The basic characteristics of patients with infectious endophthalmitis recruited in this retrospective study are shown in Figure 1. Altogether, 359 eyes from 359 patients were diagnosed as showing infectious endophthalmitis, and the patients were treated at our hospital within 5 years. In the five-year case counts, the overall trend showed an increase, but with a small drop in 2016 and 2017 (2014: 62, 2015: 76, 2016: 73, 2017: 68, and 2018: 80).
Age, gender, marital status, occupation and eye characteristics
The average age of the injured patients was 48.0±18.27 years (range, 4 to 86 years), and the median age was 48 years. The 46-60-year age group contained the most cases (108, 30.1%), followed by the 31-45-year group (92, 25.6%). Of the 359 patients, 283 were male (78.8%) and 76 were female (21.2%). The male-to-female ratio was 3.7:1. In the injured population, 77.4% (278) of the patients were married, 15.6% (56) were single, 2.5% (9) were divorced, and 4.5% (16) were widowed. The patients were mainly involved in five kinds of occupations: farmer, worker, student, retired, office clerks, and others, with the corresponding incidence rates being 39.3%, 32.9%, 6.1%, 6.1%, 4.2%, and 11.4%, respectively. Among the 359 patients, 51.0% (176) showed only unilateral right eye involvement, 49.0% (183) showed only unilateral left eye involvement, and no one showed bilateral involvement.
Causative organisms
Following the onset of endophthalmitis, 316 patients (88.0%) underwent diagnostic tapping of ocular specimens for microbiological investigations, including vitreous tap and/or aqueous tap. All 316 diagnostic taps were performed at the time of initial presentation of ocular symptoms and prior to commencement of intravitreal antimicrobial therapy. The incidence of negative cultures was 55.4%, while that of positive cultures was 44.6%. Gram-positive organisms were found to be the causative microorganism in 115 eyes (81.6%), Gram-negative organisms in 16 eyes (11.3%), mixed bacterial populations in 4 eyes (2.8%), and fungi in 6 eyes (4.3%), Common pathogenic bacteria are shown in Table 1, the most common microorganisms were Staphylococcus epidermidis (59 cases), followed by S. aureus (11 cases).
Etiological classification
Etiological classification of endophthalmitis were demonstrated (Table 2). Table 2 shows that ocular trauma, especially open globe injury, was the most frequent cause of the disease, accounting for 68.52% of all patients, Among the 246 posttraumatic cases, 87 (35.37%) involved IOFB, 71 (28.86%) involved metals, and 16 (6.50%) involved non-metallic items, while 197 (80.08%) had zone I injuries, followed by zone II (44, 17.89%) and zone III (5, 2.03%). Post-surgical cases were the second most common, accounting for 22.6%, and included post-cataract (62.96%), post-glaucoma (24.69%), post-PPV (11.11%), and post-IVI (1.23%). Other causes were corneal ulcer-associated (CA) (6.69%) and endogenous (2.22%).
Statistical analysis of the factors among the four groups showed in Table 2. There was no significant difference in every year and affected eye among the four groups (χ2 test: P>0.05),but patients with endogenous endophthalmitis are more susceptible to the left eye. Age demonstrated a significant difference among the four groups (F=43.04, P < 0.001), and the posttraumatic group (41.67±15.51 years) was younger than the other groups, while the endogenous group (67.13±16.13 years) had the oldest age. There were significantly more males (89.02%) in the posttraumatic group compared with the other groups, while the cornea ulcer-associated endophthalmitis tended to affect females(75.00%) more ( χ2 = 84.69, P < 0.001). There was a significant difference in occupation distribution among the four groups, and the posttraumatic group tended to have more workers (43.90%), while the other groups tended to have more farmers (38.27%, 66.67% and 50.00%, respectively, P < 0.001). Patient visit time in different group was statistically significant difference(χ2=47.41, P<0.0001). Regarding therapy modalities, PPV were the major in posttraumatic and postsurgical endophthalmitis, while enucleation accounted for the most in cornea ulcer-associated and endogenous endophthalmitis (χ2=137.3, P<0.0001). The causative organisms had significant difference among the four groups (χ2=40.20, P = 0.0004). Most of endogenous endophthalmitis tended to be caused by Gram‑positive organisms (100%), while as fungus in cornea ulcer-associated group(42.86%). The pre-therapy visual acuity and post-therapy visual outcome compared in the four groups were significant difference(F=7.198, F=15.82, respectively, P<0.0001). The visual acuities were improved in 253 cases (60.2%), stable in 58 (13.8%), and worse in 109 (26%). The improvement rate of visual acuity was highest in postsurgical endophthalmitis(81.48%), while the visual acuity did not improve significantly in cornea ulcer-associated (79.17%) and endogenous (75.00%) endophthalmitis.
