- Research
- Open access
- Published:
Effectiveness of remdesivir for hospitalized COVID-19 patients depending on the severity of respiratory status: a multicenter retrospective study in Japan
BMC Infectious Diseases volume 25, Article number: 1016 (2025)
Abstract
Background
Remdesivir, an antiviral nucleotide analog prodrug, is approved for clinical use against COVID-19 worldwide. However, its effectiveness varies depending on the respiratory failure status of patients. This study aimed to evaluate the effectiveness of remdesivir treatment based on the severity of respiratory failure, as indicated by the oxygen demand upon hospital admission. Subgroups analyses were conducted to identify patient groups that might benefit from remdesivir treatment.
Methods
This retrospective observational study (the J-RECOVER) enrolled patients with COVID-19 from 64 institutions in Japan between January 1 and September 30, 2020. Patients aged ≥ 18 years who were administered remdesivir within 3 days of hospital admission were included. A total of 3,591 patients were included, and propensity score overlap weighting analysis was used to compare in-hospital mortality based on respiratory failure status at admission between remdesivir and control groups. Subgroup analyses identified specific patient populations that may benefit most from remdesivir treatment, considering factors such as respiratory status and renal function.
Results
The overlap weighting (OW)-adjusted odds ratio (OR) for mortality in overall cohort, mild cases without supplemental oxygen, moderate cases requiring supplemental oxygen, and severe cases requiring ventilation was (OR, 0.65 (95% confidence interval (CI), 0.36–1.19; P = 0.16), 0.11 (95% CI, 0.01–1.03; P = 0.05). 0.82 (95% CI, 0.31–2.16; P = 0.69), and 0.78 (95% CI, 0.28–2.17; P = 0.63), respectively. A trend toward improvement in mortality was observed in respiratory indicators, such as SpO2 ≥ 94% (OR, 0.43; 95% CI, 0.19–0.99; P = 0.04), oxygen support with FiO2 < 0.5 (OR, 0.40; 95% CI, 0.16–0.97; P = 0.04), and PFR ≥ 300 (OR, 0.17; 95% CI, 0.03–0.94; P = 0.04). Subgroup analyses indicated improved mortality in patients with an estimated glomerular filtration rate (eGFR) of > 60 mL/min per 1.73 m2 (OR, 0.29; 95% CI, 0.09–0.94; P = 0.03), with a p-value for interaction of P = 0.18.
Conclusion
Remdesivir treatment may reduce the risk of in-hospital mortality in patients with mild respiratory distress. Subgroup analysis suggested that remdesivir treatment may improve mortality in patients with eGFR ≥ 60 mL/min per 1.73 m2.
Background
In December 2019, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which causes coronavirus disease 2019 (COVID-19), was first reported in Wuhan, Hubei Province, China [1] and rapidly spread worldwide, causing a pandemic in various regions. By June 26, 2022, the World Health Organization reported > 541 million confirmed cases and 6.3 million deaths due to COVID-19 worldwide [2]. The COVID-19 pandemic continues, with new waves of outbreaks and multiple variants of SARS-CoV-2 emerging.
Remdesivir, a nucleotide analog prodrug with demonstrated antiviral activity against SARS-CoV-2 in vitro, has been approved for clinical use against COVID-19 in many countries [3, 4]. The ACTT-1 trial [5] reported no improvement in mortality in severe cases that required ventilation or high-flow oxygen, whereas an improvement in mortality was observed in cases of moderate respiratory failure that required low-flow oxygen. However, subsequent randomized controlled trials (RCTs) have reported neutral results in patients requiring oxygen support [6, 7]. In cases of mild respiratory failure where hospitalization was not required, the PINETREE trial reported that remdesivir prevented the progression of the disease [8]. Based on these findings, we hypothesized that remdesivir would be less effective in patients with severe respiratory failure at the time of treatment initiation. Additionally, a possible reason for the different treatment outcomes in moderate cases requiring supplemental oxygen may be the heterogeneity of remdesivir efficacy in the treatment groups, which means that other factors may have modified the treatment effect of remdesivir. This study aimed to address an inconsistency in effectiveness of remdesivir treatment in moderate cases requiring supplemental oxygen, which has not been conclusively established in previous studies. Additionally, this study sought to evaluate the effectiveness of remdesivir treatment based on the severity of respiratory failure, depending on the oxygen demand upon hospital admission. Subgroup analyses were conducted to identify specific patient groups that might benefit from remdesivir treatment.
