INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The rate at which an antituberculosis drug kills Mycobacterium tuberculosis in the cavities of pulmonary tuberculosis during the first 2 d of therapy is termed the early bactericidal activity (EBA) of the drug (1). EBA is thought to reflect the efficacy of the agent against rapidly metabolizing and multiplying bacilli present in the well-aerated walls of tuberculosis cavities, since organisms that multiply slowly would also be killed slowly, and would not contribute appreciably to the early bacterial death rate. The EBA can be used to compare the potencies of a new drug and an established drug, and to determine the lowest effective dose size of both drugs (2).

Since its introduction as an antituberculosis agent in 1952, isoniazid (INH) has been the principal component of all antituberculosis regimens for use in patients with drug-sensitive organisms. It is particularly valued for its efficacy in treating tuberculosis, for preventing the emergence of drug-resistant organisms, and for its low toxicity and cost. In several studies, INH has been shown to have a greater EBA than any other known antituberculosis agent (1). No association could, however, be shown between dose sizes of INH in the range of 150 to 600 mg and EBA (1), suggesting that the conventional INH dosage of 300 mg was well above the minimal effective dose. In the study described here we explored the EBA of a range of INH dosages in patients with pulmonary tuberculosis and its relationship to a number of clinical factors and to the acetylator genotype of the patient.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Most of the patients admitted to the study were of mixed race (black with additional Malay, European, and Khoi-San parentage), and a minority were black. The patients' ages ranged from 18 to 60 yr; they weighed between 40 and 60 kg and had newly diagnosed pulmonary tuberculosis confirmed by the presence of acid-fast bacilli (AFB) on direct sputum-smear microscopy. They had not been previously treated for tuberculosis, and antituberculosis treatment is not available outside hospitals and local health authority tuberculosis clinics in South Africa. Patients with other complicating medical conditions, those in poor general condition, and pregnant women were excluded from the study. Patients were admitted to Tygerberg Hospital, Cape Town, for the 3 d of the study, and after discharge were given standard chemotherapy then recommended by the South African Tuberculosis Control Programme, consisting of INH, rifampin, and pyrazinamide in a fixed combination tablet. The study protocol was approved by the Ethics Committee of the Faculty of Medicine of the University of Stellenbosch, and all patients gave written informed consent for their inclusion in the study.

INH Regimens

The study was undertaken in three overlapping phases lasting 19 mo in all. In Phase 1, patients were randomly assigned to receive INH in a dosage of 150 mg, 75 mg, 37.5 mg, or 18.75 mg. In Phase 2, patients were randomly assigned to receive 9 mg or 600 mg INH. In Phase 3, which formed part of another study (4), but was conducted at the same time as Phase 2, patients were randomly assigned to receive 300 mg INH or no treatment (nil group). INH was administered at least 60 min before breakfast in the morning following an overnight fast.

Sputum Samples

Patients were actively encouraged to cough, and a 16-h collection of sputum was made between 4:00 P.M. on the day of admission and 8:00 A.M. the next day (S1 sputum sample). INH was administered soon after 8:00 A.M. on the following 2 d, and the sputum-collection procedure was repeated to give S2 and S3 sputum samples following the first and second doses of INH, respectively. Sputum specimens were sent by air courier to Pretoria for analysis in the laboratories of the South African National Tuberculosis Research Programme. The results for any patient producing less than 5 ml of sputum in a collection were excluded from further analysis.

Microbiologic Evaluation

Smear examination, culture, and sensitivity testing were done by conventional methods. Counts of colony-forming units (cfu) were done as described previously (2), except that after mixing 2 ml of homogenized sputum with 3 ml 1:10 dithiothreitol (DTT; Sputolysin; Hoechst, Cape Town, South Africa), 20 µl of serial 10-fold dilutions were spread on thirds of triplicate plates of selective 7H10 medium (Ditco, MI). INH resistance did not develop in the organisms from any patient during the 3 d of study.

