Early treatment of COVID-19 with anakinra guided by soluble urokinase plasminogen receptor plasma levels: a double-blind, randomized controlled phase 3 trial

Early increase of soluble urokinase plasminogen activator receptor (suPAR) serum levels is indicative of increased risk of progression of coronavirus disease 2019 (COVID-19) to respiratory failure. The SAVE-MORE double-blind, randomized controlled trial evaluated the efficacy and safety of anakinra, an IL-1α/β inhibitor, in 594 patients with COVID-19 at risk of progressing to respiratory failure as identified by plasma suPAR ≥6 ng ml−1, 85.9% (n = 510) of whom were receiving dexamethasone. At day 28, the adjusted proportional odds of having a worse clinical status (assessed by the 11-point World Health Organization Clinical Progression Scale (WHO-CPS)) with anakinra, as compared to placebo, was 0.36 (95% confidence interval 0.26–0.50). The median WHO-CPS decrease on day 28 from baseline in the placebo and anakinra groups was 3 and 4 points, respectively (odds ratio (OR) = 0.40, P < 0.0001); the respective median decrease of Sequential Organ Failure Assessment (SOFA) score on day 7 from baseline was 0 and 1 points (OR = 0.63, P = 0.004). Twenty-eight-day mortality decreased (hazard ratio = 0.45, P = 0.045), and hospital stay was shorter.

suPAR levels to first identify patients at risk of progressing to severe respiratory failure or death and then to initiate early targeted treatment with anakinra, a recombinant IL-1 receptor antagonist that blocks the activity of both IL-1α and IL-1β. The open-label, phase 2 SAVE study was conducted as a proof of concept for this approach 5 . Results showed a 70% decrease in the relative risk of progression to severe respiratory failure and a significant reduction in 28-d mortality with anakinra treatment compared to standard of care.
The SAVE-MORE study (suPAR-guided Anakinra treatment for Validation of the risk and Early Management Of seveRE respiratory failure by COVID-19) is a pivotal, confirmatory, phase 3, double-blind randomized controlled trial that evaluated the efficacy and safety of early initiation of anakinra treatment in hospitalized patients with moderate or severe COVID-19. The primary objective was to evaluate the efficacy and safety of early anakinra administration on the 11-point ordinal WHO-CPS on day 28 from the start of treatment.

Results
Patients. From 23 December 2020 to 31 March 2021, 1,060 patients with molecular definition of COVID-19 by positive polymerase chain reaction (PCR) for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) were screened, and 606 were randomized at 37 study sites (Methods). The main reason for exclusion from the study was suPAR <6 ng ml −1 . Twelve patients withdrew consent and requested removal of all data, leaving a final intention-to-treat (ITT) analysis cohort of 594 patients. All patients received standard of care; 189 patients were allocated to the placebo arm; and 405 patients were allocated to the anakinra arm. Only one patient was lost to follow-up ( Fig. 1). Baseline characteristics and co-administered treatments were similar between the two treatment arms (Table 1). Overall, 91.6% of patients had severe pneumonia as defined by the WHO classification for COVID-19. Patients not on dexamethasone at baseline started dexamethasone in compliance with standard of care after the start of the study drug.
Primary and secondary endpoints. The distributions of patient scores on the 11-point WHO-CPS in the two treatment arms at day 28 (primary outcome) are shown in Table 2 and Fig. 2a. In brief, 50.4% (204/405) of patients receiving anakinra had fully recovered with no viral RNA detected on day 28 compared to 26.5% (50/189) of patients receiving placebo, and 3.2% (13/405) and 6.9% (13/189) of patients in the anakinra and placebo arms, respectively, died. Overall, the unadjusted proportional odds of having a worse score on the 11-point WHO-CPS at day 28 with anakinra was 0.36 versus placebo (95% confidence interval (CI) 0.26-0.49, P < 0.0001; ordinal regression analysis) (Fig. 2a). The assumptions of the ordinal regression analysis-that is, the goodness-of-fit test and the parallel lines test-were not statistically significant, indicating that there was a high likelihood of the treatment effect size being homogeneous for all 11 points of the WHO-CPS.