Methods
Design and setting
Japanese multicenter research on COVID-19 by assembling real-world data (J-RECOVER study) was designed to establish a database of COVID-19 cases for research on unresolved clinical questions. The protocol for this study has been published previously [9]. This study used electronic medical records organized by the Diagnosis Procedure Combination (DPC) payment system. This study was conducted in accordance with the principles of the Declaration of Helsinki. The study protocol was registered with the University Hospital Medical Information Network (UMIN) Clinical Trials Registry (registration no. UMIN000047056). The registration date for this study was March 2, 2022. This study was approved by the Institutional Review Board of Nippon Medical School Musashikosugi Hospital (approval no. 561-2-26), which waived the requirement for informed consent because of the anonymous nature of retrospective data. The study complied with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for reporting [10].
Study population
This retrospective observational study enrolled patients with COVID-19 laboratory-confirmed SARS-CoV-2 infection among those discharged from 64 institutions in Japan between January 1 and September 30, 2020. All institutions included in this analysis were acute care hospitals with intensive care units, and approximately half were university hospitals (27/58 institutions). A flowchart of the patient selection process is shown in Fig. 1. Eligible patients were aged ≥ 18 years and admitted to the hospital with a diagnosis of COVID-19, regardless of admission to the intensive care unit (ICU). Additionally, only patients who were started on remdesivir within 3 days of admission were included because antiviral therapy is considered were also excluded. This was because we could not obtain information on the covariates at the time of administration, and determined that they were inappropriate for analysis. Patients with missing or erroneous blood collection data upon admission were excluded. After manually reviewing the dataset, we excluded cases with clearly erroneous key laboratory values (e.g., white blood cell count not measured or sodium level recorded as zero), as well as cases in which the majority of laboratory data were either missing or zero.beneficial in the early phase of infection [11]. Patients with missing respiratory status data regarding oxygen administration at the time of hospitalization were excluded from the analysis. Patients for whom remdesivir was administered before the date of admission.
Data collection
The baseline patient characteristics of the remdesivir treatment and control groups are presented in Supplemental Table 1. Clinical information was obtained from the DPC data and medical records. Essential clinical information that was unavailable from the DPC data was obtained from medical records. In this study, a specialized application developed by the principal investigator, called the DPC hash application, was used at the participating institutions for eligible patient selection and anonymization of DPC data.
The DPC is a case-mix patient classification system developed in Japan that uses an electronic billing system for patients admitted to acute care hospitals [12]. To ensure coding validity, physicians participating in this study recorded the diagnoses based on medical charts. The DPC data included sex, age, height, weight, current pregnancy, smoking index, Japan Coma Scale scores on admission, admission date, discharge date, primary diagnoses, concomitant diagnoses, complications, admission route, admission status, medical and surgical procedures, prescriptions, drug administration, discharge destination, and discharge status. Diagnoses resulting in hospitalization, requiring the first or second most medical resources, and complicated diagnoses during hospitalization were recorded based on International Classification of Diseases Tenth Revision (ICD-10) codes.
For each approved research proposal in this study, the researchers compiled a list of essential clinical data not included in the DPC system and systematically collected this information manually. Data collection procedures were standardized using a dedicated electronic application. The following clinical information was obtained manually by the researchers: body temperature (BT), systolic and diastolic blood pressure (sBP and dBP), pulse rate (PR), respiratory rate (RR), saturation of percutaneous oxygen (SpO2), Glasgow Coma Scale (GCS), date of admission, date of COVID-19 symptom onset, presence of pneumonia on initial chest X-ray or computed tomography, oxygen flow rate, fraction of inspiratory oxygen (FiO2), partial pressure of oxygen in arterial blood (PaO2), and lactate on admission; date of tracheal intubation; date of ICU admission; therapeutic drugs for COVID-19 such as favipiravir and remdesivir, use of steroids, vasopressors and heparin, mechanical ventilation used, initial ventilator settings, extracorporeal membranous oxygenation (ECMO) device used, complications during hospitalization, such as cerebral infarction, pulmonary thromboembolism, deep vein thrombosis, bleeding complications (requiring therapeutic intervention such as transfusion or hemostasis).
All blood test data, including complete blood count, blood cell profile, blood sugar, biochemistry, and blood coagulation test results on admission and during hospitalization were collected as electronic information from each participating institution. Variations in electronic blood test data among institutions were normalized and stored in a database by a principal researcher.
The following information regarding each institution was collected: type of institution, such as university hospital; number of hospital beds; ICU and emergency center beds; and presence of emergency department specialists and intensivists.
Dealing with missing data
Missing data is an inevitable problem in observational studies and, when not treated properly, can decrease statistical power and lead to biased estimates, thus weakening the generalizability of the findings [13]. Regarding patient characteristics and clinical information after the inclusion and exclusion criteria, there were missing values that varied from 0 to 46.7% of the total observations. To address missing data, we used a multiple imputation method, assuming that the missing data mechanism was missing at random (MAR). We also included auxiliary variables that may correlate with responses or explanatory variables in the imputation model as far as possible. Supplemental Table 2 presents the variables used in the imputation model.
