Comprehensive findings on clinical, bacteriological, histopathological and therapeutic aspects of cutaneous tuberculosis

Summary

Objective  To define the bacteriological and histological correlates of the three predominant clinical forms of cutaneous tuberculosis and to evaluate the efficacy of a 9-month daily regimen containing rifampicin and isoniazid.

Methods  In the dermatological clinics of two major teaching hospitals in Chennai, 213 patients with suspected clinical manifestations of cutaneous tuberculosis underwent examination and a skin biopsy for bacteriological and histological tests. They were treated with a daily regimen of rifampicin and isoniazid for 9 months and follow-up for 3 years.

Results  Bacteriological and/or histological confirmation of tuberculosis was obtained in 88% of the cases. Lupus vulgaris lesions were seen mainly in the extremities and verrucosa cutis occurred predominantly on the sole and foot, while the cervical and axillary regions were the commonest sites for scrofuloderma. Ninety-two per cent of the patients showed resolution of the lesions within the first 6 months of chemotherapy; 1% failed to respond to this regimen. There was no relapse in any of the cases during the follow-up period of 3 years.

Conclusions  Clinical findings were adequate to identify major forms of cutaneous tuberculosis as evidenced by bacteriological and histopathological examination. A daily regimen of rifampicin and isoniazid for 9 months was effective in treating cutaneous tuberculosis.

Objectifs  Définir les corrélations bactériologiques et anatomopathologiques des trois formes cliniques prédominantes de la tuberculoses cutanée et évaluer l'efficacité d'un traitement journalier de 9 mois contenant de la rifampicine et de l'isoniazide.

Méthodes  A partir des consultations de dermatologies de deux grands hôpitaux universitaires à Chennai, 213 patients avec une suspicion clinique de tuberculose cutanée ont bénéficié d'un examen clinique et d'une biopsie cutanée pour des examens bactériologiques et anatomopathologiques. Ils ont été traités par un traitement journalier de rifampicine et d'isoniazide durant 9 mois et suivis pendant 3 ans.

Résultats  la confirmation bactériologique et/ou anatomopathologique de tuberculose était obtenue dans 88% des cas. Des lésions de lupus vulgarisétaient présentes le plus souvent aux extrémités et de verrucosa cutis de façon prédominante sur la plante et sur le pied, alors que les régions cervicales et axillaires étaient les sites habituels pour les gommes tuberculeuses. 92% des patients ont vu leur lésion disparaître dans les 6 premiers mois du traitement ; 1% n'ont pas répondu à ce traitement. Il n'y a eu aucune rechute dans les 3 années de suivi.

Conclusions  Les éléments cliniques sont adéquats pour identifier les formes majeures de tuberculose cutanée comme en témoigne l'examen bactériologique et anatomopathologique. Un traitement journalier de rifampicine et d'isoniazide de neuf mois est efficace dans le traitement de la tuberculose cutanée.

Mots clefs tuberculose cutanée , rifampicine , isoniazide , diagnostic bactériologique/anatomopathologique , diagnostic clinique , Inde

Objetivo  Definir la correlación bacteriológica e histológica de las tres formas clínicas predominantes de la tuberculosis cutánea, y evaluar la eficacia de un régimen diario de 9 meses de duración de rifampicina e isoniazida.

Métodos  En las clínicas dermatológicas de dos importantes hospitales universitarios en Madrás (Chennai), 213 pacientes con sospechadas manifestaciones clínicas de tuberculosis cutánea fueron examinados, y se les realizaron biopsias de piel para análisis bacteriológicos e histológicos. Fueron tratados con un régimen diario de rifampicina e isoniazida durante 9 meses, y controlados durante 3 años.

Resultados  Se obtuvo una confirmación bacteriológica y/o histológica de tuberculosis en 88% de los casos. Lesiones de Lupus vulgar fueron observadas principalmente en las extremidades, y cutánea verrucosa predominantemente en la planta y pie, mientras que las regiones cervicales y axilares fueron los lugares más comunes de escrofuloderma. 92% de los pacientes mostró una resolución de sus lesiones dentro de los primeros 6 meses de quimioterapia; un 1% no respondió a este tratamiento. No hubo recaída en ninguno de los casos durante el siguiente período de control de 3 años.

Conclusiones  Los hallazgos clínicos fueron adecuados para identificar las formas más importantes de tuberculosis cutánea, como quedó evidenciado por los exámenes bacteriológicos e histológicos. Un régimen diario de rifampicina e isoniazida durante nueve meses fue efectivo en el tratamiento de la tuberculosis cutánea.

