INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Full supervision of the taking of medicament has been recognized for many years as necessary to obtain high cure rates in the chemotherapy of tuberculosis (1) and more recently has been accorded high priority as directly observed treatment (DOT) by the World Health Organization. But it is difficult and expensive to carry out, particularly in rural areas. The longer the interval between doses of drug, the easier is the task of supervision so that an effective once-weekly regimen would have substantial operational advantages. Rifapentine, a rifamycin that has complete cross-resistance with rifampin, was as effective as daily rifampin when given once weekly in experimental murine tuberculosis (2, 3) and therefore offers the prospect of effective once-weekly dosage in the treatment of human tuberculosis. It has a lower minimal inhibitory concentration (MIC) against Mycobacterium tuberculosis than rifampin (4) and a much longer half-life, such that peak concentrations of 20 µg/ml or more occur at 6 h and about 6 µg/ml is found at 48 h after a 600-mg dose (5, 6). Rifapentine was first developed in 1965 by Lepetit, but for various commercial reasons, including competition with rifampin, has only recently been developed for licensing by Marion Merrell Dow (now Hoechst Marion Rousell). Clinical trials were carried out earlier in the People's Republic of China with rifapentine of Chinese manufacture (7, 8), but for various reasons, particularly the choice of regimens and the absence of good follow-up bacteriological examination of sputum, it has only been possible to conclude that its toxicity is probably no greater than that of rifampin. In 1988, because of its potential importance and because of delays in its development by Western pharmaceutical companies, the Hong Kong Tuberculosis Service (then headed by S. L. Chan) in collaboration with D. A. Mitchison decided to instigate a clinical trial of Chinese rifapentine in Hong Kong using a protocol originally developed by the British Medical Research Council's Tuberculosis and Chest Diseases Unit. Although a once-weekly regimen including rifapentine starting early in treatment would have the greatest operational advantages, the first ethically justifiable step was to explore once-weekly rifapentine and isoniazid in the continuation phase in regimens that began with a conventional 2 mo of intensive 4-drug chemotherapy. In order to assess the effects of possible poor compliance, a second regimen was included in which the once-weekly rifapentine and isoniazid was given on 2 of 3 wk. The follow-up period after stopping chemotherapy ranged from 6 to 46 mo.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

Eligible patients were Chinese, 15 yr of age or older, with newly diagnosed pulmonary tuberculosis previously treated for no longer than 2 wk or for no longer than 4 mo if treated at least 1 yr previously, and with sputum positive on culture for M. tuberculosis. Patients were excluded if they were in poor general condition, had serious nontuberculous disease, had extrapulmonary tuberculosis or large pleural effusion, or were drug addicts. All patients gave informed consent to inclusion in the study. Permission to carry it out was obtained from the Hong Kong Government Ethics Committee.

Regimens

After admission to the study, the patient's name was entered into a register held at Wanchai Polyclinic and treatment was allocated at random from slips within sealed envelopes identified externally by the registration number. Thereafter, the study was open-label. All patients were given thrice-weekly streptomycin, isoniazid, rifampin, and pyrazinamide (SHRZ₃) for the first 2 mo, which is standard therapy in Hong Kong outpatient clinics. The regimens then continued with a 4-mo period of either:

1. HR₃, the regimen of rifampin and isoniazid given thrice weekly throughout the service, standard therapy in Hong Kong.
2. HRp₁, rifapentine and isoniazid once weekly.
3. HRp₁.2/3, the HRp₁ regimen with every third dose of both drugs omitted (that is, given 2 wk out of three) to simulate poor compliance.

