Emerg Med J 29:969-971 doi:10.1136/emermed-2011-201037
  • Original articles

C reactive protein, erythrocyte sedimentation rate, or both, in the diagnosis of atraumatic paediatric limb pain?

  1. Paul Leonard2
  1. 1Emergency Department, Edinburgh Royal Infirmary, Edinburgh, UK
  2. 2Emergency Department, Royal Hospital for Sick Children, Edinburgh, UK
  1. Correspondence to Dr S Robinson, Emergency Department, Edinburgh Royal Infirmary, 51 Little France Crescent, Edinburgh, EH16 4SA UK; psrobinson{at}
  1. Contributors SR collected the data and wrote the manuscript. PL performed the statistical analysis and reviewed the manuscript.

  • Accepted 10 January 2012
  • Published Online First 14 February 2012


Objective To assess if measurement of either C reactive protein (CRP) or erythrocyte sedimentation rate (ESR) individually has an equivalent diagnostic value to measurement of both in identifying orthopaedic infection as the cause of paediatric atraumatic limb pain.

Setting Emergency department of a paediatric teaching hospital.

Study design Retrospective study of case notes for patients attending the emergency department with a complaint of atraumatic limb pain and in whom both ESR and CRP were measured at the time of presentation. Laboratory results at the time of presentation were recorded along with the final diagnosis. Receiver operating characteristic (ROC) curves were created using the data and the optimum cut-off values for each of ESR and CRP were derived using the point of best trade off between sensitivity and specificity. Likelihood ratios for ESR and CRP individually and in combination were calculated.

Results 259 patients were included in the study, of whom 17 were considered to have an orthopaedic infection. ROC curves revealed the best results were obtained using cut-off values of CRP >7 and ESR >12. The combination of a CRP >7 and an ESR >12 gave the best positive likelihood ratio at 6.26 (likelihood ratio 5.34 (CRP >7) vs 2.57 (ESR >12)). For ruling out disease, the combination of CRP ≤7 and ESR ≤12 also outperformed either variable individually (negative likelihood ratio 0.09 (CRP ≤7 and ESR ≤12) vs 0.34 (CRP ≤7) vs 0.18 (ESR ≤12)).

Conclusion Measurement of both CRP and ESR should be considered an important aid in the investigation of atraumatic limb pain.


Paediatric orthopaedic infection (septic arthritis, osteomyelitis and deep soft tissue infection) can have devastating consequences if left undiagnosed and untreated, yet is often a difficult diagnosis to make. To date, no defining clinical signs or symptoms have been identified,1 and gold standard tests, such as joint aspiration for septic arthritis or MRI for osteomyelitis, are more challenging to perform in children and often necessitate general anaesthesia.

Blood tests, especially erythrocyte sedimentation rate (ESR) and C reactive protein (CRP), have been shown to be helpful in the diagnosis of musculoskeletal infection. There is a large amount of evidence showing that ESR is raised in acute orthopaedic infection.1–14 There is also evidence that ESR is useful in the diagnostic process.15–24 This has led to its inclusion in all clinical predictive rules to date.1–6 CRP has been shown to be raised in acute orthopaedic infection,10–14 ,25 and there is a small body of evidence to support its use in the diagnostic process.1–3 19–24

Practice at our local institution has evolved to use a combination of CRP and ESR in the assessment of children with atraumatic limb pain. The reason behind this appears to be related to increased access to these investigations rather than evidence based. The time and cost implications for the emergency department of performing two tests where one may suffice has led to the questioning of current practice. Our aim is to assess whether measurement of inflammatory markers can be rationalised without impacting on diagnostic value.


A retrospective review of the case notes of patients attending the Royal Hospital for Sick Children, Edinburgh, UK, between 1 February 2009 and 31 July 2009 was performed. Patients were identified through the emergency department database which includes all electronic patient records. Details of all hospital attendances, diagnostic tests and clinic visits are recorded within the system. Patients were included if they were 13 years or younger, presented with atraumatic limb pain and had both CRP and ESR measured at presentation. Exclusion criteria were previous orthopaedic infection, chronic recurrent multifocal osteomyelitis, fractures and chronic disease likely to be the cause of pain. The search strategy employed is outlined in figure 1.

Figure 1

Study group identification. ED, emergency department.

