Table 3.
ORs for the likelihood of an article being cited, based on univariate analyses of 183 articles cited in 51 evidence assessment units from 33 reviews
| n (%) | OR (95% CI) | P value | |
| Main message of cited articles | |||
| Neutral (no directional effect or association) | 39 (21) | 1 (ref) | |
| No conclusion directly relevant to the LES–BW relationship | 26 (14) | 1.64 (0.95 to 2.84) | 0.08 |
| Decrease BW/more beneficial | 32 (17) | 1.31 (0.76 to 2.27) | 0.33 |
| Increase BW/less beneficial | 54 (30) | 1.11 (0.68 to 1.85) | 0.68 |
| Unable to draw a conclusion from the article | 9 (5) | 1.11 (0.43 to 2.50) | 0.81 |
| Evidence is insufficient to draw a conclusion | 20 (11) | 1.05 (0.53 to 2.00) | 0.87 |
| Cited article type | |||
| Systematic review with meta-analysis | 16 (9) | 1 (ref) | |
| Systematic review without meta-analysis | 9 (5) | 0.85 (0.36 to 1.90) | 0.70 |
| Randomised controlled trial | 51 (28) | 0.82 (0.48 to 1.46) | 0.48 |
| Observational study | 72 (39) | 0.65 (0.39 to 1.16) | 0.13 |
| Animal | 13 (7) | 0.63 (0.27 to 1.38) | 0.26 |
| Narrative review | 18 (10) | 0.38 (0.16 to 0.86) | 0.03 |
| Other | 1 (1) | 0.22 (0.00 to 2.67) | 0.44 |
| Cited article population‡ | |||
| Adults | 85 (46) | 1 (ref) | |
| Children | 49 (27) | 2.27 (1.59 to 3.25) | <0.0001 |
| Both | 32 (17) | 1.01 (0.60 to 1.63) | 0.98 |
| Sample size*,§ | 124 (68) | 1.00 (0.83 to 1.21) | 1.00 |
| Number of authors | 181 (99) | 1.05 (1.00 to 1.10) | 0.06 |
| Journal impact factor, current (2018)† | 179 (98) | 1.15 (1.00 to 1.31) | 0.04 |
| Journal impact factor, last 5 years† | 178 (97) | 1.13 (0.98 to 1.30) | 0.08 |
| Years since cited article was published | 183 (100) | 1.00 (0.97 to 1.02) | 0.73 |
Logistic mixed-effects regression. All analyses are additionally adjusted for overdispersion.
Bold value indicates result is statistically significant with p<0.05 or lower.
*Sample size was base 10 log-transformed, so OR is the change per 10-fold change in study population.
†Journal impact factor was base 2 log-transformed, so OR is the change per twofold change in journal impact factor.
‡Data on population were only extracted for articles considering human subjects.
§Data on sample size were only extracted for primary evidence (ie, not for reviews).
BW, body weight; CI, confidence interval; LES, low-energy sweeteners; OR, odds ratio; ref, reference variable.