Drug dosages were 600 mg isoniazid and 2 g pyrazinamide for patients under 43 kg body weight, 800 mg isoniazid and 2.5 g pyrazinamide for those of 43 to 57 kg weight, and 1,000 mg isoniazid and 3 g pyrazinamide for those over 57 kg. Doses of 1 g streptomycin, 600 mg rifampin and 600 mg rifapentine (in phase I of the study) were given irrespective of body weight. The rifapentine was from four batches produced by the Shanghai An Yan factory. Before the start of the study, the identity and purity of the rifapentine was established by high-pressure liquid chromatography and its bioavailability was found to be the same as a batch manufactured by Merrell Dow Lepetit (MDL) in volunteers (5). As reported elsewhere (6), the rifapentine content of serum specimens from 287 patients was measured in microbiologic diffusion assays since more accurate methods could not be afforded on this number of specimens. In the first batch, a randomized comparison of the curves obtained with specimens taken at 4, 6, 8, 24, and 48 h after a 600-mg dose unexpectedly showed that the mean of the areas under the curve (AUC) for 20 patients given Chinese drug was only 73% of the mean AUC of a further 20 patients given MDL drug (t test, p < 0.01). Further assays of sera taken at 8 h and 24 h after the first and second batches gave further evidence that these batches were poorly absorbed in that serum concentrations were 74% and 66%, respectively, of those obtained with the MDL batch. The first two batches were given to 417 (62%) of the patients in a dose of 600 mg in phase I, whereas the last two batches were given throughout treatment to 189 (28%) patients in phase II in doses of 750 mg (and briefly 900 mg) when the serum concentrations were 90 to 110% of those initially found with the MDL drug. The change from phase I to phase II took place on a fixed date so that 66 (10%) patients were treated initially with the low dose and then with the higher dose. The rifapentine, in gelatin capsules, was stored at 4° C until given to patients and, as shown by assays on sera taken from 91 patients throughout the life of a batch, retained its activity unchanged (6). Patients attending clinics took triple-decker fast food sandwiches, containing a fried egg and ham, before their rifapentine or, if in hospital, a meal of eggs and toast; both meals had been shown to stimulate adequate absorption (5). The taking of all drug doses was fully supervised. Urine samples were taken at 2 to 8 h after a drug dose at random during the months of treatment and also after treatment during Months 7 to 10. The taking of drugs was monitored by testing these samples for isonicotinic acid, a metabolite of isoniazid (9), and for rifampin by microbiologic assay (10).

Admission and Review of Patients

Most patients were admitted and treated in the 10 outpatient clinics (see ACKNOWLEDGMENT) but those with serious side effects were usually treated in hospital. Patients were reviewed every calendar month until 24 mo, then every 3 mo until Month 36 and every 6 mo until 5 yr after admission. Three sputum specimens were collected pretreatment and then two specimens at Months 2, 3, 4, 5, and 6 during treatment. Thereafter single specimens were obtained monthly up to 24 mo, and then pairs of specimens at clinic attendances every 3 mo and then every 6 mo to 5 yr. Sputum (never induced or obtained at bronchoscopy) was examined by fluorescence microscopy of direct smears and culture on Lowenstein-Jensen medium slopes; mycobacteria were identified and tested for sensitivity to streptomycin, isoniazid, rifampin, and rifapentine (11, 12). These examinations were usually done at the Yung Fung Shee Memorial Centre, Hong Kong but a single pretreatment culture from most patients was also sent to London for the same examinations. Chest radiographs were taken pretreatment and at 6-mo intervals thereafter up to 5 yr. Radiographic gradings and relapses were reviewed by the senior chest physician, Dr. S. L. Chan. Relapses and deterioration were defined as clear-cut increases in shadow around the original site of the lesion, whose size had decreased or remained stable at previous examinations and after exclusion of other causes for the increase. Every adverse side effect was recorded. Alanine transferase (ALT) was measured monthly during treatment to assess liver function.

