An in silico study supports the efficacy of azvudine54.

2 in vitro studies support the efficacy of azvudine54,55.

An in vivo animal study supports the efficacy of azvudine54.

Preclinical research is an important part of the development of treatments, however results may be very different in clinical trials. Preclinical results are not used in this paper.

Table 1 summarizes the results for all stages combined, for Randomized Controlled Trials, for peer-reviewed studies, after exclusions, and for specific outcomes. Table 2 shows results by treatment stage. Fig. 4 plots individual results by treatment stage. Fig. 5, 6, 7, 8, 9, 10, 11, 12, 13, and 14 show forest plots for random-effects meta-analysis of all studies with pooled effects, mortality results, ventilation, ICU admission, hospitalization, progression, recovery, viral clearance, peer reviewed studies, and long COVID.

Fig. 15 shows a comparison of results for RCTs and observational studies. Random-effects meta-analysis of RCTs shows 37% improvement, compared to 26% for other studies. Fig. 16 shows a forest plot for random-effects meta-analysis of all Randomized Controlled Trials. RCT results are included in Table 1 and Table 2.

To avoid bias in the selection of studies, we analyze all non-retracted studies. Here we show the results after excluding studies with major issues likely to alter results, non-standard studies, and studies where very minimal detail is currently available. Our bias evaluation is based on analysis of each study and identifying when there is a significant chance that limitations will substantially change the outcome of the study. We believe this can be more valuable than checklist-based approaches such as Cochrane GRADE, which can be easily influenced by potential bias, may ignore or underemphasize serious issues not captured in the checklists, and may overemphasize issues unlikely to alter outcomes in specific cases (for example certain specifics of randomization with a very large effect size and well-matched baseline characteristics).

The studies excluded are as below. Fig. 18 shows a forest plot for random-effects meta-analysis of all studies after exclusions.

Low-cost treatments were subject to bias and censorship during the pandemic. Scientific bias is seen in the design, analysis, presentation, and selective reporting of studies, which often favored negative results. A similar bias is seen in the media coverage for low-cost treatments. While broadly seen, bias was particularly notable for ivermectin and hydroxychloroquine, e.g., Scott Alexander noted that "if you say anything in favor of ivermectin you will be cast out of civilization and thrown into the circle of social hell reserved for Klan members and 1/6 insurrectionists. All the health officials in the world will shout 'horse dewormer!' at you and compare you to Josef Mengele."47.

We analyze media coverage for the 227 treatments we cover using Altmetric74, which reports the number of ~12,000 tracked news outlets that covered each study75. Studies are considered to have received significant media coverage if they were covered by at least 0.5% of the tracked news outlets. Fig. 19, 20, and 21 show the bias toward negative results for low-cost treatments, in contrast to the opposite bias for high-profit treatments. Fig. 22 shows the bias toward coverage of negative results for azvudine. This may result in widespread incorrect perceptions on the relative efficacy of high-profit and low-cost treatments. The impact is significant—increased cost limits the use of high-profit treatments and treatment equity, and high-profit treatments were also more difficult to access, especially for earlier treatment which improves efficacy and minimizes community transmission.

A combination of factors may have led to the media's suppression of low-cost treatments:

25 low-cost treatments were approved in one or more countries, yet many countries approved no low-cost treatments. The countries that did adopt low-cost treatments analyzed the evidence early and made timely approvals. With few exceptions, authorities did not change their initial views, regardless of how much evidence accumulated showing either efficacy or harm. Why?

The harms of smoking here hidden for 25 yearsC. Authorities did not analyze the data in real-time, failing to act when harm was known. Widespread acknowledgement of harm came only after attempts by two new surgeon generals, along with pressure from health advocates and a new president, and a review of 7,000 studies.