Therapy modalities
The therapy modalities are listed in Table 3. Medical therapy (MT), defined as appropriate topical, periocular, and systemic antibiotics, was used in 26 cases (7.24%). Combined vitrectomy and intraocular antibiotics were used on 243 cases (67.69%), whereas 54 (15.04%) were treated with intravitreal antibiotic injections (IVA) alone. Intravitreal antibiotics included either vancomycin and ceftazidime. Enucleation was performed in 36 cases (10.03%), ten of which had their eyes removed in a second operation. cornea ulcer-associated endophthalmitis were the most cause in the enucleation group(Table 1).
Visual outcome
Visual outcomes were assessed for 358 eyes, excluding the eye of one child who did not cooperate with the vision test. The average logMAR BCVA on pre-therapy was 2.28 ± 0.60 and the value had significantly improved to 1.67 ± 0.83 on post-therapy (t=7.161, P< 0.0001).
Pre-therapy and post-therapy VA were compared among the four groups (Gram-positive, Gram-negative, fungi, and culture-negative) are shown in Figure 2A. We observed a significant difference in visual acuity between Gram-positive(G+) and culture-negative patients at pre-therapy and post-therapy (P < 0.0001), while there was no significant difference in patients with Gram-negative(G-) bacteria and fungal infection. Figure 2B showed the statistical difference between pre-therapy and post-therapy in MT, MT+IVA and MT+IVA+PPV (P < 0.0001),while the post-therapy VA had no significant difference in these therapy modalities.
Demographic and clinical features associated with improved visual acuity are demonstrated in Table 4. There were no significant differences in average age, but significances in the segment of age and gender (P < 0.05). For eye and occupation, there are no statistically significant (P > 0.05). Patient visit time is an important statistical factor, the average visit time and time segment were statistically different in the not-improved and improved groups. It was noted that post-traumatic and post-surgical endophthalmitis had obvious improved visual outcomes compared to those having other types of endophthalmitis, and the difference was statistically significant (χ2=41.98, P<0.0001). For causative organisms, in 141 cases of positive cultures, G+ coccus showed the most favorable visual outcomes when compared to others, there was statistically significant differences (χ2=12.93, P=0.0047). Regarding therapy modalities, PPV demonstrated more improved visual outcomes, the difference was statistically significant (χ2=87.15, P<0.0001). and the pre-therapy VA display an important role in improving vision, and the patient’s visual acuity of HM (2.3 logMAR) or better trended to statistical improved (χ2=9.00, P=0.003). Among 246 patients with post-traumatic endophthalmitis, the presence of IOFB( χ2=4.841, P<0.0001) and wound location( χ2=7.398, P=0.0247) were statistically correlated with improved VA, But there is no statistical difference between metal and non-metal IOFB. The visual improvement in zone I was significantly better than that in zone II and III.
Binomial Logistic regression analysis of predictive factors of improved VA
Multivariate analysis using binomial logistic regression model was conducted to examine the predictive factors of improved VA. After adjusting for possible confounding factors, visit time>7 day (P=0.034, OR=0.522, 95%CI:0.286-0.953), pre-therapy VA≦logMAR 2.3(P=0.032, OR=1.809, 95%CI:1.052-3.110), etiology of PS (vs. PT; P=0.023, OR=2.100, 95%CI:1.109-3.974) and etiology of CA(vs. PT; P=0.005, OR=0.202, 95%CI:0.066-0.621) were significantly associated with improved VA (Table 4).