Aim To determine whether initial empiric treatment of cases with Pseudomonas aeruginosa contact lens-associated keratitis (CLAK) with chloramphenicol had an adverse effect on outcome.
Methods We retrospectively reviewed 139 cases of culture-proven P. aeruginosa CLAK seen between 2007 and 2009. We recorded chloramphenicol use prior to the prescription of a fluoroquinolone, the visual acuity (VA) when the fluoroquinolone was started and at final follow-up, complications and duration of follow-up.
Results 46 patients (33.1%) had used chloramphenicol before they were prescribed a fluoroquinolone. When we compared this group with patients who had initial treatment with a fluoroquinolone, the ulcer size was larger when a fluoroquinolone was started (Mann–Whitney, p=0.018). Although the initial VA was also worse in the chloramphenicol group (p=0.02), and complications more frequent (p=0.016), the final VA in both groups was similar (p=0.29). The chloramphenicol group had a longer median follow-up of 37 days (IQR: 9–310 days) compared with 21 days (IQR: 6–80 days) for the non-chloramphenicol group (p=0.09).
Conclusions Chloramphenicol 0.5% eye drops are available in the UK without prescription. Chloramphenicol had been used in one-third of cases of P. aeruginosa CLAK prior to the use of a broad-spectrum antimicrobial, which was associated with more complications and a longer interval to resolution, but with no adverse effect on final VA.
- Contact lens
Statistics from Altmetric.com
Bacterial keratitis is a potentially sight-threatening condition, and early treatment with an effective antimicrobial results in a better visual outcome.1 ,2 Risk factors for infection include ocular trauma, contact lens wear, surgery and ocular surface disease. Contact lens wear remains a major risk factor for bacterial keratitis especially in the working-age population in developed countries.3 Visual loss is strongly associated with keratitis caused by Gram-negative bacteria rather than by Gram-positive bacteria.4 In a recent case series of microbial keratitis from the UK, Pseudomonas aeruginosa accounted for 49% of isolates,5 and contact lens wear is a particular risk for Pseudomonas infection.6 ,7 It is therefore essential that the initial treatment of contact lens-associated keratitis (CLAK) includes antimicrobial cover for both Gram-negative (in particular P. aeruginosa) and Gram-positive bacteria. Topical fluoroquinolones are widely used as an empiric first-line treatment for suspected microbial keratitis, and randomised controlled trials have shown equivalence with fewer side effects than a combination of fortified antimicrobials (eg, cefuroxime 5% and gentamicin 1.5%).8 However, chloramphenicol, a topical antibiotic that is commonly used to treat conjunctivitis in some countries, has only limited activity against P. aeruginosa and some other Gram-negative bacteria.
Topical chloramphenicol was licensed for over-the-counter sale in the UK in June 2005 by the UK Medicines and Healthcare Products Regulatory Agency.9 ,10 It is clearly labelled that it should not be used for patients who wear contact lenses. Despite this, we were aware that some patients presenting with CLAK had used chloramphenicol prior to specialist review. In this situation, the use of chloramphenicol could allow the infection to progress until a broad-spectrum antibiotic, such as a fluoroquinolone was started. The aim of this study was to determine whether the use of chloramphenicol as primary treatment had an adverse effect on the clinical outcome of P. aeruginosa CLAK.
Patients and methods
From laboratory records, we identified all corneal cultures with Gram-negative bacterial isolates seen at Moorfields Eye Hospital from January 2007 to December 2009. The clinical records were retrieved and patients were included if they had worn contact lenses prior to onset of their keratitis. Laboratory data was recorded for the type of organism isolated, and its sensitivity to chloramphenicol and fluoroquinolones by standard disc diffusion.11 Thirty of the Pseudomonas isolates were stored on sterile beads (Protect beads; TSC, Heywood, Lancashire, UK) at −80°C for determination of fluoroquinolone minimum inhibitory concentration (MIC). The MIC of ciprofloxacin (a fluoroquinolone) was established using E-test (AB Biodisk; Solna, Sweden) plastic strips. Isolates were retrieved from storage and cultured overnight on nutrient non-selective agar (Oxoid, Basingstoke, UK). A bacterial suspension with turbidity equivalent to a 0.5 McFarland standard was prepared and inoculated onto Iso-sensitest susceptibility agar (Oxoid, Basingstoke, UK). An E-test strip was placed on the inoculated plate before incubation for 18–24 h at 36°C. After incubation, the MIC was read from the intersection of the culture with the E-test strip.
