We retrospectively reviewed 52 children treated for tuberculosis of the knee in the 21-year period 1979 to 1999.

The mean age at which the condition was diagnosed was 5.3 years (8 months to 13 years). The median duration of symptoms was four weeks (1 month to 3 years). All patients presented with swelling, mainly owing to synovitis. Pain was a symptom in only two thirds of patients.

Using Kerri and Martini’s classification of radiological appearances, 33 knees were stage I (osteopoenia), 15 stage II (osteopoenia with erosions), two stage III (joint space narrowing) and two stage IV (joint space narrowing with anatomical disorganisation). All knees had either positive histology (caseating granuloma) and/or a positive culture for tuberculosis.

Treatment was with rifampicin, isoniazide and pyrazinamide for nine months. No synovectomy was done. Of the 48 knees with stage-I and stage-II disease, 22 were immobilised for at least three months and 26 actively mobilised.

At a mean follow-up of five years (2 to 16 years), the results were classified according to Wilkinson. All stage-I and stage-II knees had an excellent result (full range of motion) or good result (more than 90° of flexion). Stage-III and stage-IV knees had a fair result (less than 30°of flexion) or poor result (ankylosis). In stage-I and stage-II knees, immobilisation did not affect outcome.

In the same period, 25 knees with a non-specific histology and negative culture presented the problem of the differential diagnosis between tuberculosis and particular juvenile rheumatoid arthritis (JRA). Of these 17 were subsequently diagnosed as JRA. A histological study assessed the value of synovial lining (SLC) hyperplasia. The sensitivity of SLC hyperplasia for JRA was only 53%. Synovial biopsies from 10 joints with tuberculosis (positive histology or culture) were subjected to the polymerase chain reaction test. The sensitivity was only 40% for tuberculosis.