Introduction

The adenosine triphosphate analogue remdesivir blocks the activity of viral polymerase and the build-up of the mRNA chain. The approved indication is for incipient forms of viral pneumonitis with a need for oxygen administration. The usual dosing scheme for severe coronavirus disease is 600 mg within 5 days (200 mg on the first day, followed by 100 mg daily), an abridged three-day dosing scheme has been approved for less severe forms in high-risk patients1,2,3,4,5.

The widely applied indication criteria for remdesivir were based on the duration of symptoms (less than 7 days), PCR positivity for SARS-CoV-2, full unlimited care with the expected ensuing quality of life corresponding to the Frailty Score 1–66,7. The risk factors supporting the indication have been lymphopenia, haemato-oncologic disease, vasculitis, obesity, diabetes, hypertension, and chronic obstructive pulmonary disease8,9,10,11. The drug should not be given to patients on high-flow nasal oxygen (HFNO) or mechanical ventilation due to anticipated advanced disease without a significant impact on therapeutic outcome7,12,13,14,15. Another relative contraindication is renal failure with respect to the result of a translational research in apes, which has not been fully validated in humans16,17.

Trials on remdesivir therapy in Covid-19 included patients with positive qualitative PCR testing and clinical symptoms within a time limit of several days. Therapy was declared to be more effective in the early stages of the disease and was thus denied in advanced forms of Covid-19 with respiratory failure or patients admitted to intensive care7,18,19,20.

However, the published data also alludes to the superior effect of virostatic therapy under the condition of high viral load21,22,23 that, in hospitalised symptomatic patients, correlates with morbidity and mortality24,25,26. This viral load can be estimated by the semiquantitative detection of SARS-CoV-2 antigen (specific nucleoproteins) using fluorescent immunochromatography of the sputum/nasal cavity, quantitative detection of serum nucleocapsid antigen27 and real-time PCR detection using the cycle threshold (Ct) to determine its positivity. Ct less than 28–30 has been correlated with a load of living virus and with the positivity of most of the quality antigen testing28. In contrast, Ct above 34 has not shown the presence of a living viral load, may indicate an early stage of the disease or represent a residual positivity after a previous viral load that was not monitored in symptomatic patients28,29,30,31.

The coronavirus mutations (beta, delta) detected during the course of the pandemic showcased prolonged replication times that often exceeded 3–5 days of the incubation period plus 7 days of clinical symptoms32,33,34. These prolonged replication times were typically reported in the immunocompromised and in patients with severe symptoms35,36,37. Thus, patients with fulminant courses of Covid-19 associating with admissions to intensive care may have high viral load and plasmatic viraemia, yet these patients should be excluded from remdesivir administration potentially due to a rapid need for invasive ventilatory support and, possibly even the ECMO therapy38,39.

The hypothesis of this retrospective study was that a long-term benefit of remdesivir in the most severe forms of Covid-19 might have associated with the presence of an active viral load, and this, to date, has been poorly validated in relation to widely available clinical testing. Patients treated with remdesivir according to the estimates of viral load using antigen tests or real-time PCR were compared with similarly positive patients where the drug was denied. The reasons why the drug was not administered despite the presence of a viral load were justified by the guidelines, for example, based on symptoms lasting for more than 7 days from the first positive test and due to advanced illness severity. The endpoints of the study were the short and long-term outcome data evaluated with confounding factors for the overall population and collected in two major Prague university hospitals. In addition, the sub-analysis of the intensive care patients and patients with immunocompromise (including oncology and haemato-oncologic patients) were aimed to clarify an impact of therapy in patients with the viral load and high illness severity. Results supporting more efficient remdesivir administration were expected with respect to a widely available semi-quantitative estimate of viral load. This research may potentially better justify antiviral therapy in the advanced stages of disease and serve as a basis for a prospective multicentric study.

Methods

In this retrospective study we compared patients with an indication for remdesivir based on the presence of a viral load with patients where the virostatic therapy was contraindicated. To underpin our hypothesis, we included Covid-19 patients in whom remdesivir therapy was withheld due to current guidelines and with, in parallel, the presence of a viral load detected by available laboratory tests, as controls.