Statistics

The clinical and laboratory data were recorded on handwritten analysis cards and entered in EPI INFO version 6.01 (CDC, Atlanta), which was used for all analyses. The Kaplan-Meier life table analysis was done with the STATA package (Stata statistical software, release 4; College Station, TX). Abbreviations for statistical terms are: 95% CL for the 95% confidence limits; ² for the Yates ² corrected for continuity, and U-test for Mann-Whitney nonparametric rank test.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Between December 1991 and July 1995, 672 patients were admitted to the study, 633 from 10 outpatient clinics and 33 from hospitals. Of these, 75 did not fulfill the admission criteria; 53 had negative pretreatment cultures, eight were infected with nontuberculous mycobacteria, eight had other serious diseases (cancer; active hepatitis B), three had extrapulmonary tuberculosis, and three were excluded for miscellaneous reasons (previous treatment; drug addiction; non-Chinese) (Table 1). A further two patients (1 HR₃ and 1 HRp₁) who died from active tuberculosis during the first month and from hemoptysis in the second month were excluded because early death cannot be considered as an event influenced by chemotherapy. There were three patients (all HRp₁) who were analyzed separately from the main population because they had organisms resistant to rifampin and rifapentine which are key drugs in obtaining satisfactory treatment; the failure and relapse rates after the end of daily short-course treatment are little altered by initial resistance to isoniazid or streptomycin but are greatly increased by initial rifampin resistance (13). The 49 patients with initial resistance to isoniazid or streptomycin were included with those having initially sensitive organisms. There remained 592 patients, 190 HR₃, 199 HRp₁, and 203 HRp₁.2/3, in various categories set out in Table 1 to be considered in the main analyses of efficacy and side effects. Included in the 592 patients were 46 who had minor irregularities in their chemotherapy (17 HR₃, 15 HRp₁, 14 HRP₁.2/3). In 15 (8 HR₃, 3 HRp₁, 4 HRp₁.2/3), treatment was stopped for periods of 3 to 5 wk and the trial regimen was then prolonged during Month 7. 

Efficacy

The condition on admission of the 592 patients (Table 2) shows that sputum culture yielded more than 100 colonies in 56% of the patients, the extent of radiographic disease was more than the area of the right upper lobe in 43%, and cavitation was present in 37%. The proportions of patients with each assessment were similar in the three treatment series as were the age and sex distributions.

An intention to treat analysis was done first using a Kaplan-Meier life table analysis (Figure 1) commencing at the start of treatment when regimens were randomized. An adverse treatment event (Table 3) was defined either as (1) failure during treatment in which at least two positive cultures were found at 5 or 6 mo or because of serious radiographic deterioration, or (2) as bacteriologic or radiographic relapse after the end of treatment. Included as a failure was an HRp₁ patient who had a radiographic deterioration at 6 mo with negative sputum cultures; his treatment was restarted thereafter and further sputum specimens were negative. The two failures in the HRp₁.2/3 series (Table 3) occurred in patients whose treatment was interrupted for several months because of drug toxicity, starting during the first 2 mo; neither received any of the rifapentine regimen. No other failures occurred during treatment. The minimum requirement for a bacteriologic relapse was two positive cultures of at least 5 colonies in two separate monthly sputum collections during a 3-mo period. The follow-up period after the end of chemotherapy in patients who continued to attend ranged from 6 to 48 mo, the median in the three regimens being 31 mo. Among patients who should have had at least a 12-mo follow-up and who did not have a relapse, there were 20 whose follow-up period was less than 12 mo, due to default or death (Table 1). The last 33 patients admitted to the study had a follow-up of less than 12 mo and the last nine had 6 mo. Adverse treatment events occurred in 7 (3.7%) of 190 HR₃ patients, in 17 (8.5%) of 199 HRp₁ patients, and in 21 (10.3%) of 203 HRp₁.2/3 patients (p = 0.04 for HR₃ versus HRp₁, p = 0.01 for HR₃ versus HRp₁.2/3, and p = 0.6 for HRp₁ versus HRp₁.2/3, log rank test).

View larger version (12K): [in this window] [in a new window]   Figure 1.   Kaplan-Meier plot of time to adverse treatment events. - - - - - - -: HR3 regimen (control); : HRp1 regimen (regular once-weekly);    : HRp1.2/3 regimen (doses on 2 of 3 wk).

