Statin use and 28-day survival in critically ill patients with non-acute myocardial infarction cardiogenic shock: A retrospective cohort study accounting for treatment timing.
The apparent survival advantage associated with conventional time-fixed statin exposure was not reproduced after treatment timing was addressed, and these findings do not support an independent 28-day survival benefit from statin administration within the first 24 hours of ICU admission.
Key Findings
Results
Conventional time-fixed analysis suggested lower 28-day mortality among statin users in non-AMI cardiogenic shock patients, both before and after propensity-score matching.
In the conventional cohort, 475 patients received a statin during the ICU stay and 516 did not.
Before PSM, adjusted HR for 28-day mortality was 0.703 (95% CI, 0.539–0.918; P = .010).
After 1:1 PSM (255 patients per group), adjusted HR was 0.709 (95% CI, 0.507–0.993; P = .045).
Exposure was defined as receipt of any statin at any point during the ICU stay, regardless of timing.
Results
After applying a 24-hour landmark analysis to account for treatment timing, the apparent survival benefit of statin use entirely disappeared.
In the 24-hour landmark cohort, 311 patients received a statin within 24 hours and 680 did not.
Before PSM, adjusted HR was 1.127 (95% CI, 0.854–1.487; P = .398).
After 1:1 PSM (255 patients per group), adjusted HR was 1.190 (95% CI, 0.833–1.701; P = .339).
The landmark design classified exposure by statin administration during the first 24 hours and began follow-up at 24 hours, thereby eliminating immortal-time bias.
Results
In 4-category analyses examining new initiators versus prior users, the apparent benefit seen among new initiators in conventional models disappeared when the 24-hour landmark definition was applied.
The 4-category analysis distinguished new statin initiators from patients already on statins prior to ICU admission.
The conventional time-fixed model produced an apparent benefit specifically among new initiators.
This benefit was not reproduced under the landmark exposure definition, suggesting the finding was an artifact of treatment-timing bias.
This pattern highlights treatment-selection bias as a potential confounding influence in conventional analyses.
Results
Atorvastatin was the predominant first statin agent in both matched exposed groups across the conventional and landmark cohorts.
Statin agent, first recorded dose, route, and time to first administration were descriptively summarized in matched cohorts.
Atorvastatin was identified as the predominant first statin in both the conventional and 24-hour landmark matched exposed groups.
Other statin agents were also represented but were less common.
Specific mechanical-support, anti-inflammatory, and vasoactive co-treatments were also descriptively summarized.
Methods
The study used MIMIC-IV data from 991 adults with non-AMI cardiogenic shock to evaluate statin exposure and 28-day mortality.
Data were drawn from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database.
The total cohort included 991 adults with non-acute myocardial infarction cardiogenic shock.
Statin exposure was ascertained from medication records including atorvastatin and other statin agents.
Separate 1:1 propensity-score matching procedures and Cox proportional hazards models were performed for each exposure definition.
The study design was a retrospective cohort study.
Discussion
The authors identified immortal-time bias, treatment-selection bias, and co-treatment bias as potential influences explaining the spurious benefit observed in conventional analyses.
Immortal-time bias arises in conventional time-fixed analyses when patients who survive long enough to receive a treatment late in their ICU stay are classified as 'exposed,' making the pre-treatment survival time incorrectly attributed to the treatment.
Treatment-selection bias may occur if clinicians preferentially administer statins to patients who are more hemodynamically stable or have better prognoses.
Co-treatment bias refers to the potential confounding effect of concurrent treatments such as mechanical support, anti-inflammatory agents, and vasoactive drugs.
The authors stated these findings 'highlight the potential influence of immortal-time, treatment-selection, and co-treatment biases.'
What This Means
This research examined whether taking statins (cholesterol-lowering medications) during an ICU stay improves survival for critically ill patients experiencing cardiogenic shock not caused by a heart attack. Using a database of nearly 1,000 ICU patients, researchers found that when they used a standard approach—simply comparing patients who ever received a statin during their ICU stay to those who did not—statin users appeared to have significantly better 28-day survival odds. However, this standard approach has a well-known flaw: patients who survive long enough to receive a statin days into their stay are automatically counted as 'survivors' for that early period, making the drug look more beneficial than it really is.
To address this flaw, the researchers used a 'landmark analysis,' which only looked at whether patients received a statin within the first 24 hours of ICU admission and then tracked survival from that 24-hour point onward. With this corrected approach, the apparent survival benefit completely vanished—statin users and non-users had essentially identical 28-day mortality rates. This held true both before and after the researchers used statistical matching to make the two groups as comparable as possible.
This research suggests that the previously observed association between statin use and better ICU survival in this patient group was likely a statistical artifact caused by treatment-timing bias rather than a true drug benefit. The findings do not support giving statins specifically to improve short-term survival in this type of critically ill patient, and they underscore the importance of accounting for when a treatment is given—not just whether it was given—when studying drugs in observational ICU data.
Cheng T, Wang P, Ren J, Ji S, Zhang G, Liu Z, et al.. (2026). Statin use and 28-day survival in critically ill patients with non-acute myocardial infarction cardiogenic shock: A retrospective cohort study accounting for treatment timing.. Medicine. https://doi.org/10.1097/MD.0000000000050754