Split-drug regimens for the treatment of patients with sputum smear-positive pulmonary tuberculosis – a unique approach


  • Tuberculosis Research Centre (Indian Council of Medical Research), Chennai, India

P. R. Narayanan (corresponding author), Tuberculosis Research Centre, Mayor V. R. Ramanathan Road, Chetput, Chennai 600 031, India. Tel.: +91 (44) 28 36 2432; Fax: 91 (44) 28 36 2528/28 36 2529; E-mail: nrparanj@md2.vsnl.net.in


Objective  To evaluate the efficacy of split-drug regimens for treatment of patients with sputum smear-positive pulmonary tuberculosis in south India.

Design  Randomized controlled clinical trial where eligible patients were randomly allocated to: (i) 2RE3HZ3(alt)/4RH2 (split I): rifampicin plus ethambutol given on one day and isoniazid plus pyrazinamide the next day for first 2 months followed by rifampicin plus isoniazid twice weekly for 4 months, or (ii) 3RE3HZ3(alt)/3RH2 (split II): similar to regimen 1, except duration was 3 months in each phase, or (iii) 2REHZ3/4RH2 (control): rifampicin, isoniazid, ethambutol and pyrazinamide, given thrice weekly for 2 months followed by isoniazid and rifampicin twice weekly for 4 months. All patients were followed up clinically and bacteriologically every month up to 2 years and every 6 months for up to 5 years.

Results  A favourable response (cultures negative for Mycobacterium tuberculosis during the last 2 months of treatment) was observed in 91% of 407 patients in split I, 94% of 415 in split II and 89% of 418 in the control regimen. Ninety-one per cent of 370 patients in split I, 93% of 389 in split II and 90% of 370 in control regimens had quiescent disease at the end of 60 months. Gastrointestinal symptoms were more frequent under the control regimen (P = 0.01).

Conclusion  Split-drug regimens were as effective as the control regimen in terms of favourable response at the end of treatment and quiescent disease at 5 years, and caused fewer gastrointestinal side-effects.


In controlled clinical trials, fully supervised, thrice weekly intermittent regimens using isoniazid (H), rifampicin (R), pyrazinamide (Z) and ethambutol (E) or streptomycin (S) are found to be as effective as daily regimens consisting of the same drugs (HKCS/BMRC 1982, 1991; Prabhakar 1986; Rani 1991; Cao et al. 1998). In these trials, all oral drugs were given together in a single dose. However, in vitro studies have shown that a split-drug combination of RE and HZ does not affect the bacteriological action of these regimens (Paramasivan et al. 1993, 1994). Split-drug regimens have been shown to be as effective as the conventional intermittent four-drug regimen in controlling experimental murine tuberculosis (TB) (Guy et al. 1993). The benefit of using a split regimen would be a reduction in the bulk of drug intake and possibly fewer side-effects for patients.

We conducted a randomized controlled clinical trial to evaluate the efficacy of two split regimens for the treatment of patients with sputum smear-positive pulmonary TB. We studied a 2- or 3-month intensive phase of a split regimen in which two of the four drugs (R and E) were given on one day and H and Z the next day, and a control regimen where all the four drugs were given together thrice weekly in a single dose during the initial 2-month period. During the continuation phase of 4 months, all patients received R and H twice weekly. As the study was planned and executed in the pre-DOTS period, we used a twice weekly continuation phase as had been used in the earlier controlled clinical trials at our centre [Tuberculosis Research Centre (TRC) 1983, 1997; Prabhakar 1986].


The study protocol was approved by the scientific advisory and the ethics committees of the TRC, Chennai, South India.

Patients and enrolment

Between October 1990 and November 1995, all patients referred to the TRC from neighbouring chest clinics at Chennai and Madurai cities of south India were assessed for their eligibility to participate in the study. Patients were enrolled if they were at least 12 years old, had two sputum cultures positive for Mycobacterium tuberculosis (allocation to a regimen was based on the smear results and were considered for analysis only if cultures yielded growth for M. tuberculosis), were willing to attend the clinic for supervised chemotherapy for a period of 6 months, and did not have extra-pulmonary TB or other serious non-tuberculous disease. Prior anti-tuberculosis treatment received by patients was not a criterion for admission to the trial and about one-third of the patients included in the analysis had received treatment from other sources before admission to the trial. Baseline investigations of each patient included a chest X-ray and examination of two overnight and two spot sputum specimens by direct smear microscopy and culture for mycobacteria (Allen & Baker 1968). Cultures were identified and those positive for M. tuberculosis were tested for sensitivity to H and R (Allen & Baker 1968; Canetti et al. 1969). Evaluation of baseline hepatic and renal functions and serological testing for HIV was carried out for all patients. Those who were HIV positive and had defective vision were not included in the analysis.