In addition to the primary analysis the defined subgroups were patients with similar high illness severity, e.g., ICU patients, and patients with an immune deficit. These patients had history of immunosuppressive therapy including combined immunosuppression, with recent or ongoing oncology treatment including haemato-oncology therapy and on chronic dialysis.

The endpoints of the study were the short and long-term outcomes adjusted for anamnestic, diagnostic, and therapeutic factors. The comparisons of outcomes were assessed in patient subgroups with high illness severity, e.g., ICU patients, and patients with immunosuppression.

The patients were identified from the hospital database of laboratory results and patients who tested positive (implicating a presence of viral load, e.g., an abundance of living coronavirus either in blood or in the mucosal specimens) were included. The tests used were quantitative detection of a nucleocapsid antigen (NAg) in serum (SARS-CoV-2 Antigen Quantitative Assay Kit, Biohit Healthcare (Hefei) Co., ELISA, cut-off: 2.97 pg/ml, from 26.02. 2021 till 21.09. 2021 and SARS-CoV-2 Antigen Quantitative Test, Biohit Healthcare (Hefei) Co., fluorescence immunochromatography test, cut-off 8.92 pg/ml, after 22.09. 2021), semi-quantitative SARS-CoV-2 antigen (specific nucleoproteins) detection by fluorescent immune-chromatography in sputum/nasopharyngeal samples (cut-off-index (COI) 1, Standard F Covid-19 Ag FIA, SD Biosensor) and real-time PCR using cycle threshold for positivity (Ct) in sputum/nasopharyngeal samples. Patients included in the study had a diagnosis of a viral load confirmed by one of the aforementioned tests. The cut off for the NAg in serum was set to > 20 pg/ml, to COI > 10 for the immunofluorescent antigen in either the nasopharyngeal or sputum samples, and the cycle threshold of the real-time PCR on the nasopharyngeal or sputum samples had to be lower than 28. Antigen tests have been available for routine use since 2021 and the patients were recruited over the period of 16 months between the 1st of February 2021 and 31st of May 2022.

Patients who tested positive for the presence of a viral load were divided into a group treated with remdesivir and a group where the drug was denied/withheld due to delayed admission after an initial positive test or due to the advanced illness severity where the therapy was deemed futile.

The patient´s data included age, gender, body mass index (BMI), length-of-stay in the ICU and in the hospital. The acute physiology and chronic health evaluation score IV (APACHE IV) was calculated as a mortality predictor in every ICU patient on admission. Illness severity was characterised by the sequential organ failure score (SOFA), the degree of respiratory insufficiency and the dependence on oxygen represented by an application of nasal oxygen prongs, face mask, high-flow nasal oxygen (HFNO), non-invasive ventilation (NIV), intubation with invasive mechanical ventilation (IPPV), or the need for any form of the extracorporeal membrane oxygenation (ECMO)40. The available inflammatory markers (CRP, PCT, leucocyte/lymphocyte ratio) were recorded at start of therapy as well as plasmatic creatinine and the need for acute renal replacement therapy during hospital stay. The therapeutic factors potentially influencing the impact of remdesivir were concomitant administrations of monoclonal antibodies (Celltrion, RegnCov, Bamlanivimab, Evusheld), molnupiravir, and corticosteroids. These were recorded as either 6–8 mg of dexamethasone per day (e.g. low dose steroids), or more than 6–8 mg of dexamethasone or more than 40 mg of methylprednisolone per day (e.g. high dose steroids). Other factors with potential impact were coronavirus re-infection (i.e., previous test positivity), therapy with isoprinosine, tocilizumab, baricitinib, and any type of nationwide certified vaccination started at least 3 weeks before admission.

Anamnestic risk factors evaluated were hypertension, diabetes, ischaemic heart disease, asthma and COPD, chronic liver disease, haemato-oncologic history, history of immunosuppressive medication (corticosteroids, combined immunosuppressive therapy, biologic therapy), oncologic medical history incl. chemotherapy or radiotherapy in the last 6 months. The outcome data were ICU survival, hospital survival, and 1 year survival.

