Final Results of a Randomized, Placebo-Controlled, Two-Arm, Parallel Clinical Trial of Proxalutamide for Hospitalized COVID-19 Patients: A Multiregional, Joint Analysis of the Proxa-Rescue AndroCoV Trial

Background The role of androgens on COVID-19 is well established. Proxalutamide is a second-generation, non-steroidal antiandrogen (NSAA) with the highest antiandrogen potency among NSAAs and concurrent regulation of angiotensin-converting enzyme 2 (ACE2) expression and inflammatory response. Proxalutamide has been demonstrated to be effective to prevent hospitalizations in early COVID-19 in randomized clinical trials (RCTs). Conversely, in hospitalized COVID-19 patients, preliminary results from two different arms of an RCT (The Proxa-Rescue AndroCoV Trial) also demonstrated a reduction in all-cause mortality. This study aims to report the final, joint results of the two arms (North arm and South arm) of the Proxa-Rescue AndroCoV trial of the two arms (North and South arms) combined, and to evaluate whether COVID-19 response to proxalutamide was consistent across different regions (Northern Brazil and Southern Brazil). Materials and methods Upon randomization, hospitalized COVID-19 patients received either proxalutamide 300mg/day or placebo for 14 days, in addition to usual care, in a proxalutamide:placebo ratio of 1:1 in the North arm and 4:1 in the South arm (ratio was modified due to preliminary report of high drug efficacy). Datasets of the South and North arms were combined, and statistical analysis was performed for the overall study population. Proxalutamide was compared to placebo group for 14-day and 28-day recovery (discharge alive from the hospital) and mortality rates, and overall and post-randomization hospitalization stay. Results of proxalutamide and placebo groups were also compared between the North and South arms. Analysis was also performed stratified by sex and baseline WHO COVID Ordinary Score. Results A total of 778 subjects were included (645 from the North, 317 from the proxalutamide group and 328 from the placebo group; 133 from the South arm, 106 from the proxalutamide group and 27 from the placebo group). Recovery rate was 121% higher in proxalutamide than placebo group at day 14 [81.1% vs 36.6%; Recovery ratio (RecR) 2.21; 95% confidence interval (95% CI), 1.92-2.56; p<0.0001], and 81% higher at day 28 (98.1% vs 47.6%; RecR, 1.81; 95% CI, 1.61-2.03; p<0.0001). All-cause mortality rate was 80% lower in proxalutamide than placebo group at Day 14 [8.0% vs 39.2%; Risk ratio (RR), 0.20; 95% CI, 0.14-0.29; p<0.0001], and 78% lower at Day 28 (10.6% vs 48.2%; RR, 0.22; 95% CI 0.16-0.30). Post-randomization time-to-discharge was shorter in proxalutamide [median, 5 days; interquartile range (IQR), 3-8] than placebo group (median, 9 days; IQR, 6-14) (p<0.0001). Results were statistically similar between North and South arms for all measured outcomes. Males and females presented similar results in all outcomes. Patients that did not require oxygen use (scores 3 and 4) did not present statistically significant improvement in recovery and mortality rates, whereas scores 5 and 6 presented significant improvements in all outcomes (p<0.0001 for all). Conclusion Proxalutamide increased recovery rate, reduced mortality rate and shortened hospital stay in hospitalized COVID-19 patients. Results were similar between the two different arms, providing further consistency for the efficacy of proxalutamide when used in late-stage COVID-19.


Introduction
Early in 2020, when the pandemics of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and its resulting disease, coronavirus disease 2019 (COVID- 19), was announced, early reports already indicated that males were an independent risk factor for severe COVID-19 and death, since the overrepresentation of males was not fully justified by the presence of comorbidities, body mass index (BMI), habits or age [1,2].

