a Medical Research Council, Tygerberg, South Africa; b Stellenbosch University, Cape Town, South Africa; c Kenya Medical Research Institute, Nairobi, Kenya; d Tuberculosis Research Centre, Madras, India; e Hong Kong Government Tuberculosis Service; f Pathology Service, Hong Kong Government; g Ruttonjee Hospital, Hong Kong; h St George's Hospital Medical School, London, UK
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
In 1993, at the suggestion of the WHO Steering Committee on Treatment of Mycobacterial Diseases (THEMYC), a multi-centre study of EBA was started. The four collaborating centres agreed to use the same protocol, and central co-ordination was provided at St George's Hospital Medical School. The aims of the study were (i) to compare estimates of the EBAs obtained by viable counting with those obtained by counting total numbers of acid-fast bacilli (AFB), using drugs and dosages with a wide expected range of EBAs; (ii) to see whether estimates of the EBAs obtained by total counting could be used to correct and make more accurate the EBAs obtained by viable counting; (iii) to see whether the EBAs were similar in the different collaborating centres; (iv) to see whether extending the treatment period from 2 days to 5 days gave additional information; and (v) to compare the variation between patients in the centres to test the hypothesis that precise estimates are only obtained with Africans.
![]() |
Materials and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The protocol stated that patients should be aged 18 years or more, have newly diagnosed pulmonary tuberculosis with smear-positive sputum and should never have received any previous antituberculosis chemotherapy. They should weigh 4060 kg, so as to be able to state the dose size approximately as mg drug/kg body weight. Drug addicts, pregnant women, diabetics and patients in poor general condition were excluded. Ethical permission for the study was gained from local committees at each centre. Patients gave informed consent. The presence of HIV infection was not examined at all centres for cost reasons. About 36% of patients in Nairobi would have been infected7 whereas the prevalence of infection was estimated to be much lower amongst those in Cape Town (3.1% in a survey8) and Madras (0.5% in 800 tuberculosis patients; P. R. Narayanan, personal communication). In Hong Kong <0.5% of patients would have been HIV seropositive.9
Anti-tuberculosis drugs
Patients were randomly allocated, by the use of treatment slips inside envelopes, to daily dosage with either isoniazid 300 mg (6 mg/kg), isoniazid 18.5 mg (0.35 mg/kg), rifampicin 600 mg (12 mg/kg), ofloxacin 800 mg (16 mg/kg) or no drug. These treatment groups are termed INH 300, INH 18.5, RMP 600, OFL 800 or Nil, respectively. The drug dosages used cover a wide range of expected EBA values. The dose sizes of rifampicin and ofloxacin are those usually used in the treatment of tuberculosis; 600 mg rifampicin has been found to yield an EBA lower than that of 300 mg isoniazid1,2 and only one study, with wide confidence limits, has been done previously on ofloxacin by a group in Durban, South Africa.4
Timetable
Drugs were given daily soon after 8 am for 5 days. Collections of sputum were made over a 16 h period from 4 pm to 8 am the next day before the first drug dose (S1 collection), before the third dose (S3 collection) and after the last dose (S6 collection). Only S1S3 estimates were obtained in the Nil group in Cape Town. Routine short-course chemotherapy was then started.
Sputum collections
Sputum was collected in wide-mouthed 200 mL polystyrene, screw-capped honey-pots' (Medfor, Farnborough, UK) and the volume of the collection was measured by comparison with a calibrated honey-pot. The sputum was homogenized by adding three to six glass beads 10 mm in diameter, and shaking vigorously. Portions of the sputum from the S1 and S6 collections were examined by smear and culture; positive cultures were tested for their sensitivity to isoniazid, rifampicin and to ofloxacin in some centres.
Cfu counts
The method described elsewhere3 has been modified slightly to improve antifungal activity. Plates of 7H11 oleic acid-albumin agar were made selective by the addition of special batches of Selectatabs' (Mast, Bootle, UK) so that final concentrations of antibiotics in the medium were: polymyxin B sulphate 200000 units/L, carbenicillin 100 mg/L, trimethoprim (as lactate) 20 mg/L and amphotericin B 100 mg/L, added without bile salt.