The electronic database was hand searched to identify the study group. All patients presenting to our institution with atraumatic limb pain have blood tests performed on first presentation. The database was used to retrieve the data of all patients attending the emergency department who had blood tests performed. The study group was then identified using the presenting complaint entered on the system (this is a hand typed description of the complaint, given by the patient or parent on arrival, entered on the database by reception staff). For each patient meeting the inclusion criteria, the initial test results were recorded, as were the results of further investigations, treatment received and any further presentations over the subsequent 6 month period.

As gold standard diagnostic tests are rarely performed for this set of orthopaedic infections, a pragmatic approach to classification was taken. The criteria for a patient to be recorded as having an orthopaedic infection were that they were referred to the orthopaedic team, considered to have an infection by the orthopaedic consultant and treated with antibiotics.

ESR and CRP values were used to construct Receiver operating characteristic (ROC) curves. These curves were used to derive the optimum cut-off values for each of ESR and CRP to give the best trade off between sensitivity and specificity. These cut-off values enabled the calculation and comparison of positive and negative likelihood ratios for each test individually and for the two tests in combination.


During the trial period, 313 patients attended the emergency department with a complaint of atraumatic limb pain and had blood tests performed. Two hundred and fifty-nine of these patients had both their CRP and ESR measured at presentation; these patients make up our study group. Within the study group, 17 patients were considered to have an infective cause for their limb pain by the orthopaedic team and treated accordingly. Of this group, 11 were diagnosed with osteomyelitis, five with septic arthritis and one with a flexor sheath infection.

ROC curves of their CRP and ESR values were constructed and are shown in figure 2. From these curves it was determined that the optimum cut-off value for CRP was >7 (sensitivity 70.6%; specificity 86.8%) and for ESR >12 (sensitivity 88.2%; specificity 65.7%). These values gave, for CRP, a positive likelihood ratio of 5.34 and a negative likelihood ratio of 0.34. For ESR, the positive likelihood ratio was 2.57 and the negative likelihood ratio 0.18. The combination of a raised CRP and a raised ESR gives a positive likelihood ratio of 6.26 and having CRP ≤7 and ESR ≤12 gives a negative likelihood ratio of 0.09.

Figure 2

Receiver operating characteristic curve analysis of erythrocyte sedimentation rate (ESR) and C reactive protein (CRP) values.

Of the 17 patients treated for an orthopaedic infection, 12 had a CRP >7, 15 an ESR >12 and 16 an ESR >12 and/or CRP >7 (11 had ESR >12 and CRP >7). One patient, with a diagnosis of sacral osteomyelitis, had normal ESR and CRP values. One hundred and fifty-one of the 242 patients with no infective cause had both ESR ≤12 and CRP ≤7.


This was a retrospective study with the inherent problem of data loss. Fifty-four patients were lost due to incomplete blood test results. Of these patients, three were treated for orthopaedic infection.

The gold standard for this study, ‘considered to have an infection by the orthopaedic consultant and treated with antibiotics’, while pragmatic is a significant source of bias.

The ROC curve was used to give the best trade off between sensitivity and specificity so that it was possible to gauge how the ESR and CRP values compared with their diagnostic value. Different cut-off levels may on further analysis maximise the value of the two tests in combination.


The results of our study reflect those of other studies which have found that both ESR and CRP are useful indicators of orthopaedic infection.1–3 19–24 In common with other studies, we found a raised CRP to be a stronger predictor of infection. Direct comparisons between studies are difficult as the cut-off levels for CRP and ESR values vary considerably. The values we derived from the ROC curves were generally lower than those quoted in other studies, especially ESR. This variation obviously leads to very different results.

In terms of minimising further clinical intervention, the likelihood ratio of 0.09 given by the combination of low CRP and ESR values offers a level of risk which is acceptable to most emergency department physicians. It is however important to realise that clinical judgement must always take precedent. Using the values calculated, one case of orthopaedic infection would have been missed in the study group. It is unlikely that blood tests alone will ever be sufficient to rule out orthopaedic infection.


Measurement of both CRP and ESR should be considered an important aid in the investigation of atraumatic limb pain.


  • Competing interests None.

  • Ethics approval This was a service evaluation and therefore ethics board approval was not required.

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


Free sample
This recent issue is free to all users to allow everyone the opportunity to see the full scope and typical content of EMJ.
View free sample issue >>

Don't forget to sign up for content alerts so you keep up to date with all the articles as they are published.


Among patients with minor TBI (GCS 13-15) getting CT scans ≥ 24 hours after injury, what proportion have a traumatic finding?


0.5% - 43% response rate
3% - 41% response rate
10% - 16% response rate

Related original article: PCT head imaging in patients with head injury who present after 24 h of injury: a retrospective cohort study