TUBERCULOSIS OF THE HIP IN CHILDREN

TUBERCULOSIS OF THE HIP IN CHILDREN J. A. B. CAMPBELL, E. B. HOFFMAN From Maitland Cottage Hospital, Cape Town, Republic of South Africa We have reviewed 74 tuberculous hips in 73 children treated from i950 to i99i. From 1979 to 1991 we treated 28 patients with rifampicin, isoniazid and pyrazinamide given for nine months (series A), using active mobiisation for the more recent cases. Before this, 46 hips had been treated with streptomycin and isoniazid with or without para-aminosalicyclic acid given for a mean of 18 months (series B), and all these patients were immobilised for a mean of 2.2 years. The radiological appearances at presentation as classified by Shanmugasundaram (i983) predicted the outcome. Most hips were of the normal type (50% and 59% of series A and B respectively) followed by the dislocating type (25% and i3%) and the atrophic type (8% and 9%). There were good or excellent results in 93% of the normal type. All the atrophic type had poor results. The dislocating type had a poor result if the joint space was narrow after reduction of the hip. Early mobilisation had no effect on the outcome of the normal type of disease. The newer drug regimens allowed for shorter courses of treatment, but did not necessarily give a better outcome. J Bone Joint Surg [Br] 1995;77-B:319-26. Received 6 May 1994; Accepted 24 June /994 Before the introduction of chemotherapy, tuberculosis of the hip often ended in a fibrous ankylosis (Hatcher and Phemister 1937; Ponseti 1948). Streptomycin and isoniazid were introduced in the 1940s, and treatment at that time was by prolonged bed-rest and prolonged chemotherapy. Surgery, however, was advised by some authors. Roberts and Webster ( I 95 1) performed early extra-articular arthrodesis and Marmor et a! (1969) and Wilkinson (1969) used synovectomy with or without curettage of the bone lesions. The introduction of rifampicin in 1967 allowed a shorter course of chemotherapy, and there was a tendency to earlier mobi!isation, J. A. B. Campbell, FCS SA(Orth), Registrar E. B. Hoffman, FCS SA(Orth), Consultant Orthopaedic Surgeon Department of Orthopaedics, University of Cape Town. Observatory 7925. Republic of South Africa. Correspondence should b sent to Dr E. B. Hoffman at 7 Marne Avenue. Newlands 7700, Cape Town, Republic of South Africa. 1995 British Editorial Society of Bone and Joint Surgery 0301-620X/95/2923 $2.00 either actively or with continuous passive motion (Salter et a! 1980). We have reviewed 74 hips in 73 children treated for tuberculosis over the 41-year period from 1950 to 1990. This spanned two eras of different treatment and we have therefore reviewed them as two series, considering first (series A) our most recent experience. In series A, treated from 1979 to 1990, we used rifampicin, isoniazid and pyrazinamide in short courses and in the later years tended towards earlier mobilisation. In series B, treated from 1950 to 1978, the methods were prolonged immobilisation and a long course of streptomycin, isoniazid and para-aminosalicylic acid. The aims of our review were to determine whether short drug regimens and early mobilisation altered the outcome, and to assess the value of the clinicoradiological classification of Shanmugasundaram ( 1983) in predieting the result. PATIENTS AND METHODS Series A. Twenty-seven patients with 28 infected hips presented during the I 2 years from 1979 to 1990. There were 14 boys and 13 girls; 16 were black and 11 of coloured race. Their average age was 5 years ( I year to 13 years 1 0 months). All case records and serial radiographs were available, as were chest radiographs and the results of ESR and Mantoux skin tests. All patients had had open biopsy. and had been treated with rifampicin. isoniazid and pyrazinamide for nine months. The involved hip was either immobilised in a spica or on a frame, or was mobilised actively on skin traction, with or without continuous passive motion (CPM). The radiological appearance of the hip at presentation was recorded by the 1983 classification of Shanmugasundaram (Fig. I ). At review all patients were assessed Table I. Grading of results of treatment of tuberculosis of the hip, modified from Shanmugasundaram (1983) Grade Range of flexion (degrees) Radiography Result Ia Full Normal Excellent lb > 90 Ovoid/irregular Good II 35 to 9() Destroyed Fair III < 35 Fibrous ankylosis Poor IV Nil Bony ankylosis Poor VOL. 77-B. No. 2, MARCH 1995 319