Multivariate ordinal regression analysis of the primary outcome was performed using the stratification factors for randomization as independent variables-that is, disease severity, intake of dexameth-1,060 were assessed for eligibility 454 were excluded • suPAR < 6 ng ml -1 (n = 405) • pO 2 /FiO 2 < 150 mmHg (n = 23) • Withdrew consent before randomization (n = 12) • Unwillingness not to remain pregnant during the study period (n = 3) • Age < 18 years (n = 2) • Anti-cytokine biologicals the last month (n = 2) • Stage IV solid tumor malignancy (n = 2) • Absence of radiological findings of pneumonia (n = 1) • Primary immunodeficiency (n = 1) • Neutrophils < 1,500 per mm 3 (n = 1) • Transfer to another hospital before randomization (n = 1) • Oral or IV ≥ 0.4 mg kg -1 prednisone for >15 last days (n = 1) asone, body mass index (BMI) >30 kg m −2 and country (Fig. 2b). Results of the univariate analysis showed that dexamethasone treatment and baseline severe COVID-19 were associated with higher odds, and anakinra treatment with lower odds, of a worse (higher) WHO-CPS score at day 28. Of note, the observed higher odds of a worse outcome with dexamethasone likely reflects the administration of dexamethasone to patients with severe pneumonia rather than a true detrimental effect of dexamethasone. In the multivariate analysis, treatment with anakinra was the only independent variable associated with the primary outcome. Compared to placebo, the adjusted proportional odds of having a worse score on the 11-point WHO-CPS at day 28 with anakinra was 0.36 (95% CI 0.26-0.50, P < 0.0001) (Fig. 2b). Three pre-specified confirmatory analyses of the primary endpoint were performed: (1) comparison of patients' WHO-CPS score distributions in the anakinra versus placebo arms at day 14; (2) analysis of the proportion of patients with persistent disease at day 28 in each treatment arm (no disease persistence defined as full recovery with no viral RNA detected, WHO-CPS score of 0; persistent disease defined as WHO-CPS score ≥1) and of the proportion of patients with severe disease or who were dead at day 28 in each treatment arm (WHO-CPS score ≥6); and (3) analysis of the progression to severe respiratory failure (defined as respiratory ratio <150 mm Hg, necessitating high-flow oxygen (HFO), NIV or MV) or death by day 14. All three confirmatory analyses fully supported the clinical benefit of anakinra treatment. Compared to placebo, the unadjusted proportional odds of having a worse score on the 11-point WHO-CPS at day 14 with anakinra was 0.57 (95% CI 0.42-0.77, P < 0.0001) (Extended Data Fig. 1). After multivariate adjustment using the stratification factors for randomization as covariates, the OR was 0.58 (95% CI 0.42-0.79, P = 0.001) (Supplementary Table 1), showing that treatment with anakinra was an independent variable associated with clinical benefit compared to placebo at day 14. Multivariate logistic regression analysis of persistent disease at day 28 indicated that severe COVID-19 at baseline significantly increased the risk of disease persistence, whereas treatment with anakinra significantly reduced the risk (OR of WHO-CPS ≥1, anakinra versus placebo: 0.36; 95% CI 0.25-0.53, P < 0.0001) (Supplementary Table 2). In addition, treatment with anakinra significantly reduced the risk of severe disease or death at day 28 (OR of WHO-CPS ≥6, anakinra versus placebo: 0.46; 95% CI 0.26-0.83, P = 0.010) (Supplementary Table 2). The third confirmatory analysis validated the results of the phase 2 SAVE trial. A lower proportion of patients treated with anakinra progressed to severe respiratory failure or death by day 14 compared to placebo (20.7% versus 31.7% in the anakinra and placebo arms, respectively; hazard ratio = 0.62, 95% CI 0.45-0.87, P = 0.005) (Extended Data Fig. 2 and Supplementary Table 3). Finally, survival analysis showed that anakinra treatment significantly reduced the risk of death by day 28 compared to placebo (3.2% versus 6.9% in the anakinra and placebo arms, respectively; hazard ratio = 0.45, 95% CI 0.21-0.98, P = 0.045) (Extended Data Fig. 3).