Eligible patients were stratified on the basis of sputum smear grading of the first overnight specimen (0 or 1+; 2+ or 3+) and duration of previous chemotherapy (<15 days; ≥15 days). They were then randomly allocated to one of the following three fully supervised regimens of 6 months’ duration using the sealed envelope technique.

  • 1Split I (2RE3HZ3[alt]/4RH2): R and E were given on one day and H and Z on the next day, thrice weekly for the initial 2 months (intensive phase), followed by R and H twice weekly for the next 4 months (continuation phase).
  • 2Split II (3RE3HZ3[alt]/3RH2): R and E were given on one day and H and Z on the next day, thrice weekly for the initial 3 months (intensive phase), followed by R and H twice weekly for the next 3 months (continuation phase).
  • 3Control (2REHZ3/4RH2), i.e. all four drugs given together in a single dose thrice weekly for 2 months (intensive phase) followed by R and H twice weekly for the next 4 months (continuation phase).

The drug dosages were: H 600 mg, R 450 mg, E 1000 mg and Z 1.5 g for patients weighing 40 kg or less; for those weighing more than 40 kg, the doses were 600 mg, 600 mg, 1200 mg and 2 g, respectively, for all the three regimens. Fixed-drug combinations were not employed. Every dose was administered under direct observation of a health staff at the clinic.

All patients received treatment under supervision during the intensive and the continuation phases. If a patient missed an appointment, a home visit was made the next day by a health worker, and subsequently by a social worker and a medical officer, in that order. These visits were continued until the patient was either retrieved or ‘lost’; loss being defined as continuously defaulting for more than 1 month. Any complaint suggestive of an adverse reaction volunteered by a patient during treatment was recorded and a clinician elicited further details by interviewing these patients. If the complaints were related to the anti-tuberculosis drugs administered, suitable action was taken wherever necessary.


All patients were monitored at monthly intervals for the first 2 years and once every 6 months for the next 3 years. Follow-up examinations included clinical assessment, regularity in drug taking and monitoring of adverse drug reactions. Three sputum specimens during treatment and two specimens during follow-up were examined at monthly intervals up to 2 years and once in 6 months thereafter by microscopy and culture. If a culture grew M. tuberculosis, at least one culture each month was tested for sensitivity to H and R to monitor the emergence of drug resistance. A chest X-ray of all patients was taken at 1 month, at the end of chemotherapy (6 months) and at 12, 24, 36, 48 and 60 months; and at any other time considered necessary by a physician. If any culture was reported as positive for M. tuberculosis during the 6-month follow-up, additional sputa were examined at monthly intervals for 6 months.

Outcome measures

A patient was considered to have had (i) a favourable response, if all the six cultures were negative during the last 2 months of treatment, or one culture was positive at 5 or 6 months but became negative subsequently without additional chemotherapy; (ii) an unfavourable response if two or more cultures were positive in the last 2 months of treatment of which at least one grew 20 colonies or more, or if a patient died of tuberculosis during treatment, had change of treatment for clinical/or radiographic deterioration or for persistent culture positivity; (iii) a bacteriological relapse if two or more cultures and at least one smear was positive in a 2-month period during follow-up among patients who had a favourable response at the end of treatment; and (iv) quiescent disease if they were culture-negative during the follow-up period of 5 years.

Criteria for inclusion in the analysis

In all, 6275 patients were assessed for eligibility and 1269 (420 split I, 424 split II, 425 control) were allocated to the study. After excluding 13, nine and seven patients who were ineligible for entry into the study as per the criteria of the protocol, there remained 407, 415 and 418 patients for intention to treat analysis in split I, split II and control regimens, respectively (Figure 1). After excluding 37, 26 and 48 patients with unfavourable response at the end of treatment, there remained 370, 389 and 370 patients in the split I, split II and control regimen, respectively, for follow-up analysis up to 5 years.

Figure 1.

Schematic presentation of the randomized controlled clinical trial to compare the effectiveness of three anti-TB regimens, Chennai, South India.

