of globe wall closed tightly by themselves without primary repair. Intraocular tissue prolapse means that the intraocular tissues (including vitreous body, uvea, retina and lens) slipped out through wounds partly and were incarcerated there. The diagnosis of injuries was finally made by determining the mode of injury and by the findings during operation. Diagnosis of endophthalmitis was made chiefly by the clinical characteristic symptoms and signs. Culturing of intraocular contents was made only in certain cases. The outcomes of culturing were not considered in the diagnosis of endophthalmitis.

A standardised data sheet was completed for each case of ocular trauma. Information such as patient age, gender, occupation, medical and ophthalmic history, previous vision, if known, circumstance and timing of injury occurrence, clinical presentation, the way the monitoring system was implemented and outcomes at discharge or follow-up were collected in detail.

All data were collected in an electronic database and cross-checked for errors. Statistical analysis was performed using the SPSS version 17.0 (SPSS, Chicago) data-analysis software package. Categorical variables were analysed using the χ² test. Continuous variables were examined for normality, and means were compared using the t test. A further multiple logistic regression analysis was conducted in order to predict the independent factors affecting occurrence of endophthalmitis postopen globe injury. The critical value of significance was set up at p<0.05 for all tests.

Results

This study identified 571 eyes (571 patients) with endophthalmitis in all the open globe injury cases with an incidence of 11.91% (table 1). The median age was 26.19 (16.57) years (range 15 months to 78 years) and less than that of patients without endophthalmitis (28.05 (15.02) years; u = 15.47, p<0.01). The ratio of males to females was 5.72:1, and 311 (54.47%) left eyes were involved. Apart from 42 (7.36%) eyes with rupture, mechanisms in the other eyes with laceration which included penetration (281 eyes, 49.21%), IOFB (232 eyes, 40.63%) and perforation (16 eyes, 2.80%) are listed in table 2.

Univariate analysis identified six clinical presentation factors associated with development of endophthalmitis. Retaining foreign body, penetration, perforation, primary repair of globe wounds more than 24 h after trauma, no intraocular tissue prolapse, I or II zone of wounds and self-sealing of wounds were statistically significant factors. However, gender, injury side and lens breach were not significant factors (table 3). Logistic regression analysis was also performed on the data set. Laceration was found to be an independent risk factor, and primary repair within 24 h, intraocular tissue prolapse and self-sealing of wounds were independent but significant protective factors against occurrence of post-traumatic endophthalmitis. However, age and posterior zone of wounds were not significant (table 4).

Other clinical findings included hypopyon in 223 eyes (39.05%), vitritis in 450 eyes (78.81%), retinal necrosis in 74 eyes (12.96%) and occlusion of retinal vessels in 67 eyes (11.73%). There were four (0.70%) cases that subsequently developed bilateral sympathetic ophthalmia. For 53 cases (9.28%) with comparatively severe vitritis, intravitreal antibiotic and corticosteroid therapy, by means of pars plana injection or vitreous cavity irrigating during vitrectomy, was administered. Intraocular specimens obtained during pars plana vitrectomy of 42 eyes with endophthalmitis yielded positive cultures in 25 eyes (59.52%), which included Streptococcus species in seven eyes, Staphylococcus species in 12 eyes, Bacillus species in four eyes and Pseudomanas aeruginosa and fungus respectively in one eye. Systemic, subconjunctival and topical antibiotic and/or corticosteroid therapy were administered routinely. Retrobulbar rejection of antibiotic and corticosteroid was applied for some eyes with comparatively severe vitritis. Vancomycin hydrochloride, tobraymycin, ceftazidime or ceftriaxone and dexamethasone were the commonly used antibiotics. A total of 305 (53.42%) eyes underwent vitrectomy for reasons of injury, vitritis or both.

At discharge or follow-up, 96 eyes (16.81%) with endophthalmitis underwent enucleation or evisceration totally. The proportion of eyes with endophthalmitis was higher than that (368 eyes, 8.71%) of other open injured eye globes without endophthalmitis (χ² = 13.3286, p = 0.0003). The proportion of eyes with endophthalmitis with a final visual acuity (VA) of NLP (no light perception) and LP-4/200 decreased, and the proportion of eyes with endophthalmitis with a final VA of 5/200 or greater almost doubled significantly (u = 6. 291, p<0.01; table 5).

