Design characteristics of neonatal trials that included multiple births (n=62)
| Characteristic | N=62* |
| Proportion of women with multiple births, % | 9 (2, 28) |
| Unknown | 40 (65%) |
| Proportion of infants from a multiple birth, % | 26 (19, 31) |
| Unknown | 3 (5%) |
| Reported information enabling assessment of balance of multiples between groups | |
| No | 2 (3%) |
| Yes | 60 (97%) |
| Highest order multiples included in the trial | |
| Twins | 23 (37%) |
| Triplets | 6 (10%) |
| Higher order multiples | 2 (3%) |
| Unclear | 31 (50%) |
| Who was randomised | |
| Site was randomised | 2 (3%) |
| Mother (cluster randomised†) | 22 (35%) |
| Infant | 38 (61%) |
| If infant was randomised, how were multiples randomised | |
| Infants randomised independently | 14 (37%) |
| Randomised to same group (cluster randomised†) | 9 (24%) |
| Unclear | 15 (39%) |
| (Infant not randomised) | 24 |
| Method of randomisation | |
| Stratified blocks (covariate-adaptive randomisation‡) | 39 (63%) |
| Minimisation (covariate-adaptive randomisation‡) | 9 (15%) |
| Simple | 5 (8%) |
| Other§ | 7 (11%) |
| Unclear | 2 (3%) |
| If covariate-adaptive randomisation‡ was used, was multiple births a balancing factor | |
| No | 38 (79%) |
| Yes | 10 (21%) |
| (Covariate-adaptive randomisation not used) | 14 |
| Who received the intervention | |
| Mother | 22 (35%) |
| Infant | 39 (63%) |
| Both infant and mother | 1 (1.6%) |
| On what level is the outcome defined | |
| Cluster level | 11 (18%) |
| Infant level | 51 (82%) |
| Did the sample size calculation account for clustering due to multiples | |
| No | 54 (87%) |
| Yes | 8 (13%) |
*Median (IQR); n (%).
†Indicates cluster randomised trials (n=31).
‡Indicates covariate balancing method was used (n=48).
§Other randomisation methods include unstratified blocks and stratified random number lists (with no blocking).