321) J. A. B. CAMPBELL. E. B. HOFFMAN Type 1 Normal Type 2 Travelling acetabulum Type 3 Dislocating Type 4 Perthes Type 5 Protrusio-acetabuli Type 6 Atrophic Type 7 Mortar and pestle Fig. 1 Radiological types of tuberculosis of the hip, redrawn from Shanmugasundaram (l983. clinically and radiologically at an average follow-up of 3 years 9 months (2 years to 6 years 8 months). The results were graded as shown in Table I, modified from Shanmugasundaram (1983). Series B. From 1956 to 1978, 86 patients had been seen at Maitland Cottage Hospital with the presumed diagnosis of tuberculosis of the hip. No hip had been biopsied, and only 46 were considered to be suitable for detailed review. Of the 40 hips excluded, 1 1 did not have the radiological features of tuberculosis and 29 had been treated elsewhere for at least one year before referral. There were I 8 boys and 28 girls; 1 2 were black and 34 were of coloured race. Their average age was 7. 1 years (2 to 14). All 46 patients had been treated with streptomycin and isoniazid, sometimes with the addition of para-aminosalicylic acid. for a mean of 18 months (9 to 24). All had been immobilised in a spica or on a frame for a mean of 2.2 years (1 to 4). These patients were not available for follow-up, and assessments were made from case records and serial radiographs from admission to the time of discharge from hospital. RESULTS Series A. Chest radiographs were positive for active or healed pulmonary tuberculosis in 42%, and the mean ESR on admission was 70 mm in the first hour (25 to 130). The Types Fig. 2 Distribution of radiological types in series A. Mantoux skin test was positive in 26 of the 27 patients. All hips had a definite or probable diagnosis of tuberculosis. The presence of acid-fast bacilli on direct microscopy, or on histological examination or positive Kirschner culture was recorded for 16 hips. Twenty-six hips had histological findings of caseating granulomas. The distribution of the different Shanmugasundaram types is shown in Figure 2. Half of the hips ( 14) showed normal -type radiographs. seven were of the dislocating type and four of the Perthes type. There were no cases showing a travelling acetabu!um or protrusio acetabuli. THE JOURNAL OF BONE AND JOINT SURGERY

TUBERCULOSIS OF THE HIP IN CHILDREN 321 16 14 12 10 8 6 4 2 0 3 4 6 7 Types Fig. 3 Results br each radiological type in series A. The results were excellent in 43% of hips, good in 28.5%. fair in 3.5%, and poor in 25% (Tables I and II) and these are shown in relation to each radiological type in Figure 3. Of the 14 normal -type hips (Fig. 4) six patients had been on frames and eight had been mobilised, six actively and two by CPM. Eleven of these patients had an excellent result, two had a good and one a poor result. Of the seven dislocating or subluxing hips, two were immobilised, one by spica and one on a frame. The other five were mobilised, four actively and one by CPM. One patient had an excellent result, three had a good and three a poor result. The three dislocating hips with poor results all had a joint space reduced to 3 mm or less after reduction (Fig. 5). Radiograph of a 3-year-old girl with the normal type of tuberculosis of the right hip. showing osteopenia and acetabular cysts (a). She had an excellent result at 2-year follow-up (b). Table II. Results in series A and B by number and percentage Result Series A (n = 28) Series B (n = 46) Excellent 12 43 19 41 Good 8 28.5 9 /9.5 Fair 1 3.5 3 6.5 Poor 7 25 15 33 All the four Perthes-type hips (Fig. 6) were mobilised actively; three had a good result and one a poor result. Both atrophic-type hips, with joint spaces 3 mm, had poor results (Fig. 7). The single mortar-and-pestle-type hip was immobilised and had a fair result (Fig. 8). Series B. The radiological types are shown in Figure 9. Again. most were of the normal type (27/46). with six of the dislocating type (6/46) and six of the mortar-and-pestle type. There was one Perthes-type hip. The results, as shown in Tables I and II, were excellent in VOL. 77-B. N. 2. MARCH 1995

