Table 3.
Primary and Secondary Outcomes in the Noncritically Ill Per-Protocol Population
| Therapeutic-Dose Anticoagulation | Usual-Care Thromboprophylaxis |
Adjusted Difference in Risk (95% CrI), Percentage Points | Adjusted OR (95% CrI) | Probability of Superiority of Therapeutic-Dose Anticoagulation, % | |
|---|---|---|---|---|---|
| Primary outcome, organ support-free daysa | |||||
| Overall groupb | 743/919 (80.8)c | 640/839 (76.3)c | 5.1 (1.2-8.6) | 1.36 (1.07-1.74) | 99.3 |
| D-dimer groupd | |||||
| High level | 208/259 (80.3)c | 159/223 (71.3)c | 7.9 (1.9-13.9) | 1.53 (1.10-2.32) | 99.7 |
| Low level | 384/474 (81.0)c | 341/422 (80.8)c | 2.6 (−3.5 to 6.5) | 1.19 (0.81-1.63) | 81.7 |
| Unknown level | 151/186 (81.2)c | 140/194 (72.2)c | 7.6 (1.3-14.5) | 1.52 (1.07-2.51) | 99.1 |
| Secondary outcomes | |||||
| Survival until hospital dischargea | 858/919 (93.4) | 771/839 (91.9) | 2.0 (−0.7 to 3.0) | 1.35 (0.91-2.00) | 93.7 |
| All thrombotic events or deathe | 67/921 (7.3) | 85/839 (10.1) | −3.0 (−5.0 to −0.4) | 0.68 (0.48-0.96) | 98.6 |
| Major bleedingf | 20/921 (2.2) | 8/839 (1.0) | - | - | - |
Values are n/N (%) unless otherwise indicated. The primary outcome of organ support-free days (OSFDs) consisted of death during index hospitalization, censored at 90 days, and the number of days free of cardiovascular or respiratory organ support among survivors. OSFD was modeled as an ordinal outcome and counted up to day 21 among patients who survived to hospital discharge. Results for this outcome are presented in the entire noncritically ill per-protocol analysis population and in each prespecified D-dimer cohort. The primary adaptive model estimated treatment effects through the use of a Bayesian hierarchical approach in critically and noncritically ill patients, the latter stratified on the basis of baseline D-dimer. Additionally, a model assuming a single treatment effect in all noncritically ill patients in the per-protocol analysis regardless of their baseline D-dimer was evaluated. This primary model for OSFDs permitted dynamic borrowing across illness-severity and D-dimer groups (similar treatment effects are shrunk together on the basis of their degree of similarity). Results from a sensitivity analysis without dynamic borrowing are provided in Supplemental Table 3. Survival until hospital discharge and the composite of thrombotic events or death were modeled as dichotomous outcomes, without borrowing. Models were adjusted for age, sex, trial site, D-dimer group, and enrollment period. The ORs summarize the comparison of therapeutic-dose anticoagulation group vs usual-care thromboprophylaxis.
CrI = credible interval.
For these outcomes, an OR >1.00 indicates benefit with therapeutic-dose anticoagulation.
This model assumes a single treatment effect in all noncritically ill patients in the per-protocol analysis regardless of their baseline D-dimer. Dynamic borrowing of information on treatment effect from critically ill patients was permitted, which occurred to an extent reflective of how similar the treatment effects were. In the context of divergent treatment effects by disease severity, little dynamic borrowing occurred between critically ill and noncritically ill patients, and the results from a sensitivity analysis assuming independent treatment effects were similar. In the overall trial modified intention-to-treat population of noncritically ill patients,5 the OR was 1.27; 95% CrI: 1.03-1.58; probability of superiority of therapeutic-dose anticoagulation: 98.6%.
The median value for OSFD was 22 in both the therapeutic-dose anticoagulation and usual-care thromboprophylaxis arms in all noncritically ill patient groups. Accordingly, the proportion of patients who survived to hospital discharge without receipt of organ support (22 on the ordinal scale–the most common value) is reported.
The noncritically ill primary adaptive stopping groups evaluated the treatment effect based on baseline D-dimer groups, defined as high D-dimer (≥2 times local upper limit of normal for assay), low D-dimer (<2 times local upper limit of normal for assay), and unknown D-dimer.
For this outcome, an OR <1.00 indicates benefit with therapeutic-dose anticoagulation.
No modeling was performed for the outcome of major bleeding because of the low number of events.