Okay so...I'm actually in a rare position where I might be able to contribute to discussion on this! Brief caveats: am undergrad pre-med, but also took an infectious diseases grad class at Penn that spent a week diving into one of the papers cited in this article (when the authors mentioned the two compounds "showed no statistically significant difference in potency"). I think it's worth exploring what the authors of that paper actually meant to assess for. (also @HN please shoot me down if anything I say is inaccurate, also I do think the authors make other valid points, just not so much in reference to this particular aspect)
The paper in question is here (https://mbio.asm.org/content/9/2/e00221-18) and they're looking at table 1. First of all, the paper didn't set out to actually determine whether GS-441524 is better than GS-5734 (aka remdesivir). It mainly set out to show that remdesivir could work against coronaviruses in general (this was written back in 2018, so they were focusing on SARS and MERS only). Furthermore, all of these experiments were in vitro with a model betaCoV, murine hepatitis virus, so results shouldn't necessarily mean much for human populations. But down to the nitty-gritty:
In figure 1 of the paper, the authors do in vitro experiments to show that both GS-441524 and remdesivir reduce their model CoV viral titer. However, they show a major difference. Remdesivir has an EC50 (dose at which 50% of viral replication is inhibited) that is much lower than GS-441524. This is great because you don't need as much remdesivir to inhibit viral replication. One area where GS-441524 did beat remdesivir was cytotoxicity (CC50 = dose at which 50% of cells w/the drug is killed). You want as high a CC50 as possible. So that would mean that remdesivir is technically more toxic to cells; however, it's effective dose is so low that you wouldn't need to give a dose large enough to kill off any cells. You can use a handy metric known as the selective index (CC50/EC50) to measure the tradeoff. High selective index = high efficacy without killing off your own cells. In this case, remdesivir has a higher selective index compared to GS-441524.
Now onto what the article referred to. The authors made their in vitro studies slightly more relevant to humans by using human epithelial cells and this time actual SARS CoV and MERS CoV virus and measured EC50 and CC50 again for both GS-441524 and remdesivir. The authors report some +/- standard deviation ranges. From intro stat, I think one of the conclusions about looking at overlapping standard deviation bars is that no conclusion can be made about whether or not the differences are statistically significant or not! I encourage people to take a look at the table itself and make your own conclusions (https://mbio.asm.org/content/9/2/e00221-18#T1)
The rest of the paper focused on remdesivir and showed a number of things. It showed it was most effective early post infection (in vitro) and it highly inhibited viral replication when viral exonuclease was knocked out (indicating that remdesivir messes with the function of that). It also showed (via serial passage experiments...basically think about it as mini-artificial selection) that nucleoside analog drugs will tend to favor the selection for two RNA-dependent RNA polymerase mutant viruses that confer resistance to the nucleoside drugs! But they showed in the next figure that these mutant viruses have a reproductive disadvantage compared to their wild type CoV.
Now for my 2 cents: I think that remdesivir was favored primarily because it was being used for anti-viral treatment in Ebola virus trials...but it was falling out of favor for Ebola treatment because of other monoclonal antibody cocktail treatments that show better survival rates (ref PALM trial: https://www.nejm.org/doi/full/10.1056/NEJMoa1910993). Sooo I guess gilead had this drug lying around, and they had a good idea as to its mechanism of action...so why not apply it for Sars CoV2?
The paper in question is here (https://mbio.asm.org/content/9/2/e00221-18) and they're looking at table 1. First of all, the paper didn't set out to actually determine whether GS-441524 is better than GS-5734 (aka remdesivir). It mainly set out to show that remdesivir could work against coronaviruses in general (this was written back in 2018, so they were focusing on SARS and MERS only). Furthermore, all of these experiments were in vitro with a model betaCoV, murine hepatitis virus, so results shouldn't necessarily mean much for human populations. But down to the nitty-gritty:
In figure 1 of the paper, the authors do in vitro experiments to show that both GS-441524 and remdesivir reduce their model CoV viral titer. However, they show a major difference. Remdesivir has an EC50 (dose at which 50% of viral replication is inhibited) that is much lower than GS-441524. This is great because you don't need as much remdesivir to inhibit viral replication. One area where GS-441524 did beat remdesivir was cytotoxicity (CC50 = dose at which 50% of cells w/the drug is killed). You want as high a CC50 as possible. So that would mean that remdesivir is technically more toxic to cells; however, it's effective dose is so low that you wouldn't need to give a dose large enough to kill off any cells. You can use a handy metric known as the selective index (CC50/EC50) to measure the tradeoff. High selective index = high efficacy without killing off your own cells. In this case, remdesivir has a higher selective index compared to GS-441524.
Now onto what the article referred to. The authors made their in vitro studies slightly more relevant to humans by using human epithelial cells and this time actual SARS CoV and MERS CoV virus and measured EC50 and CC50 again for both GS-441524 and remdesivir. The authors report some +/- standard deviation ranges. From intro stat, I think one of the conclusions about looking at overlapping standard deviation bars is that no conclusion can be made about whether or not the differences are statistically significant or not! I encourage people to take a look at the table itself and make your own conclusions (https://mbio.asm.org/content/9/2/e00221-18#T1)
The rest of the paper focused on remdesivir and showed a number of things. It showed it was most effective early post infection (in vitro) and it highly inhibited viral replication when viral exonuclease was knocked out (indicating that remdesivir messes with the function of that). It also showed (via serial passage experiments...basically think about it as mini-artificial selection) that nucleoside analog drugs will tend to favor the selection for two RNA-dependent RNA polymerase mutant viruses that confer resistance to the nucleoside drugs! But they showed in the next figure that these mutant viruses have a reproductive disadvantage compared to their wild type CoV.
Now for my 2 cents: I think that remdesivir was favored primarily because it was being used for anti-viral treatment in Ebola virus trials...but it was falling out of favor for Ebola treatment because of other monoclonal antibody cocktail treatments that show better survival rates (ref PALM trial: https://www.nejm.org/doi/full/10.1056/NEJMoa1910993). Sooo I guess gilead had this drug lying around, and they had a good idea as to its mechanism of action...so why not apply it for Sars CoV2?