RAndomized Clinical Trial Of NAfamostat Mesylate, A Potent Transmembrane Protease Serine 2 (TMPRSS2) Inhibitor, in Patients with COVID-19 Pneumonia

Even though SARS-CoV-2 was declared by WHO as constituting no longer a public health emergency, the development of effective treatments against SARS-CoV-2 infection remains a critical issue to prevent complications, particularly in fragile patients. The protease inhibitor nafamostat, currently used in Japan and Korea for pancreatitis, owing to its anticoagulant properties for disseminated intravascular coagulation (DIC), is appealing for the treatment of COVID-19 infection, because it potently inhibits the transmembrane protease serine 2 (TMPRSS2) that, after virus binding to ACE-2, allows virus entry into the cells and replication. Moreover, it could prevent the DIC and pulmonary embolism frequently associated with COVID-19 infection. The goal of the RAndomized Clinical Trial Of NAfamostat (RACONA) study, designed as a prospective randomized, double-blind placebo-controlled clinical trial, was to investigate the efficacy and safety of nafamostat mesylate (0.10 mg/kg/h iv for 7 days), on top of the optimal treatment, in COVID-19 hospitalized patients. We could screen 131 patients, but due to the predefined strict inclusion and exclusion criteria, only 15 could be randomized to group 1 (n = 7) or group 2 (n = 8). The results of an ad interim safety analysis showed similar overall trends for variables evaluating renal function, coagulation, and inflammation. No adverse events, including hyperkalemia, were found to be associated with nafamostat. Thus, the RACONA study showed a good safety profile of nafamostat, suggesting that it could be usefully used in COVID-19 hospitalized patients.


Introduction
The COVID-19 pandemic outbreak that emerged in China in 2019 caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has infected more than 769 million people and killed over 6.9 million worldwide (https://www.who.int,accessed

Study Design and Aims
The RACONA is a randomized, double-blind, group-sequential parallel-arm, placebocontrolled trial (allocation ratio 1:1) testing efficacy and safety of nafamostat mesylate on top of best standard of care in hospitalized patients with severe COVID-19 infection.The impact of nafamostat mesylate on several inflammation biomarkers was also assessed as a secondary endpoint.
The trial was registered at clinicaltrial.gov (Identifier: NCT04352400) and was approved by the Ethics Committee of INMI Lazzaro Spallanzani (IRCCS) and the Italian Medicines Agency (AIFA).
To enforce control over bias, an algorithm was specifically created for this study.The algorithm uses a permuted block randomization sequence with stratification (see Supplemental Data for details).Strata were defined by the cross-combination of use of oxygen therapy (nasal duct, mask, etc.) and ongoing treatment with inhibitors of the renin-angiotensin-aldosterone system, as these drugs were suggested to affect outcomes of COVID-19 patients.Investigators and patients were blinded to the treatment administered.
Vital signs, including body temperature, heart rate, respiratory rate, systolic and diastolic blood pressure (BP), oxygen saturation, were monitored daily.Serum K + levels were measured after the first 6 h of infusion and daily during the 7 days of drug administration.
The anonymized data were entered and stored securely in an ad hoc created webbased collection data form (Zucchetti SpA, Lodi, Italy), and stored securely in a server protected with firewalls and passwords.
The primary efficacy outcome was defined as the time-to-clinical improvement, defined as the time from randomization to an improvement of two points (from the status at randomization) on a seven-category ordinal scale.The outcome followed the recommendations of the WHO R&D Blueprint expert group (https://iris.who.int/handle/10665/330680,accessed on 5 August 2023).

Statistical Analysis 2.3.1. Main Analysis
Main analysis was conducted as "as treated".Results are expressed as mean ± SD, or median and interquartile range, as appropriate.In the case of a skewed distribution, log-transformed data were used.Comparisons were performed with parametric or nonparametric tests (Wilcoxon Mann Whitney), as appropriate; Pearson χ 2 test was used for analysis of categorical variables.Generalized Linear Models (GLMs) were used to evaluate the relationship between treatments, and time.SPSS for Mac (version 28 for Mac, IBM-SPSS, Bologna, Italy) was used for the statistical analysis and significance was set at p < 0.05.

Outcome Evaluation
Bayesian analysis of the response rate was prespecified to estimate the probability of treatment benefit considering the recommendations for trials conducted with a limited sample size in frequentist design [18].A Beta-binomial model was used to analyze the difference in response rate between arms [18].
The posterior distribution for the difference in proportions outcome requires the estimation of the posterior distribution of the response rate in each arm, separately.It has been computed with the following resampling procedure [19]: A first resampling of the response rate π * treat from π treat |X treat , which is the posterior distribution for the treatment group.