Plasma Concentrations of INH

The concentration of INH in plasma was measured with a high-performance liquid chromatographic method (5), using plasma samples taken from patients at 3 h after dosing.

Genotyping of Patients' Acetylator Status

Genotyping of the pNAT2 gene, mutations in which control the rate of acetylation of INH, was done on samples of patients' blood through methods described elsewhere (6). In brief, mutations in the gene were detected by polymerase chain reaction (PCR) amplification of a 1,000-bp fragment of the gene with the Nat-Hul4/Hul6 primer pair (Genosys Biotechnologies Inc., Cambridge, UK), followed by cleavage with the restriction enzymes Msp I, Fok I, Kpn I, Taq I, Dde I, and BamHI, and allele-specific amplification. Fragments were resolved in 3% Metaphor agarose (FMC Bioproducts, Rockland, ME), and were classified according to the system of Vatsis (7). The alleles for rapid acetylation were NAT2*4, NAT2*12A, and NAT2*13, whereas those for slow acetylation were NAT2*5A, NAT2*5B, NAT2*6A, NAT2*6B, NAT2*6C, NAT2*7B, NAT2*14A, and NAT2*14B. Homozygous rapid acetylators had two rapid alleles (RR); heterozygous rapid acetylators had one rapid and one slow allele (RS), and slow acetylators had two slow alleles (SS).

Statistical Evaluation

The cfu counts on sputum, calculated as log₁₀ cfu/ml sputum, were expressed as scores, z₁, z₂, and z₃ for the S1, S2, and S3 collections, respectively. The significance of differences z₁  z₂, z₂  z₃ and Q = (z₁  z₂)  (z₂  z₃) were tested with one-sample t tests. The data fitted the hypothesis that E(Q) = 0, which is essentially a test of the linearity of the trend in z during the 2-d period of patient evaluation, supporting the calculation of EBA as (z₁  z₃)/2. A regression model of EBA on log dose (D) was developed that satisfied the constraints that when D = 0, E(EBA) should also = 0, and that E(EBA) should be bounded. This implies that the regression curve should have the characteristic shape of a logistic or probit curve, passing through the 0,0 coordinate. Although there is a suggestion that the variance of EBA in the groups decreased with the mean EBA, only the variance at 18.5 mg INH is significantly smaller than at other dose levels. There also appears to be an increase in variance between the 18.75-mg and the 9-mg dose groups, but this may have been a characteristic of unusual wide scatter of results in the 9-mg group (see RESULTS and DISCUSSION). Bartlett's test for homogeneity of variance of the seven groups gave a nonsignificant result (p = 0.2), although the test is insensitive to changes in variance. The best assumption to be made in fitting a curve was that variance is homogeneous. The curve was then fitted by the method of least squares. The fitted model is E(EBA) = a[exp (bD)]. Its parameters were estimated by the method of least squares. A one-way analysis of variance (ANOVA) of the EBA values in the seven groups given doses of 9 mg to 600 mg isoniazid gave an interpatient, intragroup variance of 0.046 on 68 degrees of freedom (df). The estimated residual variance from the least-squares fit of the model is 0.0498. An approximate F test for goodness of fit gives F = 1.90 with 5 df and 67 df, whence p = 0.11, indicating an acceptable fit. A multiple regression analysis was done with the Epi Info 6.02 package (U.S. Centers for Disease Control and Prevention) (8).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