A subanalysis was carried out on the 534 patients who received the full trial regimen, including the 46 who had minor irregularities in their therapy. This gives a better estimate of the efficacies of the chemotherapy regimens, as it does not include the 2 HRp₁.2/3 patients whose treatment was changed during the first 2 mo for toxicity. They would not have been included in the intention to treat analysis had randomization taken place after the first 2 mo (as would have been done in a more appropriate trial design). At 2 mo when all patients had received the same treatment, 11.9% of the 534 patients had a positive sputum culture. At 3 mo, a positive culture was obtained from 5 (2.9%) of 172 HR₃ patients and from 4 (1.1%) of 372 patients on the two rifapentine regimens, a nonsignificant difference. Of the nine patients with 3-mo positive cultures, three (1 HR₃, 2 HRp₁) had bacteriologic relapses at 16, 14, and 9 mo, respectively. Thereafter only three patients had positive sputum cultures; two (HR₃) had isolated positive cultures at 4 and 5 mo while one (HRp₁.2/3) converted slowly with positive cultures of 2 colonies and 1 colony at 4 and 5 mo but did not relapse during 27 mo of follow-up. All of the 42 relapses occurred after the patients had received the allocated trial regimen, in seven (4.1%) of the 172 HR₃ patients, in 16 (8.9%) of the 179 HRp₁ patients, and in 19 (10.4%) of the 183 HRp₁.2/3 patients (Table 3). A high proportion (74%) of relapses occurred during the first 6 mo of follow-up (Figure 2), the numbers being 5, 10, and 16 in the HR₃, HRp₁, and HRp₁.2/3 series, respectively. There is a strong suggestion that they occurred more rapidly in the two rifapentine series than in the HR₃ series. Thus, among relapses during the first 12 mo of follow-up, the median months from the end of chemotherapy to relapse were 4.0 in the HR₃ series and 2.5 in the combined rifapentine series (U-test, p = 0.07) while if the HRp₁ failure patient who had a radiographic relapse at 6 mo is included, the medians are unchanged but the difference is close to statistical significance (U-test, p = 0.06).

View larger version (36K): [in this window] [in a new window]   Figure 2.   The speed with which relapses occurred during follow-up in the three treatment series. Months of follow-up are calculated by subtracting 6.5 from the month of observation in which the relapse was first noted either by sputum culture taken at the time or radiographically.

A direct comparison in the rifapentine series of relapse rates in phase I with those in phase II, when the dose of rifapentine had been increased, is inadvisable because the comparison is not concurrent or randomized. However, a comparison of the rates in the two rifapentine series with those in the control HR₃ series, in which the rifampin dose remained unaltered (Table 4) shows that the relative risk of a relapse or failure was 3.77 in phase I and 2.25 in phase II when calculated by proportions and 4.37 and 2.36, respectively, when calculated as rates (in life table terms as patient months, omitting the period between the two phases). The suggestion that the higher dose diminished the chance of relapse or failure was not however statistically significant as is evident from the substantial overlap of the 95%CLs.

A comparison of the drug sensitivity patterns of the 40 relapse cultures with the pretreatment cultures showed that 36 patients had the same pattern, while in four patients one or more of the relapse cultures had additional resistances, in two to isoniazid and in two to streptomycin.

Justification for the inclusion of patients with initial resistance to isoniazid and streptomycin in the relapse analysis is provided by the finding that the relapse rate was not altered by initial resistance to isoniazid or streptomycin or both drugs (Table 5). Thus, in the HR₃ series, relapse occurred in 7 of 156 patients with initially sensitive organisms and in none of 16 with initial resistance, while in the two rifapentine series combined, 33 (10.0%) of 329 patients with sensitive organisms relapsed or failed as compared with 3 (9%) of 33 with initial resistance to isoniazid or streptomycin. For comparison with other studies, the relapse rates in patients with initially sensitive organisms were 4.5% in the HR₃ series, 8.7% in the HRp₁ series, and 10.7% in the HRp₁.2/3 series.

The proportions with marked radiographic improvement were 22% of 169 HR₃ patients, 25% of 177 HRp₁ patients, and 30% of 181 HRp₁.2/3 patients (²_[2] = 2.4, p = 0.3). Only 3.6% of patients failed to show improvement during the 6 mo of treatment.

Regularity of drug taking was confirmed by the results of urine tests. Positive results on the 776 specimens taken during treatment were found in 98.1% of tests for isoniazid and in 89.0% of tests for rifampin. The proportions of negative results on 999 specimens taken after treatment were 98.7% for isoniazid and 96.6% for rifampin.

Patients with Organisms Initially Resistant to Rifampin

There were three patients (all HRp₁) with initial rifampin resistance who were considered a separate group in the efficacy analysis (Table 1). The treatment of one of these, who had a strain resistant to isoniazid and streptomycin as well, was changed in the third month with an eventual successful outcome. In a second patient, high ALT values led to a change of treatment for drug toxicity. The third patient, who had initial resistance to rifampin (and rifapentine) only, failed during treatment with both cultures at 6 mo positive on smear and culture, and subsequently at each month until 10 mo, when retreatment was successful.