3.
A posterior distribution, for the parameter related to the difference in proportions, has been obtained by calculating π * treat − π * control from the previously resampled distributions [20].

Sensitivity Analysis
The inference was expected to be seriously conditioned by the prior choice, because the data points available to estimate the likelihood were only a few.Hence, a sensitivity analysis was performed to assess the robustness of the inferential conclusion with respect to the different prior choices.
Different levels of penalization (discounting) are provided for the historical information using a power prior approach [23] in order to perform a sensitivity analysis on the prior choices.

•
Power Prior without discounting (Informative).A Beta (46,6) for the treatment arm and a Beta (40,10) for the control arm were derived considering the number of successes reported in the literature in a similar research framework [24].

•
Power Prior 50% discounting (Low Informative).The second scenario provides the same informative Beta as previously indicated, also applying a 50% down weight on parameters, to control the effect of prior information on final inference as indicated in the literature [25].

•
Power Prior 100% discounting (Uninformative).A Beta(1, 1) prior in both treatments arms has been considered.The inference results have been analyzed evaluating the hypothesis that π Treat − π Control < 0.

Results
Patient recruitment started on 4 June 2021, and stopped on 1 August 2022.The results reported herein refer to a safety ad interim analysis performed after recruitment stopped, before opening of the treatment codes, when the investigator was still blinded to randomization.After closing the study on 11 May 2023, we learned that blinded treatments 1 and 2 corresponded to nafamostat and placebo, respectively.
The baseline demographic and disease characteristics were well balanced in the groups, thus testifying to the effectiveness of the randomization (Table 1).
A total of 131 hospitalized patients with COVID-19 were screened, of whom 62 were found to be eligible for the study (Figure S1).After exclusion of 69 patients because of comorbidities, and/or older age, and/or no need of O 2 supplementation (see Table S2 for exclusion causes), 15 patients were randomized to group 1 (n = 7) and group 2 (n = 8).One patient withdrew his consent and, therefore, was excluded from the study (Figure S1).No violation or deviation from the protocol was reported.

Safety Outcomes
At the end of treatment (day 7), the clinical and biochemistry parameters showed no differences and no abnormal values (Table 2A).At day 14 of follow-up, higher values of S-creatinine and lower values of eGFR were observed in nafamostat group (p = 0.016 and p = 0.029, respectively; Table 2B).At GLM analysis, a decrease of SBP was found in the placebo group at days 1 (−25 mmHg) and 3 (−15 mmHg) (day 1, p = 0.02; day 3, p = 0.01; Figure 1C), followed by a transient increase of DBP at day 10 (Figure 1D).
Trends of pO 2 /FiO 2 and SaO 2 were similar, with progressive increase with time (Figure 2A,B).The SOFA score decreased with time in both groups, but a transient increase was seen at day 5 in the placebo group (p = 0.032) (Figure 2C).The respiratory rate similarly decreased in both treatment groups (Figure 2D).
The eGFR decreased slightly in the nafamostat group from day 6 to 10 (day 6, p = 0.013; days 10 and 14, p < 0.001), but returned to normal values.Trends of AST and ALT (Figure 3A,B), CRP (Figure 3C), and S-procalcitonin values were similar between groups.Clinically irrelevant fluctuations were observed in the platelets count in the placebo group (Figure 3D).

Outcomes
Rapid deterioration of lung function leading to death was seen in a 74-year-old female patient, who was initially referred to the Infectious Diseases Unit and the transferred to Pneumology Intensive Unit for dyspnea and interstitial pneumonia requiring oxygenation via high flow nasal cannulae, remdesivir, azithromycin ceftriaxone, steroids, and enoxaparin.Nine days after admission cardiac arrest occurred followed by successful cardiopulmonary resuscitation.An increase of D-dimer (1842 μg/L on day 4, CRP (180 mg/L) on day 5, and troponin I on day 8 (1173 ng/L) was seen, alon with apical akinesia with severe depression of LV ejection fraction (EF = 30%) at trans thoracic echocardiogram.Death occurred on day 10, due to pulseless electrical activit (PEA) unresponsive to CPR and acute kidney failure.
Necropsy findings showed anterior myocardial infarction due to severe narrowin of the left anterior descending coronary artery (LAD), interstitial pneumonia with lymphocytic infiltration, marked injury of pneumocytes with lymphocytic vasculitis and neoangiogenesis, in addition to pronounced damage of the renal tubule-interstitium adrenal medulla, and liver.Hence, the cause of death was attributed to PEA due t severe hypoxia in the setting of a (previously unknown) tight proximal narrowing of th LAD.
After unblinding, we found that the patient had been allocated to group 2, i.e., th placebo group, but had received treatment for only one day, before the transfer to the ICU No drug adverse events were observed during treatment in any patients.I particular, no patient developed cardiovascular disease, including arrhythmia and myocardial infarction, hemorrhages, or hyperkalemia or hyponatremia, defined as S-K + 5.0 mmol/L and Na + < 130.0 mmol/L, respectively.No differences were found in S-K + or S Na + average levels between groups (Figure 1A,B).