Palabras clave tuberculosis cutánea , rifampicina , isoniazida , diagnóstico bacteriológico/histológico , diagnóstico clínico , India

Introduction

Tuberculosis continues to be a major public health problem worldwide and especially in the developing world (Walker & Stevens 2003). The lung is affected predominantly and accounts for approximately 80% of all forms of tuberculosis (Balasubramaniam & Ramachandran 2000) while other organs are involved less frequently. However, skin appears to be relatively more resistant to infection with Mycobacterium tuberculosis compared with other organs and is estimated to constitute <2% of all cases of tuberculosis (Gawkrodger 1998). Although the proportion of cutaneous tuberculosis is reported to be approximately 0.15% of all skin out patients in India (Sehgal et al. 1987), reports on the bacteriological and histological correlates of the predominant forms of cutaneous tuberculosis are scanty and need to be defined more clearly.

With our extensive experience in evaluating the clinical and laboratory correlates of extra pulmonary tuberculosis such as tuberculosis of the lymph node (Jawahar et al. 1990), abdomen (Balasubramanian et al. 1989), spine (Parthasarathy et al. 1999), brain (Rajeswari et al. 1995) and tuberculous meningitis (Ramachandran et al. 1989), we undertook the present study with the aim of establishing the diagnosis either by histopathology and/or by bacteriology among the clinically diagnosed cutaneous tuberculosis patients and studied the efficacy of a 9-month rifampicin and isoniazid regimen.

Patients and methods

Patients living in and around Chennai with clinically diagnosed cutaneous tuberculosis were referred by the dermatologists of the two major teaching hospitals of Chennai to Tuberculosis Research Centre. After obtaining informed consent, all eligible patients were registered for the study. The study was approved by the Institutional Ethical Committee.

Pre-treatment assessment and investigation

Patients were clinically evaluated and then postero-anterior radiograph of the chest was taken and tuberculin skin test with one tuberculin unit of purified protein derivative RT 23 with Tween 80 was performed. If the chest radiograph was abnormal, four sputum specimens were collected for examination by smear for acid-fast bacilli and culture for M. tuberculosis. Their renal and hepatic functions and haematological status were evaluated. Those with renal, hepatic or haematological abnormalities or with bacteriologically confirmed pulmonary tuberculosis or who had received >1 month of prior treatment for tuberculosis were excluded from the study. Clinical photographs were taken at the time of presentation and 1, 3, 6, 9, 12, 24 and 36 months after initiating treatment.

Cutaneous biopsy was performed from the most active part of the lesion and the biopsy specimen was divided into two parts: one was embedded in paraffin wax and processed for histopathological examination. The other bit was collected in Kirchner's liquid medium (Vanajakumar et al. 1997). The biopsy sample was aseptically homogenized using a sterile Teflon grinder and then inoculated on Lowenstein Jensen (LJ) slope, LJ with pyruvate and 7H11 agar. The Kirchner's medium in which the biopsy sample was initially collected was incubated as such. To the remaining homogenate, 5 ml of sterile Kirchner's medium was added and incubated. In all, each specimen was inoculated onto three solid media and two liquid media. A smear from the homogenate was also examined for the presence of acid-fast bacilli.

Treatment

Based on clinical diagnosis, patients were treated with rifampicin and isoniazid for 9 months with once a fortnight collection. Adults were given rifampicin 450 mg and isoniazid 300 mg. For children, rifampicin was given at a dose of 10 mg/kg of body weight and isoniazid 5 mg/kg.

Assessment during treatment and follow-up

Clinical progress was monitored every month by a Dermatologist and a physician from Tuberculosis Research Centre during the treatment period, every 3 months up to 24 months and every 6 months up to 36th months.

Outcome measures

A favourable response to treatment was defined as clinical resolution of the lesion as determined by the dermatologist. An unfavourable response was defined as a persisting lesion at the end of treatment or deterioration during treatment. Repeat biopsy was performed for those with unfavourable response and the specimen was examined for histopathological and bacteriological features.