Statistical analysis
In the North arm, intention-to-treat (ITT) protocol was employed, whereas in the South arm a modified ITT (mITT) was employed. All subjects that remained at least 24 hours after the first dose of proxalutamide or placebo were considered for the mITT. The difference occurred because in the North arm sites were unable to communicate the discontinuation of the drug within less than 24 hours after randomization, which did not allow the employment of the mITT, whereas this communication was instantaneous in the South arm. Indeed, in the North arm, ITT and mITT would not lead to any differences.
To determine the efficacy of proxalutamide compared to usual care, combined proxalutamide groups (i.e., proxalutamide group from the South and North arms together) were compared to combined placebo groups (i.e., placebo group from the South and North arms together). Cox proportional hazard model was employed to calculate the hazard ratio (HR) and 95% confidence interval (95% CI), for 14-Day and 28-Day all-cause mortality rate, and the rate ratio (RR) of the 14-Day and 28-Day recovery rate. The Wilcoxon Rank Sum and Wilcoxon-Mann-Whitney U tests were employed to compare the differences of the ordinal scale scores at Day 14, length of hospitalization stay, and post-randomization time-to-discharge alive from the hospital, and the following baseline characteristics: median age and interval between hospitalization and randomization. Chi-Square Test was employed to compare the other baseline characteristics and TEAEs. XLSTAT 2021.5 (Addinsoft, Paris, France) and Easy Med Stat were used to run the statistical analysis.
To evaluate the level of similarity between the North and South arms, baseline characteristics of each arm were compared using the same statistical tools as used to compare proxalutamide and placebo groups. To assess the consistency between arms, the outcomes of the active and placebo groups from the North arm were compared to the respective groups from the South arm using Cox proportional hazard model and the TEAES were compared using Chi-Square Test.
To assess whether there were differences between the level of efficacy of proxalutamide arms between regions, the estimated number of additional subjects to be benefited from the treatment if all participants were treated with proxalutamide were calculated for all outcomes in the North and in the South arm, and then compared using Cox proportional hazard model. Similar statistical tools were employed for the comparative analysis of sex-and baseline COVID ordinary score-specific subpopulations for the corresponding outcomes and TEAEs.

Results
A total of 778 subjects were included in the study, including 645 from the North arm (317 from the proxalutamide group and 328 from the placebo group), and 133 from the South arm (106 from the proxalutamide group and 27 from the placebo group) of the Proxa-Rescue AndroCoV trial. Figure 2 depicts the CONSORT flowchart. No transmen, transwomen, non-binary, or intersex subjects were included.  Table 1 describes the baseline characteristics of overall, proxalutamide, and placebo groups resulted from the joint analysis of the North and the South arms. Except for the use of medications, none of the characteristics was statistically significantly different between groups. However, participants from the proxalutamide group were marginally older (p=0.062) and had marginally more type 2 diabetes (p=0.055) than the placebo group. None of the groups presented patients with chronic kidney disease (CKD), since this was criteria for exclusion per the protocol.

Characteristic
Overall N=778   Table 2 depicts the results of the outcomes in the combined overall, combined proxalutamide, and combined placebo groups. Figure     Recovery rate over 14 days after randomization, as the primary outcome of this study, was 121% higher in the proxalutamide group (81.1%) than in the placebo group ( Hospitalization stay was shorter in the proxalutamide group (median, 8 days; IQR, 6-13) than in the placebo group (median, 12 days; IQR, 8-18) (p<0.0001). Conversely, post-randomization time-to-discharge alive from the hospital was shorter in the proxalutamide group (median, 5 days; IQR, [3][4][5][6][7][8] than in the placebo group (median, 9 days; IQR, 6-14) (p<0.0001). Figure 4 illustrates the proportion of patients discharged alive in males (A) and in females (B), and Kaplan-Meier survival estimate of males (C) and females (D), when evaluated according to sex, recovery rate at Day 14 between males and females (115% and 131% higher recovery rate in the proxalutamide group, respectively), as well as at Day 28 (81% and 80%, respectively) (p<0.0001 between proxalutamide and placebo groups for all). All-cause mortality rates in males were similar to females at Day 14 (79% and 82% improvement, respectively) and at Day 28 (75% and 82% improvement, respectively) (p<0.0001 between proxalutamide and placebo groups for all).    A. Proportion of patients discharged alive from the hospital in subjects with baseline scores 3 and 4; B. Proportion of patients discharged alive from the hospital in subjects with baseline score 5; C. Proportion of patients discharged alive from the hospital in subjects with baseline score 6; D. Kaplan-Meier survival probability hospital in subjects with baseline scores 3 and 4; E. Kaplan-Meier survival probability hospital in subjects with baseline score 5; F. Kaplan-Meier survival probability hospital in subjects with baseline score 6. Table 3 describes the number and percentage of subjects that required the use of concurrent medications during the period of the study. All antibiotics, including cephalosporins (ceftriaxone, cefepime, cefuroxime), macrolides (azithromycin, clarithromycin), vancomycin, meropenem, and piperacillin/tazobactam, were used in a larger percentage of participants from the placebo group than from the proxalutamide group (p<0.0001 for all). Colchicine was also used in a larger percentage of patients from the placebo group (p<0.0001).       Table 6 compares results between proxalutamide groups and between placebo groups from the South and the North arm. Figure 6 and Figure 7 illustrate, respectively, recovery rates and mortality rates in the North arm, South arm, and overall population. Recovery rate at Days 14 and 28, and mortality rates at Days 14 and 28 were statistically similar between proxalutamide groups and between placebo groups. Hospitalization stay was slightly longer in the active group in the North compared to the South (p=0.05) while was lower in the placebo group in the North compared to the South (p=0.045). Post-randomization length of hospital stay was significantly shorter in the placebo group of the North arm (median, 10 days; IQR, 6-15) than the placebo group of the South arm (median, 12 days; IQR, 9-15) (p=0.005).