The endpoints of the study, namely the outcome of patients, as well as selected clinical records (e.g. vaccination status, time to ICU admission after vaccination, etc.) were controlled against population registries of the National Health Information System which is a database collecting all insurance data with 100% coverage of the Czech population.

The patients were included in two university hospitals in Prague. The General University Hospital is a prominent cardiac and ECMO centre, oncology centre, vasculitis centre, and it concentrates patients with multiple risk factors for the development of severe Covid-19. The University Hospital Kralovske Vinohrady is a complex trauma centre and intensive care department that covers a large area of eastern Prague. The ethical boards of both hospitals approved the study and confirm that all research was performed in accordance with the Declaration of Helsinki. A need for written informed consent was waived due to the retrospective nature of the study.

Statistical analysis

All analyses and data processing were performed in R 4.3.341 and RStudio 2023.12.141,42. Because most of the continuous data did not meet the criteria for normal distribution (Shapiro–Wilk test) they are expressed as median with 25th and 75th percentiles (interquartile range). Continuous parameters were compared by using the Wilcoxon test, the categorical parameters expressed as counts and percentages were compared using the chi-square or Fisher exact test. The primary and secondary outcomes were analysed using the chi-squared test, time to event analysis (Kaplan–Meier analysis with log rank test) and the univariate and multivariate Cox regression. The initial set of confounders for multivariate Cox regression (N = 19) was selected based on expert judgment. Subsequently, feature selection was performed using LASSO regularization with tenfold cross-validation using the glmnet 4.1-8 package43. For the multivariate model the missing values were input using the missForrest algorithm44. A p-value < 0.05 was considered statistically significant.

Results

Altogether, 339 patients (189 males, 55.8%), aged 71 (59; 77) years, were included in the analysis due to the detected presence of a viral load. The diagnosis was made by detecting NAg in the serum of 155 patients with a median of 217 (28; 1524) pg/ml, antigen detection in the sputum of 18 patients with a median of 18 (4.6; 32) COI, antigen detection in the nasopharyngeal samples of 44 patients with a median of 17 (8; 35) COI and the real-time PCR performed on nasopharyngeal samples of 122 patients with a median Ct of 21 (18; 27) (Fig. 1).

Fig. 1
figure 1

339 patients were selected according to the presence of viral load based on the positivity of the antigen tests or the real-time PCR. They were divided into the group without remdesivir therapy (199 patients, out of them 82 admitted to the ICU and 28 immune-supressed) and those treated with remdesivir (140 patients, out of them 103 admitted to the ICU and 57 immune-supressed).

Full size image

140 of the 339 patients were treated with remdesivir, of those 103 in the ICU, and 57 of the 140 were treated with immune suppression. The median number of administered 100 mg remdesivir vials was 5.46 (1.08) per patient, which confirms a full 5-day treatment in the absolute majority of patients. Four patients (2.9%) of the remdesivir group presented as re-infections. The control group consisted of 199 patients who were denied therapy, of whom 82 were treated in the ICU and 28 of the 199 with immune suppression. Eight patients (4%) of the control group were re-infections.

Remdesivir patients were by 3 years younger (p = 0.005), more frequently with a history of immune suppression (40.7% vs. 14.1%, p < 0.001) and required more oxygen therapy (74.6% vs. 61.8%, p = 0.016, Table 1). Their inflammatory markers (CRP, PCT) and leucocyte counts were mildly higher (p < 0.001), their lymphocyte counts were lower resulting in a higher leucocyte-to-lymphocyte ratio (p < 0.001, Table 1). Regarding therapy, corticosteroids were administered more frequently in the remdesivir group (70% vs. 52.3%, p = 0.001), as well as monoclonal antibodies (39.3% vs. 8.5%, p < 0.001) and the immune stimulant isoprinosine (41.4% vs. 17.6%, p < 0.001). Patients on remdesivir were more frequently vaccinated more than 3 weeks prior to admission (42.9% vs. 11.6%, p < 0.001, Table 1). The recorded types of vaccines were Comirnaty (Pfizer, n = 69, 83%) or Vaxzevria (Astra-Zeneca, n = 14, 17%). Their hospital and long-term outcomes did not differ from the control group (Table 1).

