Panton-Valentine leukocidin secreted by Staphylococcus aureus is known to cause severe skin, soft tissue and lung infections. However, until recently it has not been described as causing life-threatening musculoskeletal infection. We present four patients suffering from osteomyelitis, septic arthritis, widespread intravascular thrombosis and overwhelming sepsis from proven Panton-Valentine leukocidin-secreting Staphylococcus aureus. Aggressive, early and repeated surgical intervention is required in the treatment of these patients.

The Panton-Valentine leukocidin toxin not only destroys host neutrophils, immunocompromising the patient, but also increases the risk of intravascular coagulopathy. This combination leads to widespread involvement of bone with glutinous pus which is difficult to drain, and makes the delivery of antibiotics and eradication of infection very difficult without surgical intervention.

A delay in the diagnosis of paediatric acute and subacute haematogenous osteomyelitis can lead to potentially devastating morbidity. There are no definitive guidelines for diagnosis, and recommendations in the literature are generally based on expert opinions, case series and cohort studies.

All articles in the English literature on paediatric osteomyelitis were searched using MEDLINE, CINAHL, EMBASE, Google Scholar, the Cochrane Library and reference lists. A total of 1854 papers were identified, 132 of which were examined in detail. All aspects of osteomyelitis were investigated in order to formulate recommendations.

On admission 40% of children are afebrile. The tibia and femur are the most commonly affected long bones. Clinical examination, blood and radiological tests are only reliable for diagnosis in combination. Staphylococcus aureus is the most common organism detected, but isolation of Kingella kingae is increasing. Antibiotic treatment is usually sufficient to eradicate the infection, with a short course intravenously and early conversion to oral treatment. Surgery is indicated only in specific situations.

Most studies were retrospective and there is a need for large, multicentre, randomised, controlled trials to define protocols for diagnosis and treatment. Meanwhile, evidence-based algorithms are suggested for accurate and early diagnosis and effective treatment.