95%CI = 95% confidence interval
As shown in Table 8, there were no significant differences between proxalutamide groups and between placebo groups in SAEs (grades 3 to 5), except for a marginally increased chance of disease progression in females of the placebo group from the South arm (80.0%) compared to the North arm (50.7%) (p=0.055). Diarrhea was more commonly present among participants of the proxalutamide group from the South arm (24.5%) than in the North arm (16.1%) (p=0.048). Irritability and spontaneous erection were more commonly observed among overall and males of the proxalutamide group from the South arm (4.7% and 6.6%, respectively) than from the North arm (1.3% and 1.6%, respectively (p=0.046 and p=0.052, respectively). Vomiting, dyspepsia, or palpitations 0 (0.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%) n/a n/a n/a n/a

Discussion
This final, joint analysis of the Proxa-Rescue Trial including both South and North arms with proxalutamide for hospitalized COVID-19 patients demonstrated that proxalutamide is effective for COVID-19 in later stages of the disease and that benefits of proxalutamide for this population were consistent across different regions with distinct demographical characteristics.

In the South as in the North: successful reproduction of the findings from the Proxa-Rescue AndroCoV trial
The fact that this RCT demonstrated a large efficacy of proxalutamide in patients hospitalized due to COVID-19 in Northern Brazil, we had to address the question of whether these findings would be reproducible in a population with distinct characteristics in a different region.
The important improvements in clinical outcomes [30], including reductions in mortality rate and hospital length stay and increase in recovery speed rate, observed in the first study on hospitalized patients, were replicated in the Proxa-South Rescue Trial [31]. The all-cause mortality rate ratio between the proxalutamide arm and the placebo arm at Day 14 was similar between studies: 0.22 (95%CI, 0.09 -0.56) in the South arm and 0.21 (95%CI, 0.14 -0.32) in the North arm. At Day 28, all-cause mortality ratio rate was slightly higher in the South arm (0.28; 95%CI, 0.13 -0.63) than in the North arm (0.22; 95%CI 0.16 -0.31). The reduction of all-cause mortality at both Days 14 and 28 was between 78% and 72%, almost identical between the studies in the North and in the South.
Overall and post-randomization hospital stay until discharge in the placebo group was shorter in the North arm (Overall hospital stay: median, 12 days; 95%CI, 9 -16; post-randomization hospital stay: median: 10 days; 95%CI, 6 -14) than in the South arm (Overall hospital stay: median, 14 days; 95%CI, 12 -18; postrandomization hospital stay: median: 12 days; 95%CI, 8 -18) (p = 0.12 and 0.025, respectively). This is possibly resulted from the increased need to discharge patients from the hospitals in the North arm, due to the collapsed situation in hospitals, which was worse in Northern than Southern Brazil.
The sole non-severe TEAE more commonly present among patients from the proxalutamide group was diarrhea, which corresponds to the same sole TEAE reported in the study conducted in the North.
In the North arm, patients were younger, had fewer males and had higher proportion of subjects without comorbidities than in the South arm. This is in accordance with the regional demographic between regions in Brazil [35]. In the case of higher prevalence of male sex in the South arm, approximately 45% of the participants were originally from a military hospital, where the vast majority of patients are males. Conversely, there were approximately 20% more patients in severe state (score 6) the in North arm than in the South arm (p<0.001). While patients in Southern Brazil had a higher risk of dying from COVID-19, patients in Northern Brazil presented more severe COVID-19 states, possibly due to the collapse in the overall health system in the state of Amazonas during the trial, which hampered non-severe COVID-19 patients from being hospitalized, even fulfilling criteria for hospitalization.
The higher percentage of participants that used antibiotics in the North arm compared to the South arm may reflect the higher proportion of patients in the placebo group in this arm (1:1) than in the South arm (4:1), and the higher percentage of severe patients in the North arm compared to the South.
Recovery and mortality rates presented better results in the North than in the South arm: recovery rate 128% versus 67% higher at Day 14 (p=0.009) and 81% versus 51% at Day 28 (non-significantly; p=0.66), and all-cause mortality rate 83% versus 78% lower at Day 14 (marginally significant; p=0.073) and 78% versus 72% at Day 28 (p=0.004). The difference in the size of efficacy, which shows a more prominent response with proxalutamide in the North arm, can be fully explained by the higher proportion of patients in score 6 in the North (approximately two thirds) than in the South arm (approximately half of participants), since when stratified by baseline score, all differences vanish.