322 J. A. B. CAMPBELL. E. B. HOFFMAN Fig. Sa Radiograph of a 6-year-old boy with the dislocating type of disease (a). After open biopsy and relocation, a radiograph in abduction traction shows a normal joint space (b). The final result was good. Fig. Sb 41% of hips, good in 19.5%, fair in 6.5% and poor in 33%; these are shown for each radiological type in Figure 10. Most excellent or good results were seen in normal -type hips, with most poor results in dislocating, atrophic, and mortar-and-pestle DISCUSSION types. We have confirmed the findings of Shanmugasundaram ( I 983) that the radiological appearance of the hip at presentation accurately predicts the final outcome. The normal type of disease has mainly synovial involvement. There may be cysts or cavities in the femoral head, neck, or acetabulum, but there is no gross destruction of subchondral bone and the joint space is normal. Shanmugasundaram had eight such hips in his series of 3 1 in children under 14 years of age; six of the eight had an excellent or good result. In our series 41 of 74 hips (55%) were of the normal type: all but three had excellent or good results. In the Perthes-tvpe hip the femoral head is sclerotic at presentation; only one of our five patients with this type had an isotope bone scan, which appeared cold or avascular. Shanmugasundaram reported three Perthes-type hips all in patients under five years of age, and all with poor results; he stated that the distinction from true Perthes disease may be extremely difficult. Four of our five patients with Perthes-type hips (aged 1 year 9 months to 8 years 4 months at presentation) had good results. Because of the high prevalence of tuberculosis in South Africa we perform ESR and Mantoux tests on all patients suspected of Perthes disease. Our tuberculous Perthes-type hips all had radiological features which differentiated them from true Perthes disease: the whole head was always involved, but there were none of the metaphyseal changes seen in true Perthes disease (Catterall THE JOURNAL OF BONE AND JOINT SURGERY

TUBERCULOSIS OF THE HIP IN CHILDREN 323 Fig. 6a Radiograph of a 2-year-old boy with Perthes type of disease (a). An isotope bone scan showed a cold hip. At two years there is coxa magna with good range of movement (b). Fig. 6b 1 97 1 ). Acetabular osteopenia and cysts are not seen in true Perthes disease (Fig. 6a). Those with good results all had an ovoid femoral head suggesting recurrent vascular insults with a tendency to revascularisation. Shanmugasundaram did not state whether his patients had open biopsies, but he did suggest that surgery provided little benefit. All our patients had an open biopsy, which may have resulted in some decompression of the hip. In the dislocating type, the dislocation or subluxation in series A was seen in all cases at open biopsy to be due to capsular laxity and synovial hypertrophy and not to accumulation of pus. Shanmugasundaram had poor results in all six cases in children. Four of the seven dislocating hips in series A had excellent or good results, but in series B only two of the six hips had a good result. The better results in series A were probably due to open relocation of the hip, as opposed to frame traction in series B. A narrow joint space after reduction ( 3 mm) predicted a poor result, suggesting that there was also an atrophic element. The trat elling-acetabulum, protrusio, and?nortar-andpestle type hips result from erosion of subchondral bone. The travelling-acetabulum hip was common in Shanmugasundaram s series (9 of 3 1 hips). We saw only two cases and as in his study they both had a poor result. The mortar-and-pestle type was seen by Shanmugasundaram