For each patient, clinical data was recorded on gender, age at presentation, type of contact lens worn, risks for CLAK (eg, swimming or showering with lenses), and date of onset of symptoms. We also recorded whether the patient had used chloramphenicol drops prior to treatment with a fluoroquinolone, the interval from onset of symptoms to treatment with a fluoroquinolone, visual acuity (VA) at presentation, ulcer size (maximum and minimum diameter) and location (central, paracentral, peripheral), final VA, complications (vascularised corneal scar, keratoplasty), and interval from presentation at this hospital to final follow-up. The final dimensions of any scars were poorly documented (recorded in 13 out of 139 cases) and this was therefore not included in the analysis.
A Mann–Whitney test and Fisher's Exact test were used to assess evidence for a difference in continuous and categorical putative risk factors between patients who did and did not receive chloramphenicol. These tests were also used to see whether there was any difference in outcome between the two patient groups. This study was exploratory or hypothesis generating, so no adjustment for multiple testing was made. Statistical analysis was conducted in Stata 10 for Windows, all tests were two-tailed and a p value of 0.05 was taken as statistically significant.
Over the 3-year study period, there were 149 cases (90 female, 60%) of CLAK with culture-proven Gram-negative organisms (139 Pseudomonas and 10 Serratia species). The great majority (94%) of cases with a Pseudomonas isolate in which the lens type was recorded had worn a contact lens to correct refractive error. Six patients (4.5%) had used a therapeutic contact lens for ocular comorbidity (eg, bullous keratopathy, neurotrophic ulcer, recurrent erosion syndrome). Of the 139 cases with a Pseudomonas isolate, 46 (33.1%, 95% CI 25.3 to 41.5) had received chloramphenicol drops prior to their presentation at this hospital (chloramphenicol group). In this group, there was inevitably a delay in the initiation of a fluoroquinolone antimicrobial from the onset of symptoms (median 2 days) compared with the group that had not received initial treatment with chloramphenicol (median 1 day, p=0.002) (table 1).
The differences in the outcome variables between groups are shown in table 2. At presentation to this hospital, the ulcer size was larger in the chloramphenicol group (Mann–Whitney, p=0.018). This group also had a worse VA than the non-chloramphenicol group (Mann–Whitney, p=0.02), but any difference in VA at final review was not statistically significant (p=0.29). The median interval to final follow-up for the chloramphenicol group was 37 days (IQR: 9–310 days) compared with 21 days (IQR: 6–80 days) for the non-chloramphenicol group (p=0.09). There were six complications in the chloramphenicol group compared with two in the non-chloramphenicol group (Fischer's exact test, p=0.016) (table 2).
According to the laboratory reports, all 139 P. aeruginosa isolates were resistant to chloramphenicol but sensitive to a fluoroquinolone. The median ciprofloxacin MIC of the 30 stored isolates was 0.11 mg/l with a range of 0.047 mg/l to 0.25 mg/l.
In developed countries, the incidence of ulcerative keratitis increased dramatically following the introduction of contact lenses. A retrospective cohort study from Minnesota reported a 435% increase in the annual incidence of ulcerative keratitis between the 1950s and the 1980s to 11.0 per 10 000 residents, by which time 52% of cases were associated with contact lens wear.12 Another retrospective cohort review of medical records of patients living in northern California in 1999 reported a much higher incidence of corneal ulceration in contact lens wearers of 130.4 per 100 000 person-years (95% CI 111.3 to 151.7) although, surprisingly, no Pseudomonas isolates were recorded.13 A retrospective population-based survey in Australia in 2004 reported a penetrance of contact lenses in the 15–64 year age group of 5.01% (CI 4.78 to 5.24) with an annualised incidence of microbial keratitis in this population of 4.2 (CI 3.4 to 5.5) cases per 10 000 population.14 Thus, contact lens wear is now the major risk for microbial keratitis, and Pseudomonas is the most common pathogen.
To the best of our knowledge, this is the first study to look at the effect of initial treatment with chloramphenicol on the outcome of CLAK caused by P. aeruginosa. Chloramphenicol is not effective against P. aeruginosa, and delay in treatment of keratitis with a broad-spectrum antibiotic is known to be associated with more severe disease.2 It was a concern that treatment of these cases with topical chloramphenicol could adversely affect outcome. Our results show that patients who were initially treated with chloramphenicol had more complications and required a longer follow-up than a group that had primary treatment with a fluoroquinolone, but despite this, both groups had a similar final VA.