Results

During the period of intake (from 27 December 1993 to 17 November 1998), 330 patients presented with clinical features consistent with a diagnosis of cutaneous tuberculosis. This constituted 0.125% of the total out patient attendance in the Dermatology Departments of the two hospitals during this period. Of these, 99 patients were unwilling to participate in the study. Of the 231 patients registered, four had carcinoma, two had sputum smear positive pulmonary tuberculosis, two were HIV positive, two died because of non-tuberculous causes. Eight patients who presented with crops of hard, dusky, non-itchy symmetrical, multiple papules were diagnosed as having papulo necrotic tuberculid. These patients were excluded as histopathology and culture for M. tuberculosis were negative and molecular biological and/or immunological tests were not performed. Thus, a total of 213 patients remained in the study.

Patients’ characteristics

Basic features

Of the 213 patients, 160 (75%) were male. Among male patients, 77 (48%) had lupus vulgaris, 68 (43%) had verrucosa cutis, 15 (9%) had scrofuloderma. Among 53 (25%) female patients, 33 (62%) had lupus vulgaris, 10 (19%) had verrucosa cutis and 10 (19%) had scrofuloderma. The overall male:female ratio was 3:1. It was 2.3:1 for lupus vulgaris, 6.8:1 for verrucosa cutis and 1.5:1 for scrofuloderma (Figure 1). The age of the patients ranged from 3 to 65 years. Fifty-two patients were <12 years of age. Of these, 26 had lupus vulgaris, 14 had verrucosa cutis and 12 had scrofuloderma. The mean weight of children under 12 was 18.1 kg (males) and 19.2 kg (females); and 39.4 kg (males) and 37.8 kg (females) in adolescents between 12 and 20 years; in patients older than 21 years, 53 kg (males) and 45.7 kg (females).

Figure 1.  The distribution of the different types of cutaneous tuberculosis based on age and sex.

Disease duration

The duration of disease was <1 year for 97 (45%) of 213 patients and 5–10 years for 27 (13%; Figure 2). Of the 97 patients reporting with <1-year duration, 45 (41%) of 110 were lupus vulgaris patients, 30 of 78 were verrucosa cutis, 22 of 25 were scrofuloderma. Eighteen patients with lupus vulgaris and nine patients with verrucosa cutis had reported 5–10-years duration (Figure 2).

Figure 2.  Duration of the disease at presentation in different types of cutaneous tuberculosis.

Site of lesion

Limbs were involved predominantly in lupus vulgaris (46 lower limb and 15 upper limb of 110) followed by gluteal region 17 and axilla 10 and only in face 9. In verrucosa cutis, of the 78 patients, the main sites of lesion were observed in foot (27) and sole (23) and scrofuloderma was seen mainly in the inguinal region (7) and neck (6).

X-ray

In six patients with chest X-ray suggestive of pulmonary tuberculosis, the smear and cultures were negative. They were routinely checked for sputum AFB and culture and none of them developed pulmonary tuberculosis during the entire period of the study.

Mantoux results

Of the 213 patients, Mantoux results were not available for 18 and 13 patients, showed no reaction. Of the remaining, in 79 (41%), the induration was > 20 mm. The distribution of the induration was similar in all the three types of cutaneous tuberculosis (Table 1). The median induration was 20 mm in lupus vulgaris and verrucosa cutis and 22 mm in scrofuloderma.

Table 1.   Mantoux results

	LV (%)	VC (%)	SD (%)
Mantoux test was performed by intradermal injection of PPD RT 23 0.1 ml in the volar surface of the forearm and the induration was measured after 48 h in lupus vulgaris (LV), verrucosa cutis (VC) and scrofuloderma (SD).
>5 mm	94 (94)	65 (83)	19 (86)
>10 mm	86 (86)	62 (79)	17 (77)
>15 mm	73 (94)	51 (70)	17 (77)
>20 mm	37 (37)	30 (38)	12 (48)

Bacteriology

Out of 213 patients, for 10 (5%), bacteriological investigations could not be performed. Culture was positive for M. tuberculosis in 112 (55%) patients (Table 2). The culture positivity was similar for all the three types of lesions: 60 of 106 (57%) for lupus vulgaris, 40 of 73 (55%) for verrucosa cutis and 12 of 24 (50%) for scrofuloderma. Drug susceptibility results were available for 110 out of 112 positive cultures; 96 (87%) were susceptible to streptomycin, isoniazid and rifampicin and 14 (12.7%) showed resistance to one or more of the drugs tested.

