Characteristics of observational analyses
| Overall (n=100) | Studies emulating a real target trial (n=18) | Studies emulating a hypothetical target trial (n=58) | No target trial specified (n=24) | |
| Eligibility criteria | ||||
| Users of the intervention | ||||
| New users | 86 (86%) | 14 (78%) | 52 (90%) | 20 (83%) |
| Prevalent users | 14 (14%) | 4 (22%) | 6 (10%) | 4 (17%) |
| Eligibility defined on characteristics observed during follow-up | 22 (22%) | 5 (28%) | 10 (17%) | 7 (29%) |
| Look-back period to assess eligibility (in years) | 54 (54%) | 10 (56%) | 36 (62%) | 8 (33%) |
| Median(Q1;Q3) | 1(1 1) | 1(0.49;1) | 1(1 1) | 1(0.44;1) |
| Minimum/maximum | 0.08/9.5 | 0.49/9.5 | 0.25/3 | 0.08/1 |
| Intervention strategy* | ||||
| Complex strategy | 33 (33%) | 9 (50%) | 19 (33%) | 5 (21%) |
| Dynamic | 4 (4%) | 0 (0%) | 4 (7%) | 0 (0%) |
| Joint strategies | 17 (17%) | 8 (44%) | 6 (10%) | 3 (12%) |
| Grace period | 16 (16%) | 2 (11%) | 12 (21%) | 2 (8%) |
| Pharmacological intervention | 67 (67%) | 15 (83%) | 36 (62%) | 16 (67%) |
| Comparator strategy* | ||||
| Nature of comparator | ||||
| Active comparator | 46 (46%) | 14 (78%) | 23 (40%) | 9 (38%) |
| No intervention or usual care | 40 (40%) | 2 (11%) | 24 (41%) | 14 (58%) |
| Other comparator strategies | 14 (14%) | 2 (11%) | 11 (19%) | 1 (4%) |
| Pharmacological intervention | 46 (46%) | 15 (83%) | 25 (43%) | 6 (25%) |
| Outcome† | ||||
| Efficacy and/or safety assessment | ||||
| Efficacy | 50 (50%) | 9 (50%) | 22 (38%) | 19 (79%) |
| Safety | 13 (13%) | 1 (6%) | 11 (19%) | 1 (4%) |
| Both efficacy and safety | 37 (37%) | 8 (44%) | 25 (43%) | 4 (17%) |
| Nature of outcome | ||||
| Binary | 19 (19%) | 4 (22%) | 6 (10%) | 9 (38%) |
| Continuous | 5 (5%) | 0 (0%) | 2 (3%) | 3 (12%) |
| Time-to-event | 76 (76%) | 14 (78%) | 50 (86%) | 12 (50%) |
| Follow-up | ||||
| Length of follow-up (in years, if applicable) | ||||
| Median(Q1;Q3) | 2(0.5;6) | 2.21 5 | 2(0.43;5.75) | 1(0.28;5.75) |
| Minimum/maximum | 0/20 | 0.3/11.6 | 0/20 | 0/18 |
| Causal contrast(s) reported by authors | 70 (70%) | 14 (78%) | 46 (79%) | 10 (42%) |
| Observational analogue of ITT effect | 43 (61%) | 9 (64%) | 25 (54%) | 9 (90%) |
| Observational analogue of PP effect | 32 (46%) | 4 (29%) | 26 (57%) | 2 (20%) |
| Both observational analogue of ITT and PP effects | 12 (17%) | 1 (7%) | 9 (20%) | 2 (20%) |
| Other causal contrasts | 10 (14%) | 4 (29%) | 5 (11%) | 1 (10%) |
| Design for emulation | ||||
| Sequential trials analysis with emulation several times of a trial | 18 (18%) | 0 (0%) | 12 (21%) | 6 (25%) |
| Cloning, censoring and weighting | 10 (10%) | 1 (6%) | 8 (14%) | 1 (4%) |
| Statistical methodology | ||||
| A priori sample size or computation of anticipated power | 5 (5%) | 3 (17%) | 2 (3%) | 0 (0%) |
| Propensity score analysis (with baseline confounders) | 58 (58%) | 16 (89%) | 26 (45%) | 16 (67%) |
| Use of g-methods‡ | 20 (20%) | 3 (17%) | 13 (22%) | 4 (15%) |
| Exploration of residual confounding | ||||
| Controls for confounding | 21 (21%) | 5 (28%) | 14 (24%) | 2 (8%) |
| Reporting of E-value | 13 (13%) | 0 (0%) | 8 (14%) | 5 (21%) |
*First reported ones.
†Primary outcome or first reported outcome.
‡g-formula, inverse probability weighting with time-varying weights and/or doubly robust methods; studies that used inverse probability of treatment weighting only with a single weight at baseline were not counted in g-methods. Statistical techniques were not mutually exclusive (eg, a sequential trials analysis with the use of a propensity score weighting in each sequential trial).
ITT, intention-to-treat; PP, per protocol; Q1;Q3, first and third quartiles.