Table 1 Characteristics of 140 patients treated with remdesivir compared to 199 patients without treatment (rates and percentages for the categorical data and medians with interquartile ranges for the continuous parameters).
Full size table

The analysis of the ICU subgroup showed that patients treated with remdesivir had slightly higher morbidity scores (SOFA 8 (5;10) vs. 5 (3;9), p = 0.039, Table 2). They more often had a history of immune suppression (35.9% vs. 4.9%, p < 0.001), and their highest requirement for therapy of the respiratory insufficiency aligned more frequently with nasal oxygen prongs (13.6% vs. 1.2%, p = 0.002) or oxygen face mask (16.5% vs. 1.2%, p < 0.001). Their IPPV rates were similar to those of the control group, yet they were weaned more often with a tracheostomy than in the control group (42.5% vs. 22.4%, p = 0.01, Table 2). Their therapy frequently involved monoclonal antibodies against SARS-CoV-2 (39.8% vs. 7.3%, p < 0.001), isoprinosine (43.7% vs. 22%, p = 0.002) and prior vaccination (43.7% vs. 7.3%, p < 0.001, Table 2). Their ICU mortality (25.5%) was significantly lower than that of the control group (47.6%, p = 0.002), similarly the hospital mortality was lower in the remdesivir group (31.1% vs. 47.6%, p = 0.022, Table 2).

Table 2 Characteristics of 103 ICU patients treated with remdesivir compared to 82 ICU patients without treatment (rates and percentages for the categorical data and medians with interquartile ranges for the continuous parameters).
Full size table

Analysis of the subgroup with immunosuppression showed that the remdesivir patients were by 7 years younger (p = 0.008), fewer had a history of oncology disease (29.8% vs. 85.7%, p < 0.001) and fewer were after radiotherapy (1.8% vs. 17.9%, p = 0.014, Table 3) than the control patients. In contrast, there were more haemato-oncology patients (45.6% vs. 14.3%, p = 0.005) and more patients on chronic corticosteroids (38.6% vs. 10.7%, p = 0.008) in the remdesivir treated group (Table 3). The immunosuppression group treated with remdesivir had higher inflammatory markers (CRP, leucocytes, p < 0.03) than the control group (Table 3) and a lower lymphocyte count (p = 0.008) associating with a significantly higher leucocyte-to-lymphocyte ratio (p = 0.01). Regarding therapy, patients on remdesivir were more frequently concomitantly treated with low dose steroids (68.4% vs. 35.7%, p = 0.004, Table 3) and monoclonal antibodies (49.1% vs. 3.6%, p < 0.001). They were also more often vaccinated than the controls (54.4% vs. 10.7%, p < 0.001). The hospital and 12-month outcomes of the immunosuppression patients treated with remdesivir did not differ from the controls (Table 3).

Table 3 Characteristics of the 57 immune supressed patients treated with remdesivir versus 28 immune supressed patients without treatment (rates and percentages for the categorical data and medians with interquartile ranges for the continuous parameters).
Full size table

Regression analysis

The univariate Cox regression (Table 4) demonstrated significant positive impacts on the adverse outcome of the male gender (HR 1.518, p = 0.027), admission to intensive care (HR 2.674, p < 0.001), intubation and IPPV (HR 2.966, p < 0.001). In contrast, maintaining patients with severe Covid-19 only with oxygen nasal prongs was associated with a better outcome (HR 0.49, p = 0.015) as well as the use of oxygen face mask (HR 0.548, p = 0.034). The multivariate adjusted Cox regression (LASSO) selected covariates significantly influencing outcomes (Table 4). These were remdesivir (HR 0.66, p = 0.039, Fig. 2), admission to the ICU (HR 2.43, p = 0.006) and intubation with IPPV (HR 3.04, p < 0.001). Nine missing CRP values, ten leukocyte/lymphocyte ratio values and four creatinine values were imputed using the missForrest algorithm. The other most significant confounders were age, ICU admission, mechanical ventilation, leucocyte/lymphocyte ratio and admission creatinine.