Table 2.   Bacteriology results

Culture	LV	VC	SD	Total
Tissues obtained from the active edge of the lesions from lupus vulgaris (LV), verrucosa cutis (VC) and scrofuloderma (SD) were cultured as detailed in multiple media. *NTM, non-tuberculous mycobacteria.
Positive	60	40	12	112
Negative	36	22	11	69
NTM*	9	10	1	20
Contamination	1	1	0	2

Resistance to isoniazid either alone or in combination with other anti-tuberculous drugs was observed in 8 (7.2%), resistance to streptomycin was found in 7 (6.3%) and multidrug resistant tuberculosis was observed in 2% (Table 3).

Table 3.   Drug sensitivity results

	N	%
Drug susceptibility pattern of M. tuberculosis was tested in 110 of the 112 cases where the bacilli were isolated. S, streptomycin; H, isoniazid; R, rifampicin.
Sensitivity results available	110	98
Sensitive to all drugs	96	87
Resistance to S	5	4
H	4	4
R	1	1
SH	2	2
HR	2	2

Non-tuberculous mycobacteria were grown in twenty patients. Of these, four (two each from lupus vulgaris and verrucosa cutis) were rapid growers and eight (six from verrucosa cutis and two from lupus vulgaris) were non-pigmented. In the remaining, further characterization was not carried out.

Histopathology

Of the 213 patients, histopathological examination could not be performed for 11(5%) patients. Of the remaining 202, 175 (86%) patients had histopathology suggestive of tuberculosis 82% of 108 in lupus vulgaris, 90% of 72 in verrucosa cutis and 95% of 22 in scrofuloderma. There was a marked hyperplasia of the epidermis in both lupus vulgaris and verrucosa cutis. In verrucosa cutis, there was additionally hyperkeratosis and para keratosis and in many cases, there was a pseudo epitheliomatous hyperplasia of the epidermis. Further in 33% of 65 verrucosa cutis lesions, neutrophilic micro abscess were present in the epidermis. The granuloma in all three types of cases consisted of epithelioid cells, macrophages and Langhans giant cells. Plasma cells were seen in moderate numbers in lupus vulgaris and scrofuloderma and minimally in verrucosa cutis. While caseation necrosis was present predominantly in all cases of scrofuloderma, it was seen in only three of 108 lupus vulgaris cases and in none of the verrucosa cutis lesions. In none of the 202 cases, acid-fast bacilli could be demonstrated in the section.

Histopathology and bacteriology correlation

Among the 193 patients from whom both histopathology and bacteriology results were available, both were positive for 96 (50%) patients, histopathology alone was positive for 78 (41%) and bacteriology alone for 16 (8%; Table 4).

Table 4.   Correlation of the bacteriological and histopathological findings

Bact.	Histopathology		Total
	Negative	Positive
Negative	13	53	66
Positive	11	96	107
NTM	2	18	20
Total	26	167	193

Response to treatment

The clinical resolution of the lesions in the three types of cutaneous tuberculosis at various time points from presentation up to 24 months is depicted in Photo 1. The lesions resolved clinically within 3 months in 101 (47%) of 213 patients and it resolved by the sixth month in a further 95 cases (Table 5). This included one patient who had verrucosa cutis with rifampicin resistance. In the seventh to ninth month period the lesion resolved in 11 (5%). These consisted of four cases of lupus vulgaris (one patient with rifampacin and isoniazid resistance) and seven of verrucosa cutis. The lupus vulgaris patient with rifampicin and isoniazid resistance was biopsied at the end of treatment and this showed complete clearance both bacteriologically and histologically. However, this patient was lost to follow-up as she migrated.

Figure Photo 1.  Lupus vulgaris: (1a) Hyperpigmented infiltrated plaque studded with multiple nodules over the upper margin and atrophy in the lower margin. Size 5 cm × 5 cm. At presentation. (1b) The infiltrated nodular plaque has subsided considerably. After 6 months of chemotherapy. (1c) Only atrophic scar is seen without any activity. After 9 months of chemotherapy. (1d) Atrophic scar. After 24 months of chemotherapy. Verrucosa cutis: (2a) Two well-defined asymptomatic verrucous plaque involving dorsum of right index finger and palmar extension. Size: little finger 4 1/2 cm × 4 cm; index finger 5 cm × 4 cm. At presentation. (2b) Verrucous plaque has flattened considerably; areas of depigmentation seen. After 6 months of chemotherapy. (2c) All the verrucous lesion has resolved leaving behind depigmentation. After nine months of chemotherapy. (2d) Verrucous lesion has completely resolved leaving minimal depigmentation. After 24 months of chemotherapy. Scrofuloderma: (3a) Two ulcers overlying an enlarged lymph node on the medial side of the right thigh with draining sinus. At presentation. (3b) Lesions and the sinus have resolved completely after 6 months of chemotherapy leaving a puccured scar. (3c) Complete resolution at 9 months. (3d) Complete resolution at 24 months without any recurrence.

