| Potential source of heterogeneity | Subgroup;* no. of studies | RR (95% CrI) in subgroups | Meta-regression coefficient (95% CrI) | p value (meta-regression coefficient > 0)† |
|---|---|---|---|---|
| Study quality | B or C: 6 A: 4 | 0.17 (0.03 to 0.64) 0.27 (0.07 to 0.71) | 0.4 (-1.3 to 2.4) | 0.70 |
| Type of probiotic | Mixture: 5 Lactobacillus only: 5 | 0.31 (0.07 to 0.68) 0.17 (0.04 to 0.57) | -0.56 (-2.1 to 1.3) | 0.26 |
| Probiotic dosage,‡ dichotomous§ | < 50 × 109 CFUs: 3 ≥ 50 × 109 CFUs: 6 | 0.18 (0.03 to 0.67) 0.27 (0.07 to 0.68) | 0.4 (-1.5 to 2.3) | 0.68 |
| Support from manufacturer | No: 4 Yes: 6 | 0.21 (0.02 to 0.67) 0.26 (0.07 to 0.65) | 0.16 (-1.3 to 2.9) | 0.60 |
| Proportion of cases of CDAD in placebo group, dichotomous¶ | < 6%: 5 ≥ 6%: 5 | 0.46 (0.12 to 1.05) 0.17 (0.05 to 0.41) | -1.0 (-2.4 to 0.5) | 0.08 |
| Proportion of cases of CDAD in placebo group, continuous | - | - | -3.2 (-11.5 to 7.6) | 0.21 |
| Probiotic dosage,‡ continuous** | - | - | 0.06 (-0.02 to 0.06) | 0.81 |
Note: CDAD = Clostridium-difficile-associated diarrhea, CFU = colony-forming unit, CrI = credible interval, RR = risk ratio.
*Subgroup listed first was the control group for the comparison, e.g., P (RR in studies of quality A > studies of quality B or C) = 0.7
†p values (regression coefficient > 0) close to 0 or 1 indicate a strong association between the risk ratio and the covariate.
‡For studies that used a mixture of different types of Lactobacillus probiotics, we estimated the average dosage assuming all types of probiotics were in equal proportions.
§Missing for 1 study (Heimburger and colleagues31).
¶Used to represent population risk (in inpatients given antibiotics) for development of C. difficile-associated diarrhea.
**One outlying study (Selinger and colleagues25) was omitted because the extremely high dosage reported resulted in model convergence problems.