Table 4 The univariate Cox regression (left) and the multivariate adjusted Cox regression (LASSO) selected covariates (right) significantly influencing outcomes of the 339 patients with viral load.
Full size table
Fig. 2
figure 2

Univariate Cox regression (A) and the multivariate adjusted Cox regression (LASSO) (B) comparing 12-month survival of all the patients with remdesivir to controls. The univariate 12-month survival analysis (C), and multivariate adjusted Cox regression (LASSO) (D) for remdesivir treated subset of ICU patients versus controls.

Full size image

For the subset of 185 ICU patients the univariate Cox regression (Table 5) demonstrated significant impacts on the adverse outcome of hypertension history (HR 1.65, p = 0.029), history of immune suppression (HR 1.62, p = 0.038), intubation and IPPV (HR 2.15, p = 0.004). In contrast, maintaining patients with severe Covid-19 on nasal oxygen prongs improved the outcome (HR 0.29, p = 0.037) as well as the use of oxygen face mask (HR 0.24, p = 0.016). The remdesivir therapy improved the outcome (HR 0.65, p = 0.043). The multivariate adjusted Cox regression (LASSO) selected covariates significantly influencing outcomes (Table 5). These were remdesivir (HR for adverse outcome 0.49, p = 0.006, Fig. 2) and intubation with IPPV (HR 2.55, p = 0.001). Eight missing CRP values, eight leukocyte/lymphocyte ratio values and two creatinine values were imputed using the missForrest algorithm. The other most significant confounders were age, leucocyte/lymphocyte ratio, admission creatinine and immunosuppression.

Table 5 The univariate Cox regression (left) and the multivariate adjusted Cox regression (LASSO) selected covariates (right) significantly influencing outcomes of the 185 ICU patients with viral load.
Full size table

For the subset of 85 patients with immune suppression the univariate Cox regression (Table 6) demonstrated significant impacts on the adverse outcome of the male gender (HR 2.31, p = 0.017), history of COPD (HR 2.37, p = 0.021), ICU admission (HR 4.48, p < 0.001), intubation and IPPV (HR 3.03, p = 0.001). The multivariate adjusted Cox regression (LASSO) selected covariates significantly influencing outcomes (Table 6). These were a history of COPD (HR for adverse outcome 2.81, p = 0.012) and ICU admission (HR 6.01, p = 0.005). The impact of remdesivir did not reach statistical significance. One leukocyte to lymphocyte ratio value was imputed using the missForrest algorithm.

Table 6 The univariate Cox regression (left) and the multivariate adjusted Cox regression (LASSO) selected covariates (right) significantly influencing outcomes of the 85 patients with immune suppression and with viral load.
Full size table

Discussion

Although remdesivir was administered more frequently to patients with immune suppression including with haemato-oncology diseases who required more oxygen therapy than the controls, were more lymphopenic and had a more unfavourable course of the inflammatory markers, the drug has proven its benefits in patients with a high viral load. The impact on short and long-term outcomes was remarkable especially in patients with higher illness severity, e.g., in the intensive care unit, which contrasts with the guidelines for the administration of remdesivir applied throughout the pandemic.

Interestingly, patients on remdesivir were more often managed without mechanical ventilation, e.g., on nasal oxygen cannulas or face masks, which could have brought an associated prognostic benefit in comparison to intubation and mechanical ventilation which is in line with published data1,7,40,45. The adverse prognostic impact of an ICU admission and mechanical ventilation has been confirmed particularly in the immune compromised patients37,39. This study confirmed unfavourable impacts of the male gender, history of hypertension, COPD, or immune suppression on the short and long-term outcomes of severe Covid-19. Of the therapeutic factors, the group with remdesivir therapy was more frequently vaccinated (e.g., more than 3 weeks after the first jab), and this did not transmit in any favourable outcome parameter. The associated illness severity and the high proportion of patients with immune suppression with a poor previous immune response to vaccination could help to explain its limited protective effects on patient outcomes46,47. Similarly, the application of the monoclonal antibodies against the coronavirus spike protein, which was more frequent in the remdesivir group, was not associated with any outcome data. These results allude to a high illness severity of the intensive care patients which was likely less in the Recovery trial admitting a benefit of the monoclonal antibodies even for patients admitted to the hospital with Covid-1948. The immunity stimulant isoprinosine found more often in the remdesivir group has not shown any significant impact on the outcomes in the multivariate regression analysis. The number of re-infections among the patients on remdesivir and in the control group were negligible.