Table 5.   Clinical resolution of the lesions in cases of cutaneous tuberculosis after starting chemotherapy

Type	Month of resolution (m)						All
	1	2	3	4–6	7–9	>9
LV, lupus vulgaris; VC, verrucosa cutis; SD, scrofuloderma.
LV	12	20	20	51	4	3	110
VC	14	10	13	32	7	2	78
SD	3	2	7	12	0	1	25
All (%)	29 (13)	32 (15)	40 (19)	95 (45)	11 (6)	6 (3)	213

In the remaining six (lupus vulgaris three, verrucosa cutis two, scrofuloderma one), the lesion did not resolve even at the end of treatment. Of the three lupus vulgaris patients, two were irregular for the initial treatment and the same treatment was continued and one patient had resolution of the lesion within 3 months of regular intake of treatment and another patient who was resistant to rifampicin and isoniazid, died of non-tuberculous cause after restarting one month of treatment. The third patient in whom the lesion persisted clinically and histologically had a change of treatment after which it resolved completely at third month and histopathology showed only fibrous tissue. However, the treatment was continued for 9 months and follow up of 36 months with no relapse.

Of the two patients with verrucosa cutis, one had a change of treatment to ethambutol and isoniazid because of hypersensitivity reaction to rifampicin and the lesion resolved at the end of 12 months. The other patient developed hypersensitivity reaction to isoniazid and rifampicin but treatment was continued under cover of antihistamine. However, the patient defaulted after treatment and was lost for further follow-up.

One patient with scrofuloderma had a lesion, which persisted clinically and histologically at the end of treatment. Hence, this patient was retreated with an alternative regimen and the lesion resolved completely at the end of 6 months.

Discussion

The proportion of cutaneous tuberculosis amongst the outpatients attending the dermatology departments of the major teaching hospitals in Chennai appears to be extremely low – 0.125% in this study. This is similar to the figures reported by Mammen and Thumbiah (1974) in the same area approximately 30 years ago. This is also in agreement with that reported by others (Sehgal et al. 1987; Kumar & Muralidhar 1999).

It is generally believed that in extra pulmonary tuberculosis, there is a greater preponderance in adult females as observed in tuberculous lymphadenitis (Ramanathan et al. 1999). However, in cutaneous tuberculosis, there appears to be a greater prevalence in men (3:1) as seen in data from the study and also as reported by others (Sehgal et al. 1987; Kumar & Muralidhar 1999). This is similar to pulmonary tuberculosis (Khatri & Freiden 2000). However, this ratio varied from type to type. In patients with lupus vulgaris, the ratio was 2.3:1, in verrucosa cutis 6.8:1 and in scrofuloderma 1.5:1. The male preponderance in verrucosa cutis was significantly higher compared with scrofuloderma (χ² = 7.285; P < 0.01) and lupus vulgaris (χ² = 6.693; P < 0.01). However, the distribution of the sexes was comparable in lupus vulgaris and scrofuloderma (P > 0.05). It was found that the proportion of children (below12 years) was 24.4% (52 of 213) which is similar to that reported by Arora and Agarwal (2005). Further, in children below 12 years, scrofuloderma was significantly more than lupus vulgaris (P < 0.05) and verrucosa cutis (P < 0.01) irrespective of the sex.

We found larger proportions of individuals with verrucosa cutis compared with other Indian studies in both adults and children. This could be due to the fact that these lesions were seen mainly in the sole and foot and could be attributed to the practice of many individuals walking bare foot in this part of the world.

Patients with scrofuloderma presented earlier compared with lupus vulgaris and verrucosa cutis. This could be due to the fact that in cutaneous tuberculosis, none of the patients had any constitutional symptoms and locally in lupus vulgaris and verrucosa cutis, the lesions were non-itchy and painless. However, in scrofuloderma, the discharging sinus was physically uncomfortable and occasionally painful and hence these patients could have reported earlier.

