Clinical Review

Acute haematogenous osteomyelitis in children

BMJ 2014; 348 doi: (Published 20 January 2014) Cite this as: BMJ 2014;348:g66

This article has a correction. Please see:

  1. Andrea Yeo, specialty registrar,
  2. Manoj Ramachandran, consultant paediatric orthopaedic surgeon
  1. 1Centre for Orthopaedics, The Royal London and Barts and The London Children’s Hospitals, Barts Health NHS Trust, London E1 1BB
  1. Correspondence to: M Ramachandran manoj.ramachandran{at}
  • Accepted 23 December 2013

Acute osteomyelitis is an uncommon but important disease that affects previously healthy children. A high index of suspicion is required as early treatment is essential for a good outcome. In the past decade, rapid changes in the epidemiology of the condition, in particular of infections as a result of meticillin resistant Staphylococcus aureus (MRSA), and advances in diagnostics have highlighted a need to change practice based on current evidence. We review the pertinent aspects of acute osteomyelitis, highlighting the pitfalls in diagnosis and providing a framework for management.

Summary points

  • Acute osteomyelitis in children is an infection of the bone of less than two weeks’ duration, which typically spreads haematogenously

  • Meticillin resistant Staphylococcus aureus osteomyelitis in particular is on the increase

  • Children may present insidiously, hence careful history and clinical examination with a high index of suspicion are required

  • A delay in diagnosis may cause growth disturbance, deformity, or even death

  • Treatment includes intravenous antibiotics administered promptly

  • A multidisciplinary approach involving primary, secondary, and tertiary care teams is essential to ensuring a good outcome

What is acute osteomyelitis?

Osteomyelitis is the inflammation of bone caused by pyogenic organisms. In the acute setting, the duration of symptoms is less than two weeks. The major sources of infection are haematogenous spread, tracking from adjacent foci of infection, and direct inoculation from trauma or surgery.1 Haematogenous spread is the most common source of infection in children, typically affecting the long bones. The infection seeds in the metaphysis, where blood flow is rich but sluggish.2 The femur and tibia are most commonly affected (27% and 26%, respectively, fig 1).3

Sources and selection criteria

We searched Medline and Google Scholar using the terms “p(a)ediatric”, “acute osteomyelitis”, “bone infection”, “septic arthritis”, “antibiotics” and “imaging”. Wherever possible we used evidence from randomised controlled trials, systematic reviews (including Cochrane reviews), and expert review articles published in the past five years, to provide an up to date summary. We also consulted guidelines from the British Orthopaedic Association and the British Society for Children’s Orthopaedic Surgery (2013).


Fig 1 Skeletal distribution of osteomyelitis. Incidence rates are shown at commonly affected sites, as calculated in a recent systematic review.3 The femur and tibia are the most commonly affected bones

In long bones where the metaphysis is intracapsular (the shoulder, ankle, hip, and elbow, in decreasing order of incidence),4 the infective foci may extend into the joint space, resulting in concurrent septic arthritis.2 Similarly, in children under 18 months of age, anatomical transphyseal vessels facilitate translocation of bacteria from the metaphysis to infect the epiphysis and adjacent joint, increasing the risk of concurrent septic arthritis.2

Who gets osteomyelitis and why is it important?

The incidence of osteomyelitis varies between 1 and 13 per 100 0003 5 and accounts for around 1% of all hospital admissions in children.6 Boys are nearly twice as commonly affected as girls,7 with 50% of cases occurring in those under 5 years old,8 peaking in children under 1 year.7 In most cases the lesion is solitary but can be multifocal (7% of children and 22% of neonates).9 10

Recent studies have reported an increasing incidence and, more worryingly, worsening severity of musculoskeletal infection in children.10 11 12 A recent large case-control study from the United States reported a 2.8-fold increase in the incidence of osteomyelitis over the past 20 years.11 Over this same period, the incidence of septic arthritis has remained unchanged, occurring half as frequently.11 There has been a concomitant increase in MRSA as the causative organism in most cases of complicated osteomyelitis,10 12 with one group reporting a 30% incidence.11