Although previous, mostly retrospective studies reported a benefit of remdesivir in slowing the progression of severe Covid19 and avoidance of admission to the ICU, mechanical ventilation, and death, this research suggests a benefit specifically for patients in the ICU with very severe disease and detected load of a living virus. In general, the viral load is a key to understanding the illness severity and the immune response. A high load of living virus has been correlated with the probability of intubation, mechanical ventilation, and death45,49,50,51. It is also associated with an immune response, seroconversion, and the native immunisation with coronavirus of the general population29,52.

This has a substantial implication for adequate testing. A qualitative PCR from a nasopharyngeal or sputum sample may represent only mucosal infection or colonisation that has been associated with approximately 50% chance for triggering of a systemic immune response, thus resulting in a similar probability of seroconversion and a build-up of the native immunity which is associated with outcomes52,53. In other words, patients with only qualitative PCR mucosal positivity without considering any clinical symptoms may have about 50% chance for re-infection due to the absence of systemic immune response52. Therefore, assessing the viral load with some of the available tests would be crucial for the prognosis and for the indication of antiviral medication which would be more targeted and justified than in subjects tested only by means of a qualitative PCR.

In 64% of our selected cohort of 339 patients, the viral load was detected by suitable antigen tests performed in certified laboratory settings. Since early 2021 the antigen tests have been tested as accurate not only for symptomatic patients but also for oligosymptomatic and asymptomatic subjects with possible implications for population screening27,47,54. The specificity of antigen tests for detecting the presence of a viral load makes them a great tool for the detection of high-risk patients and the target population for the antiviral therapy. The requirements for antigen tests have been published27 and their growing popularity has also been supported by financial savings compared to a real-time PCR. The economic aspect of our study is not only related to testing, but also to the costly administration of remdesivir, which may become more medically and economically efficient when targeting patients with a viral load detected by modern antigen tests55,56,57.

The limitations of the study are multiple. This is a retrospective study on selected severe Covid-19 patients with a detected living virus—in contrast to other studies working with only qualitative PCR positivity which, we believe, may not be sufficient for the administration of the medication against the coronavirus. Our diagnosis of viral load are only estimates based on the certified laboratory tests and not controlled by a qualitative viral analysis. By selecting a rather large group that tested positive for viral load, we cannot exclude the effects of remdesivir on another subset of patients, for example, less burdened with an immune compromise and a cumulation of risk factors. Specifically, for patients with immune compromise, the study was likely underpowered, and we cannot exclude the benefits of remdesivir with respect to the reported prolonged replication of the coronavirus in these patients10,11,28,37. There is also a potential problem with the inclusion of all oncology treated patients, those on immune suppressants and chronic dialysis into one group as these diseases are heterogenous in their response to coronavirus, immune interventions, and a vaccination35,36,46. The median remdesivir dosage represents a full five-day therapy in the absolute majority of patients included, we cannot dismiss an impact of a shorter therapy indicated in less severe patients with a presence of a viral load. For comparison, we show the differences between the univariate and the multivariate regression analysis. Regardless of the LASSO multivariate Cox analysis we cannot fully exclude a bias given by the complex therapeutic interventions to severe Covid-19 patients, especially in the ICU.

Conclusions

Despite advanced Covid-19 and high illness severity of the intensive care patients the therapy with remdesivir associated with the outcome benefits in patients with a presence of viral load, which was estimated by widely available antigen tests in majority of patients or less frequently by the real-time PCR with determination of the cycle threshold. The result may serve as a basis for further prospective research to better justify antiviral therapy in the advanced stages of the disease.