Objectives: The Sequential Organ Failure Assessment (SOFA) score originated as a tool for assessing organ dysfunction in critical illness but has expanded to become an outcome measure in clinical trials. We aimed to assess how the SOFA score was used as the primary or secondary endpoint of major randomized controlled trials (RCTs). Data Sources: Independent reviewers searched MEDLINE/PubMed, Scopus, and Embase databases. Study Selection: Articles were selected when they fulfilled: 1) RCT; 2) SOFA score was primary or secondary endpoint; and 3) published in the Lancet, New England Journal of Medicine, or Journal of the American Medical Association. Data Extraction: Data collection included study details, outcomes, statistical differences in SOFA score, choice of score statistics, timepoints of SOFA reporting, and how missing data and competing risks analysis were managed. Data Synthesis: Twenty-three RCTs had SOFA score as outcome measure, eight used it as primary endpoint. Daily maximum SOFA was the key statistic in 11 RCTs, delta SOFA was used in eight, and mean SOFA in four. Mean SOFA was most frequently chosen as primary endpoint (4/8, 50%). There were 18 different outcome assessment timepoints, ranging from 1 to 28 days. Three RCTs reported statistically significant difference in SOFA between groups. Handling of missing SOFA scores was not described in ten of 23 RCTs. When described, it varied from study to study with variable imputation methods and variable accounting for the competing risk of mortality and ICU discharge. Conclusions: There is major variability in the choice of summary statistic for SOFA score analysis and assessment timepoints, when using it as outcome measure in RCTs. There was either no information or great variability in the handling of missing values, use of imputation, and accounting for competing risk. The current use of SOFA scores in RCTs lacks sufficient reproducibility and statistical and methodological robustness.
Mastering the Sequential Organ Failure Assessment Score: Critical Choices of Score Statistic, Timing, Imputations, and Competing Risk Handling in Major Trials—A Systematic Review / Marmiere, Marilena; D'Amico, Filippo; Monti, Giacomo; Landoni, Giovanni. - In: CRITICAL CARE MEDICINE. - ISSN 0090-3493. - (In corso di stampa). [Epub ahead of print] [10.1097/ccm.0000000000006532]
Mastering the Sequential Organ Failure Assessment Score: Critical Choices of Score Statistic, Timing, Imputations, and Competing Risk Handling in Major Trials—A Systematic Review
Marmiere, MarilenaPrimo
;D'Amico, Filippo;Monti, GiacomoPenultimo
;Landoni, GiovanniUltimo
In corso di stampa
Abstract
Objectives: The Sequential Organ Failure Assessment (SOFA) score originated as a tool for assessing organ dysfunction in critical illness but has expanded to become an outcome measure in clinical trials. We aimed to assess how the SOFA score was used as the primary or secondary endpoint of major randomized controlled trials (RCTs). Data Sources: Independent reviewers searched MEDLINE/PubMed, Scopus, and Embase databases. Study Selection: Articles were selected when they fulfilled: 1) RCT; 2) SOFA score was primary or secondary endpoint; and 3) published in the Lancet, New England Journal of Medicine, or Journal of the American Medical Association. Data Extraction: Data collection included study details, outcomes, statistical differences in SOFA score, choice of score statistics, timepoints of SOFA reporting, and how missing data and competing risks analysis were managed. Data Synthesis: Twenty-three RCTs had SOFA score as outcome measure, eight used it as primary endpoint. Daily maximum SOFA was the key statistic in 11 RCTs, delta SOFA was used in eight, and mean SOFA in four. Mean SOFA was most frequently chosen as primary endpoint (4/8, 50%). There were 18 different outcome assessment timepoints, ranging from 1 to 28 days. Three RCTs reported statistically significant difference in SOFA between groups. Handling of missing SOFA scores was not described in ten of 23 RCTs. When described, it varied from study to study with variable imputation methods and variable accounting for the competing risk of mortality and ICU discharge. Conclusions: There is major variability in the choice of summary statistic for SOFA score analysis and assessment timepoints, when using it as outcome measure in RCTs. There was either no information or great variability in the handling of missing values, use of imputation, and accounting for competing risk. The current use of SOFA scores in RCTs lacks sufficient reproducibility and statistical and methodological robustness.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.