Similarly for COVID-19, most authorities and experts did not proactively analyze data in real-time. This guarantees delayed recognition of efficacy or harm, by which time moral, legal, career, and reputational liabilities strongly disincentivize any admission of error. Claims of no efficacy (for effective treatments) or safety (for harmful treatments) were often made prior to strong data being available. Correction would require admitting to errors that increased mortality, which is unlikely with the same generation of officials.

Analysis of potential treatments was rarely done, and when done these were typically minimal efforts. For example, NIH reviews were highly delayed, cover only a tiny fraction of treatments, reference only 2% of studies for the treatments covered, and include no quantitative analysis. They appear as rarely updated side projects from external panels implicitly tasked with justifying prior failures. As with smoking, the thousands of studies could (and should) have been analyzed and acted on in real-time.

A key structural improvement, applicable to all current and future diseases, is for authorities to implement real-time proactive analysis of clinical evidence. This does not remove all bias, but does make it possible to act on evidence, whereas delayed action may be unlikely due to moral, legal, career, and reputational liabilities.

Heterogeneity in COVID-19 studies arises from many factors including:

Azvudine is an effective treatment for COVID-19. Significantly lower risk is seen for mortality, ICU admission, hospitalization, progression, and viral clearance. 32 studies from 22 independent teams in 2 countries show significant benefit. Meta-analysis using the most serious outcome reported shows 27% [19‑33%] lower risk. Results are similar for Randomized Controlled Trials, higher quality studies, and peer-reviewed studies.

Results are very robust—in worst case exclusion sensitivity analysis 22 of 42 studies must be excluded before statistical significance is lost.

Azvudine is a nucleoside analog with potential host cell and organ toxicity. Studies to date highlight liver, reproductive, and hematologic toxicity - there is not the same volume of evidence for harm as there is with remdesivir, but there is also less data overall. Therefore, results may be less favorable with longer follow-up, and long-term morbidity could be an issue. COVID-19 studies show significantly increased risk of liver injury1,2.

5 other meta-analyses show significant improvements with azvudine for mortality3-7, mechanical ventilation3, clinical improvement3, progression7, and viral clearance3,5,6.

Retrospective 207 COVID-19 patients in China, showing azvudine associated with faster viral clearance, with azvudine-treated patients obtaining a negative PCR test result 1.7 days faster on average compared to supportive care alone after adjusting for age and sex. Submit Corrections or Updates.

PSM retrospective 4,868 hospitalized COVID-19 patients with hypertension showing reduced mortality with azvudine.

For composite disease progression, the azvudine group had more raw events (330) than the control group (320) out of identical starting populations (2434 per group). Yet, the study reports a significantly protective Hazard Ratio (0.84). The data suggests a large differential in person-days driven by informative censoring (e.g., early discharge of control patients), biasing the Cox model. The person-days calculated for the composite progression endpoint drastically differ between groups (approx 28,100 for azvudine vs 22,700 for control) despite starting with identical sample sizes and following a 30-day max horizon. This suggests control patients were systematically censored earlier, creating a survival bias in the time-to-event analysis. Authors also acknowledge that the composite disease progression finding was not robust across sensitivity analyses.

The mortality result should have higher reliability, with consistent results across all five sensitivity analyses - probabilistic PSM, mean imputation, excluding first-day discharges, doubly robust estimation, and IPTW. However, differential censoring likely affects the results here as well. The raw counts in the PSM groups shows a more modest ~15% benefit.

In Table 3, the reported percentages do not match the counts. For example, serum phosphorus in the control group reports 74 cases out of 2223 (which is 3.3%) but shows 23%. Table 3 reports 70 Grade >= 3 events for hypertriglyceridemia in the control group and 100 in the azvudine group, however these are much larger than the number of events for all grades. Table 3 lists available data for hypertriglyceridemia as control = 2465 and azvudine = 2454. These numbers exceed the post-matching cohort size of 2434 patients per group.

The text states the mortality rate in the control group was 8.79% (n = 214), but later claims there were 216 cases. Submit Corrections or Updates.