As mentioned, compared to Northern Brazil, hospitals in Southern Brazil were slightly less, although overwhelmingly occupied. As a result, criteria for hospitalization were adapted to encompass patients that needed medical assistance most in both regions, but patients at a higher severity were to be hospitalized in the North than in the South, which explains the differences in the proportion of patients in score 6.
Differences in mortality rate in the placebo group between Northern and Southern Brazil were nonsignificant and reflected regional differences observed in Brazil [34,35]. In both studies, the all-cause mortality rate in the placebo groups was smaller or similar to the COVID-19 in-hospital mortality rate in the respective regions [34] -of approximately 50% in the North arm and 35% in the South arm [35].
Overall, results were exceptionally similar between the populations in the North and in the South, despite the slight demographic differences, indicating a high consistency of the findings when validated externally within the study. Of note, in both cases, P.1 (gamma) variant was the cause of COVID-19 in virtually all patients, and standard of care was in general similar between all hospitals.
The fact that comparisons between active groups and between placebo groups in all major outcomes showed similar responses between regions reinforces the efficacy of proxalutamide in terms of recovery and mortality rates.

Overall results
When analyzed for overall participants from the proxalutamide group were marginally older than participants from the placebo group because, as mentioned, in the South, where the randomization ratio was 4:1, patients were older than patients in the North. Conversely, older patients tend to present more comorbidities, and T2DM was marginally more present in the proxalutamide group. This is a conservative bias that may underestimate the efficacy of proxalutamide.
Disease progression occurred approximately 4.1 times more frequently in the placebo group than in the proxalutamide group, while progression to shock requiring vasopressors was 4.4 times more frequent in the placebo group (p<0.0001 for both). Accordingly, the proportion of patients in the placebo group with renal injury and liver damage was 4.5 and 3.2 times higher than patients in the proxalutamide group (p=0.0008 and p=0.003, respectively). The remarkable reduction in disease progression and liver and kidney injuries reinforces the efficacy of proxalutamide to interrupt the COVID-19 disease course during the second and third phases of the disease. The resulted reduction in mortality rate likely reflects not only the improvement of COVID-19 per se, but also the numerous secondary complications caused by the progression of the disease.
Reduction in hospitalization stay was 33.3% (p<0.0001) with proxalutamide, particularly when proxalutamide is initiated earlier in hospitalization, since post-randomization time-to-discharge alive from the hospital was almost 50% lower in the proxalutamide group than in the placebo group (p<0.0001). This is particularly important during outbreaks and collapses of hospitals, when earlier discharge may allow undertreated patients to be hospitalized.
The larger proportion of patients from the placebo arm that used different classes of antibiotics reflect the progression of the disease, with deterioration of the pulmonary status and occurrence of secondary bacterial infections, that occurred more commonly in the placebo group. As reinforced by the lower use of antibiotics, proxalutamide may have prevented the occurrence of secondary bacterial infection, commonly seen in severe COVID-19. Proxalutamide also demonstrated theoretical reduction of the COVID-19-triggered endothelial dysfunction, observed through substantial decrease in D-dimer levels. To which extent antiinflammatory, anti-thrombotic and immunomodulatory effects of proxalutamide were directly mediated or indirectly induced by reduction in viral load is not clarified. The prevailing SARS-CoV-2 variant during the course of the trials was the P.1 (gamma), which, unlike other variants, seemed to present persistent viral activity after seven to 14 days of the disease, as observed through the persistent, unexplained and refractory increase of inflammatory and thrombotic markers during hospitalization [36][37][38][39]. Figure 8 summarizes the critical mechanisms in late-stage COVID-19 that lead to death, and the proposed mechanisms of protection conferred by proxalutamide.