The MissForest algorithm is a nonparametric imputation method based on random forest, which essentially constitutes a multiple-imputation scheme. The MissForest algorithm has been shown to perform as well as the Multivariate Imputation by Chained Equations (MICE) algorithm, a common multiple imputation method, in handling mixed data composed of continuous and categorical variables [14]. In this analysis, we used the R package “missForest” to predict and impute missing values in the dataset. The imputation process was performed under the following settings: the maximum number of iterations was set to 10 and the number of trees to grow in each forest was 500. A complete case analysis was conducted for the sensitivity analysis.
Statistical analysis
We examined the effectiveness of remdesivir in reducing in-hospital mortality, depending on the severity of respiratory status at the time of admission. The severity of respiratory failure was defined as mild respiratory failure requiring no oxygen support, moderate respiratory failure requiring oxygen support, or severe respiratory failure requiring ventilation upon admission. With respect to the severity of respiratory failure, the number of days from hospital admission to remdesivir treatment initiation was as follows: 1 day (IQR: 0–2) for mild cases, 0 days (IQR: 0–1) for moderate cases, and 0 days (IQR: 0–0) for severe cases. The primary outcome was in-hospital mortality in COVID-19 patients. The secondary outcome was length of hospital stay (LOS), which was defined as the number of days between admission and discharge from the DPC data, regardless of discharge status and destination.
In this analysis, we performed a propensity score overlap weighting (PSOW) analysis to adjust for differences in measured characteristics between the remdesivir-treated and control groups, because significant differences in respiratory status and coexisting illnesses between the groups may be estimated. The PSOW method can achieve a good statistical balance by addressing the extreme distributions of the propensity score (PS) between the remdesivir treatment and control groups that cannot be balanced by PS methods, such as inverse probability of treatment weighting (IPTW) or PS matching [15, 16]. The PSOW method employed in our analysis differs from PS matching and represents an enhanced version of IPTW. Unlike PS matching, which excludes unmatched cases and therefore reduces the sample size, both PSOW and IPTW retain all cases by assigning weights based on propensity scores, thereby maintaining statistical power and generalizability. Individual PS for the receipt of remdesivir treatment was estimated using a multivariable logistic regression model with the following baseline covariates, which were included based on previous studies and clinical importance: sex, age, body mass index (BMI), Brinkman index, coexisting hypertension, cardiac infarction, heart failure, cancer, diabetes, asthma, chronic obstructive pulmonary disease, GCS, BT, sBP, RR, PR, PaO2, FiO2, PaO2/FiO2 ratio (PFR) on admission, Charlson Comorbidity Index (CCI), the total Sequential Organ Failure Assessment (SOFA) score and SOFA sub-scores, National Early Warning Score (NEWS2), white blood cells (WBC), platelets, bilirubin, alanine aminotransferase, aspartate aminotransferase, creatinine, blood urea nitrogen, estimated glomerular filtration rate (eGFR), lactate dehydrogenase, C-Reactive Protein (CRP), procalcitonin, blood glucose level, serum sodium, serum potassium, prothrombin time and international normalized ratio (PT-INR), activated partial thromboplastin time (APTT) sec, D-dimer, Fibrinogen, hemoglobin A1c (HbA1c), Favipiravir, Steroids, Vasopressors, Heparin, and mechanical ventilation. The standardized mean difference (SMD) was used to evaluate the performance of PS adjustment, with an SMD of < 10% considered to indicate a proper balance between groups [17]. In addition, subgroup analyses were conducted to assess the heterogeneity of remdesivir effectiveness on in-hospital mortality for each variable, such as age (< 70 and ≥ 70 years), sex, BMI (< 25 and ≥ 25), SpO2 (< 94 and ≥ 94%), PaO2 (< 90 and ≥ 90 mmHg), FiO2 (< 0.5, ≥ 0.5), PFR (< 300 and ≥ 300), respiratory status, NEWS2 (< 7 and ≥ 7), SOFA score (< 2 and ≥ 2), eGFR (< 60 and ≥ 60 mL/min per 1.73 m2), and HbA1c (< 6.5 and ≥ 6.5%). In our analysis, we recalculated the propensity scores separately for the overall cohort and for each subgroup (mild, moderate, and severe cases) to ensure appropriate background adjustment within each comparison. In the subgroup analysis of mild respiratory failure cases, the covariate of “FiO2” was excluded from the multivariate logistic regression model for PS estimation because of its SMD of 39%. Using the PSOW method, we adjusted the distributions of PS between the remdesivir treatment and comparison groups and performed overlap weighting (OW)-adjusted univariable logistic regression, multivariable logistic regression for interaction analysis, and multiple regression analysis. We calculated OW-adjusted odds ratios (ORs) and 95% confidence intervals (CI) to estimate the association between remdesivir use and in-hospital mortality. In addition, we applied the OW-adjusted Kaplan–Meier method for survival curve analysis to compare survival probabilities between remdesivir recipients and controls using the OW-adjusted log-rank test. The data of study patients were censored at discharge.
