Author: Claude (AI), from PubMed abstracts · Reviewed by: Dr Mina Abdelmseih · Reviewed: 2026-11-15 · Review due: 2027-11-15

Educational content only. Not personalised medical advice. Clinical decisions remain the responsibility of the treating clinician.

Bacterial keratitis

Bacterial keratitis is a sight-threatening corneal infection that, if not diagnosed and treated promptly, can cause corneal scarring, visual impairment and blindness [S11]. Nearly half of microbiologically confirmed infectious keratitis worldwide is bacterial, with trauma and contact lens wear as the main reported risk factors [S12]. Fungal keratitis is covered in its own article.

Key points - Broad-spectrum topical antibiotics are the primary treatment; local organism and resistance patterns should guide the choice [S11] [S12]. - Fluoroquinolone monotherapy and fortified antibiotic combinations gave similar cure rates in a meta-analysis of 9 RCTs [S9]; in a network meta-analysis ciprofloxacin healed slowest [S11]. - Adjunctive topical corticosteroids did not improve 3-month vision overall in the largest RCT (SCUT) [S1]; possible benefit appeared only in subgroups, and Nocardia ulcers did worse on steroids [S3] [S4]. - Where effective antibiotics are not available, povidone-iodine 1.25% was not significantly different from commonly available antibiotics in an RCT in the Philippines and India [S6]. - Adjuvant corneal cross-linking (CXL) is unproven for bacterial keratitis [S7] [S8].

Evidence last checked: 2026-10-17 (PubMed abstracts).

Red flags

• Corneal infiltrate with an epithelial defect in a contact lens wearer or after trauma. These are the leading reported risk factors [S12].
• Suspected Pseudomonas. Pseudomonas aeruginosa made up 17% of isolates in a global review, which advises empirical therapy that covers it [S12].
• Possible MRSA or fluoroquinolone resistance. About 20% of S. aureus isolates were MRSA and fluoroquinolone susceptibility of S. aureus was about 77% in a global meta-analysis [S12].
• Steroids started without antibiotic cover or in Nocardia keratitis. Corticosteroids were associated with larger scars in Nocardia ulcers [S3].
• Worsening despite treatment. In one RCT 18.2% and 18.5% of ulcers worsened in both treatment arms [S2].

Recognise it

• Who is at risk. Reported risk factors include trauma (about 44%) and contact lens wear (about 23%) [S12]; a Cochrane review also lists ocular surface disease and previous ocular or eyelid surgery [S5].
• Burden. The pooled global proportion of bacterial keratitis among infectious keratitis was 44% (95% CI 39 to 49%), ranging from about 61% in Oceania to about 29% in Africa [S12]. Infectious keratitis rates are higher in low- and middle-income countries [S11].
• Organisms. Of about 25,000 isolates, 72% were Gram-positive; dominant species were S. epidermidis (15%), S. pneumoniae (13%), S. aureus (10%) and P. aeruginosa (17%) [S12].

Confirm it

• Corneal scraping for smear and culture identifies the organism so that the correct antibiotic can be started [S10].
• Instrument. In an RCT of 115 patients, a 26-gauge needle and a Kimura spatula gave comparable sample adequacy, staining, culture and safety; the needle gave more fully adequate samples in subgroup analysis and is cheap, disposable and needs no sterilisation unit [S10].
• Susceptibility. Gram-positive isolates showed more than 95% susceptibility to vancomycin and Gram-negative isolates more than 90% to aminoglycosides and cephalosporins [S12].

Manage it

Before acting on this section, check the national and local guidelines, product information and regulatory approvals that apply where you practise. See the medical disclaimer.

Tier A (primary care / no specialist)

• Recognise the red flags and refer early (see "Refer when").
• Where effective topical antibiotics are not an option, povidone-iodine 1.25% showed no significant difference from neomycin-polymyxin B-gramicidin (Philippines) or ciprofloxacin 0.3% (India) for time to presumed cure in a randomised, investigator-masked trial of 172 patients [S6]. The authors suggest it can be considered when antibiotic treatment is not practical [S6]. This is a single trial with wide confidence intervals [S6].
• Dosing schedules are not covered by these abstracts: follow local formularies.

