Description of studies
In addition to the 10 trials included in the original systematic review, we initially identified a possible four new trials that met the inclusion criteria. Of these, one trial (Grant 2005) was later excluded because it assessed the effect on incidence of TB at a clinic that offered INH under routine conditions and participants were not randomized to receive INH or placebo. The study by Lim 2006 (Lim 2006) was an additional publication from the Whalen trial (Whalen 1997), which was already included in the review. Lim et al reported data on the rate of disease progression to AIDS or death, with results reported for PPD positive persons and those with anergy (Lim 2006).Long-term follow-up results of Mwinga 1998 were published in Quigley 1998.
This current update thus includes 12 trials with a total of 8578 randomized participants. Study size varied from 98 to 2018 participants. The two new trials included in this update are Rivero 2007 and Mohammed 2007. The study by Rivero et al. that was included in the original review recruited anergic individuals Rivero 2003. The study identified in this update recruited PPD+ individuals Rivero 2007.
Full details of the studies included in this review are provided in the table of included studies. In addition ongoing trials are summarised in appendix 1 (Churchyard GJ 2007).
Summary of included studies:
The trials were conducted in both developed and developing countries: Haiti (Fitzgerald 2001; Halsey 1998; Pape 1993), Uganda (Whalen 1997; Lim 2006), Kenya (Hawken 1997), Zambia (Mwinga 1998), South Africa (Mohammed 2007), Spain (Martinez 2000; Rivero 2003; Rivero 2007) and the USA (Gordin 1997). One multi-national study included participants from Mexico, USA, Haiti and Brazil (Gordin 2000).
Study participants were 13 years of age or older (mean age 33 years). 47% were female, with a range of 20% to 69% across trials. Mean duration of follow up ranged from 1 to 3 years.
Some reports included only individuals who were PPD+ (Gordin 2000; Halsey 1998; Rivero 2007) and others involved only those known to be anergic (Gordin 1997; Rivero 2003; Mohammed 2007). One trial included both PPD+ and anergic individuals as separately randomized groups (Whalen 1997; Lim 2006) and another as one group with stratification by PPD status in the analysis (Martinez 2000). A further trial among PPD negative individuals did not test for anergy (Fitzgerald 2001). The remaining trials included individuals regardless of PPD status (Hawken 1997; Mwinga 1998; Pape 1993) with only two of these providing stratified data (Hawken 1997; Pape 1993)
In all, 4811 individuals were PPD+, 2030 were PPD- (of which 1640 were known to be anergic) and in 1737 individuals the PPD status was unknown.
Five trials compared isoniazid (INH) with placebo (Fitzgerald 2001;Gordin 1997 Hawken 1997 Mohammed 2007; Mwinga 1998), three with rifampicin (RIF) (Gordin 2000; Halsey 1998; Martinez 2000), one with pyridoxine (Pape 1993), one to either placebo or INH+RIF (Whalen 1997) and two to RIF or RIF + PZA (Rivero 2003; Rivero 2007)
Treatment dosage varied among the trials: INH 300 mg or 600 mg for daily regimens; 600 mg or 900 mg for twice weekly regimens; RIF 450 mg or 600 mg; and PZA 20 mg/kg body weight to a total of 3500 mg. Dosage frequency was daily except in two trials (Halsey 1998; Mwinga 1998) which offered treatment twice a week. The duration of INH treatment varied as follows: 12 months (Fitzgerald 2001; Gordin 2000; Martinez 2000; Pape 1993 ; Mohammed 2007) and 6 months (Gordin 1997; Halsey 1998; Hawken 1997; Mwinga 1998; Rivero 2003; Whalen 1997 ; Rivero 2007).
We collected data for active TB diagnosed either by culture or other methods of diagnosis as defined by the study authors (confirmed, probable and possible). Active TB data were available for all participants.
Effects of interventions
Preventive therapy (any anti-TB drug) versus placebo reduced the risk of active TB by 32% (10 trials; 5762 participants; RR 0.68, 95% CI 0.54 to 0.85). There was a small amount of statistical heterogeneity among the trials (I2 = 31%; Chi 2 =18.93, p = 0.13). For confirmed (culture-proven) TB, the result was similar (4 trials; 2573 participants; RR 0.73, 95% CI 0.49 to 1.08) although not statistically significant.