Discussion

This large-sample retrospective review was based on a preformulated eye injury registry. The details of every patient of open globe injury were recorded in a standardised data sheet during admission to the hospital. This ensured that the important data for statistical analysis were not missed out as far as possible. The limitation to our retrospective study lies in a lack of uniformity of 15 hospitals in spite of the fact that they are all tertiary referral hospitals with equivalent medical level. The control respective trial may yield more persuasive data.

The present study found an 11.91% incidence of post-traumatic endophthalmitis. This is consistent with values of 2.6–54.16% published previously for open globe injury7 8 9 10 12 but is higher than 6.8% reported in another study of inpatients with a much smaller sample size11 (table 1). The incidence of endophthalmitis after penetrating ocular injury has been reported to be as low as 3.3% to as high as 30%2 4 13 14 and that after an IOFB to be 1.3% to 61%.15 16 17 18 The results from the current study are in agreement with the above related studies respectively (table 3). On the other hand, the frequency of endophthalmitis after open globe injury without IOFB in this study (339/3411, 9.94%) is much higher than the 4.4% reported by Essex et al.11 The statistical differences in data can likely be attributed to differences in composition and size of samples.

As for the demographic data, male preponderance was similar to the previous report by Alfaro et al,6 and the mean age was much lower than that reported by Essex et al,11 which could be attributed to different populations of eye injury. In the current series, patients with endophthalmitis are comparatively younger in a statistically significant manner by univariate analysis, which is contrary to the results of Essex et al.11 However, age was not found to be an independent risk factor by multivariate analysis. Thus, we only can conclude that age is a meaningless factor on development of traumatic endophthalmitis.

The mechanisms by which eyes with endophthalmitis sustained injuries were by the use of a sharp instrument or IOFB, or the injuries were sustained through occupational hazards. Among these, it is worthy of note that the risk of eye injuries to children from disposable injection needles cannot be ignored. In China, the rules for safe disposal of disposable injection needles are not implemented effectively in some rural areas and hence remain a major problem. Children do not have sufficient awareness for self-protection, and some routine sharp objects, such as tips of pencils or pens, could prove to be safety hazards. Effective adult supervision of children, education regarding the spectrum of ocular injuries and the risk of endophthalmitis following ocular trauma may help prevent the same.

Different opinions exist about the factors that influence endophthalmitis development following open globe injury. In the univariant analysis of this study, retaining IOFB was found to be a risk factor for the development of endophthalmitis, which is similar to most other reported studies.11 15 16 Although the incidence of endophthalmitis is commonly reported to be higher in eyes with IOFBs than in eyes without IOFBs,1 5 7 11 14 16 17 only a few studies16 18 have been able to associate IOFB with endophthalmitis in a statistically significant fashion by multivariable analysis. However, using a variance analysis, we found there was no significant difference in frequency of endophthalmitis among eyes with IOFB, where penetration and perforation were much higher than that of eyes with rupture (p<0.001). Additionally, laceration (including penetration, IOFB and perforation) was verified by logistic regression analysis to be an independent risk factor. On this basis, we assume that it is the nature of the injury itself, that is intraocular intrusion of foreign vulnerants, rather than retention of IOFBs that truly increases the risk of development of endophthalmitis after open globe injury. Intrusion of vulnerants which break the intraocular physiological barrier and the foreign contaminants which infect the sterile inner environment might be the common risk of laceration which is markedly different from rupture.

Another independent risk factor that we identified was the timing of primary repair, namely, the time interval between injury and treatment at a hospital ⩾24 h, which was in accordance with the rules of the United States Eye Injury Registry.12 Narang et al11 concluded that delayed primary repair beyond 24 h after open globe injury was a significant risk factor for the occurrence of endophthalmitis and even poorer final visual acuity. Additionally, the limit delay of primary repair reported was 12 h by Essex et al11 and 36 h by Gupta et al.19 Although endophthalmitis may not occur when primary repair is delayed in certain cases of open globe injuries, and a delay of 12 h and 24 h can be tolerated, primary repair after open globe injury should be performed as soon as possible because this is the link that can be most easily controlled in the prophylactic treatment of endophthalmitis.