Delays in treatment may lead to complications such as concurrent septic arthritis, subperiosteal abscess, pyomyositis, deep vein thrombosis, permanent impairment (longitudinal growth arrest with subsequent discrepancy in limb length, angular deformity, chronic infection), septicaemia, multiorgan failure, and death.2 Osteomyelitis and concurrent septic arthritis have been reported at a rate of 3% to 33%.11 13 14 In a cohort of 212 children with osteomyelitis, up to 8% presented with deep musculoskeletal infections of subperiosteal abscess, pyomyositis (2%), or all four (1%)—osteomyelitis, concurrent septic arthritis, pyomyositis, and subperiosteal abscess.11 As can be expected, the more extensive the infection, the greater the clinical manifestation and severity of the illness, with increased requirements for intensive care support, longer hospital stays, and higher complication rates, including deep vein thrombosis (33%).11

The profile of the condition has changed noticeably with vaccinations and the use of antimicrobial treatment; mortality rates of about 50% in the preantibiotics era have fallen to less than 1%, and it is usually from overwhelming septicaemia with involvement of multiple organs.2 5 10 15 With prompt, effective treatment, prognosis is generally good, with cure rates of more than 95%.7 Management goals have therefore changed from survival to limb preservation to maintenance of normal limb development and function.15

What are the risk factors for osteomyelitis?

Up to half of cases have no risk factors at all, and minor trauma, such as an innocent fall on to the knee, may be the presenting history in up to 30% of cases.8 However, a particular subset of children are more susceptible to developing acute haematogenous osteomyelitis and hence warrant greater vigilance by clinicians.

Immunocompromised children (for example, those with diabetes, malignancy, HIV, requiring steroid therapy, or who are malnourished) are generally more susceptible to infections. Likewise, premature infants are predisposed owing to their immature immune systems and exposure to repeated invasive procedures (for example, umbilical artery catheterisation), providing a portal for haematogenous spread. Such children often show systemic upset and pronounced leucocytosis.16 Children with chronic illness requiring frequent venesection (for example, haemodialysis catheters) are also at increased risk.

Sickle cell disease is a haemoglobinopathy with a prevalence of 1 in 2000 in the United Kingdom.17 Microvascular obstruction by the sickled red blood cells result in tissue ischaemia and infarction. This, in combination with impaired immune function and splenic dysfunction, leads to an increased risk of osteomyelitis, with Salmonella the most common causative organism.18 19 20 Osteomyelitis is, however, still rare, being 50 times less common than a vaso-occlusive crisis.18 However, it can be difficult to differentiate a vaso-occlusive crisis from osteomyelitis as both often present with fever and a painful, swollen limb with restricted joint motion.20 No definitive feature exists in the clinical history, physical examination, laboratory test, or radiological investigation that can reliably distinguish the two; hence most patients are empirically treated for both. However, one of the largest case-control series comparing children with sickle cell disease who had osteomyelitis with those who had vaso-occlusive crisis identified swelling at a single site (odds ratio 8.4), prolonged fever (probability increased 80% for each day of fever before presentation), or pain (probability increased 20% for each day a child had pain before presentation) as being predictive of osteomyelitis.20

How is osteomyelitis diagnosed?

The classic florid picture of an unwell, febrile child with a high white cell count is increasingly uncommon.10 This is speculated to relate to improved standards of living and hygiene; as such, fewer virulent pathogens lead to a more subacute presentation.21 In subacute osteomyelitis, the balance between the microbe and host is such that the destructive and repair processes do not take place as rapidly as in acute infection.2

The onset of osteomyelitis can be insidious, the clinical presentation variable, and the physical findings non-specific.22 No single test can confirm or rule out acute osteomyelitis. A combination of careful history and examination, accompanied by a high index of clinical suspicion, and followed by laboratory and imaging studies are key parts of the clinical investigations. It should always be borne in mind that childhood malignancies (for example, leukaemia, lymphoma, and neuroblastoma) can present with non-specific musculoskeletal symptoms.15 Box 1 lists other differential diagnoses.

