Logistic regression analysis of effect of treatment by culture result.a
| Group | n | Odds ratio (95% CI) | NNT (benefit) (95%CI)b | NNT (harm)a (95%CI)b |
|---|---|---|---|---|
| All patients | 163 | 1.25 (0.64 to 2.39) | 18.97 (NNT[harm] 8.92 to∞ to NNT[benefit] 5.47) | – |
| Culture positive patients | 50c | 2.58 (0.79 to 8.42) | 4.36 (NNT[harm] 17.09 to∞ to NNT[benefit] 2.36) | – |
| Culture negative patients | 112c | 0.85 (0.37 to 1.86) | – | 25.25 (NNT[harm] 4.10 to∞ to NNT[benefit] 7.97) |
↵a Corrected for age.
↵b The odds ratios obtained from the regression analysis were converted into relative risks. The relative risks were converted into risk differences and numbers needed to treat using the clinical outcome rate in the placebo group as the background risk. The numbers needed to benefit and the numbers needed to harm and their 95% confidence intervals are presented as recommended by Altman.23
↵c The culture result of one patient was missing. P-value of the interaction term is 0.122. NNT (benefit) = number needed to benefit. NNT (harm) = number needed to harm.