Background: The LOVIT (Lessening Organ Dysfunction with Vitamin C) trial is a blinded multicenter randomized clinical trial comparing high-dose intravenous vitamin C to placebo in patients admitted to the intensive care unit with proven or suspected infection as the main diagnosis and receiving a vasopressor.
Objective: We aim to describe a prespecified statistical analysis plan (SAP) for the LOVIT trial prior to unblinding and locking of the trial database.
Methods: The SAP was designed by the LOVIT principal investigators and statisticians, and approved by the steering committee and coinvestigators. The SAP defines the primary and secondary outcomes, and describes the planned primary, secondary, and subgroup analyses.
Results: The SAP includes a draft participant flow diagram, tables, and planned figures. The primary outcome is a composite of mortality and persistent organ dysfunction (receipt of mechanical ventilation, vasopressors, or new renal replacement therapy) at 28 days, where day 1 is the day of randomization. All analyses will use a frequentist statistical framework. The analysis of the primary outcome will estimate the risk ratio and 95% CI in a generalized linear mixed model with binomial distribution and log link, with site as a random effect. We will perform a secondary analysis adjusting for prespecified baseline clinical variables. Subgroup analyses will include age, sex, frailty, severity of illness, Sepsis-3 definition of septic shock, baseline ascorbic acid level, and COVID-19 status.
Conclusions: We have developed an SAP for the LOVIT trial and will adhere to it in the analysis phase.
International registered report identifier (irrid): DERR1-10.2196/36261.
Keywords: ascorbic acid; binomial distribution; critical care; health data; infection; intensive care; intensive care unit; intravenous; mortality; organ; organ dysfunction; patient; patient outcome; sepsis; statistical analysis; statistical framework; trial database; vasopressor; vitamin C.
©Neill KJ Adhikari, Ruxandra Pinto, Andrew G Day, Marie-Hélène Masse, Julie Ménard, Sheila Sprague, Djillali Annane, Yaseen M Arabi, Marie-Claude Battista, Dian Cohen, Deborah J Cook, Gordon H Guyatt, Daren K Heyland, Salmaan Kanji, Shay P McGuinness, Rachael L Parke, Bharath Kumar Tirupakuzhi Vijayaraghavan, Emmanuel Charbonney, Michaël Chassé, Lorenzo Del Sorbo, Demetrios James Kutsogiannis, François Lauzier, Rémi Leblanc, David M Maslove, Sangeeta Mehta, Armand Mekontso Dessap, Tina S Mele, Bram Rochwerg, Oleksa G Rewa, Jason Shahin, Pawel Twardowski, Paul Jeffrey Young, François Lamontagne, LOVIT Investigators. Originally published in JMIR Research Protocols (https://www.researchprotocols.org), 20.05.2022.