The aim of this study was to produce estimates of specificity and predictive value of presenting symptoms and signs of paediatric bone cancer, a rare and frequently misdiagnosed condition, to aid clinical decision-making in primary care.

A systematic literature review plus questionnaire to primary care physicians were carried out to determine frequency of bone cancer symptoms in both cancer and the benign conditions as which cancer is misdiagnosed. Literature sources – Ovid MEDLINE (1950-May 2008), EMBASE (1980-May 2008) and AMED (Allied and Alternative Medicine) (1985-May 2008). Literature review methods – We included systematic reviews, cohort studies or case series (where n ≥ 10), reporting frequency of symptoms and signs at initial presentation, as originally recorded in case notes or observed by the authors, in subjects aged 0–18 years. Disease incidence data was taken from retrospective and prospective studies from 1980 onwards which recorded incidence over a defined time period, in a large pre-defined population within Europe, North America or Australia. Questionnaire respondents – 32 general practitioners and paediatric Accident & Emergency physicians throughout Scotland and England.

Positive predictive values (PPVs) for bone cancer symptoms range from 0.003 to 0.034% (percentage of children presenting with symptom who have cancer). Specificity (percentage of children without cancer who do not have the symptom) varies considerably between symptoms and ranges from 24% (tenderness) to 95% (weight loss). Specificity can be improved by looking for combinations of symptoms. Weight loss and fever are the features with both highest specificity and highest PPV.

Bone cancer symptoms, even those with high specificity for cancer, have low positive predictive value. We suggest that diagnosis based on initial presentation to primary care is intrinsically difficult and that delay in diagnosis is not unreasonable if it is to make use of time as a diagnostic aid.