A portion of homogenized sputum was mixed with an equal volume of dithiothreitol as Sputasol (Oxoid, Basingstoke, UK) or Sputolysin (Hoechst, Hounslow, UK) in a screw-capped container and vortex mixed for 20 s. The sputum was then placed on a rotating mixer for 3060 min until well digested. Serial 10-fold dilutions from undiluted to 105 were prepared in sterile 2% horse serum or in sterile 0.2% bovine albumin by adding 100 µL from a Gilson pipette to 900 µL diluent or by using separate pipettes. Duplicate or triplicate segments of selective 7H11 medium plates were inoculated with 100 µL of each dilution. The plates were placed in polythene bags together with a plate inoculated with Mycobacterium phlei, to provide CO2, and were incubated at 37°C for 3 weeks. Colonies were counted a day after addition of a square of blotting paper soaked in 40% formaldehyde to the lids of the plates. Counts were calculated per mL sputum. S1S3 EBA = (log S1 countslog S3 counts)/2. S3S6 EBA = (log S3 countslog S6 counts)/3.
Total counts
Counts were done within 8 mm diameter clear circles in slides otherwise coated with polytetrafluorethylene (Hedley, Loughton, UK). To these circles were added either 30 µL undiluted sputum homogenate, or 10 µL undiluted homogenate or 10 µL from the 101 or 102 dilutions made for the cfu counts. The drops of sputum dilutions were spread evenly within the clear circles, allowed to dry, and were then fixed and stained with auramine O. AFB were counted under epifluorescence using a 40x or 60x objective. The number of fields examined and the total number of bacilli counted were noted until either 200 bacilli had been counted in all or, where few bacilli were seen, at least 20 bacilli were found in at least 200 fields. The area of a field scanned by the particular lens combination used for counting was measured with a stage graticule engraved with 0.1 mm rulings. The number of bacilli per mL sputum was then obtained as:
![]() |
Statistical procedures
Patient details and the sputum counts obtained were entered into EXCEL worksheets (Microsoft Corp) from which were obtained the mean, range, standard deviation (S.D.) and 95% confidence limits (95CL). Further statistical analyses were done with EPI INFO 6,10 using one-way analysis of variance (ANOVA), KruskalWallis non-parametric analyses where indicated by Bartlett's tests and regression analyses. The mean EBA weighted for differences in variance was calculated as Ym = wiYi /
wi where wi = 1/Vi and Yi and Vi are individual means and variances, respectively. Homogeneity of the means was tested with k2 degrees of freedom as X2 =
wiYi2 (
wiYi)2/
wi where k = number of groups. If homogeneous, the 95CI = Ym ± 1.96
(1/
wi).
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
The study started after about 1 year of preparation in May 1995, ran more slowly than had initially been envisaged and was closed in June 1998 with only part of the planned intake of 70 patients per centre completed in three of the four centres. Of the total of 262 patients admitted, 63 were studied in Cape Town, 58 in Nairobi, 75 in Madras and 66 in Hong Kong (Table I). The protocol unfortunately failed to include mandatory urine tests for isonicotinic acid to provide evidence that patients had not received antituberculosis chemotherapy before admission. The content of isoniazid in the sputum collections was, however, measured at Cape Town and six patients were excluded because it was found. None of the centres kept the weight range of patients within the limits of 4060 kg specified by the protocol though deviations were not large and no exclusions for out of range weight were made. The entire results from patients with initial resistance to their study drug have been excluded. Particular items of data are missing or have been excluded for a variety of reasons such as failure to obtain a sputum collection, contamination, freezing of sputum before cfu counting or because the volume of the sputum collection was less than 5 mL. In such cases, the remaining data for the patient were used. In Cape Town, total counts were not done on a high proportion of the sputum collections from patients admitted towards the end of the study, nor for ethical reasons were patients in the Nil group studied for more than the initial 2 days. In none of the patients did drug resistance arise during the 5 days of monotherapy.
|
Highly significant correlations between the initial S1 cfu and total counts were found in the four centres (Table II). The cfu count was slightly lower than the total count in all centres except Madras. The high cfu counts in Madras, the poor correlation and a lack of proportionality in the cfu counting data (not tabulated here) indicated a problem in the cfu counting technique at Madras probably due to carry-over from lower to higher dilutions in pipettes.
|
There are many reasons why cfu EBAs may vary from one patient to another in the same drug dosage group, such as previous chemotherapy, differences in individual response to drugs, faulty collection or labelling of sputum collections. Although these type I variations can affect total as well as cfu counts, there is no way in which the cfu counts can be corrected by the total counts. In type II variation, sputum collections may vary in the proportion of purulent cavitary material containing viable bacilli to the content of mucus from bronchial walls and saliva, as emphasized by Sturm.4 Type II variation can be corrected by using total counts since both will be altered to the same extent by dilution of the purulent material. An adjustment was therefore made to estimate the proportion of total bacilli (t) that are viable (v), so that a new EBA can be calculated, for instance over the S1S3 interval, as the subtraction EBA (Sub EBA):
![]() |
![]() |
![]() |
An example of the adjustment process using Hong Kong data sets that have complete cfu and total counts for the S1S3 period is shown in Table III. Hong Kong results were chosen because of the high correlation (r = 0.88, Table II
) between the total and cfu counts, which suggest that adjustment should be successful in reducing the S.D. In comparing the Sub EBA with the original cfu EBA, the same means are obtained but there is a fall in S.D. within each treatment group, indicating the success of the adjustment process. The overall discrimination between the treatment groups measured by F increased from 5.1 (P = 0.002) to 10.5 (P = 0.00003) and the pooled S.D. decreased, correspondingly, from 0.39 to 0.27 as a result of the adjustment.