FIGURE 8: Proposed mechanisms for proxalutamide against late-stage COVID-19.
The unexpected spontaneous prolonged erection among males in the proxalutamide group may have occurred due to a short-term rise in testosterone levels, as seen in our previous study in male outpatients [29].

In females as in males: an unexpected balance in the results
While it would be expected that males would present a more substantial improvement in clinical outcomes due to the higher circulating testosterone and dihydrotestosterone (DHT) levels compared to women, reduction in overall mortality rate and increase in recovery rate with proxalutamide were unexpectedly observed both in males and females. Responses not only occurred in a non-gender specific manner but were more marked in women in both recovery rate ratio at Day 14 (4.0 vs 2.5 in men) and at Day 28 (2.2 vs 1.4 in men) and mortality risk ratio at Day 14 (96% reduction vs 66% reduction in men) and at Day 28 (83% reduction vs 67% reduction in men). This finding is consistent with the findings in the North arm, where both males and females responded similarly to proxalutamide, when compared to the placebo arm.
The unexpected response in women may be justified by the fact that although females have lower androgen circulating levels, they tend to be more sensitive to androgens due to potential higher sensitivity of the androgen receptor (AR) and inherently upregulation of the AR, leading to increased density of AR in the cytoplasm. The hyperresponsiveness to androgens in a manner that slightly increased testosterone levels may be sufficient to become an independent risk factor for severe COVID-19 [40]. Consequently, antagonism of AR or blockage of more active androgen hormones may lead to more prominent responses.
Some of the differences between females of the placebo groups can be justified by the small number of participants in the group from the South arm. This peculiarity should be considered in comparisons that encompass female placebo groups.
The importance of the analysis according to the baseline clinical score is to predict subjects that should present better responses. From this analysis, apparently the patients that are most benefited from proxalutamide are those that required oxygen supplement but still not needing higher oxygen flow or additional non-invasive respiratory devices, i.e., in score 5, and in any case when oxygen is needed (scores 5 and 6) compared to cases when oxygen is not needed (scores 3 and 4).