All statistical inferences were made using a 2-sided p value at the 5% significance level. In the subgroup analysis, a P-value ≤ 20% of interaction was considered significant. ORs with 95% CI were used as the estimated measures of association reported in these analyses. All statistical analyses were performed using R software version 4.1.3. (R Foundation for Statistical Computing, Vienna, Austria).
Results
Study population
A total of 3,591 patients were included in the study between January 1 and September 30, 2020 (Fig. 1). The patient characteristics of each group after PS adjustment are shown in Table 1 (all covariates are listed in Supplemental Table 3). Among these patients, 183 were in the remdesivir treatment group and 3,408 were in the control group (patients who did not receive remdesivir). In-hospital mortality was observed in 22 of the 183 patients (12.0%) in the remdesivir treatment group and 185 of the 3,408 patients (5.4%) in the control group. The remdesivir treatment group consisted of older men with a higher BMI than the control group. Blood samples showed significantly higher WBC and CRP levels, indicating a high inflammatory response. Diabetes mellitus, hypertension, and hyperlipidemia were the most common comorbidities in the remdesivir group. Regarding respiratory status, lower PaO2 and PFR were observed at the time of admission. The total SOFA score, particularly the respiratory SOFA score, was higher and the total NEWS2 score was significantly higher. There were more intubation cases on admission, and steroids and heparin were more commonly used.
Results after propensity score adjustment
The area under the curve (AUC) was 0.95, indicating an extreme distribution of PS between the treatment and comparison groups. After PS adjustment for extreme distributions using the OW method, the SMD was < 10% for all measured covariates, achieving a good statistical balance.
Mortality with respiratory status
Overall, we performed a logistic regression analysis with the PSOW method, showing an OW-adjusted OR for mortality of 0.65 (95% CI, 0.36–1.19; P = 0.16) in the remdesivir treatment group compared to the control group (Fig. 2). Based on the severity of respiratory failure, 50 of 2,773 patients (1.8%) with mild cases, 84 of 520 patients (16.1%) with moderate cases, and 73 of 298 patients (24.4%) with severe cases died. In the analyses performed according to the severity of respiratory failure status between the groups, the effects of remdesivir treatment on mortality were estimated in mild (OW-adjusted OR, 0.11; 95% CI, 0.01–1.03; P = 0.05), moderate (OW-adjusted OR, 0.82; 95% CI, 0.31–2.16; P = 0.69), and severe cases (OW-adjusted OR, 0.78; 95% CI, 0.28–2.17; P = 0.63). Remdesivir treatment tended to decrease the odds ratio for mortality in mild cases compared with that in moderate or severe cases. A trend toward improvement in mortality was observed with mild respiratory distress, and similar trends in reduction of mortality were observed in other respiratory indicators, such as SpO2 ≥ 94% (OW-adjusted OR, 0.43; 95% CI, 0.19–0.99; P = 0.04), required oxygen support with FiO2 < 0.5 (OW-adjusted OR, 0.40; 95% CI, 0.16–0.97; P = 0.04), and PFR ≥ 300 (OW-adjusted OR, 0.17; 95% CI, 0.03–0.94; P = 0.04). Compared to the odds ratio for mortality of NEWS2 ≥ 7 (OW-adjusted OR, 1.03; 95% CI, 0.42–2.51; P = 0.93), which was classified as high risk, that of NEWS2 < 7 (OW-adjusted OR, 0.35; 95% CI, 0.14–0.89; P = 0.02), which was classified as low or medium risk, decreased.
Forest plot of the comparison: remdesivir-treated group vs. control group: in-hospital mortality. In the subgroup analysis of respiratory status, p values for the interaction between mild and moderate/severe cases are shown. OR, odds ratio; CI, confidence interval; BMI, body mass index; SpO2, saturation of percutaneous oxygen; PaO2, partial pressure of oxygen in arterial blood; FiO2, fraction of inspiratory oxygen; PFR, partial pressure of oxygen in arterial blood/fraction of inspiratory oxygen ratio; NEWS2, National Early Warning Score 2; SOFA, Sequential Organ Failure Assessment; eGFR, estimated glomerular filtration rate
Survival analysis
Survival analyses adjusted for OW were performed based on the severity of respiratory failure on admission. There was a significant difference in the LOS between the remdesivir treatment and control groups in all cases (–5.25; 95% CI, − 8.08 − 2.43; P < 0.01). Although LOS did not differ significantly between the two groups for moderate cases (–1.96; 95% CI, − 6.04–2.11; P = 0.34), a reduction in LOS was observed in mild (–2.47; 95% CI, − 4.93 to − 0.02; P = 0.04) and severe cases (–11.09; 95% CI, − 18.37 to − 3.81; P = < 0.01). As these results included patients discharged due to death, caution is warranted to avoid over-interpretation of the findings.