In all, 105 patients were entered into the study, of whom 20 (19%) were subsequently excluded. Three patients had bacterial contamination of their sputum cultures, nine patients showed no growth or very poor bacterial growth, three patients had cultures of M. tuberculosis initially resistant to INH, four patients produced less than 5 ml sputum on one or more occasions, and one patient's specimens were lost in transit. The remaining 85 patients had a mean age of 30.9 yr and a mean weight of 50.9 kg. Eighty-one patients (95%) had multicavitary disease on chest radiography, and 80 (94%) had disease that occupied an area the size of the right upper lobe or greater. Table 1 summarizes the patients' age, gender, and weight; radiographic extent and severity of disease; and smear positivity of sputum in the three phases of the study. No statistically significant differences in any of the characteristics were found in the different dosage groups between or within the three phases of the study. Table 2 shows the S1, S2, and S3 counts of cfu/ml sputum in the dosage groups. The dose is expressed as its size in milligrams and also as mg/kg body weight, using the mean weight of each dosage group. The S1 counts ranged from 4.531 log₁₀ cfu/ml (3.40 × 10⁴) to 7.724 log₁₀ cfu/ml (5.30 × 10⁷), with similar SDs in each group and a pooled intragroup SD of 0.656. The patient's combined S1 counts were unrelated to gender (males: 4.656 log₁₀ cfu/ml, SD = 0.83; females: 4.952 log₁₀ cfu/ml, SD = 0.69) or body weight, but were related to the radiographic extent of disease (p < 0.001) and cavity size (p < 0.001). Mean sputum volumes for the S1 specimens (27.3 ml), S2 specimens (27.4 ml), and S3 specimens (26.2 ml) did not differ significantly. In the seven groups given isoniazid, the mean changes between the S1 and S2 cfu counts and between the S2 and S3 counts (Table 2) did not differ significantly in any of the dosage groups or overall. The S1 count was significantly lower in the 9-mg INH group than in any of the other seven groups, whose S1 counts appeared homogeneous.

Early Bactericidal Activity

Table 3 and Figure 1 show the EBAs in the different INH dosage groups, calculated as the mean daily decrease in cfu counts/ml sputum during the 2-d period (i.e., [S1 log₁₀ cfu/ml  S3 log₁₀ cfu/ml]/2). The EBA increased with each 2-fold increment in dose size up to 300 mg INH, but no further increase occurred with the 600-mg dose. The results with the 9-mg dose are surprising in that the mean EBA of 0.129 was below the value of 0.011 for the nil group. Furthermore, there were an unusual number of large negative EBA values in the 9-mg dose group; EBA values of less than 0.2 (0.72, 0.45, and 0.37) were found for three of 11 patients but for none of the 10 nil-group patients. There was also a trend for the variation in the groups, shown by the standard deviation (SD), to decrease with dose size from a value of 0.23 for the pooled results in the groups given 150 mg, 300 mg, or 600 mg INH to 0.05 for the nil group, as was found in an earlier study (2). However, the SD of 0.29 in the 9-mg group was an exception to this trend, again suggesting that the response to this low dose was unusual. Small, nonsignificant associations were found between the EBA and the S1 count in each of the seven groups given from 18.75 mg to 600 mg INH. A multiple regression of log INH dosage and log S1 count on the EBA in these seven groups indicated a highly significant effect of INH dosage (p < 0.000001), and a significant (p < 0.01) effect of the S1 count, the slope of the regression line indicating a rise of 0.094 (95% confidence limits of 0.029 to 0.158) in the EBA for each 10-fold increase in the S1 count. The pooled SD of the variation between patients, within dosage groups, was 0.21 (68 df). As shown in Figure 2, a curve was fitted to the results in the groups given INH, with its lowest value forced through the 0,0 coordinate of the graph because the mean for the nil group had only a very small negative value, the wide 95% confidence limits for the 9-mg group included zero, and it is reasonable to suppose that, with increasingly low dose size, the EBA would tend to zero. Figure 2 also shows the plotted mean EBAs found with doses of 600 mg, 300 mg, and 150 mg rifampin, obtained in an earlier study (2). Although the curve for these points is approximately parallel to the INH EBA-dose curve, the rifampin EBAs are much lower than those obtained with the same dose sizes of INH. Doses of 150 mg rifampin and 18.75 mg INH are similar in that both were the lowest that produced a definite positive EBA; a further 2-fold reduction in the dose of either drug would seem to extend the curve to an EBA of 0 or less. These concentrations, producing the lowest detectable EBAs, are for INH at some 16 times less than the usual therapeutic daily dose of 300 mg, whereas for rifampin they are only four times less than the usual dose of 600 mg. Thus, one can define a "therapeutic margin" of 16 for isoniazid and 4 for rifampin.