only in adults, but we had seven such hips, most early in series B. Shanmugasundaram classified his results as good or excellent, because of the good range of movement. We have modified his classification. however, (Table I), regarding a destroyed femoral head with a good range of movement as a fair result (2 of the 7 hips). The other five hips had fibrous ankylosis and therefore poor results. We had no hips of the protrusio-acetabuli type. Shanmugasundaram reported poor results for the atrophic-type hip, seen almost exclusively in adults. We saw six atrophic-type hips in patients under 14 years of age, and had poor results in all six. Initial radiographs showed a joint space of 3 mm or less, but at open biopsy the cartilage VOL. 77-B. No. 2. MARCH 1995

-. 324 J. A. B. CAMPBELL, E. B. HOFFMAN g- 2 Fig. 7a Radiograph of a 10-year-old boy showing an atrophic type of hip at presentation, with the joint space almost obliterated (a). A radiograph at two years shows fibrous ankylosis (b). Fig. 8 Radiograph of a 6-year-old girl with mortar-andpestle type of disease of the left hip. looked normal and glistening. In one case histological examination of the cartilage was normal (Fig. 7). This suggests that articular cartilage is not destroyed by surface pannus, but by subchondral erosion which appears as jointspace narrowing on radiographs, as reported by Phemister (1925) and Phemister and Hatcher ( 1933). These hips inevitably progress to fibrous ankylosis despite treatment by vigorous active movement, with CPM in later cases; this supports the supposition that there is subchondral erosion. Shanmugasundaram reported that the mean duration of symptoms in normal, Perthes, dislocating and atrophic types varied from 4 to 7 months, and for the travellingacetabulum, protrusio-acetabuli and mortar-and-pestle types, for 10 to 14 months. We had an accurate record of THE JOURNAL OF BONE AND JOINT SURGERY

TUBERCULOSIS OF THE HIP IN CHILDREN 325 Types Fig. 9 Distribution of radiological types of disease in series B. 30 25 20 15 10 1 2 3 4 6 7 Types Fig. 10 The results br each radiological type in series B. duration of symptoms for 20 of the 28 patients in our series A. Normal. Perthes, dislocating and atrophic types showed a mean duration from 2 to S months. This suggests that duration of symptoms may be a factor: in travellingacetabulum, protrusio-acetabuli and mortar-and-pestle types with a longer duration, there is time for significant local subchondral bone destruction to occur. Patients with a similar duration of symptoms, however, may present with a normal type of hip and have a good outcome. or with an atrophic hip and have a poor outcome. In a normal hip, the disease is mainly localised to the synovium. Pannus does not proliferate over the area of joint contact and caseation is a form of local control and necrosis of the disease (Phemister 1925: Phemister and Hatcher 1933). By contrast, there is significant subchondral erosion in the atrophic hip. It is now accepted that the immune mechanism in tuberculosis may either confer protection or lead to tissuedestroying hypersensitivity (Ainslie and Bateman 1991). Protective immunity is related to granuloma formation and is seen in the normal, synovial type of tuberculous hip disease. It seems possible that tissue-destroying hypersensitivity, causing subchondral erosion, may be responsible for the atrophic hip and possibly for the travelling-acetabulum, protrusio-acetabuli, and mortar-andpestle types. The closely similar results in our series A and B (Table II) confirm that the newer drugs rifampicin, isoniazid and pyrazinamide allow shorter courses of treatment than the older drug regimes using streptomycin, isoniazid and paraaminosalicyclic acid, but do not necessarily give improved results. Similar findings have been reported for the treatment of tuberculosis of the spine (MRC 1973, 1993). Both Marmor et al (1969) and Wilkinson ( 1969) reported improved results for synovectomy with or without curettage of bony lesions. We consider that surgery has no major role in the early management