Evaluation of Thiamine as Adjunctive Therapy in COVID-19 Critically Ill Patients

A Two-Center Propensity Score Matched Study

Khalid Al Sulaiman; Ohoud Aljuhani; Maram Al Dossari; Asma Alshahrani; Aisha Alharbi; Rahmah Algarni; Majed Al Jeraisy; Shmeylan Al Harbi; Abdulmalik Al Katheri; Fahad Al Eidan; Abdulkareem M. Al Bekairy; Nouf Al Qahtani; Mashael Al Muqrin; Ramesh Vishwakarma; Ghassan Al Ghamdi


Crit Care. 2021;25(223) 

In This Article


Study Design

A retrospective study of critically ill patients admitted to intensive care units (ICUs) with a confirmed diagnosis of COVID-19 in two-tertiary care centers in Saudi Arabia between March 1 and December 31, 2020. All patients who met the inclusion criteria during the study period were included. Eligible patients have been divided into two groups based on thiamine use as adjunctive therapy during ICU stay; there were no pre-defined criteria at the two centers for thiamine initiation. Intravenous (IV) or enteral) thiamine was given empirically and not based on baseline thiamine levels. Patients were observed during their hospital stay until discharge or in-hospital death, whichever occurred first. The study was approved by the Ministry of National Guard Health Affairs-Institutional Review Board (IRB), Riyadh, Saudi Arabia (Study Number: RC20/589/R).


Patients who were 18 years of age or older and admitted to ICU with confirmed COVID-19 by reverse transcriptase-polymerase chain reaction (RT-PCR) on nasopharyngeal or throat swabs were eligible for inclusion. Patients were excluded if the ICU length of stay (LOS) was less than a day or labeled as "Do-Not-Resuscitate" code status within 24 hours of ICU admission.


This study was conducted in two large, tertiary governmental hospitals; King Abdulaziz Medical City, Riyadh (KAMC-RD) and King Abdulaziz University Hospital, Jeddah (KAUH-JD). The ICUs admit medical, surgical, trauma, burn and transplant patients and operate as a closed unit with 24/7 on-site coverage by critical care board-certified intensivists. The distributions of total enrolled patients were 77% and 23% in KAMC-CR and KAUH-JD, respectively. The primary site was KAMC-RD.

Data Collection

Data gathered from the patients' electronic medical records included demographic data (see Additional file 1), thiamine use, acute physiology and chronic health evaluation II (APACHE II), sequential organ failure assessment (SOFA) and nutrition risk in critically ill (NUTRIC). Comorbidities, vital signs, laboratory tests, the needs for mechanical ventilation (MV), MV parameters (e.g., PaO2/FiO2 ratio, FiO2 requirement) and inflammatory markers (C-reactive protein (CRP), procalcitonin) within 24 hours of ICU admission were recorded. Additionally, ICU complication (s) during ICU stay (e.g., Acute Kidney Injury (AKI), thrombosis/infarction), ICU length of stay (LOS), hospital LOS, mechanical ventilation (MV) duration and ICU/in-hospital mortality were collected for eligible patients.


The primary endpoints were determining the association between using thiamine as adjunctive therapy with the in-hospital and 30-day mortality in critically ill patients with COVID-19. The secondary endpoints include evaluation of MV duration, length of stay and complication (s) during ICU stay (i.e., acute kidney injury, acute liver injury, respiratory failure and thrombosis/infarction).


  • The acute kidney injury was defined using the AKIN definition.[18]

  • Thrombosis/infarction was defined using ICD10-CM code (i.e., Myocardial infarction (MI), ischemic stroke, pulmonary embolism, deep vein thrombosis).[6]

  • Respiratory failure was defined as either hypoxemic respiratory failure (PaO2 < 60 mm Hg with a normal or low arterial carbon dioxide tension (PaCO2) or hypercapnic respiratory failure (PaCO2 > 50 mm Hg) that requires mechanical ventilation.[16]

  • Acute liver injury, defined as alanine aminotransferase (ALT) exceeds three times the normal upper limit (ULN) or doubled in patients with elevated baseline ALT.[6]

Data Management and Statistical Analysis

Categorical variables were reported using numbers and percentages, whereas continuous variables reported using means with standard deviation (SD) or medians with interquartile range (IQR) when appropriate. The normality assumptions were assessed for all numerical variables using a statistical test (i.e., Shapiro–Wilk test) and graphical representation (i.e., histograms and Q–Q plots). We compared categorical variables using the Chi-square or Fisher exact test, normally distributed numerical variables with the Mann–Whitney U test.

Model fit was assessed using the Hosmer–Lemeshow goodness-of-fit test. Multivariable logistic regression and negative binomial regression were used to determine the relationship between thiamine use and different outcomes considered in this study. The odds ratios (OR) and estimates with the 95% confidence intervals (CI) were reported for the associations. No imputation was made for missing data as the cohort of patients in our study was not derived from random selection.

Propensity score matching procedures (Proc PS match) (SAS, Cary, NC) were used to match patients who received thiamine with patients who did not based on patient's baseline severity scores (i.e., APACHE II, SOFA, and NUTRIC scores), systemic use of corticosteroids and study centers. A greedy nearest neighbor matching method was used in which one patient in the control group was matched with each patient in the thiamine (treated) group. This eventually produces the smallest within-pair difference among all available pairs with treated patients. These patients were matched only if the difference in the logits of the propensity scores for pairs of patients from the two groups was less than or equal to 0.5 times the pooled estimate of the standard deviation. We considered a P value of < 0.05 statistically significant and used SAS version 9.4 for all statistical analyses.