Outcomes
Rapid deterioration of lung function leading to death was seen in a 74-year-old female patient, who was initially referred to the Infectious Diseases Unit and then transferred to Pneumology Intensive Unit for dyspnea and interstitial pneumonia, requiring oxygenation via high flow nasal cannulae, remdesivir, azithromycin, ceftriaxone, steroids, and enoxaparin.Nine days after admission cardiac arrest occurred followed by successful cardiopulmonary resuscitation.An increase of D-dimer (1842 µg/L) on day 4, CRP (180 mg/L) on day 5, and troponin I on day 8 (1173 ng/L) was seen, along with apical akinesia with severe depression of LV ejection fraction (EF = 30%) at trans-thoracic echocardiogram.Death occurred on day 10, due to pulseless electrical activity (PEA) unresponsive to CPR and acute kidney failure.
Necropsy findings showed anterior myocardial infarction due to severe narrowing of the left anterior descending coronary artery (LAD), interstitial pneumonia with lymphocytic infiltration, marked injury of pneumocytes with lymphocytic vasculitis and neoangiogenesis, in addition to pronounced damage of the renal tubule-interstitium, adrenal medulla, and liver.Hence, the cause of death was attributed to PEA due to severe hypoxia in the setting of a (previously unknown) tight proximal narrowing of the LAD.
After unblinding, we found that the patient had been allocated to group 2, i.e., the placebo group, but had received treatment for only one day, before the transfer to the ICU.
No drug adverse events were observed during treatment in any patients.In particular, no patient developed cardiovascular disease, including arrhythmia and myocardial infarction, hemorrhages, or hyperkalemia or hyponatremia, defined as S-K + > 5.0 mmol/L and Na + < 130.0 mmol/L, respectively.No differences were found in S-K + or S-Na + average levels between groups (Figure 1A,B).
Fourteen recruited patients completed the follow up for the study outcome (seven in the treatment group and seven in the control group), with the response rates shown in Table 3.The Bayesian evaluation of potential differences in outcome is shown in Figure 4.The Bayesian analysis was conducted to evaluate the probability that the treatment with nafamostat would have a beneficial effect, based on observed data.To evaluate this probability in a Bayesian framework, prior information on the effect of the treatment should be considered in the analysis.Such prior information was incorporated in the analysis by depicting three scenarios, one corresponding roughly to the classical non-Bayesian framework, and two other scenarios where the a priori expectations were mildly or more strongly in favor of the treatment.Prior probabilities were also adjusted in a sensitivity analysis, to correct for a potential over-optimism in the expectations, via down weighting (i.e., penalizing a priori probabilities).Such prior distributions are plotted for clarity on the right side of Figure 4.The distribution functions of the probabilities that the difference in event rate is less than 0 have then been plotted.The estimated curves, which are "blending" prior information with observed data, show a clear signal for a beneficial effect of the nafamostat treatment.In the uninformative scenario the probability that the treatment is effective (difference not zero) is 69%.When prior expectations are taken into account, probabilities that nafamostat treatment is effective are stronger (in low informative prior scenario probability of an effective treatment is 88%, and in the informative prior scenario with down weight is 81%).
Fourteen recruited patients completed the follow up for the study outcome ( the treatment group and seven in the control group), with the response rates sh Table 3.The Bayesian evaluation of potential differences in outcome is shown in F The Bayesian analysis was conducted to evaluate the probability that the treatme nafamostat would have a beneficial effect, based on observed data.To evalu probability in a Bayesian framework, prior information on the effect of the tr should be considered in the analysis.Such prior information was incorporated analysis by depicting three scenarios, one corresponding roughly to the classic Bayesian framework, and two other scenarios where the a priori expectations wer or more strongly in favor of the treatment.Prior probabilities were also adjus sensitivity analysis, to correct for a potential over-optimism in the expectations, v weighting (i.e., penalizing a priori probabilities).Such prior distributions are plo clarity on the right side of Figure 4.The distribution functions of the probabilities difference in event rate is less than 0 have then been plotted.The estimated curves are "blending" prior information with observed data, show a clear signal for a be effect of the nafamostat treatment.In the uninformative scenario the probability treatment is effective (difference not zero) is 69%.When prior expectations are ta account, probabilities that nafamostat treatment is effective are stronger informative prior scenario probability of an effective treatment is 88%, and informative prior scenario with down weight is 81%).