Kaplan–Meier curves adjusted for OW were drawn for the overall and each respiratory severity group. Overall, the survival curves of mild and moderate cases crossed at 3–4 weeks, and those of severe cases almost overlapped during the early observation period (Fig. 3A–D). As the proportional hazard assumption was not satisfied by the Kaplan–Meier curves, the log-rank test was deemed unsuitable.
Kaplan–Meier Survival curves for remdesivir recipients and controls at each severity of respiratory status A overall. B Mild respiratory failure. C Moderate respiratory failure. D Severe respiratory failure The number “98” under “Number at Risk” represents the PS-adjusted patient count for each group. Since the propensity scores were recalculated separately for the overall cohort and each respiratory severity subgroup, the number of adjusted patients shown in panels B–D does not correspond to the total number shown in panel A
Subgroup analyses
Subgroup analyses revealed no significant differences in the interaction of mortality between remdesivir treatment and control groups. Remdesivir treatment improved mortality in patients with eGFR values above 60 mL/min per 1.73 m2 (OW-adjusted OR, 0.29; 95% CI, 0.09–0.94; P = 0.03). The following subgroups could be associated with the variation in remdesivir effectiveness: eGFR ≥ 60 and FiO2 < 0.5, with p-values for interaction of P = 0.18 and P = 0.22, respectively.
Sensitivity analyses
For sensitivity analysis, we conducted a complete case analysis by excluding cases with missing values that could not be used to calculate the PS from the original data. The effects of remdesivir treatment on mortality were estimated in all cases (OW-adjusted OR, 0.67; 95% CI, 0.36–1.25; P = 0.21), mild cases (OW-adjusted OR, 0.11; 95% CI, 0.01–1.04; P = 0.05), moderate cases (OW-adjusted OR, 0.82; 95% CI, 0.31–2.16; P = 0.69), and severe cases (OW-adjusted OR, 1.01; 95% CI, 0.34–2.95; P = 0.98). Similar to the primary analysis results with the missing-value imputation, no significant differences were observed. Although not statistically significant, a numerical reduction in mortality was observed in patients with mild respiratory distress, suggesting a potential benefit that warrants further investigation. Similar reductions in mortality were also seen in other respiratory indicators, such as PaO2 ≥ 90 (OW-adjusted OR, 0.1; 95% CI, 0.01–0.96; P = 0.04) and PFR ≥ 300 (OW-adjusted OR, 0.1; 95% CI, 0.01–0.88; P = 0.03). These findings are generally consistent with those of the main analysis and may serve to support its conclusions. Details of the sensitivity analysis are presented in Supplemental Fig. 1.
Discussion
In this study, we did not observe any significant differences in all-cause mortality between the remdesivir treatment and control groups in all patients. Subgroup analysis for renal function showed a treatment effect modification of remdesivir treatment. Analysis of the severity of respiratory failure showed a trend toward improvement in mortality in mild cases, but not in moderate or severe cases. Patients with milder respiratory failure who received remdesivir had a lower odds ratio of mortality than those with more severe respiratory failure.
The pathogenesis of COVID-19 is divided into two stages. The early phase, a few days after disease onset, is characterized by a high viral load and an innate immune response. The late phase, approximately one week after onset, is characterized by a low viral load and an adaptive immune response that triggers systemic inflammation in response to viral infection by host immunity [18]. The pathological findings of COVID-19 show a chronological series of diffuse alveolar damage (DAD) as follows: exudative phase (days 1–7), proliferative phase (one–a few weeks), and fibrotic phase (3–4 weeks or longer). The progression of pulmonary damage is generally associated with disease severity and is the main cause of mortality [19]. Although the spatiotemporal heterogeneity of COVID-19 characterized by the simultaneous occurrence of three DAD phases in various lung areas has been reported, the different DAD appearances might be influenced by the patient’s comorbidities and not by COVID-19 infection [20]. With respect to the timing of treatment initiation, an animal experiment reported that remdesivir treatment initiated early in infection has the clinical benefit of decreasing lung viral loads and reducing lung damage [21]. In the late phase, COVID-19 patients may miss the timing when remdesivir treatment is effective if their respiratory status has deteriorated owing to a systemic inflammatory response, resulting in pulmonary damage. In the survival analysis, the overall Kaplan–Meier curve was crossed at 3–4 weeks. This could be explained by the difference in pathology between the early and late phases of infection, which may be associated with increased lung fibrosis due to disease progression, as indicated by the pathological findings of previous studies [19]. However, patients with prolonged hospitalization are expected to become critically ill, and their outcomes may have influenced the results. In addition, the crossing survival curves should be interpreted with caution because the number of patients decreased due to censoring by discharge in the late phases. The observed reduction in LOS among severe cases may reflect early mortality rather than the therapeutic efficacy of remdesivir. This interpretation is supported by our analysis, which suggests that remdesivir had a little clinical benefit in patients with severe cases. Although LOS appeared shorter in the remdesivir group overall, this finding should be interpreted with caution, as it may have been influenced by early deaths in severely ill patients.