Box 1: Differential diagnoses for acute osteomyelitis

  • Vaso-occlusive disease—for example, bone infarct secondary to sickle cell disease

  • Septic emboli

  • Chronic recurrent multiple osteomyelitis

  • Septic arthritis

  • Stress fracture

  • Osteoid osteoma

  • Acute lymphoblastic leukaemia

  • Eosinophilic granuloma

  • Metastatic neuroblastoma

  • Ewing’s sarcoma

  • Osteosarcoma

Clinical features and severity may vary greatly depending on the site of infection, age of the child, and the species of the responsible pathogen.1 23 In a recent systematic review involving over 12 000 patients with acute and subacute osteomyelitis, the commonest presenting features were pain (81%), swelling and erythema (70%), fever (62%), reduced joint movement or pseudoparalysis (50%), and reduced weight bearing or a limp (49%).3 Typically, a young child may present with little more than irritability or refusal to use a limb. Neonates may present with a much more complex clinical picture as their immune systems are less well developed and their inflammatory response is not well localised.24 Fever in this group is not a major symptom, and the infection, particularly in premature infants, may be multifocal.15 Multifocal disease is rare but presents as multiple acute foci of infection, affecting mainly the femur and tibia, or the tibia and humerus simultaneously.25 Box 2 summarises the clinical features to look for in a child with suspected osteomyelitis.

Box 2: Clinical features to look for in a child with suspected osteomyelitis

  • Listen to the child and observe his or her interaction with the parents (there could be other causes for pain or a limp26)

  • Ask about the site of pain (beware referred pain—children with hip disease often complain of knee pain instead)

  • Ask about the duration of symptoms (fever for >7 days or localised symptoms for >10 days are suggestive of a complicated course)

  • Ask about prodromal symptoms (for example, recent fever, cough, cold, diarrhoea)

  • Ask about changes in behaviour (in a young child, subtle features such as irritability and being quieter than usual may be the only presenting sign)

  • Ask about recent trauma (30% of patients have this as a feature)

  • Ask about chronic illnesses (especially sickle cell disease or diabetes)

  • Take the child’s temperature (in 40% of cases, children may be afebrile)

  • Observe the movement of all four limbs (pseudoparalysis may be the only feature in an infant)

  • Observe the resting position of the child’s limb—children may hold their hip in a flexed and externally rotated position, or the knee in a flexed position (fig 2)


Fig 2 A normally healthy toddler with suspected osteomyelitis of the distal femur who presented with a temperature of 38°C, reluctance to weight bear, and, according to her mother, being “more clingy and quieter than usual.” Her inflammatory markers were unremarkable (white cell count 8×109/L, C reactive protein 17 nmol/L). These pictures show subtle fullness of the left knee, which is kept in a flexed position while at rest. Clinically, the patient had a restricted range of movement of 30° to 90°. Subsequent ultrasonography and bone scan were positive for osteomyelitis in the distal femur

  • If the child is of walking age, look for a limp or reluctance to weight bear

  • Look for swelling and erythema, which can be subtle (the femur and tibia are the most commonly affected sites)

  • Look for open wounds

  • Look for localised tenderness and restricted range of joint motion

  • Examine the rest of the child, as osteomyelitis can be multifocal or present in unusual sites (see fig 1)

  • Most importantly, have a high index of suspicion, and, if in doubt, refer acutely to paediatrics, paediatric infectious disease, or paediatric orthopaedics

What investigations can help confirm osteomyelitis?

A systematic review found that only 36% of children will have a raised white cell count on presentation, 91% an increased erythrocyte sedimentation rate, and 81% an increased C reactive protein value.3 The sensitivity is highest when both the erythrocyte sedimentation rate and C reactive protein are increased (98%).27 C reactive protein values >100 mg/L are particularly significant for concomitant septic arthritis and are also the best predictor of a complicated course and the need for prolonged intravenous antibiotics.28 C reactive protein has a short half life (19 hours) and hence is also useful for monitoring response to treatment.29

The British Orthopaedic Association and the British Society for Children’s Orthopaedic Surgery recommend that specimens should be taken, when possible, before starting antibiotic treatment. This recommendation should not, however, lead to a prolonged delay in starting antibiotics, especially in very sick children.15 Blood cultures may be positive in only 50% of cases, but should still be taken before antibiotics are started as it may be the only positive source of identification of the pathogen.15 Bone or joint aspirates can have a higher yield (70%).15

Specimens can be obtained through interventional radiology (for example, guided by ultrasound or computed tomography) or surgery; and should be sent for urgent Gram stain and culture. Culture of joint aspirates in commercially available blood culture bottles helps to improve the yield of organisms that are difficult to culture.30 Additionally, specimens should always be sent for histology as several childhood malignancies may present with similar clinical features.15

What imaging studies are useful in suspected osteomyelitis?

