Over the past four decades, methicillin-resistant Staphylococcus aureus (MRSA) has spread throughout the world and become highly endemic in many geographic areas. This pathogen causes severe morbidity and mortality in hospitals worldwide 123. In France, 30 to 40% of S. aureus strains are methicillin-resistant and the median incidence of MRSA in clinical specimens is about 0.8 per 1000 patient-days in acute care facilities and 3.42 in intensive care units (ICUs) 4. French intensive care experts had recommended that barrier precautions should be implemented for patients colonised or infected with MRSA, but there is no consensus about an MRSA screening programme and MRSA nasal decolonisation. Systematic MRSA screening on admission and preventive isolation have been shown to be cost-effective 567 but the usefulness of mupirocin-based nasal decolonisation remains a matter of debate. Indeed, mupirocin has emerged as the topical antibacterial agent of choice for the elimination of S. aureus nasal carriage. In hospitals in which MRSA is epidemic, the intra-nasal administration of mupirocin to both patients and personnel colonised with MRSA is considered to be appropriate 8. Furthermore, mupirocin has been used successfully to prevent staphylococcal infections in surgical and haemodialysis patients 9. Although the available literature does not support routine use of intranasal mupirocin to prevent subsequent infections, there may be a role in some selected cases, such as those involving critically ill patients 9. Thus, it was reported that intranasal mupirocin could decrease the occurrence of S. aureus pneumonia in ICU patients 10. In 1994, an MRSA control programme was implemented in all high-risk units of our hospital, including adult ICUs. Thus, all patients were screened for MRSA on admission and during hospitalisation; MRSA-positive patients were kept in isolation and positive patients were prescribed nasal mupirocin ointment. In June 2001, clinicians from the medical ICU decided to stop using mupirocin for MRSA nasal decolonisation. We conducted a retrospective study to determine whether intranasal mupirocin prophylaxis is useful in preventing ICU-acquired infections with MRSA.

During the first study period with mupirocin use, 912 patients were admitted to the medical ICU, 775 were screened for MRSA (84.5%), 38 were MRSA carriers and 9 were infected with MRSA. During the second period without mupirocin use, 987 patients were admitted, 819 were screened (83.0%), 49 were MRSA carriers and 17 were infected with MRSA (Table 1). The number of endogenous acquired infections was significantly higher during the second period than during the first (11 versus 1; P = 0.006, relative risk = 8.53 (CI95% [1.15–63.21])) although there was no significant difference in the total number of patients infected with MRSA between the two periods. The 12 endogenous acquired infections detected are described in Table 2. In the second period, the delay between the detection of MRSA carriage and the date of the first positive clinical sample was sufficient to implement nasal mupirocin ointment for 7 of the 11 infected patients. Three patients died during the 2 weeks after the MRSA bacteraemia or pneumonia.

We next evaluated the efficiency of nasal mupirocin for MRSA eradication. The rate of nasal MRSA decolonisation among the patients hospitalised for more than 14 days after the discovery of MRSA carriage (delay necessary to evaluate the efficacy of mupirocin) reached 86.4% in period 1 and just 46.4% in period 2 (P = 0.002). Three of the 19 patients who were successfully decolonised during the first period received concomitant treatment with antimicrobials effective against MRSA (namely vancomycin).

The total amount of antibiotics used remained stable during the study period (1,700 DDD per 1,000 hospitalisation days) and the use of glycopeptides did not significantly vary (104 DDD per 1,000 hospitalisation days in the first period and 91 in the second period).

Our observational study suggests that nasal mupirocin can effectively prevent the occurrence of endogenous acquired MRSA infections in ICUs. The increase in endogenous ICU-acquired MRSA infections in the second period occurred although different indicators such as mean length of stay, mean gravity scores of fatality rates did not significantly vary between the periods (Table 1). In 7 of the 11 MRSA infections observed in the period 2, the delay between the two samples was sufficient to implement mupirocin treatment; that is, at least 5 days. Our findings stand in contrast to double-blind randomised, placebo-controlled trials that included patients hospitalised in different type of unit 17. However, in this study, the number of infections was three in the mupirocin group and seven in the placebo group. It is likely that the enrolment of more than 98 patients would show a significant effect of mupirocin in preventing subsequent MRSA infection. Moreover, the only randomised study applied to ICU patients also concluded that mupirocin was effective in reducing the occurrence of S. aureus pneumonia 10.

Some limitations of our study have to be addressed. First, we compared historical groups, whereas randomised, double-blind placebo controlled trials are more powerful. However, MRSA infection remains a rare event in our medical ICU and a very large study period would be needed to show that a programme including mupirocin use is beneficial. We simply reported the evolution of the infections with MRSA between two periods during which only one element had changed: the use of mupirocin for nasal MRSA decolonisation.

Second, we have not evaluated mupirocin resistance in this study. Low-level mupirocin resistance has been identified as a risk factor for a failure to eradicate MRSA 18, and numerous authors have reported the emergence of mupirocin resistance in settings where mupirocin is commonly used 19. Concerns over the development of resistance have dissuaded many hospitals from using mupirocin in this manner, and recent recommendations from the guidelines of the Society for Healthcare Epidemiology of America stated that 'widespread use, prolonged use or both of decolonization therapy should be avoided' 20. Others studies have reported that relapse of MRSA carriage was not associated with the development of resistance to mupirocin 2122.

The body of the literature currently does not support routine intranasal mupirocin prophylaxis to all inpatients to decrease the rate of clinical infection. However, there might be as yet unidentified patient populations that could benefit. One can speculate that a significant effect should be seen for patients at high risk of infection, such as patients admitted to ICUs 23. Indeed, our data showed that the infectious risk in our ICUs is high; during period 2 (mupirocin-free), 25.6% of carriers became infected (Table 1). A recent meta-analysis reported a significant increase in mortality associated with MRSA infection (odds ratio 1.93) and most studies depicted a crude mortality rate between 20% and 40% 3.

Our results concerning MRSA nasal eradication are in agreement with those observed by several authors in different settings, Kalmeijer and colleagues reported that MRSA was eradicated from 83.5% of patients admitted for orthopaedic surgery 24, and Mody and colleagues also showed that treatment with mupirocin had decolonised 93% of residents in long-term care facilities 21. However, they differed from those obtained in other randomised trials. This difference was probably due to the status of the patients included, regardless of the type of setting.

Our results suggest that the use of mupirocin to prevent MRSA infections in ICUs has to be evaluated. Further double-blind, randomised, placebo-controlled clinical trials are needed to demonstrate its cost-effectiveness and its impact on mupirocin resistance.

• MRSA causes severe morbidity and mortality in ICUs worldwide.

• Mupirocin is regarded as the topical antibacterial agent of choice for the elimination of S. aureus nasal carriage.

• The body of the literature does not currently support routine intranasal mupirocin prophylaxis.

• Our observational study suggests that nasal mupirocin can effectively prevent the occurrence of endogenous acquired MRSA infections in ICUs.

• Further double-blind, randomised, placebo-controlled clinical trials are needed to confirm our findings in ICUs.

DDD = defined daily doses; ICU = intensive care unit; MRSA = methicillin-resistant Staphylococcus aureus.

The author(s) declare that they have no competing interests.

AM collected the data and performed the statistical analysis. AP recorded the clinical information. XB and DT composed the writing committee. EB and GC supervised AP for the record of clinical information. All authors read and approved the final manuscript.