Chloramphenicol is active against most Gram-positive bacteria, with resistance rates as low as 10%, but for Gram-negative organisms the resistance rates are much higher.5 ,15 ,16 The proportion of Gram-negative isolates from cases of bacterial keratitis in the UK with in vitro resistance to chloramphenicol increased from 46.3% between 1999 and 2004 to 75% between 2004 and 2009.5 ,15 ,16 However, it is important to distinguish the antimicrobial sensitivity based on systemic breakpoint concentrations reported in these studies, from sensitivity based on the higher corneal concentration achieved when an antimicrobial is applied topically.17 This difference between systemic and topical breakpoint concentrations applies in particular to some Gram-negative bacteria, such as Serratia species that may be susceptible to chloramphenicol when it is applied topically to the cornea. By contrast, chloramphenicol has no activity against P. aeruginosa due to poor permeability into the bacterial cell and an active efflux pump,18 and chloramphenicol would not be expected to have any therapeutic activity even at the high concentrations achieved when it is applied topically to the cornea. As expected, all the P. aeruginosa isolates that we tested were susceptible to ciprofloxacin but resistant to chloramphenicol. The ciprofloxacin MICs of the 30 random isolates were comparable to those described for other collections of P. aeruginosa from bacterial keratitis,19 and were all below the systemic breakpoint for this agent.11
Chloramphenicol was made available in the UK for over-the-counter sale without prescription for the treatment of conjunctivitis in 2005.9 When used for bacterial conjunctivitis, it has a small effect on the time to clinical resolution of symptoms.20 After 2005, there was a large increase in the use of topical chloramphenicol, with an estimated 3.4 million prescriptions and over-the-counter sales in England in 2007,21 with the potential for increased rates of resistance. Despite clear labelling warning against the use of chloramphenicol if a contact lens had been worn, approximately one-third of the cases of P. aeruginosa CLAK in our series were initially treated with chloramphenicol, which inevitably resulted in a delay before the start of appropriate treatment with a broad-spectrum antibiotic (eg, a fluoroquinolone). Because P. aeruginosa keratitis can progress very rapidly, any delay in appropriate treatment could have an adverse effect on outcome. We found that although the mean ulcer size was larger in the group initially treated with chloramphenicol by the time a fluoroquinolone was started, any associated delay in starting a fluoroquinolone did not, on average, affect the final VA. However, when we used the duration of hospital follow-up as an index of disease severity and response to treatment, patients initially treated with chloramphenicol had a worse outcome with almost double the median follow-up prior to discharge from the hospital service. Complications (vascularised corneal scar, keratoplasty) were also more frequent in the group initially treated with chloramphenicol. Although we presume that these effects are the result of a delay in effective treatment, it is possible that prior exposure to chloramphenicol has a sublethal effect on P. aeruginosa that either alters the susceptibility to other antimicrobials or the host immune response.
This study has the limitation of being retrospective with possible recording bias. For example, it is not known whether some mild cases of P. aeruginosa keratitis resolved following treatment with chloramphenicol. Also, variables such as the VA prior to presentation with keratitis and the size of the ulcer at the start of any treatment could not be included in the analysis. The best-corrected VA was also not evaluated at a uniform time interval after the resolution of infection when the healing process could be considered to be complete. However, we have shown that in England, chloramphenicol is used inappropriately in a significant proportion of cases of CLAK, when its use delays recovery. Although it is unlikely that further antibiotic classes will be made available in the UK without prescription, it is our view that any topical antimicrobial marketed for ophthalmic use without prescription should be effective against both Gram-positive and Gram-negative bacteria.
This research is supported by the National Institute for Health Research (NIHR) Biomedical Research Centre based at Moorfields Eye Hospital NHS Foundation Trust and UCL Institute of Ophthalmology.
Contributors ST, SK and TN devised the overall concept and design of the manuscript. RB extracted the data from the clinical records. RB and ST wrote the first drafts. CB and JS provided interpretation of the data. All authors read and approved the final document.
Disclaimer The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.
Competing interests None.
Ethics approval Moorfields Eye Hospital local Research Ethics Committee.
Provenance and review Not commissioned; externally peer reviewed
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.