Plain radiographs of the affected part often do not show any abnormality in the early phase of infection. However, these are useful to exclude other conditions such as fractures or malignancy.24 Acute skeletal changes (periosteal elevation, bone destruction, fig 3) are not visible before 5-10 days, but subtle features of soft tissue swelling may be seen (fig 4).7


Fig 3 Anteroposterior and lateral radiographs of a toddler’s forearm (top) with clear features of extensive osteomyelitis. The whole length of the radius is grossly expanded, with areas of lysis (arrows). Complete resolution of the infection in the radius after treatment (bottom), which now looks radiologically normal (asterisk)


Fig 4 Anteroposterior radiograph of the chest in a severely unwell toddler with concurrent osteomyelitis and septic arthritis of the shoulder. The radiograph shows the likely septic focus in the right proximal humerus, with more subtle features of mottling of the proximal humerus (arrows), soft tissue swelling, and fullness (asterisk) of the shoulder joint

Magnetic resonance imaging is the preferred modality for initial clinical investigations,31 with high sensitivity (82-100%) and specificity (75-99%).7 This type of imaging is helpful in identifying the location and extent of disease and offers more detailed evaluation of the adjacent structures in suspected complicated cases, such as pyomyositis, joint effusion, and subperiosteal abscesses.23 It can also help in assessing difficult sites of infection (for example, pelvic osteomyelitis) and planning surgical intervention.15 Magnetic resonance imaging is safe as it carries no risk from radiation; however, young children often require sedation or general anaesthesia before the procedure can be carried out. It also has the disadvantages of higher costs, prolonged imaging time, and difficulties with access, especially out of hours. Whole body magnetic resonance imaging is increasingly used to evaluate multifocal osteomyelitis or in cases where localisation of symptoms is doubtful.24

Computed tomography provides excellent multiplanar image reconstructions, allowing delineation of subtle osseous changes; however, its role in the assessment of acute haematogenous osteomyelitis is limited owing to poor soft tissue contrast and high exposure to ionising radiation (compared with magnetic resonance imaging). It is more useful in chronic osteomyelitis, where computed tomography can show sclerotic changes, abnormal thickening of bone, extent of disease, and chronic draining sinuses.32 Computed tomography can be used for evaluation of complications of osteomyelitis if magnetic resonance imaging is not available or is contraindicated.24

Bone scans may also be useful in cases of poorly localised infection—for example, in younger children who cannot verbalise their site of pain, or in multifocal disease.15 The overall sensitivity and specificity of bone scans are 73-100% and 73-79%, respectively.24 In neonates, however, the sensitivity decreases (32-87%).24 It is not clear why the sensitivity decreases in neonates. However, this may relate to false negative scans where “cold” rather than “hot” spots are seen in severe infection.33 34 New bone formation and blood flow are important requirements for the uptake of radionuclide on to bone; hence any mechanism that interferes with hyperaemia or osteoblastic activity (for example, thrombosis and tamponade of the affected bone) is likely to produce an image of decreased activity.34

Ultrasound cannot evaluate bone marrow (ultrasonic waves cannot penetrate bone cortex) and hence its use in the diagnosis of osteomyelitis is limited. Nevertheless, ultrasonography is a useful investigation for visualising subperiosteal collections and joint effusions, and it can aid in the aspiration of soft tissue fluid and joint.24 The key role of ultrasonography is to support the suspected clinical diagnosis.15 It is cheap, safe, non-invasive, and portable. Hence it can be utilised in cases where access to other modalities is not readily available or is contraindicated, such as a child in intensive care where transfer to a magnetic resonance imaging scanner is unsafe.

Which organisms cause acute osteomyelitis?

Staphylococcus aureus is the most common pathogen in acute osteomyelitis, being cultured in 70-90% of culture positive cases,1 followed by streptococcal (S pyogenes and S pneumoniae) and Gram negative organisms. Salmonella is an important pathogen in children with sickle cell disease, especially the non-typical serotypes (S typhimurium, S enteriditis, S choleraesius).6 It has been suggested that tiny infarctions in the gastrointestinal tract lead to salmonella (and other enteric Gram negative) bacteraemia and ultimately to infection.35 Children with sickle cell disease and suspected osteomyelitis may require a completely different antibiotic regimen, and local microbiological advice should be sought. A recent Cochrane review found no randomised trials on antibiotic approaches for osteomyelitis in people with sickle cell disease.36

Over the past few decades the pattern of causative organisms has been changing, with more resistant strains emerging. Haemophilus influenzae, previously the most common Gram negative organism in paediatric osteomyelitis, is now rare as a result of the vaccination programmes of the early 1990s.23 Conversely, the incidence of community acquired MRSA is increasing in many parts of the world.1 10 11 12 29 37 The MRSA epidemic has not only altered antibiotic treatment regimens, but also affected the severity of disease.10 11 12 37 MRSA is a causative agent in 9-30% of children with osteomyelitis,10 29 and cases have been documented of PVL-MRSA (Panton-Valentine Leukocidin MRSA), an extremely virulent strain.