The ACTT-1 trial [5] reported that remdesivir improved in-hospital mortality in COVID-19 patients requiring supplemental oxygen, but did not improve the requirement for high-flow oxygen devices or noninvasive ventilation. In contrast, subsequent RCTs found negative results in patients requiring similar oxygen support [6, 7]. The present analysis also showed that remdesivir was ineffective in moderate cases that required supplemental oxygen. The largest RCT, the Solidarity Trial, also found no significant difference in in-hospital mortality for moderate cases (OR, 0.85; 95% CI, 0.66–1.09) [22]; however, it included patients receiving low-flow or high-flow oxygen support who were not classified by the administered oxygen concentration. From the present analyses, demonstrating a trend toward improvement in mortality with remdesivir in patients with lower respiratory failure status and an improvement in mortality with remdesivir treatment at SpO2 ≥ 94% and FiO2 < 0.5, mild respiratory distress requiring low-flow oxygen may have a better prognosis than those requiring high-flow oxygen. In addition, possible reasons for the discrepancy in remdesivir treatment effectiveness in moderate cases among previous trials include differences in the background study population and treatment details, which may be associated with the heterogeneity in remdesivir treatment effectiveness observed in the present subgroup analysis. As treatment data for remdesivir have accumulated globally, it may be possible to estimate remdesivir heterogeneity by analyzing real-world data using appropriate statistical methods such as the OW method to adjust the PS distribution. This may contribute to the selection of patients that can be efficiently treated with remdesivir.
The trials presented subsequently indicate the possibility that early administration of remdesivir may improve the clinical status. The ACTT-1 trial [5] reported that the effectiveness of remdesivir in terms of time to recovery was higher when remdesivir was administered within 10 days of symptom onset than when it was administered > 10 days after symptom onset. In subgroup analyses of the DisCoVeRy trial, the longer the duration from symptom onset to assignment, the lower the odds ratio for the treatment effect on clinical status. This trial concluded that no clinical advantage of remdesivir was observed in patients who required oxygen support for > 7 days after symptom onset [7]. In the present study, the median time from symptom onset to remdesivir administration in the treatment group was 8 days (IQR, 5–10 days), which was comparable to that reported in previous trials [5, 7]; further, no treatment advantages were found in patients requiring oxygen support. Based on the present results, if the effectiveness of remdesivir is reduced by increasing the demand for oxygen with disease progression over time, remdesivir administration as early as possible after symptom onset may provide clinical benefits [23].
Our study has several limitations. First, the clinical information recorded in the DPC database may have included data errors because the clinical database recorded for administrative claims generally shows lower accuracy than that recorded for prospective studies. In addition, misclassification, underestimation, or overestimation of underlying conditions may have occurred. Second, 95% of the patients in the study were Japanese, which could have led to a bias. Third, standardization of the duration and dose of remdesivir administration in the treatment groups was not possible, owing to the retrospective observational nature of the study. The database lacks information on respiratory status at the exact time of remdesivir administration. Consequently, some cases initially classified as mild or moderate at admission may have included patients whose condition had progressed to a more severe stage by the time treatment was initiated. Fourth, the OW method was adjusted for the extreme distribution of PS to achieve a good statistical balance; however, unmeasured confounding factors were not adjusted. The following potential biases existed: The study was conducted during the early and highly atypical phase of the pandemic, when decisions regarding remdesivir administration were likely influenced by non-standardized and emergency-driven circumstances. Fifth, multiple imputations were performed for the missing values under the assumption of the missing-at-random (MAR). Overall, the missing data rate was 6.9% in the control group and 4.8% in the remdesivir group. However, some variables had larger differences—for example, procalcitonin was missing in 48% of controls vs. 28% of remdesivir patients. While we assumed an MAR mechanism, we acknowledge the potential for missing-not-at-random (MNAR) bias. Although no statistical method has been established to deal with the missing data mechanism in MNAR, including auxiliary variables as much as possible, it might change the missing data mechanism from MNAR to MAR [24, 25]. A complete case analysis was conducted as a sensitivity analysis and no substantial differences were observed in the results of the main analysis. Finally, owing to time constraints, this may not reflect the current dominant variants of SARS-CoV-2. Although further validation of its clinical effectiveness is needed, remdesivir has been reported to inhibit viral replication in an in vitro study of BA.5 and BA.2 variants and their subvariants, such as BQ.1.1 and XBB, which are currently prevalent worldwide [26].
