ACTIV-6 And TOGETHER Bear Strangely Similar Design Fingerprints

A while ago, I wrote an article enumerating ten questionable features of the TOGETHER trial on ivermectin, conducted in Brazil. What are the chances those same features would make an appearance in the ACTIV-6 trial on ivermectin, conducted by the NIH in the USA? Pretty good, apparently, because that is exactly what seems to be happening.

#1 - Randomization Anomalies

While the shape of the issue is different, the result is the same: the randomization of the trial is under question due to the fact that the treatment and control groups of the trial were drawn from different populations. As the ACTIV-6 trial participant I spoke to revealed, he was asked to pick which drug study he would be enrolled in. In contrast, the placebo group contained patients assigned to other drugs, most or all of whom chose to be randomized into that different drug. This is a systematic difference between the populations not accounted for in the trial.

Alexandros Marinos @alexandrosM

ACTIV-6 ivermectin paper out... looking at the randomization section... WHAT?! Participants could opt out of a drug if they didn't FEEL it works. Or the site investigator could opt them out. "Here's our site menu, what drug would you like to randomize for?" This is not an RCT.

#2 - Dosing

The ACTIV-6 trial takes the flawed dosing of the TOGETHER trial and makes it worse. This is one of those cases where a picture is worth 1000 words:

Not only does ACTIV-6 underdose those at highest risk the most, it also underdoses people with lower weights, effectively falsifying the dose given to patients.

#3 - Exclusion for Prior Use of Ivermectin

ACTIV-6, like TOGETHER, did not exclude patients who had used ivermectin from the trial. Instead, the ACTIV-6 protocol allowed them to randomize into other drugs. As a result, the patients allocated to fluvoxamine and fluoxetine treatment may also have been taking ivermectin.

#4 - Variable Placebo

The subdivision of the placebo group into smaller subgroups corresponding to each treatment arm (3-day pills, 10-day pills, inhaler...) was very similar to TOGETHER. Placebo patients from all arms were commingled for each study:

Similarly, the authors cite no theoretical justification for using this scheme. Why was the variable placebo approach introduced? What was done to avoid the potential unblinding of the investigators and other clinical staff that the TOGETHER trial suffered from?

Read more about this issue here

#5 - Novel Endpoint

The trial used a composite secondary endpoint intermingling ER visits, hospitalizations, and deaths:

This is despite the fact that the use of ER visits was heavily criticized by regulators for the TOGETHER trial. In fact, TOGETHER at least had a 6-hour minimum to count an ER visit as an “event,” whereas this trial did not set any limit, making the endpoint even less clinically relevant. Also, similarly to TOGETHER, this trial does not report COVID-19 hospitalizations separately.

#6 - Missing Data

While imputation was not a particular culprit in the case of ACTIV-6, other potentially concerning information was made to disappear in this case. In particular, the subgroup analysis blatantly omits to report patient populations for most subgroups:

#7 - How Was the Dose Decided?

While this trial didn’t suffer from unexplained dose changes like TOGETHER did, the determination of the dose administered in this trial is shrouded in mystery, with a study author attributing the decision implausibly to the FLCCC. It should be noted that this author is also an author of the TOGETHER trial.

#8 - Placebo Adherence Unknown

At the very least, TOGETHER gives us some semblance of information about adherence, even if it is obviously incomplete. In the case of ACTIV-6, no information about adherence or per-protocol analysis is given. We are only shown the modified Intention-To-Treat analysis, despite the fact that this was not the pre-registered analysis in the protocol. We are given no information as to the number of those patients that actually adhered to the protocol and took their drugs as described.

#9 - Independent Monitoring Board Not Independent

The Data and Safety Monitoring Committee (DSMC) is a key decision-making body in trials like this. The various publications clearly state that it is independent. Yet, according to ivmmeta:

The IDMC vice chair was reportedly on the NIH panel that did not recommend treatment despite strong evidence, and provided no quantitative analysis, no reference to the majority of the research, and no updates for new research for a very long period

This echoes the DSMC independence issues found in the TOGETHER trial. In addition, the organization that runs the ACTIV-6 trial is a public-private partnership that contains numerous conflicted pharmaceutical companies in its management committee and membership.

#10 - Failure to Share Data

The data sharing statement published with the study paper reads as follows:

To this day, I am not aware of the data having been made available to any researcher, for any reason. The number of hoops written into the data sharing statement make it unlikely we will ever see this trial audited, putting it in the same category as TOGETHER and COVID-OUT. If there’s anything new on this front, I will update this section. I hope the data finds its way out of the trial team’s basement, but I would not bet on it.