All drug regimens (regardless of type, frequency or duration of treatment) reduced the incidence of active TB compared with placebo:
In trials that directly compared drug regimens we found no differences in effectiveness:
INH vs. RIF+PZA: 6 trials; RR 1.03, 95% CI 0.75 to 1.40 Analysis 6.1
INH vs. INH+RIF: 5 trials; RR 0.97, 95% CI 0.52 to 1.83 Analysis 7.1
INH + RIF vs. RIF+PZA: 2 trials; RR 2.64, 95% CI 0.71 to 9.8 Analysis 8.1
INH vs. INH+RIF+PZA: 1 trial; RR 0.60, 95% CI 0.23 to 1.57 Analysis 9.1
INH + RIF vs. INH+RIF+PZA: 1 trial; RR 0.75, 95% CI 0.31 to 1.82 Analysis 10.1
We detected no heterogeneity in the outcome across the trials.
We found no trials that compared the effects of different drug dosages, treatment frequency or duration of therapy on clinical outcomes. Current trials do not provide sufficient data to assess the impact of preventive therapy on interval to active TB.
We assessed the influence of various factors on the incidence of active TB:
Among individuals who were tuberculin skin test positive, preventive therapy reduced the risk of active TB by 62% (4 trials; 2378 participants; RR 0.38, 95% CI 0.25 to 0.57). There was no evidence of effect for individuals with a negative tuberculin test (7 trials; 2822 participants; RR 0.89, 95% CI 0.64 to 1.24).Analysis 1.1
Stage of HIV disease at baseline
We found limited data stratified by stage of HIV/AIDS at baseline. In Gordin 1997, which compared INH to placebo, the relative risk (95% CI) for the development of confirmed TB was 3.42 ( 0.14 to 82.33) for those with AIDS and 0.32 (0.06 to 1.54) for those without AIDS; neither of these findings being statistically significant. Similarly, in Gordin 2000 comparing INH with RIF+PZA, the risk of confirmed TB was not statistically significant in subgroups defined by AIDS status: AIDS RR 0.97, 95% CI 0.28 to 3.43; no AIDS RR 1.42, 95% CI 0.74 to 2.74.
Time since treatment
We found limited information on the duration of the protective effect of preventive therapy. Mwinga 1998 provided data for a median follow-up of 1.8 years in a mixture of PPD positive and negative people found a reduction in the risk of active TB in the intervention groups versus placebo (INH: RR 0.62, 95% CI 0.39 to 0.97; RIF plus PZA: RR 0.57, 95% CI 0.35 to 0.91). In a subsequent report (Quigley 1998) presenting findings after a mean follow-up of 3 years, Kaplan Meier analysis demonstrated a diminishing effect over time. Nevertheless, compared to placebo the reported cumulative risk in the first 2.5 years remained lower for INH (RR 0.52, 95% CI 0.27 to 1.00), for RIF plus PZA (RR 0.58, 95% CI 0.30 to 1.09) and for both intervention arms combined (RR 0.55, 95% CI 0.32 to 0.93).
In another study (Whalen 1997) involving PPD positive individuals, INH (RR 0.29, 95% CI 0.12 to 0.67), INH + RIF (RR 0.36, 95% CI 0.17 to 0.77) and INH + RIF + PZA (RR 0.48, 95% CI 0.23 to 1.00) were each shown to significantly lower the risk of active TB after a mean of 15 months. This benefit remained statistically significant on long-term follow-up for the rifampicin containing regimens but not for INH alone (Johnson 2001). Based on a Cox regression analysis the adjusted relative risk at 3 years was 0.67 (95% CI 0.42 to 1.07) for INH, 0.49 (95% CI 0.29 to 0.82) for INH + RIF, and 0.41 (95% CI 0.22 to 0.76) for INH+RIF+PZA. For anergic participants the initial statistically non-significant benefit 1 year after INH treatment (RR 0.74, 95% CI 0.30 to 0.1.79) (Whalen 1997-anergy) remained at 2 years (adjusted relative risk 0.61 (95% CI 0.32 to 1.16) (Johnson 2001 -anergy).
The long-term follow-up results for the studies mentioned above should be interpreted with caution as there was substantial loss to follow-up in all trials which may have introduced bias.