Adverse Side Effects

The same 592 patients in the intention to treat analysis were included in the adverse side effects analysis. The most common adverse effects were nausea, vomiting, dizziness, rash, and pruritis. The most serious were abnormally high results in ALT tests. Most patients with side effects had several different coded entries (nausea was one and vomiting another) at each month. Of the total of 1,006 side effects noted, the great majority occurred in the first 2 mo when all patients received the same treatment, 501 in the first month and 273 in the second month, whereas the numbers in the continuation phase were only 98, 52, 53, and 29 in Months 3, 4, 5, and 6, respectively. Since many of the side effects during Months 3 to 6 started with symptoms in the first 2 mo, a side effect in Months 3 to 6 was considered to be "new" only when it had not occurred in the patient in Month 1 or 2. This procedure reduced the numbers of new side effects to 54 in the HR₃ regimen, 34 in the HRp₁, and 25 in the HRp₁.2/3 regimen. The numbers of patients (as distinct from numbers of side effects) who had side effects (Table 6) shows that side effects of any type occurred in 27 (14%) of 190 HR₃ patients, in 18 (9%) of 199 HRp₁ patients, and in 19 (10%) of 203 HRp₁.2/3 patients. The relative risk of side effects in the HR₃ series seemed greater than in the pooled HRp₁ and HRp₁.2/3 series, being 1.54 (95% CL 0.97 to 2.46, p = 0.09) just failing to achieve statistical significance. Table 6 includes a comparison of the side effects during phase I of the study when the dose of rifapentine was 600 mg (low) and phase II when the dose was increased to 750 mg and briefly to 900 mg (high) but omits patients admitted between the two phases. The comparison shows that an adverse event occurred about twice as frequently in phase II than in phase I in the HR₃ series, in which there was no change in dose size, as well as in the two rifapentine regimens. This suggests that the higher incidence of adverse events in phase II of the rifapentine regimens was probably due to factors unrelated to rifapentine dosage; the comparison between the phases was not concurrent. The "flu" syndrome attributed to rifampin dependent antibodies occurred in two patients as an isolated phenomenon, in one patient (HR₃) in the fourth month and in the other (HRp₁, phase II) in the fifth month, but treatment was not modified in either.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main finding of the study has been the unsatisfactory efficacy of the HRp₁ regimen as shown by the significant difference (p = 0.04) between the occurrence of adverse treatment events in the series as compared with the HR₃ control series. Furthermore, the only patient who failed during treatment was in the HRp₁ series, and there is a strong suggestion that relapses occurred more rapidly in the two rifapentine series combined than in the HR₃ series, suggesting that they were sometimes starting during treatment with rifapentine. The proportion of relapses after treatment was 8.9% in the HRp₁ series and 3.7% in the HR₃ series, the corresponding proportions for patients with initially sensitive organisms being 8.7% and 4.5%, respectively. The high relapse rate after the HRp₁ regimen is unacceptable in view of the rate of 4.1% of 832 patients in the most recent previous trial of antituberculosis therapy in Hong Kong (14). It is tempting to attribute these unsatisfactory results to the poor bioavailability of the rifapentine which produced serum concentration during phase I of the trial of only 66 to 74% of those obtained with a high-quality Western product (6). However, the patients regularly took meals that had been tested and found adequate to promote maximal absorption (5), while under different trial conditions or in routine practice patients might take unsatisfactory meals resulting in only about 70% of optimal serum concentrations (5) or not take a meal at all. Thus the results may well be representative of what might happen under the less rigorous conditions of mass use of drug with better bioavailability. Relevant information should also be available from results of the current Centers for Disease Control (U.S. Public Health Service) trial which is comparing regimens of HR daily and HRp once-weekly in the continuation phase. In the meantime, there is an urgent need for a study of the effects of increasing the dose size of rifapentine. In the Singapore study of intermittent isoniazid and rifampin, started after only 2 wk of daily treatment, the failure rate during 12 mo of treatment with once-weekly dosage decreased from 5% of patients getting 600 mg rifampin to only 2% when the dose size was increased to 900 mg (15). Thus, an increase in dose size of a well-absorbed preparation of rifapentine from 600 mg to only 750 mg might be adequate to ensure a satisfactorily low relapse rate even when given with meals less effective than in the present study in ensuring maximal drug absorption. This contention is supported by the lower values in phase II than in phase I of the relative risks of relapse in the HR₃ and the rifapentine/isoniazid regimens, though the phase II results are based on insufficient patients and too short a follow-up for definite conclusions. The comparison will be repeated when phase II patients have been followed further. The low incidence of adverse side effects in the rifapentine regimens in both phases of the present study suggests that a small increase in dose size would probably be well tolerated. In contrast to the differences in relapse rates, there was no suggestion of any difference between the regimens in sputum culture results at 3 mo, when the trial regimens had been given for only their first month. However, of the nine patients who had a positive culture at 3 mo, three (33%) subsequently relapsed, indicating that late conversion confers a higher than usual risk of relapse.