MRSA osteomyelitis can cause a more aggressive and complicated course, with higher inflammatory markers and fever, prolonged hospital stay, and an increased likelihood for repeated surgical debridement compared with other pathogens.10 11 37 It is also associated with many complications, including multiorgan failure, deep vein thrombosis (10%), septic pulmonary emboli, multifocal infection, subperiosteal abscesses, fractures (20%), and progression to chronic osteomyelitis.11 37 38

Kingella kingae is a common pathogen that colonises the respiratory tract and is transmitted from child to child through close contact. K kingae osteomyelitis is on the increase, with more than 95% of cases occurring in children under 3 years of age.3 This pathogen tends to present with more benign features, with only 15% of children febrile at presentation and 39% having normal inflammatory markers (C reactive protein and white cell count).39 It is, however, difficult to culture and is mainly identified by molecular techniques of polymerase chain reaction, which may not be readily available.23 39

A causative pathogen is not identified in up to 55% of cases.23 When less common organisms are suspected, such as in a child with an altered immune status who is more susceptible to different pathogens, these must be communicated to the laboratory and discussed with microbiology colleagues, as some pathogens are difficult to culture and may require specific media, growth conditions, or prolonged culture time.23 A systematic review found that culture negative osteomyelitis was successfully treated in the same way as confirmed staphylococcus disease.3

How is osteomyelitis treated?

The care of children with acute osteomyelitis is multidisciplinary, requiring communication and coordination among general practitioners, emergency departments, paediatric infectious diseases, orthopaedics, microbiology, radiology, nursing, and community teams to ensure early diagnosis and effective treatment. In a well designed case-control study, multidisciplinary management has been shown to produce more efficient clinical investigations, higher rates of identification of causative organisms, and fewer changes in antibiotics, with lower admission rates and shorter hospital stays.40

The basic goal is to deliver antibiotics in appropriate dosages according to the sensitivities of the causative organism. Antibiotic choice should be guided by culture results and local microbiology advice wherever possible to facilitate a more focused therapeutic regimen. However, to avoid treatment delays, empirical treatment should be selected to cover the most likely pathogens, which is determined primarily by local prevalence of infectious agents and resistance levels, the age of the child, and early microbiological results such as that for Gram’s stain.1 15

Little evidence, however, exists on the choice of initial antibiotic or the optimal duration of intravenous and oral treatment for acute osteomyelitis in children, as evidenced in a recent systematic review.1 The British Orthopaedic Association and the British Society for Children’s Orthopaedic Surgery recommend flucloxacillin or a cephalosporin as first line treatment owing to the dominance of S aureus15; although in the United States and Finland, clindamycin is more commonly used.41 Benzylpenicillin or a cephalosporin should be added in children not immunised against H influenzae.15

Some authors have suggested empirical cover against MRSA, especially if more than 10% of S aureus isolates are meticillin resistant1 or if the child has recognised risk factors, such as previous admission to hospital or colonisation, or is from an ethnic background in which prevalence is high—for example, Polynesian or Aboriginal. This has not been widely accepted mainly due to fear of inducing selective antimicrobial resistance.3 Broad spectrum cover is especially important in neonates, children with sickle cell disease, and immunocompromised children, who can be infected with a wider variety of organisms.