Bonus Issue: Inclusion Criteria

The ACTIV-6 trial’s inclusion criteria were quite straightforward:

The TOGETHER trial, on the other hand, was supposed to be centered on high-risk patients. Focusing on such patients is very common in clinical trials for COVID-19, as doing the opposite makes little sense: patients that will do well no matter the intervention are not going to help distinguish one treatment from another. Focusing on high-risk patients means an evidence pool where an effective treatment is more likely to make a difference.

This is where things get interesting: while the TOGETHER trial started out recruiting high-risk patients, this policy changed mid-trial. On the TOGETHER trial website, there is a protocol dated May 21, 2021 (and approved by Brazilian authorities on May 25, 2021—fast!), that adds the following enhancement factors:

• j. Patient with fever thermometry at screening > 38º C.

• k. Patients with at least one of the following symptoms: Cough, Dyspnea, ventilator-dependent chest pain or myalgias with limitation of daily activities (Criterion limited to 25% of randomizations).

An enhancement factor is a factor which qualifies patients as high-risk, the most stringent of the inclusion criteria. As a patient, having an enhancement factor means you’ve crossed the biggest hurdle to being included in the trial. So it’s a bit surprising that a fever or a cough would be listed in those factors.

If we look for the justification, in the version history of the master protocol, they write that…

If you have a cough, you are at high risk for complications due to COVID-19?

Does anyone believe this?

These are literally the main symptoms of COVID-19. In fact, the same protocol states what are considered to be the most common symptoms: fever, cough, dyspnea.

Currently, after the epidemiological knowledge of tens of thousands of cases of COVID-19, the following signs/symptoms are considered to be the most common: Fever (87.9%), Dry cough (67.7%), Dyspnea (40%) [13].

If they did, however, believe these symptoms qualified someone as being at high risk for complications, then what are we to make of this part of the criterion?

(Criterion limited to 25% of randomizations)

Why would the trial seek to put a threshold on the number of patients that would qualify under this specific enhancement factor? If, indeed, these patients “are at high risk for complications,” then why limit the number that can participate in the trial? I can’t think of any reason whatsoever, nor have I ever seen a trial place such a specific limit on an inclusion criterion.

In fact, given that Cytel—the company that “designed and led the TOGETHER trial”—bragged about its Clinical Trial optimization simulation software, it would be good to feel like someone wasn’t running interim data through simulations to figure out which kind of inclusion/exclusion criteria would give them the results they wanted.

To make matters worse, when the authors declared the change to clinicaltrials.gov, they wrote it this way:

7. Fever > 38 C at baseline

[…]

14. Patients with important limitation of daily activities due to: Dyspnea, chest pain myalgia (limited to 25% of all randomizations)

Here's the change:

They left out “cough” as a symptom from the enhancement factor. The one “new” criterion—that is the most outrageous—somehow got missed when they entered their new protocol in the non-editable government database.

Perhaps it was an honest mistake. Or perhaps they didn’t want to attract unwanted attention to how trivially easy it was to qualify for the trial in late May and beyond—or whenever the investigators started to apply this protocol. (We’re not even sure when that was.)

Were Later Patients Healthier?

From January to early April, placebo patients were in pretty bad shape (39% hypertension?!), whereas the placebo group from April to August seem like spring chickens.

The way I produced this chart was to take the metformin placebo group and subtract it from the fluvoxamine placebo group. Since both those studies started at the same time—and fluvoxamine continued until Aug 5th—this means that all the placebo patients used for metformin must also have been used for fluvoxamine. As a result, we get to see what the placebo group for fluvoxamine (and ivermectin!) looks like from April 3rd to August 5/6th.

It looks like almost all comorbidities crashed from the first to the second period.

And yet, the early placebo group has a 13.8% event rate, whereas the late placebo group has a 16.5% event rate. This must be—to some degree—related to the Gamma variant, but given that the patients are so much healthier than they were in the first period, it is a little bit confusing—to say the least.

Long story short, after the second interim analysis, the TOGETHER trial drastically changed its inclusion criteria away from high-risk patients and towards low-risk patients. That change was made in May 2021. When was the ACTIV-6 trial’s protocol first posted on clinicaltrials.gov? You guessed it, May of 2021.

Conclusion

The number of “design defects” and flawed decisions present in the ACTIV-6 trial, as well as the eery similarity of these flaws with the flaws of the TOGETHER trial raise questions about whether the design of ACTIV-6 was informed by interim results of the TOGETHER trial. Given that Dr. David Boulware is a shared author of both trials, it’s not hard to imagine that there was a line of communication between the two teams. However, any early release of TOGETHER results—that were not available to the public—would be scandalous, besides the fact that the flaws of TOGETHER appear to be amplified, not corrected, in the case of ACTIV-6.

Having seen a number of these trials now, I’m starting to sense a certain… design aesthetic… that I sincerely hope is accidental. However, the number of coincidences is stretching credulity. As always, transparency of the data would go a long way to dispel any suspicions.