The 11 points of the WHO-CPS are also grouped into five strata. Although, in the original protocol, it was considered to compare the frequency of strata between the two arms, the statistical analysis plan (SAP) developed with the COVID-19 Emergency Task Force (COVID-ETF) of the European Medicines Agency (EMA) pre-defined the use of the entire 11-point WHO-CPS. The comparison of outcomes using the WHO-CPS strata (uninfected (0), ambulatory with mild disease (1-3), hospitalized with mild disease (4,5), hospitalized with severe disease (6-9) and dead (10)) between the two study arms (Supplementary Tables 4 and 5) was also consistent with the analysis performed using the full 11-point scale of the WHO-CPS.
The rate of protocol deviations from the standard of care treatment was significantly greater in patients randomized to the placebo arm (27/189, 14.3% versus 13/405, 3.2% in the anakinra arm; P < 0.0001). Protocol deviations regarding dexamethasone use in the placebo arm were most commonly related to increasing the dose and/or lengthening the duration of administration (Supplementary Table 6). All five sensitivity analyses confirmed the analysis of the primary endpoint (Supplementary Table 7).
A significant benefit with anakinra treatment was observed for all five secondary clinical endpoints. Decreases in WHO-CPS score from baseline by days 28 and 14 and Sequential Organ Failure Assessment (SOFA) score from baseline by day 7 were significantly greater with anakinra versus placebo ( Table 2 and Supplementary  Tables 8-10). Moreover, in the anakinra group, mean time until Over-time follow-up of laboratory values showed that, in patients who were treated with anakinra compared to patients who received placebo: (1) the absolute lymphocyte count was significantly increased by day 7; (2) circulating IL-6 levels were significantly decreased by days 4 and 7; and (3) plasma CRP levels were significantly decreased by day 7 (Fig. 3).
Adverse events. The frequency of patients with at least one serious treatment-emergent adverse event (TEAE) was 21.7% (41/189) in the placebo arm and 16.0% (65/405) in the anakinra arm (Table 3  and Supplementary Table 11). The most common captured serious TEAEs were infections, but they were less frequent in the anakinra arm than the placebo arm (34/405, 8 between the two arms. The frequency of non-serious anemia was lower in anakinra-treated patients (58/405, 14.3% versus 37/189, 19.6%). There was a trend for greater frequency of neutropenia with anakinra treatment (12/405, 3% versus 1/189, 0.5% in placebo) ( Table 3 and Supplementary Table 12).
Post hoc analyses. Although no differences between the two arms were found for baseline values of IL-6 (P = 0.22), ferritin (P = 0.59) and respiratory ratio (P = 0.10), multivariate analysis of the 28-d WHO-CPS was repeated post hoc including, as independent variables, IL-6, ferritin and respiratory ratio. For this analysis, the three variables were dichotomized using their median value. The intake of remdesivir was also included as an independent variable. Results confirmed anakinra benefit (Supplementary Table 13).
One hundred forty-six patients had low baseline CRP (below the first quartile of 25.3 mg L −1 ) 6 ; several of these patients had low lymphocytes and increased IL-6, ferritin and suPAR (Supplementary  Table 14). Clinical benefit of anakinra was found in patients with low CRP (Supplementary Table 15).
Four hundred eighty-seven patients had scores of 2 or more points for the COVID-associated hyper-inflammation syndrome score; 7 multivariate ordinal regression analysis of WHO-CPS on day 28 for these patients showed overall anakinra benefit     Fig. 7). Predictors of favorable response to anakinra were defined using cutoffs of laboratory parameters that can predict progression to severe respiratory failure or death after 14 d (Extended Data Fig. 8). A combination of at least two of CRP >50 mg L −1 , neutrophil-to-lymphocyte ratio (NLR) >5.5, ferritin >700 ng ml −1 and aspartate aminotransferase (AST) > 44 U L −1 might predict this unfavorable outcome (Supplementary Table 18). Patients with levels above these cutoffs had lower odds of progression to severe respiratory failure or death if they received anakinra (Supplementary Table 19).
We collected post hoc information for the 14-d outcome of patients with suPAR <6 ng ml −1 ; only 2.9% of patients progressed to severe respiratory failure or death.