|
Tables IVVII reporting the results in the four centres have been calculated for all available results, not just for those patients who have complete sets of results. EBAs by the standard cfu counting method are set out in the top third of each table. The variation between the patients was least at Cape Town (Table IV
) with pooled S.D. values of 0.17 for both the S1S3 and the S3S6 periods. Variation was greater with S.D. values of 0.41 and 0.30, respectively, at Nairobi (Table V
), 0.64 and 0.32, respectively, at Madras (Table VI
) and 0.39 and 0.29, respectively, at Hong Kong (Table VII
). The overall ability of the EBA to discriminate between the four groups that received drugs during the S1S3 period, although statistically significant in each centre, was also greater at Cape Town (F = 19.9) than at Nairobi (F = 3.0), Madras (F = 6.0) or Hong Kong (F = 6.1). However, during the S3S6 period, as was expected, the discrimination between the treatment groups only just achieved statistical significance (F = 3.5, P = 0.02) at Cape Town but was not significant at Nairobi, Madras or Hong Kong.
|
|
|
|
|
Ofloxacin in the OFL 800 group had appreciable S1S3 EBAs, with values of 0.391 at Cape Town, 0.146 at Nairobi and 0.130 at Madras (weighted mean of 0.263). In the subsequent 3 days, it maintained bactericidal activity with EBAs of 0.165 at Cape Town, 0.293 at Nairobi and 0.236 at Madras (weighted mean of 0.202). Lower EBAs of 0.009 and 0.122 were found at Hong Kong but the results are not sufficiently precise to allow comparison with other centres.
EBAs obtained by counting total AFB
The total EBAs obtained by counting AFB in sputum are set out in the middle sections of Tables IVVII. There are no consistent patterns of results. However, of the 32 estimates during the S1S3 and S3S6 periods from the four treatment groups given drugs at the four centres, 25 were positive and only seven were negative, indicating a tendency towards a fall in counts during treatment. Of the six EBAs in the INH 300 groups (S1S3 and S3S6) at Cape Town, Nairobi and Madras, five had total EBAs > 0.2. Exposure to isoniazid is known to cause loss of acid-fastness which has been used as an endpoint in microbiological assay.11 A total EBA in the RMP 600 group of > 0.2 was also found at Nairobi (S1S3). At Hong Kong, in keeping with cfu EBAs, total EBAs in the INH 300 group were low, but in the RMP 600 group both S1S3 and S3S6 total EBAs were > 0.2. The variation between patients, estimated as the pooled S.D., was smaller when calculated as total EBAs than as cfu EBAs at Nairobi, Madras and Hong Kong but was similar at Cape Town.
Adjusted EBAs
The adjusted EBAs, calculated from all available data, are set out in the bottom section of Tables IVVII. Weighted means giving the best estimates of combined results in Cape Town, Nairobi and Madras are in Table VIII
. The success of adjustment in decreasing the S.D. from cfu EBAs to adjusted EBAs agrees with the value of r correlating cfu EBAs and total EBAs (Table IX
). The value of r was 0.01 (small, negative and non-significant) for the correlation over 02 days at Cape Town. Adjustment then increased the pooled S.D. of the cfu EBAs from 0.17 to 0.24. In the Madras data over 35 days (Table IX
), r = 0.31 and adjustment led to a small increase in the pooled S.D. from 0.32 to 0.38. With higher values of r, from 0.48 to 0.59, adjustment either did not change the pooled S.D. (Madras 02 days) or reduced it slightly (Cape Town 35 days, Nairobi 02 days and 35 days, Hong Kong 35 days). In the Hong Kong 02 days data, r had its largest value of 0.76 and the adjustment created the greatest reduction in pooled S.D. from 0.39 to 0.28. The value of F for the comparison of variation between and within treatment groups shows corresponding changes before and after adjustment.