Acute osteomyelitis has traditionally been treated with 4-6 weeks of antibiotics. However, several small cohort studies have supported shorter treatment regimens, which obviate the complications associated with intravenous lines (for example, infection and allergic sensitisation)42 43 and allow the child to be managed at home. The transition to oral step-down treatment is guided by improvement in clinical (resolution of fever and pain with restoration of function) and haematological (normalising C reactive protein levels) variables.29

The only randomised trial to deal with the duration of antibiotic treatment showed that oral antibiotics for 20 days was as effective as a 30 day course in patients who had already received intravenous antibiotics for four days, with a good clinical response.44 This has been backed up by a recent systematic review (grade 2B recommendation).1 Evidence on neonates (<3 months) is, however, insufficient to alter the current recommendation of parenteral antibiotics for four weeks owing to concerns about absorption and efficacy of oral antibiotics in this population.1

Routine exploration of acute haematogenous osteomyelitis is no longer recommended by the British Orthopaedic Association and the British Society for Children’s Orthopaedic Surgery.15 A recent systematic review showed that appropriate antibiotic therapy is sufficient and surgery does not confer any additional benefit.3 Surgery is now reserved for circumstances where medical treatment fails or when a major abscess has collected.15 The decision to drain an abscess should be made on clinical grounds (temperature, pain, reduced use of limbs, and increasing C reactive protein values), specifically the response to antibiotics.3 Drainage is required in approximately 20% of pelvis abscesses and 6% of those in long bones.3

What is the prognosis for osteomyelitis?

Acute haematogenous osteomyelitis in children is invariably curable. Recognition of the often subtle features with the help of sophisticated detection tools and early treatment with antibiotics will ensure an excellent outcome with negligible long term sequelae. However, the epidemiology of the disease continues to evolve as immunisation practices and patterns of bacterial resistance change, demanding greater vigilance from clinicians. Box 3 lists the risk factors associated with a worse prognosis.

Box 3: Risk factors for a worse prognosis (and hence a more cautious approach to treatment and follow-up)3

  • Infective organism is meticillin resistant Staphylococcus aureus, Streptococcus pneumonia, or positive for Panton-Valentine Leukocidin

  • Concurrent septic arthritis, pyomyositis, or abscess

  • Location: involvement of the hip has the highest risk of complications (40%), followed by the ankle (33%) and knee (10%)

  • Positive culture

  • Increasing C reactive protein values for four or more days

  • Younger age (possibly related to delays in presentation, diagnosis, and hence treatment)

  • Delay in treatment (especially for more than five days)

What does the future hold?

Polymerase chain reaction

Much recent attention has focused on the development of molecular diagnostic technology to identify rare pathogens.6 Polymerase chain reaction is more sensitive and identifies the organism much more quickly than conventional culture techniques do.6

Serum procalcitonin

Detection of serum procalcitonin levels is the most recent advance in diagnosing bacterial infection. It has been investigated in the critical care setting as a possible differentiator between bacterial, viral, and inflammatory processes.22 A prospective study of 44 children found that increased levels differentiated osteomyelitis and septic arthritis from other diseases (for example, arthritis, soft tissue infection, trauma) more effectively than C reactive protein, erythrocyte sedimentation rate, or white cell count.45

Positron emission tomography with computed tomography

Positron emission tomography with computed tomography has been described as superior to magnetic resonance imaging in monitoring response to treatment for osteomyelitis, as it is better at distinguishing between ongoing infection and reparative activity and has faster scanning times.31 However, exposure to radiation and limited availability reduce its practical use.

New treatments

There are concerns about the increased virulence of pathogens and the adequacy of antibiotics to combat them. Two new fifth generation cephalosporins have been recently developed with activity against MRSA (ceftaroline and ceftobiprole).46 Other advances in treatment on the horizon include use of monoclonal antibodies directed against virulence factors of the causative pathogen.6

Additional educational resources

Resources for healthcare professionals
  • Howard A, Wilson M. Septic arthritis in children. BMJ 2010;341:c4407—comprehensive overview of septic arthritis, which is closely linked with ostemyelitis and, likewise, easily missed

  • Perry DC, Harper AR, Bruce CE. A limping child. BMJ 2011;342:d3565—practical considerations and differential diagnosis of a limping child

  • UpToDate ( (subscription required)—overview of condition with links to up to date evidence based practical recommendations

Resources for patients and parents


Cite this as: BMJ 2014;348:g66


  • Contributors: MR and AY conceived and designed the article. AY performed the literature search and wrote the article. MR revised the article critically and had final approval of the version to be published. MR is the guarantor.

  • Competing interests: We have read and understood the BMJ Group policy on declaration of interests and declare the following interests: None.

  • Provenance and peer review: Not commissioned; externally peer reviewed.

  • Patient consent: obtained.


View Abstract