Table 2. Treatment Effects for Traditional and WHO Risk Level Reduction Definitions of Response During the Last 4 Weeks of Treatmenta.
| Clinical Trial | Sample Sizeb | Responder Outcome | Placebo, No. (%) | Medication, No. (%) | h (95% CI)c,d | NNT (95 % CI)d |
|---|---|---|---|---|---|---|
| Naltrexone24,e | Placebo, 302; medication, 288 | Abstinent | 85 (28.1) | 100 (34.7) | 0.142 (−0.020 to 0.303) | 16 (111.9 [H] to ∞ to 7.1 [B]) |
| No heavy drinking days | 133 (44.0) | 147 (51.0) | 0.140 (−0.021 to 0.302) | 15 (96.2 [H] to ∞ to 6.6 [B]) | ||
| WHO 2-level reduction | 197 (65.2) | 216 (75.0) | 0.214 (0.053 to 0.375) | 11 (5.8 [B] to 41.4 [B]) | ||
| WHO 1-level reduction | 238 (78.8) | 240 (83.3) | 0.116 (−0.046 to 0.277) | 23 (55.7 [H] to ∞ to 9.2 [B]) | ||
| Varenicline25 | Placebo, 101; medication, 96 | Abstinent | 4 (4.0) | 7 (7.3) | 0.146 (−0.133 to 0.426) | 30 (32.6 [H] to ∞ to 10.3 [B]) |
| No heavy drinking days | 15 (14.9) | 23 (24.0) | 0.232 (−0.048 to 0.511) | 11 (54.3 [H] to ∞ to 5.0 [B]) | ||
| WHO 2-level reduction | 42 (41.6) | 53 (55.2) | 0.273 (−0.006 to 0.553) | 8 (486.2 [H] to ∞ to 3.6 [B]) | ||
| WHO 1-level reduction | 54 (53.5) | 67 (69.8) | 0.338 (0.058 to 0.617) | 7 (3.4 [B] to 34.3 [B]) | ||
| Topiramate26 | Placebo, 185; medication, 179 | Abstinent | 5 (2.7) | 21 (11.7) | 0.369 (0.163 to 0.574) | 12 (7.0 [B] to 26.2 [B]) |
| No heavy drinking days | 24 (13.0) | 37 (20.7) | 0.207 (0.002 to 0.413) | 13 (6.5 [B] to 1581.1 [B]) | ||
| WHO 2-level reduction | 62 (33.5) | 80 (44.7) | 0.230 (0.024 to 0.435) | 9 (4.7 [B] to 81.8 [B]) | ||
| WHO 1-level reduction | 100 (54.1) | 98 (54.7) | 0.014 (−0.192 to 0.219) | 144 (10.5 [H] to ∞ to 9.2 [B]) |
Abbreviations: B, number needed to benefit; H, number needed to harm; NNT, number needed to treat; WHO, World Health Organization.
Participants with any missing drinking data were imputed as nonresponders for all outcomes. Percentages are rounded; Cohen h NNT are rounded and are based on unrounded percentages. The WHO 1- and 2-level reduction outcomes represent the percentage of participants who reduced drinking by 1 or more and 2 or more WHO drinking risk levels, respectively.
Denominator for prevalence estimates.
Cohen h = φMedication – φPlacebo, where φ = 2arcsin(percentage)½ is a measure of effect size for dichotomous outcomes, with the interpretation: small, 0.20; medium, 0.50; and large, 0.80.
For Cohen h, 95% CIs that do not include zero are statistically significant (P < .05). For NNT, statistically significant estimates are indicated by 95% CIs with lower and upper bounds both expressed as number needed to benefit. Nonsignificant estimates are indicated by 95% CIs with a lower bound expressed as number needed to harm, indicating less efficacy in the medication than the placebo group, and an upper bound expressed as number needed to benefit; infinity is included in the 95% CI because NNT is infinity when the absolute risk reduction is 0.
Participants in the WHO low-risk drinking level at baseline (n = 17) were excluded.