Latest MMWR analysis of bivalent booster is irredeemably flawed
Dangerous times for science.
There is only one right way to know who benefits from a bivalent booster, and that is a randomized trial. Take people over the age of 65, who have already gotten 3 doses of the parent vaccine, and randomize them in 3 arms to a 4th dose of Wuhan vaccine, a bivalent booster, and placebo vax, and measure severe disease and hospitalization.
Pfizer and Moderna can afford this study. It can be completed rapidly. The US FDA has a societal obligation to demand it, and yet that did not happen. This raises the question if regulators work for the public or instead plan their lucrative future consulting careers for Pfizer and wish to give them an easy market share. Remember that Scott Gottlieb former FDA commish, is now on their board of directors.
Over the last year White House officials continue to work closely with Pfizer to push bivalent boosters through based on mouse data. This has no precedent in modern regulatory history and constitutes a multi-billion dollar give-away to the company. Now the CDC seeks to perform a study to justify that action. Enter the latest MMWR study
This post has to be quick, so I will assume you understand the basics of test negative case control design. The authors make a number of analytic choices that are absolutely unjustified in their study
They assess the VE of bivalent booster vs unvaccinated people or those who received >=2 doses of mRNA. Yet, this is a farce. Bivalent boosters are largely given to people who have already gotten 3 or 4 doses. The question is: if I have gotten all the recommended doses, do I derive additional benefit from the bivalent booster? It makes no sense to compare them to unvaccinated people nor people who got 2 doses only. The analysis does not provide these results separately for those whose only difference is the bivalent booster.
The authors exclude people who got any J&J. This is unjustifiable. If I got 1 J&J and then a Moderna booster, the CDC advices I get a bivalent dose. Why exclude this? I want to know the VE in these people too.
Bivalent boosters received <7 days from symptom onset are excluded. This is guarantee time. If I get a bivalent booster and get sick tomorrow, that should count against getting a bivalent booster. Yet it won’t. The CDC’s method is baking in immortal time. There are other more complicated time issues that I am still thinking through.
“Control patients whose influenza test results were positive were excluded from the analysis because of potential correlation between COVID-19 and influenza vaccination behaviors”. This is absolutely horrific. There is no justification on planet earth for this, and it strongly makes one suspect that the original analysis, which included these people, was null or unimpressive. These patients SHOULD NOT BE EXCLUDED. This violates the very principle of test negative design.
Not everyone will appreciate that these analytic choices are deeply problematic. They are unjustified. The CDC needs to preregister their analyses because there is a deep concern that the entire purpose to return the answer that the White House overlords have asked for.
In my opinion, the CDC is no longer a scientific agency, they are a arm of political propaganda and this study fits the bill. If you run an RCT, there is a high chance the actual VE against hospitalization is 0%.