We conducted this randomized (1:1) parallel controlled superiority trial at Mansoura University COVID-19 Isolation Hospital, Mansoura city, Egypt. According to declaration of Helsinki, we obtained institutional review board approval (R.21.04.1300) on April 2nd, 2021and registered the study at clinical trial.gov (NCT04952350) on July 1st, 2021. We obtained a written informed consent from patients or the patient’s legal representative if the patient was unable to give consent. We wrote this manuscript according to the CONsolidated Standards of Reporting Trials (CONSORT) guidelines.14 For detailed methodology and related documents, the reader can find the published protocol.15
Two hundred and twenty adult patients (≥ 18 years old) with severe or critical COVID-19 who were admitted to the Mansoura University isolation hospital - from 18th August 2021 to 6th October 2021 - were included in the study. We included patients who had COVID-19 diagnosed by PCR, clinically or radiologically. PCR was then used to confirm all cases.
According to the WHO definition16, cases are classified as severe or critical, with severe cases having clinical signs of severe pneumonia and SpO2 < 90% on room air or RR > 30 breaths/min without any critical criteria. While critical cases have ARDS, sepsis, septic shock, pulmonary embolism, acute coronary syndrome, or acute stroke.
Exclusion criteria included chronic statin use, serum creatine kinase (CK) levels greater than 5 times the upper limit of normal (ULN), serum transaminases greater than 5 times the ULN, acute hepatic failure, chronic liver disease with Child-Pugh Classification C, a history of rhabdomyolysis or myopathies, or severe renal impairment not receiving renal replacement therapy (estimated Cr cl 30 ml/min). Additionally, we excluded pregnant and lactating women, patients who are expected to die within 48 hours, or patients on cyclosporine, ritonavir, or chronic colchicine.
Randomization and Allocation Concealment
A randomization table (1:1) was created by a pharmacist who was not involved in the study using computer-generated random permuted blocks (4, or 6 in each block). The same pharmacist assigned patients to either group without providing the study group with the randomization table. After evaluating the inclusion criteria, randomization was carried out within 24 hours of hospital admission.
For 28 days, patients were randomly assigned to receive either atorvastatin (Atorstat, Delta Pharm Co., Egypt) 40 mg/day orally (group A) or placebo (group B). During the hospitalization, all cases received corticosteroids and prophylactic anticoagulation. All patients received the medical care according to the local hospital protocol and on discretion of the treating physician. The placebo is similar to the original drug in terms of the drug package, tablet color, consistency, and size. Unconscious or ventilated patients received the drug through a nasogastric tube.
Blinding
Patients, caregivers, data collectors, and data analysts were all blinded to the study group. Delta Pharma (Egypt) supplied the atorvastatin and a similar placebo. The company did not participate in any aspect of the study, including participant recruitment, data collection, data analysis, or results reporting. The company prepared a similar drug and placebo package (labelled as A and B). Even the pharmacist, who is involved in treatment allocation, had no idea about what treatment group is – only A or B. In this manuscript we used “A” to describe the Atorvastatin group and “B” to describe the placebo group. This is not necessarily the drug labels used in the study.
Criteria for discontinuation
We daily monitored patients for adverse events; if the patient developed myalgia or unexplained weakness, we checked CK levels and stopped if CK was > 10 times ULN, drug-induced hepatitis, or serum transaminases were > 8 times ULN. Patients who received azole antifungals had a lower monitoring threshold for liver injury and rhabdomyolysis.
According to local hospital protocol, all patients received the standard of care. Antiviral therapy was permitted.
Data Safety and Monitoring Board
The assignment groups were hidden from the Data Safety and Monitoring Board (DSMB). The true labels and the randomization table were kept in a closed envelope for emergency unmasking by one of the directors of the Mansoura COVID-19 research council who was not participating in the study. The envelope was only supposed to be opened if the DSMB decided to unmask for safety reasons; however, unmasking was not required during this study.
Measured outcomes
The primary outcome was mortality from any cause within 28 days of randomization. Patients or their relatives were contacted by the trial staff to collect this data.