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Hong Kong results differed from those found elsewhere. There was greater variation in the S1S3 cfu EBAs amongst Hong Kong patients than amongst those from Cape Town. The pooled S.D. in Hong Kong was 0.39, almost identical to the previous estimate of 0.38.14 Using S1S3 EBAs with or without adjustment according to which had the lowest values (Table IX), the Hong Kong pooled S.D. was 0.28, >0.17 in Cape Town, substantially less than 0.36 in Nairobi and 0.57 in Madras. This must be due to characteristics of the patients since findings in both Hong Kong studies, such as the excellent correlation between the cfu and total S1 counts, the constancy of standard EBAs between studies and the low S.D. values of the total EBAs, indicate accurate work in the laboratory. Also, in the Hong Kong data, the mean S1S3 EBA was greater in the RMP 600 group (0.612) than in the INH 300 group (0.371), though the difference is not significant. Since the EBA of INH 300 was greater than the EBA of RMP 600 in an earlier study, weighted means and confidence limits were calculated to include the Hong Kong results in the present and the earlier study and the African results from the present Cape Town and Nairobi centres as well as those from previous studies at Nairobi and Cape Town4 (Table X
). The weighted means indicate lower EBAs of isoniazid and higher EBAs of rifampicin in Hong Kong than in the African studies though the differences just fail to attain statistical significance. An explanation for these findings might be that tuberculosis in Hong Kong is more chronic than in the other centres, with bacteria growing more slowly in the lesions. Slow growth would favour the sterilizing activity of rifampicin but limit the early kill due to isoniazid. Thick-walled chronic cavities might also interconnect less well than in acute disease, thus making it more difficult to sample their contents. This could be responsible for the high variability of the Hong Kong EBA results. The proposition that lesions in Hong Kong are more chronic than in African patients is supported by evidence that sterilization during clinical trials of the same chemotherapy regimen proceeds more slowly in Hong Kong patients than in African patients.15
|
A recent American study reported on the effects of various factors, such as the duration of the study period and of daily sputum collections, on the cfu and total EBAs of 16 patients treated with 300 mg isoniazid daily over a 5 day period.16 Their finding that total EBAs were measurable and started after a delay of about 2 days, justifies subtraction of the mean for the total EBAs in the adjustment procedure used in the present study. However, further findings in their study differ considerably from ours, in that the mean of the 2 day cfu EBAs derived from 12 h sputum collections (corresponding to our S1S3 period), was only 0.27 and the S.D. was high at 0.62. Furthermore, the rate at which sputum cfu counts fell persisted during the 5 day period. These discrepancies may have arisen because a proportion of their patients had taken isoniazid before the study commenced. No tests were apparently done on urine or sputum for the presence of anti-tuberculosis drugs at the start of the study. A mixture of the S1S3 cfu EBAs and S3S6 cfu EBAs in our present study would produce similar results. In Cape Town, six patients were found to have isoniazid in their sputum despite earlier assertions that it was impossible for such patients to obtain anti-tuberculosis drugs without the knowledge of the tuberculosis treatment service.2 Unfortunately, other centres did not test for the presence of isoniazid and the inclusion of patients who had actually started drug treatment is likely to be a cause of excess variation in EBA estimates
The findings of the present study with ofloxacin are of interest since little work has been done on the EBAs of quinolones. In Durban, a mean EBA of 0.36 was found in 10 patients given 800 mg ofloxacin, similar to the estimate of 0.391 at Cape Town.4 These EBAs appear higher than the 0.121 obtained with 1000 mg ciprofloxacin daily at Cape Town.17 The difference suggests that 800 mg ofloxacin would be more effective than 500 mg ciprofloxacin, though a reliable comparison cannot be made without studying a range of ofloxacin doses.
The most important implication of the study is the possibility of assessing the sterilizing activity of antituberculosis drugs by extending the dosage period of EBA studies beyond the first 2 days. This method would have advantages of speed and low cost, in assessing the sterilizing activity of new drugs, over currently available surrogate markers of relapse rates which are 2 month sputum bacteriology6 and the presence of fbpB (85B, alpha antigen) mRNA in sputum at about 15 days.18 However, it was only the Cape Town centre that was able to demonstrate the activity of rifampicin during days 25 (S3S6 EBA), because it had the least variation between EBAs of patients in the same treatment groups. The adjustment procedure might help to reduce this variation but further studies to develop the technique are necessary and should concentrate on precision.
![]() |
Notes |
---|
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
2 . Sirgel, F. A., Botha, F. J., Parkin, D. P., Van de Wal, B. W., Donald, P. R., Clark, P. K. et al. (1993). The early bactericidal activity of rifabutin in patients with pulmonary tuberculosis measured by sputum viable counts: a new method of drug assessment. Journal of Antimicrobial Chemotherapy 32, 86775.[Abstract]
3
.
