An adaptive dose-finding study can identify the most promising arm without establishing that the arm succeeded. RACE used a prespecified decision rule to make that distinction. Primary report
Study snapshot
The phase 2 trial enrolled 250 adults with septic shock and allocated them adaptively among levocarnitine regimens and placebo. The most promising regimen had a posterior probability of superiority of 0.78 for the organ-failure comparison, below the required 0.90 threshold. The investigators concluded that the trial did not demonstrate a meaningful reduction in cumulative organ failure at 48 hours. Primary report
Our appraisal: a ranking is not a success criterion
One arm will usually look better than the others even when none provides a convincing benefit. Selecting that arm for discussion does not erase the uncertainty or the threshold set before the results were known.
The posterior probability must also retain its meaning. It is a probability statement within the specified statistical model, not the percentage of patients who benefited. Describing it as a response rate would misrepresent both the analysis and the clinical outcome.
Adaptive allocation explains why group sizes differ. That imbalance should be interpreted through the design rather than assumed to show arbitrary assignment. Full review should inspect the adaptation rules, prior assumptions, and how the final analysis accounts for the allocation process.
Organ-failure scores and mortality are distinct endpoints. A development decision based on one cannot be rewritten as established survival benefit. This hospital-based trial also provides no evidence for general wellness or exercise claims. The useful conclusion is that the prespecified phase 2 criterion was not met, while future hypotheses require new testing.
Frequently asked questions
Did the most promising arm meet the required threshold?
No.
Does a 0.78 posterior probability mean 78% of patients benefited?
No. It describes uncertainty in the statistical model.
Why were group sizes unequal?
The trial used adaptive randomization.
Does best-performing mean proven effective?
No.
Is an organ-failure score identical to mortality?
No. They are different outcomes.
Can RACE support an athletic-performance claim?
No. It studied critically ill patients with septic shock.
What should full review inspect?
The decision rule, priors, adaptive algorithm, missing data, and outcome-specific estimates.