Main Secondary outcomes were six-month mortality, in-hospital mortality, need for invasive mechanical ventilation (Yes/No binary outcome), oxygen support duration (days), time to clinical improvement (defined as 2 points reduction in the WHO disease ordinal progression scale17 or discharge, whatever came first), and ICU and hospitalization length of stay in survived patients (days). Apart from the six-month mortality, these data were collected for the duration of the hospitalization.
Additional outcomes included: (1) serious adverse effect resulting in drug discontinuation; (2) CRP on days 3, 7, 14, and 28; (3) SOFA score on days 3, 7, 14, and 28; (4) COVID19 WHO disease progression scale on days 3, 7, 14, and 28; and (5) incidence of acute kidney injury (AKI), defined as an increase in Scr of 0.3 mg/dl in 48 hours or 50% increase in Scr in 7 days or Oliguria for 6 hours.18 We reported data about CRP, D-dimers, SOFA scores, WHO disease progression score, and AKI only during the hospitalization period. Venous thrombo-embolism was collected according to the suspicion of the treating physician and the subsequent appropriate investigations.
We did not change the study outcomes after the commencement of the study; however, we did not collect the long-term outcomes except the six-month mortality.
Sample size
A sample size of 97 patients in each group would achieve at least 80% power to detect a risk difference of 0.2 (20%) in the 28-day all-cause mortality (primary outcome) between the null hypothesis (both proportions are 0.6) and the alternative hypothesis (the proportion in the non-statin group would be 0.4).
We assumed a significance level (α) of 0.05 and used the Chi-square test of independent proportions in MedCalc software. To account for the estimated loss-to-follow-up, we increased the sample size in each group to 110 patients.
Statistical analysis plan
The intention-to-treat strategy with available case analysis is used in the primary analysis. Categorical variables are presented as percentages and proportions. For parametric data, continuous variables are presented as a mean (standard deviation) or as a median (25th-75th percentile) for non-parametric data. In the event of missing data, we reported the denominator or the number (percentage) in the tables or text, as appropriate. The SAMPLE guidelines were followed in reporting the statistical results.19
A univariate analysis of demographic features correlated with study groups was performed using the Chi-square test and the t-test, reporting the 95% confidence interval and the P-value. We compared the 28-day all-cause mortality rate (primary outcome) in each group using the chi-square test, reporting the rate ratio and 95% confidence interval.
SPSS version 28 was used for statistical analysis. The statistical significance level is p-value 0.05.
Interim analyses
We conducted interim analyses of in-hospital mortality after recruiting 25, 50, and 75% of the planned sample size. The DSMB and IRB members were blinded to the study allocation and were kept up to date on the results of the interim analyses. The results of the interim analyses were published with the protocol.15
Subgroup analyses and regression analyses
As pre-planned, we performed a subgroup analysis of the 28-day mortality and six-month mortality by severity, as defined 48 hours after recruitment. Subgroup analysis by invasive mechanical ventilation (IMV) was not performed due to the very low event rate of IMV after 48 hours of recruitment (eight cases in group A and ten cases in group B).
We used a multiple logistic regression to test the association of atorvastatin use and 28-day and six-month all-cause mortality, while controlling for gender, age, comorbidities (HTN, AF, IHD, DM, or COPD), and COVID-19 severity (critical or severe). Severity was defined According to the WHO definition.16 The assumptions assessed using the binary logistic regression models were the linearity of the continuous predictors using the Box–Tidwell approach, independence, absence of multicollinearity – the variance inflation factor (VIF < 5), and residual outliers. We presented the adjusted and unadjusted odds ratios (OR) in Table 3, reporting the 95% CI and R2 (the coefficient of multiple determination of the model).
We did not report the reduction in the invasive mechanical ventilation duration as survival in this group was low in both groups (11 cases in group A and 13 cases in group B).
As an exploratory analysis, we stratified the effect of atorvastatin on 28-mortality the time between symptom onset and hospitalization (within 7 days vs > 7 days).