View larger version (14K): [in this window] [in a new window]   Figure 1.   Early bactericidal activity (EBA) of isoniazid related to dose size (log10 mg). Bars indicate SEM.

View larger version (20K): [in this window] [in a new window]   Figure 2.   Fitted curves relating early bactericidal activities (EBAs) of isoniazid (INH) and rifampin to dose size and their 95% confidence limits (shaded regions).

Plasma Concentrations of INH

The means of the 3-h plasma concentrations of INH are shown in the last four columns of Table 3. They indicate a decrease in mean concentration proportional to the size of the dose, as shown by the constant value of the mean concentration divided by dose size. Since the minimum inhibitory concentration (MIC) of INH is about 0.1 µg/ml (9), it is not surprising that the dose of 18.7 mg, which produced a mean 3-h concentration of 0.12 µg/ml, was the lowest to show a detectable EBA.

Acetylator Status

Acetylator genotypes were determined in 41 patients, including all 22 from Phase 2 of the study, in which the treatment groups were given 600 mg and 9 mg INH. The EBAs in these two dosage groups are shown in Table 4. In both groups, the mean EBAs of homozygous rapid (RR) acetylators were lower than those of heterozygous rapid (RS) or in homozygous slow (SS) acetylators. The differences in EBAs between the three acetylator groups were statistically significant by analysis of variance (ANOVA) at both dosages. There were insufficient patients in any of the remaining groups (from three to five in each group) to yield further information.

Primary INH Resistance

Three patients were excluded from the present study and another two patients from an earlier study (2) because their strains of M. tuberculosis had primary INH resistance. These patients, given 37.5 mg, 150 mg, 300 mg, 300 mg, and 600 mg doses of INH, had EBAs of 0.09, 0.005, 0.09, 0.03, and 0.029, respectively. Since a positive EBA was found for all five patients and for only two of the 10 patients in the nil group (two-tailed Fisher's exact test; p = 0.007), these results suggest slight antibacterial activity. They confirm the evidence of a small response to INH in patients with primary drug resistance (10), but lend no support to the continuing use of INH in patients with acquired resistance.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Previous estimates of the EBA of 300 mg isoniazid were 0.72, 0.43, and 0.50 in studies in Nairobi (1), Hong Kong (3), and South Africa (2), respectively, for which the mean EBA of 0.55 is the same as the mean of 0.554 in the present study. Thus, estimates of the EBA for 300 mg INH are much the same in studies at different times and in different countries. It was only during the first 2 d of treatment in the present study that the daily decreases in cfu counts were so high; during the succeeding 12 d, corresponding daily decreases in cfu counts were no greater during treatment with INH (0.113) than during treatment with other drugs (overall mean = 0.129) (1). The significant association between the INH acetylator phenotype and the EBA found in patients given 600 mg or 9 mg INH needs further confirmation, but suggests that the bactericidal effect of INH during the first 2 d of treatment is critically dependent on the size of the pulse of drug that reaches the bacilli. Thereafter, the rate of kill is likely to be determined by the physiologic state of the remaining viable bacilli after the actively growing component has been killed, and in the Nairobi study (1) this rate remained the same irrespective of the drug examined or its dose size. It is of interest that a significant though small association was found between the S1 count and the EBA in groups treated with INH, as also occurred in an earlier study (2). The increase in EBA might be due to a greater rapidity of growth and therefore the bactericidal activity in cavities with large bacterial populations.

