Summary of findings for the main comparison.
Rifampicin compared to isoniazid for preventing active TB in HIV‐negative people
| Rifampicin (3 to 4 months) compared to isoniazid (6 to 9 months) for preventing active TB in HIV‐negative people | ||||||
| Patient or population: HIV‐negative people at risk of TB infection1 Intervention: Rifampicin for 3 to 4 months Comparison: Isoniazid for 6 to 9 months | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect (95% CI) | No of participants (studies) | Quality of the evidence (GRADE) | Comments | |
| Assumed risk | Corresponding risk | |||||
| Isoniazid | Rifampicin | |||||
| Active TB Follow‐up: 5 years | 150 per 1000 | 121 per 1000 (70 to 210) | RR 0.81 (0.47 to 1.4) | 332 (1 study) | ⊝⊝⊝⊝ very low2,3,4,5 | In the placebo arm of this four‐arm trial (HKCS 1992), 36/159 (23%) developed active TB. |
| Adherence6 | 690 per 1000 | 822 per 1000 (697 to 884) | RR 1.19 (1.01 to 1.3) | 1768 (5 studies) | ⊕⊕⊕⊝ moderate2,7,8,9 | |
| Treatment‐limiting adverse events | 93 per 1000 | 45 per 1000 (21 to 93) | RR 0.48 (0.23 to 1) | 1674 (4 studies) | ⊝⊝⊝⊝ very low10,11,12 | |
| Hepatotoxicity:13 Grade 3 and 4 toxicity | 46 per 1000 | 7 per 1000 (3 to 16) | RR 0.15 (0.07 to 0.4) | 1774 (5 studies) | ⊕⊕⊕⊝ moderate10 | Only one child allocated to rifampicin in Magdorf 1994 developed hepatotoxicity |
| *The basis for the assumed risk is the control group risk in single studies, and the median risk in the control group for pooled data. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio. | ||||||
| GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. | ||||||
1 The data in this table are mostly from four trials in adults (HKCS 1992; Menzies 2004; Menzies 2008; Chan 2012). In another trial in children (Magdorf 1994), data for comparative effectiveness could not be used for the outcomes of developing active TB (no cases detected in 100 children over two years of follow up), and treatment‐limiting adverse events (not reported). The data for adherence and hepatotoxicity are from all five trials. 2 No study limitations: None of the trials were judged as at high risk of bias. Not downgraded. 3 No inconsistency: Single trial (HKCS 1992) in adults with silicosis. In Chan 2012 involving adult prisoners, active TB was not detected over five years of follow up, so comparable effectiveness could not be determined. Menzies 2004, and Menzies 2008 did not report this outcome. Not downgraded. 4 Serious indirectness: This study was done over 20 years ago and only included adult men with silicosis from Hong Kong; the results are not easily generalised to other treatment groups or settings, and may not be applicable today. Downgraded by 1. 5 Very serious imprecision: The 95% CI of the effect estimate includes appreciable benefit and harm with rifampicin. The study was underpowered to be able to confidently detect differences between the two regimens. Downgraded by 2.
6 In Chan 2012, conducted in prisoners, treatment was directly observed (except when prisoners were on parole). Treatment in HKCS 1992; Menzies 2004 and Menzies 2008 were self‐administered.
7 Serious inconsistency: There was significant inconsistency in the pooled results from the five trials (I2 = 82%), but the inconsistency was largely due to the lack of difference in adherence with the two regimens in the small trial in children (Magdorf 1994), compared to greater adherence with rifampicin over INH in the four trials in adults (test for subgroup differences P = 0.00008). There also was inconsistency in the pooled results of the four trials in adults (12 = 55%), but the trials differed in the magnitude of effect estimates and not in the direction of effects. Downgraded by 1. 8 No serious indirectness: Definitions of adherence differed between the trials, and with current expectations none of the trials were conducted in high TB burden, low‐income countries, where socioeconomic circumstances may differ from those in moderate to low TB burden, high‐income countries. However, these factors may not affect the relative advantage of adherence to the shorter rifampicin regimen over the isoniazid regimen. Not downgraded. 9 No serious imprecision: Though the upper and lower limits of the 95% CI of the pooled relative risk include possibly non‐appreciable and appreciable benefits for adherence to rifampicin, the absolute increase in those adherent to rifampicin compared to INH (particularly in adults: 129 more people per 1000, 95% CI 68 to 203 more per 1000, adherent to rifampicin compared to isoniazid) is likely to represent an appreciable benefit for national TB control programmes, particularly in high TB burden countries. Not downgraded.
10 Serious study limitations: Two of the four included trials (Menzies 2004; Menzies 2008) were judged at high risk of detection bias. Downgraded by 1.
11 Serious inconsistency: The I2 value (68%) indicated significant inter‐trial variability in effect estimates. The heterogeneity was due to appreciably lower adverse events with rifampicin in Chan 2012, where prisoners were given interventions by DOT and had higher adherence (78%), thereby exposing more people to INH, than in the other three trials where self‐administered treatment resulted in lower adherence (62%) and treatment limiting adverse events did not differ significantly between both regiments. Downgraded by 1.
12 Serious imprecision: The upper and lower limits of the 95% CI of the effect estimate include appreciable benefit and no difference in treatment‐limiting adverse events with rifampicin compared to INH. Downgraded by 1. 13One trial (Chan 2012) randomized participants stratified for co‐infection with HBV and HCV; HCV infection was an independent risk factor for developing hepatotoxicity. The other three trials did not report on co‐infection with HBV or HCV.