Statistical analysis

A sample size of 1250 was estimated for all the study groups combined to assess a difference of 5% in the favourable response at the end of treatment between two treatment groups with a power of 80% assuming 90% favourable response for the control regimen. The treatment groups were compared using the chi-square test with Yate's correction. Estimated survival curves were constructed using the Kaplan–Meier method. The log-rank test was used to compare survival curves. The Cox proportional hazard model was used for multivariate analysis, adjusting for baseline characteristics given in Table 1. We included all randomized patients who had fulfilled the eligibility criterion for admission to the study as per the protocol regardless of compliance in the ‘intention to treat analysis’. Thus, patients with cultures negative for M. tuberculosis (20 patients), who tested positive for HIV (eight patients) and were identified to have visual defect after allocation (one patients) were excluded from this analysis. A subgroup analysis was carried out for patients who received more than 75% of the prescribed chemotherapy. Patients with organisms fully susceptible to the anti-tuberculosis drugs used and those with organisms resistant to one or more drugs were analysed separately. P ≤ 0.05 was considered statistically significant.

Table 1.  Baseline characteristics of patients in the randomized controlled clinical trial to compare the effectiveness of three anti-TB regimens, Chennai, South India
CharacteristicsSplit I (n = 407)Split II (n = 415)Control (n = 418)
  1. * Smear 1+: <6 bacilli per field; smear 2+: 6–100 bacilli per field; smear 3+: more than 100 bacilli per field.

  2. † Culture colonies 1–19 colonies; 1+: 20–100 colonies; 2+: innumerable discrete colonies; 3+: confluent growth.

Mean age, years (SD) 33 (11) 32 (11) 32 (12)
Mean weight, kg (SD) 41.4 (6.5) 40.5 (6.5) 40.9 (6.1)
Male, n (%)310 (76)302 (73)314 (75)
Smear,* n (%)
 Negative 1 (0.2) 1 (0.2) 0
 1+124 (30)125 (30)106 (25)
 2+221 (54)223 (54)241 (58)
 3+ 61 (15) 66 (16) 71 (17)
Culture†, n (%)
 Colonies 8 (2) 5 (1) 2 (0.5)
 1+ 21 (5) 23 (6) 18 (4)
 2+133 (33)130 (31)133 (32)
 3+245 (60)257 (62)265 (63)
Previous treatment, n (%)
 <15 days296 (73)328 (79)306 (73)
 ≥15 days111 (27) 87 (21)112 (27)


The baseline characteristics of patients admitted to the three treatment groups were similar (Table 1).

End-of-treatment response

At the end of treatment, the effectiveness of the regimens was 91% (370/407) for patients in the split I regimen, 94% (389/415) in the split II regimen, and 89% (370/418) in the control regimen, respectively (Table 2). The difference between the split II and the control regimen was statistically significant (P = 0.01).

Table 2.  Response at the end of treatment and relapse during a follow-up period of 54 months in a randomized control clinical trial comparing three anti-TB regimens, Chennai, South India
 Split ISplit IIControl
  1. Values are given as n (%).

  2. P = 0.01.

Outcome at the end of treatment
 No. of patients407415418
 Favourable370 (91)389 (94)*370 (89)*
 Unfavourable37 (9)26 (6)48 (11)
Outcome at the end of 5 years
 No. of patients370389370
 Favourable338 (91)363 (93)334 (90)
 Relapse32 (9)26 (7)36 (10)

Quiescent disease during 54-month follow-up

The proportion of patients who had quiescent disease during a 54-month follow-up period was 91% (338/370), 93% (363/389) and 90% (334/370) for patients in split I, split II and control regimens, respectively. These differences were not statistically significant.

Survival analysis

Figure 2 shows the Kaplan–Meier curves indicating the probabilities of being free from adverse events among patients in the three treatment groups. In univariate analysis, the probability of being free from adverse events was higher among patients receiving split II regimen compared with those receiving the control regimen (P = 0.002). Adjusting for baseline characteristics, in a multivariate model, the hazard ratio of split I compared with the control was 0.81 (95% CI 0.59–1.11, P = 0.06). For split II compared with the control, the hazard ratio was 0.58 (95% CI 0.41–0.81, P = 0.002).

Figure 2.

Survival curve among sputum smear-positive pulmonary TB patients treated with three different regimens, Chennai, South India.