The P.1 (Gamma) SARS-CoV-2 variant and its implications in the efficacy of proxalutamide
The P.1 (Gamma) variant of SARS-CoV-2 is a variant of concern (VOC) that, although not officially recognized, is a strong candidate to be considered the first variant of high consequence (VOHC) [36], since it has been demonstrated to lead to more severe clinical disease, increased hospitalization and increased lethality, with an up to 4-time higher risk of in-hospital death compared to the B.1.617.2/AY lineages (Delta) variant and to present increased transmissibility and reduction in the efficacy and response of the therapies against COVID-19 [37,38]. Since vaccination rates were virtually null by the time of the peak of P.1 variant, the efficacy of the vaccines to prevent infection or attenuate clinical disease for the P.1 variant is unclear.
To our knowledge, proxalutamide was the first drug tested in hospitalized patients due to COVID-19 in the P.1 (Gamma) variant, demonstrated not by the fact that the P.1 variant was present in more than 90% of tested subjects in the regions where the RCTs were conducted during the period of the studies, but essentially because the prevalence of P.1 was 100% of patients that participated in the RCTs and underwent SARS-CoV-2 genotyping presented the P.1 (Gamma) variant [30,39]. Moreover, in the city of Gramado, in Southern Brazil, we discovered the first cases of the P.1.2 subvariant, which could also interfere in the clinical course and response to proxalutamide, in addition to the changes that could have been caused by the P.1 variant, compared to RCTs conducted in prior variants [39].
The P.1 variant of the SARS-CoV-2 is highly dependent on TMPRSS2 for its cell entry, and induces dramatic dysfunctional inflammatory and immunologic responses as well as a highly pro-thrombogenic endothelial dysfunction. While a reduction in the efficacy of proxalutamide could be expected due to the P.1 variant's typical resistance to usual care for COVID-19, since all the three major aspects of the P.1 variant (high dependence on TMPRSS2 for cell entry, dysfunctional exacerbated inflammatory and immunologic reactions, and severe endothelial dysfunction) seem to be specially targeted by an anti-androgen [41,42], the high efficacy of proxalutamide in the P.1 variant could also be justified by these peculiarities of the variant present in the RCTs. However, enzalutamide, a high-potent drug from the same drug class as proxalutamide although with weaker antiandrogen activity, demonstrated to dramatically reduce SARS-CoV-2 cell entry in the lung by downregulating TMPRS22, in a non-P.1 SARS-CoV-2 variant [43].

Scientific aspects of studies on COVID-19
Some noteworthy particularities of the research on COVID-19 must be considered for the advance of the quality of the research on COVID0-19. First, the importance of employing all-cause mortality rate, not "COVID-19 related" mortality rate, as a hard outcome. The importance of this difference relies on the fact that mortality in COVID-19 is largely represented by indirect effects of COVID-19. In addition, whether deaths occurred due to COVID-19 or to a secondary complication is less critical than whether deaths occurred or not.
Second, the crisis of reproducibility in science, largely discussed in the scientific community, is particularly present in COVID-19, since mutations in the SARS-CoV-2 lead to changes in the efficacy of direct or indirect antiviral agents, as well as drugs that target the host cell. Changes in the efficacy of direct antivirals are easy to comprehend since these agents depend on specific portions of the virus to act. Conversely, mutations may also lead to changes in the level of dependency of SARS-CoV-2 on each of the steps for its cell entry, which may affect the ability of drugs that hamper one or more steps of the process of SARS-CoV-2 cell entry to prevent its successful entry. For instance, certain SARS-Cov-2 variants have a stronger dependency on the priming process that is enabled by TMPRSS2, while others tend to have their direct cell entry (not mediated by ACE2 facilitated) [36,41,42].
Due to the meaningful changes in the response to drugs according to the variant, we claim that studies on COVID-19 to test the efficacy of drugs should be considered as variant-dependent, i.e., the observed efficacy of a certain molecule for COVID-19 should be extrapolated for other variants with a cautious and lower degree of certainty, in a similar manner how vaccines for COVID-19 are analyzed.
Third, in COVID-19 studies, self-identified gender, rather than biological sex, should be considered to classify a participant of a research, in case the person is under hormonal therapy to adjust for the selfidentified gender. The adjustment for the self-identified gender when under hormonal treatment is based on the fact that COVID-19 has androgen-and sex-specific risks, as observed in higher risk in males than in females [1], higher risk in women with hyperandrogenism than women without hyperandrogenic states [10,11], higher risk when on androgen therapy [7], lower risk when in anti-androgen therapy [12][13][14][15][16][17], lower in estrogen or progesterone therapy [22,[44][45][46], and, in the case of transgenders, female-to-males (fTm), that are self-identified as males while biologically females, also termed as trans men, have higher risk than maleto-females (mTf), that are self-identified as females while biologically males, also termed as trans women [47]. This means that cis and transmen are clustered together, and in a different risk profile than cis and trans women, showing that self-identified gender, rather than biological sex, should be considered to classify within trials on COVID-19.
Forth, interference of politics in science and research on COVID-19 has negatively influenced the inherent impartiality required to evaluate and judge studies and manuscripts. In this matter, competing interests may exert indirect pressure for the acceptance or rejection of specific types of research, according to the results. Although this is speculative, inequalities in the level of severity of the review process in several scientific journals have become overwhelming. Due to the unclear interests that have increasingly influenced the scientific methodology and publication, we encourage all scientific community to be fully transparent by declaring all direct and indirect conflicts of interests, not only restricted to the current conflicts, but also extended to the previous conflicts and potential future conflicts, that could influence, at any extent, the results and analysis for both positive and negative responses, as well as decisions and evaluations when acting as reviewers or editors. We also claim that conflicts and competing interests to be more emphasized and visible within the manuscripts and publications.

