Comment on guidelines for the prophylaxis and treatment of methicillin-resistant Staphylococcus aureus (MRSA) infections in the UK.

Sir, In the Guidelines for prophylaxis and treatment of methicillinresistant Staphylococcus aureus (MRSA) infections in the UK, the Joint Working Party states for empirical treatment that ‘The prevalence level at which flucloxacillin or other penicillinasestable penicillins, in a patient group, becomes no longer the drug of choice is debatable, but 10% resistance has been used as a guide for avoiding the use of empirical gentamicin in Gram-negative infection and we would recommend the same threshold is used when contemplating treatment of staphylococcal infections with isoxazolylpenicillins or cephalosporins. This threshold may be adjusted depending on the apparent severity of infection.’ This threshold (of 10%) ignores the evidence that b-lactam drugs are more effective than glycopeptides for infections caused by methicillin-susceptible S. aureus (MSSA) in the eradication of infection, prevention of recurrence and prevention of death. Moreover, it was impossible to show an advantage of glycopeptide appropriate empirical treatment over inappropriate treatment. The 10% threshold means that in order to offer appropriate treatment (vancomycin) to 10 patients with severe infections caused by MRSA, 90 people with severe infections caused by MSSA will be given less effective treatment (vancomycin and not cloxacillin). To try and take this factor into account, we have looked at 429 patients with S. aureus bacteraemia included in the Beilinson Bacteremia Database. The fatality rate in patients given inappropriate empirical antibiotic treatment was 38% (69 of 183), versus 24% (57 of 246) in patients given appropriate treatment, P = 0.007. The multivariable-adjusted odds ratio was 1.8 [95% confidence interval (CI) 1.2–2.7]. In patients with MSSA bacteraemia, the fatality ratewas 28% (47 of 166) in patients treated with vancomycin versus 8% (4 of 48) in patients given cloxacillin,P = 0.004, univariate odds ratio of 4.3 (95% CI 1.5–12.8). The multivariable-adjusted odds ratio was larger than the univariate ones. If indeed the advantage of cloxacillin over vancomycin as empirical treatment is at least as large as that of appropriate versus inappropriate treatment, the threshold of the baseline susceptibility to methicillin for preferring vancomycin over cloxacillin in a patient suspected of harbouring a severe infection with S. aureus should be 50% rather than 10%.