of tuberculosis of the hip. In the normal type of disease chemotherapy almost invariably gives a good or excellent result. We do advocate open biopsy in all patients to provide an accurate tissue diagnosis; this may also help to decompress the avascular Perthes-type hip and to relocate the dislocating type. In the atrophic, travelling-acetabulum, protrusio-acetabuli and mortar-and-pestle types of disease, irreversible damage has occurred before presentation. The influence of mobilisation, as opposed to immobi!isation, can be compared in the!4 normal type hips in series A. All were treated in abduction traction for at least two months, but after this in the later cases active movement was encouraged and CPM was added, once the initial pain of the open biopsy had settled. Of these 14 hips, seven were immobilised and seven actively mobilised: there was no difference in the outcome. Conclusions. The radiological appearance of the tubercubus hip at presentation almost invariably predicts the outcome. Hips of normal type. Perthes type and dislocating type with a normal joint space after reduction will have a good result. Hips of atrophic, travelling-acetabulum. protrusio-acetabuli and mortar-and-pestle types will have a poor result. The newer drug regimes allow for shorter periods of treatment, but do not necessarily produce better outcome. Surgery at initial presentation should be limited to open biopsy and the relocation of a dislocated hip. Prolonged immobilisation appears to have no influence on the outcome of the radiologically normal type of disease. This study was made possible by a research grant from the Chairmans Fund Educational Trust of the Anglo-American Corporation of South Africa. The authors would like to thank Mrs V. Barrow kr the typing and Mr M. Wyeth for preparing the illustrations. No benefits in any form have been received or will be received froni a commercial party related directly or indirectly to the subject of this article. REFERENCES Ainslie GM, Bateman ED. Immunological aspects of the host response to Mvcobacterium tuberculosis infection. In: Coovadia HM, Benatar SR. eds. A century oftuherculosis: Soot/i Africa,z perspectives. Cape Town. etc: Oxford University Press: 1991:224-42. Catterall A. The natural history of Perthes disease. J Bone Joint Suq. (Br] I 97 1 :S3-B:37-53. VOL. 77-B. Nu. 2. MARCH 1995

326 J. A. B. CAMPBELL, E. B. HOFFMAN Hatcher CH, Phemister DB. The primary point of infection in tuberculosis of the hip joint. Surg Gynec Obstet 1937;65:721-40. Marmor L, Chan KP, Ho KC, Justin M. Surgical treatment of tuberculosis of the hip in children. Clin Orthop 1969:67:133-42. MRC. A controlled trial of ambulant out-patient treatment and in-patient rest in bed in the management of tuberculosis of the spine in young Korean patients on standard chemotherapy. J Bone Joint Surg (Br] 1973;55-B:678-97. MRC. Controlled trial of short-course regimens of chemotherapy in the ambulatory treatment of spinal tuberculosis: results at three years of a study in Korea. J Bone Joint Surg [Br] 1993;75-B:240-8. Phemister DB. Changes in the articular surfaces in tuberculous arthritis. J Bone Joint Surg 1925:7:835-47. Phemister DB, Hatcher CH. Correlation of pathological and roentgenological findings in diagnosis of tuberculous arthritis. Am J Roentgeno! 1933:29:736-52. Ponseti I. Evolution and treatment of tuberculosis of the hip. Surg Gynec Obstet 1948:87:257-76. Roberts WM, Webster FS. Early surgical treatment of tuberculosis of the hip in children. Surg Gynec Obstet l951;92:155-63. Salter RB, Simmonds DF, Malcolm BW, et at. The biological effect of continous passive motion on the healing of full-thickness defects in articular cartilage: an experimental investigation in the rabbit. J Bone Joint Surg [Am] 1980:62-A: 1232-S 1. Shanmugasundaram TK. Bone and joint tuberculosis. Madras: Kothandaram & Co. 1983. Wilkinson MC. Tuberculosis of the hip and knee treated by chemotherapy, synovectomy and debridement: a follow-up study. J Bone Joint Surg [Am] 1969:Sl-A:1343-59. THE JOURNAL OF BONE AND JOINT SURGERY