Remember the Ivermectin paper that was reported in the Wall Street Journal recently? It came up in this post It looks like all the usual ploys were exploited:

• Give the control group a head start. It helps to begin collecting data on the control group weeks earlier when there is a less deadly variant around. That way more of the control group “do well” which makes the test group less likely to outdo them.
• Give people ivermectin on an empty stomach so it is not absorbed well. It works against worms, but lowers the effective antiviral dose to about one fifth of the current clinical treatment protocol.
• Limit the trial to three days, just in case all the other handicaps weren’t enough.
• Start the treatment late — maybe as late as 8 days after symptoms start. It’s not clear exactly how late the treatments ended up being on average. We just have to wait for the paper, eh?
• Use a randomized control trial in a community that already knows how useful ivermectin is. That way the really sick people will rule themselves out of the trial, because they won’t want the useless placebo. By comparing groups that are already low risk, ivermectin is less likely to shine. RCT trials work best in communities where the drug is not normally known or used.
• Ivermectin was so widely known in Minas Gerais, Brazil that some of the control group may have been using ivermectin before they started the trial. No one asked. Nevermind if they were the lucky ones who actually got the right dose at the right time, we can call them “controls”.
• Use vague primary outcomes like “events” that are subject to bias and exaggeration.
• Rule out sick people (if they haven’t already ruled themselves out first) by only including patients with an “expected stay of less than 5 days. “
• Tweak the protocol in the middle of the experiment: “we hypothesize that younger patients will benefit more than older patients.” Perhaps the results weren’t going well enough?

Source: Jo Nova Website