Conclusions
The present study, using analyses with the PSOW method, showed that remdesivir treatment does not uniformly improve mortality in COVID-19 patients. However, remdesivir treatment may improve mortality in COVID-19 patients with mild respiratory distress who require low-flow oxygen therapy. The results of subgroup analysis suggested heterogeneity in the effectiveness of remdesivir treatment, particularly with respect to renal function.
Data availability
The datasets used and/or analyzed in the current study are available from the corresponding author upon reasonable request.
Abbreviations
- SARS-CoV-2:
-
Severe acute respiratory syndrome coronavirus 2
- COVID-19:
-
Coronavirus disease 2019
- ACTT:
-
The Adaptive Covid-19 Treatment Trial
- RCTs:
-
Randomized controlled trials
- DPC:
-
Diagnosis Procedure Combination
- J-RECOVER:
-
Japanese multicenter research on COVID-19 by assembling real-world data
- UMIN:
-
The University Hospital Medical Information Network
- STROBE:
-
The Strengthening the Reporting of Observational Studies in Epidemiology
- ICU:
-
The intensive care unit
- ICD-10:
-
International Classification of Diseases Tenth Revision
- BT:
-
Body temperature
- PR:
-
Pulse rate
- RR:
-
Respiratory rate
- SpO2 :
-
Saturation of percutaneous oxygen
- GCS:
-
Glasgow Coma Scale
- FiO2 :
-
Fraction of inspiratory oxygen
- PaO2 :
-
Partial pressure of oxygen in arterial blood
- ECMO:
-
Extracorporeal membranous oxygenation
- MAR:
-
Missing at random
- MICE:
-
The Multivariate Imputation by Chained Equations
- LOS:
-
The length of hospital stay
- PSOW:
-
The propensity score overlap weighting
- PS:
-
Propensity score
- IPTW:
-
Inverse probability of treatment weighting
- BMI:
-
Body mass index
- BP:
-
Blood pressure
- CCI:
-
Charlson Comorbidity Index
- SOFA:
-
The Sequential Organ Failure Assessment
- NEWS 2:
-
The National Early Warning Score 2
- WBC:
-
White blood cells
- eGFR:
-
Estimated glomerular filtration rate
- CRP:
-
C-Reactive Protein
- PT-INR:
-
Prothrombin time and international normalized ratio
- APTT:
-
Activated partial thromboplastin time
- SMD:
-
Standardized mean difference
- PFR:
-
Partial pressure of oxygen in arterial blood/fraction of inspiratory oxygen ratio
- HbA1c:
-
Hemoglobin A1c
- OW:
-
Overlap weighting
- OR:
-
Odds ratio
- CI:
-
Confidence interval
- DAD:
-
Diffuse alveolar damage
- MNAR:
-
The missing data mechanism in missing not at random
References
Zhu N, Zhang D, Wang W, Li X, Yang B, Song J, et al. A novel coronavirus from patients with pneumonia in China. N Engl J Med. 2019;382:727–33.
World Health Organization. COVID-19 weekly epidemiological update Edition 98. 2022.
Wang M, Cao R, Zhang L, Yang X, Liu J, Xu M, et al. Remdesivir and chloroquine effectively inhibit the recently emerged novel coronavirus (2019-nCoV) in vitro. Cell Res. 2020;30:269–71.
Lamb YN. Remdesivir: first approval. Drugs. 2020;80:1355–63.
Beigel JH, Tomashek KM, Dodd LE, Mehta AK, Zingman BS, Kalil AC. Remdesivir for the treatment of Covid-19– final report. N Engl J Med. 2020;383:1813–26.
Wang Y, Zhang D, Du G, Du R, Zhao J, Jin Y, et al. Remdesivir in adults with severe COVID-19: a randomised, double-blind, placebo-controlled, multicentre trial. Lancet. 2020;395:1569–78.
Ader F, Bouscambert-Duchamp M, Hites M, Peiffer-Smadja N, Poissy J, Belhadi D, et al. Remdesivir plus standard of care versus standard of care alone for the treatment of patients admitted to hospital with COVID-19 (DisCoVeRy): a phase 3, randomised, controlled, open-label trial. Lancet Infect Dis. 2022;22:209–21.
Gottlieb RL, Vaca CE, Paredes R, Mera J, Webb BJ, Perez G, et al. Early remdesivir to prevent progression to severe Covid-19 in outpatients. N Engl J Med. 2022;386:305–15.