Although the high relapse rates suggested that the HRp₁ regimen was not entirely satisfactory in sterilizing the lesions during the continuation phase, the HRp₁.2/3 regimen in which every third dose was omitted was effective in preventing failures during treatment and had a relapse rate of 10.4%, only slightly higher than 8.0% after the HRp₁ regimen. Thus once-weekly rifapentine and isoniazid appears to be a robust regimen whose efficacy is not appreciably reduced by occasional poor compliance. Provided that a higher dose size is used, it seems likely that once-weekly rifapentine will prove a welcome addition to the chemotherapy of tuberculosis by greatly reducing the number of doses that need to be supervised and thus assuring better compliance overall.

One may ask why rifapentine has performed rather less well than might be expected from its high in vitro activity against M. tuberculosis and the high serum concentrations attained in the patients. Possible reasons are the high binding of rifapentine to serum albumin (S. Gardner, Hoechst Marion Roussel, personal communication) so that only the unbound fraction of about 2% of total drug and part of the bound fraction is actually available in lesions, the much higher inhibitory concentration in macrophage cell lines than in culture (16), and the lower bactericidal activity against stationary phase cultures of M. tuberculosis than was found with rifampin and some other rifamycins (17).

An important issue is whether the high relapse rates after HRp₁ regimen were due to inadequacy of rifapentine or isoniazid. The contrast between the high efficacy of the rifapentine regimens in the patients with initial isoniazid resistance (no failures during treatment and a relapse rate similar to the rate in patients with initially sensitive organisms) and the single patient with initial resistance to rifapentine only, who failed during treatment, indicates that almost all of the antibacterial activity in the continuation phase must be due to the high sterilizing activity of rifapentine. This view is also supported by studies in experimental murine tuberculosis (2, 3) and by earlier clinical trials which indicated the importance of the rifampin component of continuation phase regimens in reducing relapses (13, 15). Nevertheless, the isoniazid component may play a minor role as it did in once-weekly regimens of isoniazid and rifampin in which all failures occurred in rapid acetylators of isoniazid (15). In Hong Kong, 78% of patients have been found to be rapid acetylators, 29% being homozygous rapid (18). A study of the acetylator status of the patients in the present study is under way and should help to assess the importance of the isoniazid component.

Finally, the prevalence of initial drug resistance in Hong Kong is noteworthy. There were 581 patients who satisfied the admission criteria, had a positive pretreatment culture of M. tuberculosis and had had no more than 2 wk treatment before admission (558 had none). Among these 581 patients, one (0.2%) had a multidrug-resistant strain resistant to isoniazid, rifampin, and streptomycin (successfully treated with a regimen of ethionamide, ofloxacin, kanamycin, pyrazinamide, and cycloserine), two patients (0.3%) had strains resistant to rifampin (and rifapentine) only, 26 patients (4.5%) had resistance to isoniazid only, 12 (2.1%) resistance to streptomycin only, and 12 (2.1%) resistance to isoniazid and streptomycin. In view of the considerable population flow between Hong Kong and mainland China, there is a remarkably low prevalence of initial resistance, particularly multidrug-resistant strains, probably due to full supervision of drug taking throughout the outpatient clinics of the Health Department. In brief, the study has shown that rifapentine appears to be a safe drug whose recommended dosage of 600 mg may well be too low to prevent an unacceptable relapse rate after the end of chemotherapy.