Subgroup analyses

Among patients who had organisms susceptible to the anti-tuberculosis drugs used and received more than 75% of the prescribed treatment, culture negativity at 2 months was 80% of 290 patients in the split I, 82% of 304 in the split II, and 79% of 273 in the control regimens (Table 3). By the third month, the negativity rates were 97% in split I and split II regimens and 92% in the control regimen. A favourable response at the end of treatment was observed in 98% of 290 patients in split I, 99% of 304 in split II and 99% of 273 in the control regimen. During a follow-up period of 54 months, 92% of 284 patients in split I, 93% of 301 in split II and 91% of 269 in the control regimen continued to have quiescent disease.

Table 3.  Culture conversion, response at the end of treatment and relapse among patients receiving more than 75% of chemotherapy in the randomized control clinical trial comparing anti-TB regimens, Chennai, South India
 No. of patients% of culture conversion atFavourable response at the end of treatmentQuiescent disease*
2 months3 months
  1. * Values are given as n (%).

Fully sensitive
 Split I290809798284 (92)
 Split II304829799301 (93)
 Control273799299269 (91)
Resistant to isoniazid
 Split I5371808344 (86)
 Split II5475898546 (91)
 Control6065787545 (87)

Among patients with organisms initially resistant to H, 83% of 53 patients, 85% of 54 and 75% of 60 in the split I, split II and control regimens, respectively, had a favourable response and 86% of 44 patients in split I, 91% of 46 in split II and 87% of 45 in the control regimen continued to have quiescent disease during the follow-up period of 54 months. Only two (one in split I and the other in split II) of 33 patients with organisms initially resistant to H and R had a favourable response.

Adverse drug reactions

Adverse drug reactions attributable to anti-tuberculosis drugs occurred in 41 (10%) of 407 patients in split I, 62 (15%) of 415 in split II and 58 (14%) of 418 in the control regimen (Table 4). Gastrointestinal reactions, such as nausea, vomiting, feeling of fullness and epigastric disturbance were more common in the control regimen than in the two split regimens combined [29 of 418 (7%) vs. 28 of 822 (3%); P = 0.01]. The occurrence of arthralgia was significantly higher among patients treated with the split II regimen than with the split I or with the control regimen (P < 0.01). There was no significant difference between the split and control regimens with respect to other adverse reactions. Chemotherapy was modified in only 16 patients due to adverse reactions, including seven for whom drugs were withheld temporarily due to the development of jaundice.

Table 4.  Adverse reactions among patients in the randomized controlled clinical trial to compare the effectiveness of three anti-TB regimens, Chennai, South India
Adverse reactionSplit I (n = 407)Split II (n = 415)Control (n = 418)
  1. Values are given as n (%).

  2. * Include peripheral neuropathy, anaphylactic reaction, glossitis, flu-syndrome, purpura and hypersensitivity reaction.

At least 141 (10)62 (15)58 (14)
Gastrointestinal14 (3)14 (3)29 (7)
Arthralgia11 (3)33 (8)13 (3)
Cutaneous10 (2)12 (3)11 (3)
Giddiness 4 (1) 4 (1) 8 (2)
Hepatitis 2 (0.5) 4 (1) 1 (0.2)
Others* 2 (0.5) 4 (1) 5 (1)

Emergence of drug resistance

Of the 13 patients with initially drug-sensitive bacilli who had an unfavourable response at the end of treatment, five developed resistance to one or more drugs; three (one split II, two control) to H, and two (one split I, one control) to H and R. None of the 52 patients who relapsed developed drug resistance.

Among 32 patients (nine split I, eight split II, 15 control) with bacilli initially resistant to H and who had an unfavourable response, 22 (six split I, five split II, 11 control) developed resistance to R. Of the 14 patients with organisms initially resistant to H and had a bacteriological relapse, only one (split II) patient had emergence of resistance to R.


This study tested a new concept in the delivery of anti-TB drugs. The findings suggest that splitting anti-TB drugs and administering them in pairs during the initial 2 months (intensive phase) is as effective, in terms of low rate of relapse and low emergence of multidrug-resistant TB, as when the drugs are given together. Moreover, it is associated with a much lower incidence of gastrointestinal side-effects than when all the four drugs are given together.