Discussion
The RACONA study was conceived in 2020, soon after the breakout of COVID-19 in China and its fast spreading that caused more than 15 thousand deaths, between 23 February and 30 April, in Italy (https://www.salute.gov.it/,accessed on 10 August 2023).Starting from the identification of the role of TMPRSS2 in virus spread and recruitment of the cell machinery, and of the effectiveness of camostat, a TMPRSS2 inhibitor in blunting SARS-CoV-2 entry into cells [6], our attention was caught by nafamostat, a protease inhibitor 10-thousand-fold more potent than camostat on TMPRSS2, which was known to be welltolerated in end stage kidney disease patients on chronic hemodialysis in Japan (Y.K., personal communication).This was also because, in contrast to camostat, which has antifibrinolytic effects and can cause eosinophilic pneumonia, nafamostat has anticoagulant properties that seemed attractive for preventing the thrombotic events associated with COVID-19 infection [26].
Based on these premises, we conceived the RACONA study to evaluate the efficacy and safety of nafamostat in COVID-19 infected patients.However, owing to several hurdles, including the fact that nafamostat had never entered the EU for clinical use before, approval of the study protocol took several months, thus allowing us to start recruitment only in May 2021.In the meantime, other treatments, including antiviral drugs, monoclonal antibodies, and vaccines, had been approved by the Italian regulatory agency AIFA.Notwithstanding the active involvement of three units, i.e., Pneumology, Infectious Diseases, and Internal Medicine, at our University Hospital, this rendered enrollment extremely difficult, and explains why we could enroll only 15 patients, a cohort that, according to our prior calculations, was underpowered to detect differences of treatment efficacy between groups.Considering, however, that nafamostat was never used before in patients outside of Japan and South Korea, we thought it important to perform a safety analysis of this drug.The latter showed that there were no major adverse events, including hemorrhages, or cardiovascular disease, including arrhythmia and myocardial infarction.
When designing the study, we paid attention to monitoring serum ion levels daily, because nafamostat inhibits the epithelial Na + channel (eNaC) and, therefore, could theoretically cause hyperkalemia and hyponatremia [27][28][29].It is worth noting that none of our patients developed any clinically relevant changes in an extended panel of biochemical markers, including serum K + and Na + levels (Figure 1A,B).This piece of information can be useful for patients who are receiving drugs, such as the mineralocorticoid receptor antagonists that act via eNaC, and, therefore, could have hyperkaliemia.
In a GLM analysis that included time in addition to treatment, we observed similar overall trends for most examined variables, with only transiently clinically differences between groups.As mentioned above, a patient of group 2 developed rapid deterioration of the lung function, which required transfer to ICU, with following death.An independent adverse event adjudication committee judged this death to be unrelated to treatment, but due to COVID-19 complications that were commonly seen at that time.After stopping the study and unblinding the treatment code, the patient was found to be allocated to the placebo group.On the whole, the data obtained support the concept that nafamostat is safe when administered in patients with no contraindications as severe CKD or low platelet count.
Interestingly, after we registered the RACONA study protocol in clinicaltrials.gov in April 2020, the interest for nafamostat rapidly grew in many countries, leading to a number of trials with a similar design in hospitalized patients for COVID-19 pneumonia.
A study with the same aims and methodology was posted in www.clinicaltrials.gov(accessed on 10 August 2023) by a Korean group (June 2020; Identifier: NCT04418128), but to date it is reported as "not yet recruiting".Six other trials have been completed, and four are still ongoing (Tables 4 and 5).In a randomized open-label trial performed in Russia in high-risk COVID-19 patients (Identifier: NCT04623021), the time to clinical improvement was reported to be shorter with nafamostat than with placebo.Phlebitis due to continuous i.v.infusion was the only side effect observed in the nafamostat group [24].
In an open label, controlled Scottish trial (DEFINE trial, NCT04473053), nafamostat was associated with more hyperkalemia (in 14.3% of patients) and deterioration of renal function [31], two adverse effects that we did not observe in RACONA.
In a retrospective cohort study performed in Japan in patients hospitalized for COVID-19 and administered with nafamostat mesylate within 2 days of admission, no differences were found in-hospital mortality between nafamostat and optimal treatment for COVID-19; no information on safety was provided [35].
Considering that use of nafamostat in COVID-19 patients has been proposed starting from solid mechanistic premises, and that the RACONA study has a strong randomized, double-blind, placebo-controlled study design, we believe that sharing our data with the scientific community is important at a time when nafamostat treatment outside of Japan and South Korea has been scarce.
The present results showed no serious adverse events with nafamostat, indicating that, when infused i.v. for 7 days, it is safe in COVID-19 patients.Moreover, results of our Bayesian analysis of efficacy evidenced a signal for a beneficial effect of the drug under different assumptions (Figure 4).Thus, these results indicate that nafamostat on top of current treatments for COVID-19 could be useful to enrich the therapeutic armamentarium against Sars-CoV-2, a disease that still affects a number of fragile people, with requirement of hospitalization and high-level health care.