Discussion
The SAVE-MORE trial evaluated a novel approach for the management of COVID-19, which relies on early identification of patients at risk for unfavorable outcome using suPAR and provision of targeted treatment with anakinra. Results showed considerable efficacy of 10-d subcutaneous administration of anakinra in patients with COVID-19 with plasma suPAR ≥6 ng ml −1 . The odds of a worse clinical outcome at day 28 with anakinra, as compared to placebo, was 0.36. The clinical benefit with anakinra treatment was already apparent from day 14, and this is of clinical importance because the first 14 d is the period during which a patient is expected to worsen; anakinra benefit was maintained until day 28. The magnitude of the efficacy of anakinra was shown in all multivariate analyses where in the presence of anakinra treatment the effect of baseline disease severity on the final outcome was lost. The proportion of patients who fully recovered exceeded 50%, and the number of patients who remained with severe disease was reduced by 54%. Most of the study population had severe COVID-19 at baseline, and 85.9% were receiving standard of care treatment containing dexamethasone. Relative decrease of mortality was 55% and reached 80% for patients likely having cytokine storm 8 . The incidence of serious TEAEs, mainly of infections, was lower in patients treated with anakinra. Patients treated with anakinra had a trend for more often non-serious neutropenia. Two main limitations of the SAVE-MORE trial need to be acknowledged: the lack of enrollment of patients with critical COVID-19 and the difficulty for application of suPAR in all hospital settings. Not all hospitalized patients with COVID-19 have as high risk of mortality as patients progressing to critical illness, who were included in the REMAP-CAP and RECOVERY trials 9,10 . This was further confirmed in the STOP-COVID 11 and CAN-COVID 12 trials where the 28-d mortality of placebo-treated patients with non-critical COVID-19 ranged from 5.5% to 7.2%. The low mortality in these studies might reflect an improvement of outcomes with the current standard of care compared to the beginning of the pandemic. Results of the SAVE-MORE trial indicated that suPAR is an indicator not only of the risk of progression to severe respiratory failure or death but also of persistence of COVID-19. Most patients who received placebo remained ambulatory with symptoms on day 28. The introduction of suPAR to guide treatment could be problematic in settings where this tool is not available. Measuring suPAR is introducing a personalized treatment approach, because early increase of suPAR is indicative of excess release of DAMPs, leading to pro-inflammatory phenomena through activation of IL-1α/ β 3 . Anakinra blocks both IL-1α and IL-1β by blocking their common receptor. The attenuation of the inflammatory responses by anakinra was shown by the decrease of IL-6 and of CRP circulating concentrations and by the increase of the absolute lymphocyte counts. Post hoc analysis revealed that CRP, NLR, ferritin and AST are predictors of favorable anakinra response and might also be used instead of suPAR.
CRP and ferritin are used to classify COVID-19 hyperinflammation and work also as predictors of treatment escalation 6 . In the SAVE-MORE study, some patients enrolled because of increased suPAR had low CRP; anakinra benefit was found even when CRP was low.
The results validate the findings of the previous SAVE open-label phase 2 trial. In SAVE, the incidence of respiratory failure after 14 d with anakinra treatment was 22.3% 5 ; in the SAVE-MORE trial, it was 20.7%. For patients in SAVE-MORE who were eventually admitted to the ICU, time until discharge was significantly shorter in the anakinra-treated group than in the placebo group, as was also observed in the SAVE trial 5 .
Since the beginning of the COVID-19 pandemic, immunomodulators have been suggested as a strategy to attenuate the exaggerated immune response of the host 13 . The most common administered drugs are anakinra and tocilizumab, targeting the IL-1 and IL-6 pathways, respectively. Anakinra use in COVID-19 has been reported in several retrospective and prospective studies, as well as in one randomized controlled trial 5,[14][15][16][17][18][19][20][21] . Although most studies report mortality benefit, it is difficult to compare findings from observational studies with those of the SAVE-MORE trial. Previous studies differed with regard to selection of patients, severity of illness and stage of the disease. Duration of treatment, dose and route of administration were also variable, and the WHO-CPS was not studied as a primary endpoint. Indeed, four of the studies were done in patients with critical illness, with plasma levels of CRP and ferritin exceeding the levels of the SAVE-MORE study population [14][15][16][17][18] . In the CAN-COVID trial, 454 patients with hypoxic COVID-19 not requiring MV and with signs of hyper-inflammation were randomized to a single injection of placebo or the anti-IL-1β monoclonal antibody canakinumab 12 . The trial failed to meet the primary endpoint-namely, the rate of survivors not in need of NIV or MV by day 29. The differential findings of SAVE-MORE could be explained by the activity of anakinra against IL-1α, which is not inhibited by canakinumab. An alternative explanation for this discrepancy is that patient stratification in CAN-COVID used CRP or ferritin instead of suPAR as biomarkers of hyper-inflammation.