The incidence of adverse drug reactions was low, occurring in 10% of patients on split I regimen, in 15% on split II and in 14% on the control regimen. This is much lower than that observed in our patients in earlier studies either with daily regimens or with thrice weekly regimens (Prabhakar 1986; Rani 1991; Paramasivan et al. 1993). Gastrointestinal problems and giddiness were twice as common with the control regimen than with the split-drug regimens probably due to the larger bulk of drugs, 7–9 in number, when all four drugs were administered together. The reduction in gastrointestinal symptoms and giddiness in the split regimens is likely to enhance compliance of patients during the intensive phase of treatment. Jaundice during treatment was observed in ≤1% of the subjects. This is much lower than that encountered during treatment with daily regimens (Parthasarathy et al. 1986). Arthralgia caused by Z was encountered in a much smaller number of patients (3–8%) with the three regimens than was observed during daily treatment (24–46%) in some of our earlier studies (TRC 1983, 1997; Nazareth et al. 1984), probably because the wider interval between the doses of Z allows excretion of excess uric acid (Ellard & Haslam 1976) or due to the concomitant administration of R, which has shown to appreciably decrease the incidence of arthralgia (Raghupati Sarma et al. 1983) by enhancing the excretion of both uric acid and pyrazinoic acid (Kannapiran et al. 1985). The higher incidence of arthralgia among those on the split II regimen than in those in the split I regimen or the control regimen could be attributed to the additional month of Z received by these patients.

In the Revised National TB Control Programme (RNTCP) of India, where the thrice weekly regimen is used, split-drug regimens may be useful in situations where patients cannot tolerate all drugs together as a result of drug side-effects. Because the split-drug regimens increase difficulties and expenses due to the need for directly observed treatment of daily dosage, the application of this concept in the RNTCP and in other higher tuberculosis prevalence countries using intermittent regimens is limited. In low tuberculosis prevalence countries, however, where the treatment policy is often self-administration for all patients whether sputum smear-positive or negative, the concept of split-drug regimens may be more relevant. In these countries, split drug regimens may improve treatment compliance because of reduced drug side-effects.

This study was designed and conducted prior to the implementation of the RNTCP in India (Parthasarathy et al. 1986) and therefore the twice weekly continuation phase as in our earlier studies was used (Allen & Baker 1968; TRC 1983; Prabhakar 1986). The control regimen tested in this study is similar to that used in the RNTCP with the exception that during the continuation phase R plus H was given twice weekly instead of thrice weekly. Another limitation of the study was that a large proportion of the patients screened for the study were excluded from being eligible for the study. Impermanent residential address and refusal of some patients to agree to home visits by study staff were the main reasons for exclusion. Strict selection of patients may have resulted in an increased likelihood of treatment completion than that expected under routine programme conditions.

To our knowledge, this is the first attempt to study the effectiveness of split drug regimens. The split drug combinations used in this study were designed to combine the practicability of daily treatment with the known advantages of fully supervised intermittent treatment. Further studies are needed to establish the applicability of the results under routine programme conditions. If found consistently effective, the split drug regimens may be useful for TB control programmes where there is a need for reducing the bulk of drugs and side-effects to promote better patient compliance – a crucial issue in programme management.


T. Santha, Fathima Rehman, D. A. Mitchison, G. Raghupathi Sarma, A. M. Reetha and R. Prabhakar prepared this report. The other staff members involved are: R. Rajeswari, B. Rani, M. S. Jawahar, K. Rajaram, M. Rema, Pauline Joseph, R. Balambal, R. Usha, Parvathy Raghavan, Ambujam Ganesh, Sudha Ganapathy, Teresa Xavier, K. N. Gopilingam (clinical division), C. N. Paramasivan, N. S. Selvakumar, P. Venkataraman (Bacteriology), G. Prema, M. Kannapiran, (Biochemistry), P. R. Somasundaram, B. Janardanam and K. Subramani (statistics). We acknowledge Prof. Wallace Fox for his contribution in the design of the study. We thank all the nursing staff, medical social workers, laboratory technicians and statisticians of the centre for their generous support of the conduct of the study. We are grateful to the Director of the Institute for Thoracic Medicine, Chennai, Superintendents of Government Thiruvatteeswarar Hospital for Thoracic Medicine, Chennai and Government Rajaji Hospital, Madurai and the Medical Officer of Tuberculosis Clinic, Pulianthope, Chennai for referring the patients for the trial. We thank Jaya Sreedhar and Bandana for their assistance in report preparation. G. Renu and Vani Kurup read and helped editing the paper. Finally, we are grateful to the patients who participated and without whom the trial would not have been possible.