The Proxa AndroCoV and Proxa-Rescue AndroCoV trials and the currently ongoing Phase 3 trials for COVID-19
In common, all RCTs with proxalutamide for COVID-19, including the RCT in outpatient settings with males, females, and in hospitalized patients for both males and females, were conducted as a sort of 'addon' therapy, since the standard of care in both outpatients and hospitalized patients, and in both Northern and Southern Brazil in hospitalized patients, included drugs with efficacy demonstrated to be at least marginally effective, if not effective, including nitazoxanide, ivermectin, antibiotics, enoxaparin, glucocorticoids and other drug classes [48][49][50][51].
COVID-19 results from a complex pathophysiology caused by the SARS-CoV-2. Like other viruses, SARS-CoV-2 has multiple sites of action in order to cause the resulting disease, COVID-19. Consequently, once the pathophysiology is comprehended, it should not be expected that monotherapy would lead to important improvements in clinical outcomes, unless a drug has multiple actions or is used in a narrow window of opportunity, in the case of direct antivirals. And, in the last case, mutations in the virus may likely result in changes in the efficacy of direct antivirals.
Conversely, the combination of drugs that act in different sites may lead to a synergistic effect against the SARS-CoV-2, enhancing the efficacy, when compared to each drug alone. The same occurred to other viruses, including the 'drug cocktail' for the HIV [52,53], hepatitis B [54], and hepatitis C [55]. Indeed, the typical poor responses to viruses observed throughout history may be a consequence of the lack of optimized treatment regimens.
Whether proxalutamide exerted synergistic effects to the existing treatments drugs -i.e., proxalutamide was not only effective, but also enhanced the efficacy of the other drugs administered -or if proxalutamide worked alone is uncertain, since not only proxalutamide, but also no other antiandrogens have been tested against COVID-19 alone to date.
Unlike the AndroCoV trials, the Phase 3 trials currently ongoing (NCT05009732, NCT04870606 and NCT04869228) are testing proxalutamide in different populations without the rationale of the combination of different drugs as being a potential critical aspect for the efficacy of the treatment against COVID-19.
A second major difference between the AndroCoV Trials, specially the Proxa-Rescue AndroCoV Trial, is that proxalutamide may have been more effective for hospitalized COVID-19 patients due to the P.1 (gamma) variant, than what it would be for other variants. The key difference is that in the P.1 (Gamma) variant viral replication may persist during the second and third stages of the disease [36][37][38][39], demonstrated by a common yet unexpected increase in usCRP levels despite the use of high-dose glucocorticoid treatment regimens, not explained by bacterial or fungal infections, and also demonstrated by a persistent increase of D-dimer despite the use of potent anticoagulants, both of which indicating direct viral activity persisting throughout all COVID-19 stages [29,31].
Because of the marked differences in the pathophysiology, pathogenicity, and clinical course between SARS-CoV-2 variants, which is particularly overwhelming in the case of the P.1 (Gamma) variant, for which proxalutamide was tested in the Proxa-Rescue AndroCoV trial, whether proxalutamide will be as effective for COVID-19 caused other variants of the SARS-CoV-2 is unknown. In addition, the currently ongoing trials are not precise replications of the design of the AndroCoV Trials, in the essential aspect of the accompanying drug therapy. Due to the lack of similarity between the AndroCoV and the currently ongoing trials and the differences in the SARS-CoV-2 variants in which proxalutamide is being tested, distinct results in terms of drug efficacy are not unexpected.
Differences between first and second-generation NSAAs, or even between second-generation NSAAs are clinically demonstrated through the fact that failure of treatment of castration-resistant prostate cancer with an NSAA is overcome with another NSAA [25,[59][60][61].