The clinical benefit of tocilizumab has been studied in six clinical trials. Τhe patient populations of four clinical trials were much similar to the population of the SAVE-MORE trial [22][23][24][25] . Clinical benefit from tocilizumab treatment was shown in only one trial 23 . Conversely, clinical benefit from tocilizumab treatment was found in the RECOVERY 9 and REMAP-CAP 10 trials, which included patients with severe and critical COVID-19. In the RECOVERY trial 9 , mortality was decreased from 35% with usual care to 31%, and, in the REMAP-CAP trial 10 , the median number of organ support-free days was increased from 0 d with usual care to 10 d with tocilizumab treatment. The benefit of tocilizumab in patients with more severe disease might be explained by the biology of the disease course. We previously showed that circulating monocytes in critical COVID-19 present with complex immune dysregulation characterized by decreased efficiency of antigen presentation and inappropriate maintenance of the potential for excess cytokine production, which were restored upon exposure to tocilizumab 26 . Our findings suggest that suPAR should be measured upon admission of all patients with COVID-19 who do not need oxygen or who need nasal or mask oxygen, and that, if suPAR levels are 6 ng ml −1 or higher, anakinra treatment might be a suitable therapy. For patients with low respiratory ratio who need NIV or MV, tocilizumab might be the most appropriate drug of choice.
In conclusion, the SAVE-MORE trial showed that early start of treatment with anakinra guided by suPAR levels in patients hospitalized with moderate and severe COVID-19 significantly reduced the risk of worse clinical outcome at day 28.

Online content
Any methods, additional references, Nature Research reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/ s41591-021-01499-z. Fig. 2 | Time to progression into severe respiratory failure. Severe respiratory failure is defined as respiratory ratio-PF < 150 necessitating high flow oxygen/non-invasive ventilation/mechanical ventilation or death) by day 14. comparison is done by univariate cox regression analysis. the hazard ratio (Hr) and the 95% confidence intervals (cIs) are provided. Abbreviations HFO: high flow oxygen; MV: mechanical ventilation; n: number; P/F: respiratory ratio; Soc; standard-of-care. Fig. 8 | AST, NLR, CRP and ferritin for prediction of response to anakinra. A, c, e, and G) receiver operator characteristics (rOc) curves of ASt, NLr, crP and ferritin to predict progression into severe respiratory failure (SrF) or death the first 14 days. the exact P-value of the rOc analysis in panel A is 9.3 ×10-8; in panel c is 5.1 ×10-5; in panel e is 2.7 ×10-7; and in panel G is 1.6 ×10-7. b, D, G and H) Diagnostic performance of ASt > 44U/l, NLr > 5.5, crP > 50 mg/l and ferritin > 700 ng/ml for prognostication of progression into SrF or death by day 14. the selected cut-offs are the Youden indexes for the best trade-off of sensitivity and specificity for each variable. Analysis involved patients for whom all four variables were available at baseline before start of the study drug. comparisons are done by Mantel-Haenszel test estimating the odds ratio and 95% confidence intervals. the exact P value of the test in panel b is 4.1 ×10-5; in panel F 5.0 ×10-6; and in H 6.5 ×10-4. Abbreviations AUc: area under the curve; ASt: aspartate aminotransferase; cI: confidence intervals; crP: c-reactive protein; n: number of patients; NLr: neutrophil to lymphocyte ratio; NPV: negative predictive value; Or: odds ratio; PPV: positive predictive value.