To our knowledge, intraocular tissue prolapse and self-sealing of wounds in the globe wall have not been previously demonstrated to be independent protective factors against post-traumatic endophthalmitis. Traditionally, people thought intraocular tissue prolapse may expose the intraocular contents to exogenous flora or organisms that increase the chances of infection. This was supported by Soheilian et al,18 who identified vitreous prolapse as a significant factor associated with the development of acute post-traumatic bacterial endophthalmitis, which is contrary to our results. On the other hand, Gupta et al19 found that the presence of uveal or vitreous prolapse did not significantly increase the risk of endophthalmitis. Further studies are needed to investigate whether intraocular tissue prolapse can facilitate or prevent invasion of foreign pollutants and whether the protective effect could be due to other causes. In addition, self-sealing of wounds is actually equivalent to primary repair in protecting the integrity of construction.

Lens breach, in our series, is an insignificant factor, which is similar to the results of some earlier studies9 15 20 and contrary to the results of others.7 11 21 In our univariate analysis, an increased rate of endophthalmitis is demonstrated in the anterior zone of wounds but was revealed not to be significant in the multivariable analysis. Schmidseder et al22 and Thompson et al16 presumed that the posterior location of wounds was a risk factor.

In the absence of a large randomised clinical trial with a long follow-up period, there are no definite or conclusive guidelines on the methods of prophylaxis and their details of administration. Prophylactic systemic, subconjunctival and topical antibiotics were traditionally recommended for the most open globe injuries as “standard care” despite the lack of strong evidence. In the clinic, the aforementioned ways were effective for most cases in which inflammation were mild and affected the anterior segment mainly, and they were usually safe. We routinely do the same, but their necessities as prophylaxis still need to be justified.

Intravitreal antibiotics and corticosteroid as a successful prophylaxis of post-traumatic endophthalmitis have been shown in animal models23 24 and clinical trials.18 25 However, the limitations included a discrepancy in the sample in the study by Narang et al25 and short follow-up period of 2 weeks compared with that of Soheilian et al.18 In fact, routine administration of prophylatic antibiotics into the vitreous is not universally accepted, which may theoretically increase the rate and speed of development of drug resistance. And the risk of retinal toxicity and complications of operation itself should not be disregarded at all. If choroidal or retinal detachment is present, the risks of intravitreal injection increase markedly. In addition, the selection of antibiotics was blind before there was any evidence of clinical infection. It is difficult to decide as to whether the benefits can outweigh risks in practice for different traumatic individuals. Therefore, the efficacy of intravitreal antibiotics and corticosteroid as a prophylaxis needs to be verified by much larger randomised prospective trials.

Vitrectomy was recommended as a therapy for all cases of post-traumatic endophthalmitis26 27 and even as prophylaxis in most cases of clinically suspected traumatic endophthalmitis.12 The theoretical benefits include reducing inflammatory contents in eyes and allowing a better distribution of intraocular antibiotics. However, vitrectomy undoubtedly would burden patients with additional surgical risk, more medical expense and mental pressure. If a definite diagnosis of endophthalmitis could not be made, most doctors and patients would have scruples about attempting vitrectomy. Besides, for mild vitritis, intravitreal or even a retrobulbar injection of corticosteroid combined with antibiotics works well. If the severe vitritis is definite or intravitreal injection of antibiotics cannot relieve inflammation, vitrectomy with intravitreal medicine treatment without doubt should be administered promptly as an appropriate and most effective way to control inflammation.

Although our results show that more eyes with endophthalmitis were removed (p<0.001), this does not mean that endophthalmitis significantly influences the frequency of enucleation/evisceration. While destruction of eyes caused by endophthalmitis itself was found to be an important factor in increasing the chances of enucleation/evisceration, it should be borne in mind that the conditions of injury were also more severe in the majority of eyes developing endophthalmitis. Essex et al11 have reported in their study that there was no difference in the frequency of enucleation/evisceration between eyes with and without endophthalmitis and that final visual acuity was worse in eyes with endophthalmitis. In our opinion, management of post-traumatic endophthalmitis is challenging, and it is difficult to evaluate the prognosis of visual acuity of eyes with post-traumatic endophthalmitis because visual prognosis is affected by a complex confluence of factors, which include varying degrees of tissue damage. However, on the whole, the final visual acuity of remained eyes improved markedly along with the recovery of anatomy and inflammation.