Donald, P. R., Sirgel, F. A., Botha, F. J., Seifart, H. I., Parkin, D. P., Vandenplas, M. L. et al. (1997). The early bactericidal activity of isoniazid related to its dose size in pulmonary tuberculosis. American Journal of Respiratory and Critical Care Medicine 156, 895900.
4 . Mitchison, D. A. & Sturm, W. A. (1997). The measurement of early bactericidal activity. In Bailliere's Clinical Infectious Diseases: Mycobacterial Diseases Part II, (Malin, A. & McAdam, K. P. W. J., Eds), pp. 185206. Bailliere Tindall, London.
5 . Mitchison, D. A. (1992). The Garrod Lecture. Understanding the chemotherapy of tuberculosiscurrent problems. Journal of Antimicrobial Chemotherapy 29, 47793.[ISI][Medline]
6 . Mitchison, D. A. (1993). Assessment of new sterilizing drugs for treating pulmonary tuberculosis by culture at 2 months. American Review of Respiratory Disease 147, 10623.[ISI][Medline]
7 . Odhiambo, J. A., Borgdorff, M. W., Kiambih, F. M., Kibuga, D. K., Kwamanga, D. O., Ng'ang'a, L. et al. (1999). Tuberculosis and the HIV epidemic: increasing annual risk of tuberculous infection in Kenya, 19861996. American Journal of Public Health 89, 107882.[Abstract]
8 . Galloway, M. R. (1998). Results of the eighth national HIV survey of women attending antenatal clinics of the public health service in South Africa in 1997. AIDS Bulletin (SAMTC) 7, 167.
9 . Report. (1997). Fourth annual report August 1996 to July 1997. Advisory Council on Aids, Hong Kong.
10 . Dean, A. G., Dean, J. A., Coulombier, D., Brendel, K. A., Smith, D. C., Burton, A. H. et al. (1994). Epi Info. Version 6: a word processing, database and statistics program for epidemiology on microcomputers. Centers for Disease Control and Prevention, Atlanta, GA, USA.
11 . Mandel, W., Cohn, M. L., Russell, W. F. & Middlebrook, G. (1957). Serum isoniazid levels and catalase activities of tubercle bacilli from isoniazid-treated patients. American Journal of the Medical Sciences 233, 668.[ISI][Medline]
12 . Mitchison, D. A. (2000). Role of individual drugs in the chemotherapy of tuberculosis. International Journal of Tuberculosis and Lung Disease. 4, 2627.[ISI][Medline]
13 . Tam, C. M., Chan, S. L., Kam, K. M., Sim, E., Staples, D., Sole, K. M. et al. (2000). Rifapentine and isoniazid in the continuation phase of a 6-month regimen. Interim report: no activity of isoniazid in the continuation phase. International Journal of Tuberculosis and Lung Disease. In press.
14 . Chan, S. L., Yew, W. W., Ma, W. K., Girling, D. J., Aber, V. R., Felmingham, D. et al. (1992). The early bactericidal activity of rifabutin measured by sputum viable counts in Hong Kong patients with pulmonary tuberculosis. Tubercle and Lung Disease 73, 338.[ISI][Medline]
15 . Fox, W., Ellard, G. A. & Mitchison, D. A. (1999). Studies on the treatment of tuberculosis undertaken by the British Medical Research Council tuberculosis units, 19461986, with relevant subsequent publications. Section 1.26.5. International Journal of Tuberculosis and Lung Disease 3, Suppl. 2, S231S79.[ISI][Medline]
16
.
Hafner, R., Cohn, J. A., Wright, D. J., Dunlap, N. E., Egorin, M. J., Enama, M. E. et al. (1997). Early bactericidal activity of isoniazid in pulmonary tuberculosis. Optimization of Methodology. The DATRI 008 Study Group. American Journal of Respiratory and Critical Care Medicine 156, 91823.
17
.
Sirgel, F. A., Botha, F. J., Parkin, D. P., Van de Wal, B. W., Schall, R., Donald, P. R. et al. (1997). The early bactericidal activity of ciprofloxacin in patients with pulmonary tuberculosis. American Journal of Respiratory and Critical Care Medicine 156, 9015.
18
.
Desjardin, L. E., Perkins, M. D., Wolski, K., Haun, S., Teixeira, L., Chen, Y. et al. (1999). Measurement of sputum Mycobacterium tuberculosis messenger RNA as a surrogate for response to chemotherapy. American Journal of Respiratory and Critical Care Medicine 160, 20310.
Received 14 April 1999; returned 15 July 1999; revised 8 December 1999; accepted 22 January 2000