Limitations of the Study and Perspectives
The major limitation of the study is the small sample size that was due to the difficulties in the enrollment and the large use of other competing treatment trials, i.e., antiviral drugs, monoclonal antibodies, and vaccines, which were meanwhile approved by AIFA for cure of the COVID-19 patients.Hence, since the RACONA study was underpowered to detect differences of treatment efficacy between groups, only a safety analysis was performed.However, following a Bayesian approach, we were able to detect a signal for a beneficial effect of the drug under different assumptions.Thus, to investigate the efficacy in more severely ill patients with COVID-19 infection, such as mechanically ventilated patients, further studies are necessary and worth pursuing.

Conclusions
Identification of new drugs, as well as repurposing of approved drugs that hold information on safety and tolerability accumulated over 30 years, are needed to contrast the changes of pandemics that follow the development of new variants and the waning of immunization.In the RACONA study, nafamostat showed a good safety profile and, therefore, could represent an effective tool, particularly against those variants that are more dependent on TMPRS2 (e.g., omicron vs. delta variant).

Institutional Review Board Statement:
The trial was registered at clinicaltrial.gov (Identifier: NCT04352400) and was approved by the Ethics Committee of INMI Lazzaro Spallanzani (IRCCS; EUDRACT 2020-002570-27) and the Italian Medicines Agency (AIFA).The study was conducted in accordance with the Declaration of Helsinki, and an independent committee was constituted for adverse event adjudication.
Informed Consent Statement: Written informed consent was obtained from all subjects involved in the study.

Figure 4 .
Figure 4. Posterior and prior distributions (blue curve is with penalization for over-optim the red curve is without it).The plots refer to the probability that the treatment is effec  −  is less than reported) under three scenarios of prior information (Info Informative with adjustment for optimism, and Uninformative).Probabilities are  = and  =  .

Figure 4 .
Figure 4.Posterior and prior distributions (blue curve is with penalization for over-optimism and the red curve is without it).The plots refer to the probability that the treatment is effective (i.e., π control − π treat is less than reported) under three scenarios of prior information (Informative, Informative with adjustment for optimism, and Uninformative).Probabilities are π 1 = π control and π 2 = π treat .

Table 1 .
Demographic and disease characteristics at baseline.
Mean ± SD, or median (range) as appropriate.Values of C-reactive protein, S-Ferritin, S-Procalcitonin, and CURB-65 refer to day 28.

Table 3 .
Number and percentages of responses in treatment and control arm.The 95% Credible Intervals are reported for the posterior distribution π Treat − π Control and for predictive posterior estimates provided in Informative, Low informative, and Uninformative scenarios.

Table 4 .
Completed trials of nafamostat mesylate for treatment of COVID-19 patients.: standard of cure; * Requiring nasal high-flow oxygen therapy and/or non-invasive mechanical ventilation; ** Evidence of acute lower respiratory disease on clinical assessment or imaging, and SpO 2 ≥ 94% on room air.Severe was defined as SpO 2 < 94% on room air, PaO 2 /FiO 2 < 300 mmHg, respiratory rate > 30 breaths/min, or lung infiltrates > 50%. SOC

Table 5 .
Ongoing trials of nafamostat mesylate for treatment of COVID-19 patients.