PURPOSE: Clinical trials using ACCP/SCCM consensus definitions of sepsis as entry criteria fail because enrolled patients whose pathophysiology cannot benefit from the study drug dilute its treatment effects to invisibility. SMART matches sepsis therapies biologically to individual patients, and thereby provides study drugs a fair chance to demonstrate efficacy. This paper describes SMART analyses of four failed sepsis investigations.
METHODS: Clinical trial databases from the E5 antiendotoxin antibody (XOMA), NORASEPT anti-TNF antibody (Bayer), IL-1ra (Synergen), and PAF acetylhydrolase (ICOS) studies were evaluated with SMART using multivariate logistic regression. From baseline data, within each study, mortality prediction models were built separately for placebo and active drug populations. Then, exploration into the relationships between SMART predictions and the outcomes of active drug and placebo arms of each study resulted in models that identified subjects among whom study drugs had maximum treatment effects, if they were to receive active drug. Baseline, pre-randomization data from all patients enrolled into each study was entered into the final models, and placebo/active drug treatment effects were recorded for consensus populations and SMART cohorts.
RESULTS: E5: consensus mortality 101/369 (27.4%) placebo, 102/390 (26.2%) E5 (1.2% absolute; 4.4% relative; p=0.747), versus SMART 32/187 (17.1%) placebo and 16/201 (8.0%) E5 (9.1% absolute; 53.2% relative; p=0.006). NORASEPT: consensus mortality 103/308 placebo (33.4%), 93/315 (29.5%) ant-TNF (3.9% absolute; 11.7% relative; p=0.20) versus SMART 52/110 (47.2%) placebo, 33/95 (34.7%) anti-TNF (12.6% absolute; 26.6% relative; p=0.03). IL-1ra: consensus mortality 101/298 (33.9%) placebo, 86/289 (29.8%) IL-1ra (4.1% absolute; 12.1% relative; p=0.2824) versus SMART 74/133 (55.6%) placebo and 43/123 (34.9%) IL-1ra (20.7% absolute; 37.2% relative; p=0.0009). PAF-AH: consensus mortality 68/304 (22.4%) placebo, 74/309 (23.9%) PAF-AH ((1.5% absolute increase; 6.7% relative; p=0.869) versus SMART 23/130 (17.7%) placebo, 35/121 (28.9%) PAF-AH (11.2% absolute increase; 63.3% relative; p=0.039).
CONCLUSION: SMART matches individual septic patients to beneficial therapies. SMART also predicts serious adverse events.
CLINICAL IMPLICATIONS: Clinical trials designed with SMART will succeed in developing life-saving sepsis therapies. SMART identification of patients subject to adverse effects can increase drug safety.
DISCLOSURE: Gus Slotman, None.