DEATH FROM ALL CAUSES
We found no evidence that preventive therapy versus placebo reduced all-cause mortality (9 trials; 5762 participants; RR 0.94, 95% CI 0.85 to 1.05) or reduced mortality among those who were PPD positive (4 trials, 2378 participants, RR 0.80, 95% CI 0.63 to 1.02), however these findings were heterogeneous (p=0.04).
The single placebo-controlled trial that assessed the effect of INH by stage of HIV/AIDS at baseline found no difference (Gordin 1997): with AIDS (RR 0.96, 95% CI 0.79 to 1.17), without AIDS (RR 1.07, 95% CI 0.84 to 1.35).
There was reduction in mortality with INH monotherapy versus placebo among individuals with a positive tuberculin skin test (RR 0.74, 95% CI 0.55 to 1.00).
We found no differences in the effect on death by study drug with the exception of INH+RIF, which was associated with a significant reduction in the risk of death (2 trials; 1179 participants; OR 0.69, 95% CI 0.50 to 0.95). Direct comparison of different drug regimens revealed no differences.Analysis 3.3.
Based on data from two trials of an INH based regimen versus placebo (Fitzgerald 2001; Pape 1993), we found no evidence of a reduction in the incidence of AIDS (RR 0.88, 95% CI 0.60 to 1.28). However, one trial (Pape 1993) found a lower risk of AIDS in PPD + individuals (RR 0.36, 95% CI 0.15 to 0.85), but not in those with a negative skin test (RR 0.78, 95% CI 0.27 to 2.20). Pape 1993 also found a significant increase in the mean time to AIDS (in months) (WMD 7.8, 95% CI 1.71 to 13.89) Analysis 2.4.
We extracted available data on adverse events deemed by the investigators as serious enough to discontinue treatment. Compared to placebo, preventive therapy led to more adverse events ( 6 trials; 5525 participants; RR 2.55, 95% CI 1.70 to 3.85) Analysis 1.5.
The likelihood of stopping treatment due to adverse effects was higher for combination therapies than for INH monotherapy:
Compared with placebo (i.e. indirect comparison)
INH: 6 trials; RR 1.66, 95% CI 1.09 to 2.51 Analysis 2.5
INH+RIF: 2 trials; RR 16.72, 95% CI 3.29 to 84.89 Analysis 3.5
RIF+PZA: 2 trials; RR 7.84, 95% CI 2.60 to 23.67 Analysis 4.5
INH+RIF+PZA: 1 trial; RR 26.11, 95% CI 3.56 to 191.63 Analysis 5.5
INH vs RIF+PZA: 5 trials; RR 0.63, 95% CI 0.48 to 0.84 Analysis 6.5
INH vs INH+RIF: 4 trials; RR 0.79, 95% CI 0.50 to 1.23 Analysis 7.5
INH+RIF vs RIF+PZA: 2 trials; RR 0.85, 95% CI 0.50 to 1.46 Analysis 8.5
INH vs INH+RIF+PZA: 1 trial; RR 0.10, 95% CI 0.03 to 0.33 Analysis 9.5
INH+RIF vs INH+ RIF+PZA: 1 trial; RR 0.42, 95% CI 22 to 0.80 Analysis 10.5
Overall, about half of the studies reported on adherence to therapy (Halsey 1998; Hawken 1997; Martinez 2000; Mohammed 2007; Mwinga 1998; Rivero 2007; Whalen 1997). Differences in the definition of adherence and the level of detail varied across the studies. There is some evidence that the length of treatment may be related to degree of adherence. Halsey 1998 reported higher rates of adherence with a 2 months course of RIF+PZA as compared to 6 months of INH. Martinez 2000 reported better adherence with 3 months INH+RIF compared to 12 months INH. On the other hand, Whalen 1997, reported no difference among treatment groups which included INH for 6 months, INH+RIF for 3 months and INH+RIF+PZA for 3 months. Hawken 1997, a placebo-controlled trial of INH for 6 months, found no difference in adherence rates between the two study arms. Mohammed 2007 reported a median adherence of 87% in the INH group and 81.2% in the placebo group while Rivero 2007 reported an adherence of 64.5%, 63.0% and 62.4% in the INH, INH+RIF and RIF+PZA treatment groups respectively. We did not have sufficient data to assess adherence as an effect modifier in the studies included in the review.