The similarity of the EBAs for INH at doses of 300 mg and 600 mg suggests that as dose size is increased, a maximum value is being reached. This is consistent with the earlier findings of similar EBAs of 0.57, 0.72, and 0.48, each based on only four patients, with doses of 150 mg, 300 mg, and 600 mg, respectively (1). The EBA in the 600-mg group in the present study might have been partly the result of a chance excess of homozygous rapid acetylators (RR) in the group (i.e., five of 11 [45%] as compared with 14 of 60 (23%) in a similar patient population [6]), and a consequent reduction of the mean for the group. Alternatively, a maximal EBA might occur because some part of the poorly understood biochemical pathway that leads to the inhibition of mycolic acid synthesis (11) was becoming saturated or because a further decrease in mycolic acid synthesis might not kill the bacterial cell more rapidly. The plateau in efficacy with high dose sizes is also evident in the treatment of pulmonary tuberculosis, since studies at Madras (12) found no difference in response between rapid and slow acetylators of INH nor when the dose size of INH alone was increased from 400 mg (8.7 mg/kg, producing a favorable response in 66% and 72% of rapid and slow acetylators, respectively) to 700 mg (13.9 mg/kg, producing a favorable response in 66% and 69%, of rapid and slow acetylators, respectively). Furthermore, there was no improvement when the dose size of INH, given with 150 mg thiacetazone, was increased from 300 mg to 450 mg (15). Even if a higher EBA were found when the dose size was increased above 300 mg, the tendency for the daily decrease in cfu counts to equalize irrespective of dose size would explain why an increase in dose size had no beneficial clinical result.

The mean EBA of 0.129 with the 9-mg dose of INH was unexpectedly low, being smaller than the mean of 0.011 for the nil group, although the difference was not significant (Mann- Whitney U test, p > 0.1). The variation within the group was also greater than in any other group. These findings might have been due to some unexplained cause of greater variability at low EBA values. Unusually high variability was not found in the 600-mg INH group in the present study, which was investigated at the same time. It seems unlikely that subinhibitory concentrations of INH might actually stimulate bacterial growth, since incorporation of INH in culture medium has been found to improve growth of M. malmoense and other nontuberculous mycobacteria but not M. tuberculosis (16, 17).

The therapeutic margin (i.e., the ratio of the usual therapeutic dose to the lowest dose that has a definite EBA) was found to be 16 for isoniazid and 4 for rifampin. The difference cannot be attributed to the sizes of the usual dose (300 mg for INH and 600 mg for rifampin) or to the MICs (0.1 µg/ml for INH and 0.3 µg/ml for rifampin [9]) of the two drugs. It seems most likely to be due to protein binding, which does not occur with INH but, at 85%, may substantially reduce the lesional concentrations of free rifampin (9). The concept that protein binding reduces the free drug concentration available in lesions is supported by results of a clinical trial showing that changing the dose of rifampin from 600 mg to 450 mg reduced the efficacy of rifampin (18). The therapeutic margin is clearly a measure of considerable value in assessing the activity of a drug, since a large margin implies that the drug will be available in adequate concentrations despite inevitable differences in the concentrations in individual lesions and in areas within lesions. INH may well prove to be uniquely valuable among antituberculosis drugs in having such a large therapeutic margin due to its low MIC, absence of protein binding, and low toxicity.

