Early pharmacologic venous thromboembolism prophylaxis and in-hospital mortality in ICU patients with intracerebral hemorrhage: a landmark, propensity-weighted cohort study.

Early pharmacologic VTE prophylaxis (within 48 hours of ICU admission) was associated with lower in-hospital mortality in ICH patients after inverse probability of treatment weighting.

• Adjusted odds ratio for in-hospital mortality was 0.72 (95% CI 0.55–0.94)
• Cause-specific hazard ratio was 0.65 (95% CI 0.50–0.86, p = 0.003)
• Weighted risk difference was -4.0%
• In-hospital death occurred in 10.8% of the early prophylaxis group vs. 17.4% of the no early prophylaxis group

Early pharmacologic VTE prophylaxis was not associated with a statistically significant difference in coded VTE events.

• Adjusted OR for coded VTE was 1.20 (95% CI 0.85–1.69; p = 0.31)
• This secondary outcome was binary and based on coded (administrative/clinical) VTE diagnoses
• The confidence interval crossed 1.0, indicating no statistically significant increase in VTE risk with early prophylaxis

The association between early prophylaxis and lower in-hospital mortality was directionally consistent across most sensitivity analyses.

• A time-varying exposure model (designed to reduce immortal-time bias) yielded HR 0.61 (95% CI 0.49–0.76)
• A competing-risk cumulative-incidence analysis was also directionally consistent with the primary result
• The 24-hour landmark estimate was not statistically significant, indicating some variability by landmark choice
• Results were described as 'directionally consistent across most sensitivity analyses'

After applying the 48-hour landmark design and exclusion criteria, 2,380 patients formed the analytic cohort from an initial pool of 2,754 ICH ICU patients.

• 2,754 ICH ICU patients were identified in MIMIC-IV v3.1 (2008–2019)
• 2,407 patients were alive and in hospital at the 48-hour landmark
• 27 patients were excluded for receiving therapeutic anticoagulation within 48 hours, leaving 2,380 in the analytic cohort
• Early prophylaxis group: n = 655; no early prophylaxis group: n = 1,725
• The no early prophylaxis group comprised both later (delayed) initiators and never-treated patients

Inverse probability of treatment weighting successfully balanced all 17 baseline covariates between the early and no-early-prophylaxis groups.

• All 17 of 17 covariates achieved a standardized mean difference of less than 0.1 after IPTW
• Covariates included the baseline Glasgow Coma Scale (GCS)
• Missing GCS values were handled using multiple imputation
• A total of 17 baseline covariates were adjusted for in the propensity model

The study used a landmark design with a fixed 48-hour time point to classify exposure and define eligibility, specifically to address immortal-time bias.

• Time zero and eligibility were defined as ICU admission plus 48 hours
• Only patients alive and in hospital at 48 hours were included, ensuring all patients had equal opportunity for early prophylaxis exposure
• Exposure was fixed at the landmark: early prophylaxis defined as pharmacologic prophylaxis started ≤48 hours; no early prophylaxis defined as not started by 48 hours
• A supplementary time-varying exposure analysis was also conducted specifically to further reduce immortal-time bias

Residual confounding by indication due to unavailable neuroimaging severity measures is a key limitation, and a causal interpretation of the mortality association is not warranted.

• Key neuroimaging severity measures were unavailable, including hematoma volume, location, intraventricular extension, and hematoma expansion
• These unmeasured variables are likely related to both the decision to initiate prophylaxis and to mortality, representing confounding by indication
• The authors explicitly state that 'a causal interpretation is not warranted'
• Randomized trials are needed to establish causality