Tier B (general eye clinic)

1. Scrape before treating, where possible, with a 26-gauge needle if no spatula is available [S10].
2. Start broad-spectrum topical antibiotics [S11]: - A meta-analysis of 9 RCTs (2024) found fluoroquinolones equivalent to fortified antibiotics for cure rate (OR 0.99, 95% CI 0.75 to 1.30) and adverse events (OR 0.75, 95% CI 0.48 to 1.17) [S9]. - In an equivalence RCT of 224 patients, moxifloxacin 0.5% healed 81.4% of moderate ulcers at 3 months versus 81.8% with fortified cefazolin 5% plus tobramycin 1.3%, as reported in [S2]. - A Cochrane network meta-analysis (2025; 23 RCTs, 2692 participants) ranked vancomycin plus ceftazidime, moxifloxacin, and cefazolin plus tobramycin as most effective for healing time, and ciprofloxacin as least effective; most evidence was of moderate to very low certainty [S11]. - Because about 20% of S. aureus isolates were MRSA, the global review recommends broad empirical therapy targeting MRSA and Pseudomonas [S12].
3. Corticosteroids: specialist decision only. SCUT (500 patients, culture-positive ulcers, at least 48 hours of moxifloxacin before randomisation to prednisolone phosphate 1% or placebo) found no difference in 3-month vision, scar size, re-epithelialisation or perforation [S1]. A Cochrane review of 4 RCTs (611 participants) found inadequate evidence for benefit or harm [S5].

Tier C (specialist centre)

• Selective steroid use. In SCUT, steroids gave better 3-month vision in eyes with counting fingers or worse at baseline (0.17 logMAR) and with completely central ulcers (0.20 logMAR) [S1]. At 12 months, steroids were associated with about 1 line better vision in non-Nocardia ulcers but larger scars in Nocardia ulcers [S3]. Adding steroids after 2 to 3 days of antibiotics gave about 1 line better vision, whereas adding them after 4 or more days did not [S4]. These are subgroup analyses and need confirmation [S1] [S4].
• Cross-linking (PACK-CXL). A Cochrane review (3 trials, 59 eyes) found very low-certainty evidence and could not show benefit [S7]. The CLAIR RCT in India (smear-positive ulcers, moxifloxacin 0.5% with or without CXL) could not confirm a benefit, although fewer perforations or therapeutic keratoplasties occurred with CXL [S8].

Refer when

Pitfalls

• Starting steroids early without a culture-proven bacterial diagnosis and antibiotic cover; SCUT enrolled only culture-positive ulcers already on moxifloxacin [S1].
• Assuming fluoroquinolone susceptibility: S. aureus susceptibility was about 77% [S12].
• Treating Nocardia ulcers with steroids [S3].
• Offering CXL as standard care: evidence is low certainty [S7] [S8].
• Skipping scraping because a spatula is unavailable: a 26-gauge needle performed comparably [S10].

Evidence notes

• In the network meta-analysis, 12 studies were at high risk of bias and most results were not of high certainty [S11].
• The global prevalence estimate had very high heterogeneity (I² about 97%) and susceptibility data were synthesised descriptively [S12].
• All Cochrane steroid trials had losses to follow-up of 10% or more and three of four were underpowered [S5].
• Steroid benefit in SCUT comes from subgroup and secondary analyses, not the primary outcome [S1] [S3] [S4].
• The fluoroquinolone versus fortified meta-analysis [S9] and the network meta-analysis [S11] differ in emphasis: one found equivalence overall, the other ranked ciprofloxacin lowest for healing time.
• Not covered by these sources: dosing frequency and duration, preparation of fortified drops, contact lens hygiene advice, surgical management of perforation, and outcomes of therapeutic keratoplasty.

Related articles

• Fungal keratitis
• Acanthamoeba keratitis
• Contact lens complications and safe wear
• Ocular antibiotics and antimicrobial resistance

References

See the numbered source list below. Each marker [S#] links to its PubMed and DOI record.