Although the lesions in all the three types of cutaneous tuberculosis were seen over the entire body, there appeared to be some marked differences in their distribution. Lupus vulgaris was observed largely over the extremities (61 of 110) and verrucosa cutis was seen predominantly in the sole and foot (50 of 78), while in scrofuloderma, it occurred mainly over the lymph nodes (17 of 25). Further, lupus vulgaris and scrofuloderma, lesions did not occur on the sole of the foot while verrucous lesions were absent over the face. The pathogenic mechanisms responsible for this differential distribution need to be explored. One possible mechanism may be infection because of direct inoculation. For example, in the present study, there was history of injury over the lesion in 14 of 110 cases in lupus vulgaris, 9 of 77 in verrucosa cutis and 3 of 25 in scrofuloderma.

On the whole, scrofuloderma lesions were much smaller than lupus vulgaris or verrucosa cutis lesions. Again this could be due to the discharging sinus which prompts the patient to seek medical help earlier. A similar association between the duration of the disease and size of the lesion in lupus vulgaris has been reported by Horwitz and Christensen (1960).

X-ray lesions suggestive of pulmonary tuberculosis were seen in six cases (one scrofuloderma, two lupus vulgaris and three verrucosa cutis), although all of them were bacteriologically negative. Further, in none of the cases, there was clinical evidence of disseminated tuberculosis. This is unlike that reported by Kumar and Muralidhar (1999) who found that scrofuloderma represents a more disseminated form of the disease. Similarly, unlike skin tuberculosis, in other forms of extra pulmonary tuberculosis, there is evidence of disseminated disease (Balasubramaniam & Ramachandran 2000)

There was no difference in the yield of culture positivity in the three types of cutaneous tuberculosis. Although it is widely believed that it is difficult to demonstrate viable bacilli in a paucibacillary condition like cutaneous tuberculosis (Michelson 1948; Sehgal et al. 1987) culture positivity rate was 55% in the present study. This could be due to the use of multiple media for culture and the use of Kirchner's liquid medium for preserving the biopsy material till they were processed as has been performed in non-cutaneous tuberculosis (Mitchison et al. 1983; Vanajakumar et al. 1997). Thus, findings from this study, therefore, belie the popular impression that culturing of cutaneous lesions for M. tuberculosis could be tedious and unrewarding (Sehgal et al. 1987).

The drug resistance pattern including initial multi drug resistance (to isoniazid and rifampicin) of the isolated tubercle bacilli in this study is similar to the resistance pattern seen among new pulmonary tuberculosis patients admitted recently to various controlled clinical trials at this centre (Tuberculosis Research Centre 2002, 2004). Although multidrug resistant tuberculosis was not observed in our earlier studies conducted on extra pulmonary tuberculosis, our recent study in patients with tuberculous lymphadenitis (Jawahar et al. 2005) shows that multidrug resistant tuberculosis is similar to the present findings of 2%.

The histopathology of the lesions seen was typical of tuberculous granuloma although caseation necrosis was absent in all cases of verrucosa cutis and in an overwhelming majority of lupus vulgaris cases. This finding is in marked contrast to that reported by Sehgal et al. (1987). The presence of moderate numbers of plasma cells in lupus vulgaris and scrofuloderma indicates that these cells as well as humoral immune response probably play a role in the pathogenesis of these types of cutaneous tuberculosis. A similar association of plasma cells in some forms of tuberculous lymphadenitis has been reported earlier (Ramanathan et al. 1999).

Correlation of histopathology with bacteriology results indicates that these two indices are complementary to each other and performing these two diagnostic procedures increases the establishment of the diagnosis of cutaneous tuberculosis by 8%.

As cutaneous tuberculosis is a pauci bacillary condition, we initiated patients on a 9 months daily rifampicin and isoniazid regimen. It is noteworthy that in nearly half the patients the lesions resolved in 3 months and in 92% by the sixth month. Further, in this series, 12 of the 14 patients who had resistance to any of the drugs either alone or in combination, had no failure or relapse of the condition during the 3-year follow-up period. However, no definite conclusion could be drawn about the effectiveness of this regimen where resistance to rifampicin was seen. In one patient who had isolated rifampicin resistance, the lesion resolved. Of the two patients who had resistance to rifampicin and isoniazid, one died of non-tuberculous cause and the other patient defaulted after completion of treatment and was lost for further follow-up. Overall, in this series it was observed that though there was bacteriological resistance, clinical resolution could be achieved even with a two drug regimen. It was also noted that irrespective of the type, size or duration of the lesion, complete resolution occurred. These findings are in agreement with those made earlier in other forms of extra pulmonary tuberculosis (Balasubramaniam & Ramachandran 2000). In view of the success of this two-drug, 9-month regimen, a further shortening of the treatment duration can be achieved by the inclusion of an additional drug.