Tagami T, Yamakawa K, Endo A, Hayakawa M, Ogura T, Hirayama A, et al. Japanese multicenter research of COVID-19 by assembling real-world data: a study protocol. Annals of Clinical Epidemiology. 2022;4:92–100.
Cuschieri S. The STROBE guidelines. Saudi J Anaesth. 2019;13(Suppl 1):S31-4.
Gonçalves A, Bertrand J, Ke R, Comets E, de Lamballerie X, Malvy D, et al. Timing of antiviral treatment initiation is critical to reduce SARS-CoV-2 viral load. CPT Pharmacometr Syst Pharmacol. 2020;9:509−14.
Yasunaga H: Chapter II. The diagnosis procedure combination database. Ann Clin Epidemiol. 2020;1:76−9. Real World Data in Japan.
Dong Y, Peng CYJ. Principled missing data methods for researchers. Springerplus. 2013;2:222.
Stekhoven DJ, Bühlmann P. MissForest–non-parametric missing value imputation for mixed-type data. Bioinformatics. 2012;28:112–8.
Li F, Thomas LE, Li F. Addressing extreme propensity scores via the overlap weights. Am J Epidemiol. 2019;188:250–7.
Thomas LE, Li F, Pencina MJ. Overlap weighting: a propensity score method that mimics attributes of a randomized clinical trial. JAMA. 2020;323:2417–8.
Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28:3083–107.
Siddiqi HK, Mehra MR. COVID-19 illness in native and immunosuppressed states: A clinical-therapeutic staging proposal. J Heart Lung Transpl. 2020;39:405–7.
Bösmüller H, Matter M, Fend F, Tzankov A. The pulmonary pathology of COVID-19. Virchows Arch. 2021;478:137–50.
Polak SB, Van Gool IC, Cohen D, von der Thüsen JH, van Paassen J. A systematic review of pathological findings in COVID-19: a pathophysiological timeline and possible mechanisms of disease progression. Mod Pathol. 2020;33:2128–38.
Williamson BN, Feldmann F, Schwarz B, Meade-White K, Porter DP, Schulz J. Clinical benefit of remdesivir in rhesus macaques infected with SARS-CoV-2. Nature. 2020;585:273–6.
WHO Solidarity Trial Consortium, Pan H, Peto R, Henao-Restrepo AM, Preziosi MP, Sathiyamoorthy V, et al. Repurposed antiviral drugs for Covid-19 - interim WHO solidarity trial results. N Engl J Med. 2021;384:497–511.
Spinner CD, Gottlieb RL, Criner GJ, Arribas López JR, Cattelan AM, Soriano Viladomiu A, et al. Effect of remdesivir vs standard care on clinical status at 11 days in patients with moderate COVID-19: a randomized clinical trial. JAMA. 2020;324:1048–57.
Collins LM, Schafer JL, Kam CM. A comparison of inclusive and restrictive strategies in modern missing data procedures. Psychol Methods. 2001;6:330–51.
Hardt J, Herke M, Leonhart R. Auxiliary variables in multiple imputation in regression with missing X: a warning against including too many in small sample research. BMC Med Res Methodol. 2012;12:184.
Imai M, Ito M, Kiso M, Yamayoshi S, Uraki R, Fukushi S, et al. Efficacy of antiviral agents against Omicron subvariants BQ.1.1 and XBB. N Engl J Med. 2023;388:89–91.
Acknowledgements
Not applicable.
Funding
Not applicable.
Author information
Authors and Affiliations
Consortia
Contributions
K.Y. and T.T. contributed to study conception and design. K.Y., M.H., A.E., T.O., A.H., Hideo.Y. T. T. designed the J-RECOVER database. I.O. and T.T. coordinated the data collection. Hidero.Y. and Y.T. performed statistical analyses. Hidero.Y., Y. T. and K. Y. interpreted the analysis results. Hidero.Y. drafted the manuscript. K.Y., A.T., S. L., and T.T. reviewed the manuscript critically. All authors have read and approved the final manuscript.
Corresponding author
Ethics declarations
Ethics approval and consent to participate
This study was conducted in accordance with the principles of the Declaration of Helsinki. The study protocol was registered with the University Hospital Medical Information Network (UMIN) Clinical Trials Registry (registration no. UMIN000047056). The registration date for this study was March 2, 2022. This study was approved by the Institutional Review Board of Nippon Medical School Musashikosugi Hospital (approval no. 561-2-26), which waived the requirement for informed consent because of the anonymous nature of retrospective data.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Yoshimoto, H., Yamakawa, K., Tanaka, Y. et al. Effectiveness of remdesivir for hospitalized COVID-19 patients depending on the severity of respiratory status: a multicenter retrospective study in Japan. BMC Infect Dis 25, 1016 (2025). https://doi.org/10.1186/s12879-025-11345-z
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s12879-025-11345-z