Proxalutamide has unique characteristics when compared to other second-generation NSAAs. It is likely the most potent, between five to ten times more potent than other NSAAs, has the ability to suppress the genetic expression of the AR gene, has important concurrent actions on the regulation of angiotensinconverting enzyme-2 (ACE2) expression, which is the gate and key regulator of SARS-CoV-2 entry in cells, and exerts important anti-inflammatory effects with dramatic blockage of tumor necrosis factor-alpha (TNF-alpha) and interleukin 1-beta (IL-1beta) [25,59,60].
Possibly, a beneficial proportion between the different levels of potency of proxalutamide when acting as an anti-androgen, ACE-2 regulator, and as an anti-inflammatory may justify its efficacy for COVID-19 throughout different stages of the disease. For instance, a strong anti-androgen without concurrent regulation of the inflammatory response and ACE-2 expression may cause more harm than good at specific stages of COVID-19. In addition, the minimum concentration to be effective as a regulator of the inflammatory response and ACE-2 modulator may be distinct from those used for castration-resistant prostate cancer. Finally, because of the specificities of each NSAA, we consider responses to a certain NSAA for COVID-19 to be considered as a partial drug-class effect. We encourage that conclusions regarding antiandrogens for COVID-19 should not be extrapolated for other NSAAs, in particular when results are negative, due to the numerous factors that may influence the efficacy of NSAA for COVID-19. Finally, researchers should avoid recommending against studies of an entire drug class in the case of antiandrogens, based on the lack of positive results of a single molecule, due to the multiple specificities and peculiarities of each anti-androgen, as described above.
Another critical aspect that must be emphasized when testing anti-androgens for COVID-19 is the treatment duration. In the case of proxalutamide, the risk of COVID-19 relapse and worsening with early interruption of proxalutamide was first identified when tested for early COVID-19 in outpatient settings [27][28][29]. Initially, the protocol was a three-day therapy of 200mg proxalutamide per day. However, two weeks after the beginning of the RCT on outpatients, the unblinded principal investigator noticed, exclusively in participants of the proxalutamide group, a remarkably faster recovery process in the first three days followed by sudden relapse and progression of the disease after discontinuing proxalutamide. This led us to increase the treatment duration from three to seven days, since its safety profile was well-established for chronic use, which was approved by the National Ethics Committee. After changing to seven days of therapy, relapses were no longer reported.
Conversely, in the Proxa-Rescue AndroCoV trial, we detected a very high mortality rate (approximately 80% mortality rate) among participants that did not comply with the 14-day treatment, significantly higher than the placebo arm [32], while treatment completers had a mortality rate lower than 15%. Oppositely, compliance in the placebo group did not determine the survival rate, of approximately 50%, whether participants complied or not. Figure 9 demonstrates the differences between proxalutamide and placebo groups depending on the compliance to treatment. The sudden loss of antiandrogen and anti-inflammatory protection may explain the dramatic relapse and progression of COVID-19. It is critical that the treatment duration of anti-androgens for COVID-19 should be strictly followed, and we highly recommend against short-duration treatments for COVID-19, of less than seven days for early, non-hospitalized COVID-19, and less than 14 days for hospitalized COVID-19.

Limitations
The Proxa-Rescue AndroCoV RCT was conducted in moderate-to-severe COVID-19 patients infected by the P.1 (Gamma) variant, when hospitals were collapsed in both Northern and Southern Brazil during the period of the study. Despite the demonstrated efficacy that was externally validated within the same study, it is unclear whether this can be reproduced under normal circumstances and in less pathogenic SARS-CoV-2 variants. The small number of subjects in the placebo group of the South arm may also have limited statistical analyses. For these reasons, we consider it indispensable that further RCTs in distinct environments should be conducted to confirm the present findings, with expected differences in terms of the level of efficacy.