Conclusions of clinical importance can be drawn from the foregoing findings. The first point concerns dose size. As considered earlier, an increase in dose size is unlikely to be clinically beneficial. However, the high therapeutic margin suggests that INH might retain its efficacy if given in a lower than usual dose size. Clinical studies have shown that when INH is given alone, its efficacy is greatly affected by dose sizes at and below 400 mg, but this seems due to the suppression of INH-resistant mutants (12, 19). When INH is given in conventional combined therapy, however, its activity against its own resistant mutants is unlikely to influence the overall efficacy of the regimen, since these mutants are inhibited by the companion drugs. Nevertheless, other factors as well as the EBA of a drug determine its efficacy. Pulses of INH have a cumulative postantibiotic effect on bacilli that depends on the size (area under the curve [AUC] of plasma drug concentration) of the pulse and inhibits growth for up to about 7 d after the pulse (20, 21). In the limiting circumstances of once-weekly dosage, treatment failure among rapid acetylators seems to occur because the pulse is not large enough to produce a maximal postantibiotic effect (14, 20). Although small doses of INH (18.75 mg) have a detectable EBA, this does not imply that the resulting pulses in rapid acetylators given daily dosing would be large enough to prevent regrowth during the interval between doses when plasma concentrations were subbacteriostatic. Thus it would be unwise to recommend a reduction in the conventional size of the dose of INH in the treatment of pulmonary tuberculosis without evidence from clinical studies that such a reduction would decrease toxicity without diminishing efficacy.

The second point concerns drug penetration into large sequestered sites, such as tuberculous abscesses in bone and joint disease, accumulations of pleural pus, and the brain and cerebrospinal fluid. As a result of slow penetration, the peak concentrations of drugs in sequestered sites are usually lower than in plasma, although the pulse is prolonged so that the total amount of drug that penetrates is not reduced (22). However, when the volume of the sequestered site is large, the peak concentration within the site may be so low that it is effectively below the MIC, and the drug may then be inactive. Such inactivity is much more likely to occur with drugs that have low therapeutic margins than with those having higher margins. Even if low INH peaks were to occur, the duration of the pulse would be prolonged, so that the AUC would remain broadly unchanged and sufficient for antibacterial action (23, 24). As an example, measurements of drug concentrations in a patient with a tuberculous empyema who had been treated with INH, rifampin, and ethambutol suggested that the peak concentrations of rifampin and ethambutol in the pleural fluid may have been too low to have had an antibacterial effect (25). No measurements were made of INH concentrations, but the strain of M. tuberculosis obtained after treatment was resistant only to INH, demonstrating its unique activity. Thus, INH in conventional dose size should retain its activity even in large sequestered lesions, so that sequestration is probably not an indication to increase the dose size.

The third point concerns the effects of individual variation in absorption or excretion of INH. The large therapeutic margin would suggest that individual variation is of little consequence in treatment provided that the treatment is supervised and uninterrupted. The greatest test of INH activity occurs if it is given at weekly intervals, since it then tends to fail in rapid acetylators (20). Even under these limiting conditions, variation in each of the acetylator groups in INH serum concentrations, at various time points after dosing, and among large numbers of patients, has not been found to be related to the efficacy of once weekly regimens of streptomycin and isoniazid (20). Consequently, there is no scientific justification for adjusting the dose of INH according to the results of measurements of plasma concentrations under routine conditions of treatment. Furthermore, little is likely to be gained by seeking variation in plasma INH concentrations as a reason for the failure of chemotherapy in individual patients, nor is accurate adjustment of INH dosage during renal failure likely to be a worthwhile exercise.

Moreover, in the development of new drugs, it is clear that estimation of the therapeutic margin should yield important information about the minimal effective dose size and its relation to the usual therapeutic dose. This has a direct bearing on the probable efficacy of the drug and the optimal dose size to be used in treatment. The margin can only be measured if the experimental design includes dose sizes ranging from the usual to a dose so small that it yields an EBA that is barely distinguishable from 0. In contrast, a single measure of the EBA obtained with only the usual dose size of a drug, as has been employed in some EBA studies (26), is of much less value. It may give information about the bactericidal activity of a drug, but this does not indicate the therapeutic margin, since one cannot assume that curves relating EBA to dose size for different drugs are parallel. In drug development one must also remember that the size of the EBA does not indicate the sterilizing activity of a drug in killing persisting organisms (25), as is shown by the contrast between the low EBAs of pyrazinamide and rifampin and these drugs' potent sterilizing activities.