16 Jan 2023 Episode 307 Controversy on mRNA boosters

Mon, 01/16/2023 - 12:45

Episode 307: Controversy on mRNA vaccine booster effectiveness in the omicron BA.5/BA.2 subvariant era.

Dear colleagues,

It is time to assess to what extent the latest boosters are contributing to a better control of the evolving omicron epidemic.  As you will see, the recent findings are confusing: with “expected” results in par 2 and some “unexpected” ones in par 3. Especially for the interpretation of par 3, Geert Molenberghs has been deeply involved.

To understand this complex situation, it is important to quickly remind what we know at present (Par 1).

Par 1 Introduction

    1. The recent virus evolution: (Ep 304)


In the US, the BA.5 is being replaced by the BQ.1 (BA.5 subvariant) and XBB.1.5 (a recombinant in the BA.2 lineage)


In Belgium, BQ and BA.2.75 are on the rise



    1. The problem of imprinting: = repeated vaccination promotes antibody (Ab) response to wild-type (WT) epitopes, while responses to new variants are suppressed.  (Adam K. Wheatley Trends in Immunology, November 2021, Vol. 42, No. 11)


Human antibodies against SARS-CoV-2 epitopes that are conserved across ancestral [wild type (WT)] and variants of concern (VOC) Spike proteins (blue) are elicited by initial WT vaccination and are likely to be boosted in response to subsequent VOC-targeted vaccines.

The preferential recall of conserved immune responses imprinted to the WT strain may limit the generation of de novo responses against VOC receptor binding domain (RBD) epitopes (red and yellow) in response to booster vaccine doses, even if they are adapted to the new VOC


    1. New “bivalent” boosters

→ Pfizer and Moderna developed Spike BA.1 and BA.5 mRNA and combined it with WT mRNA = “bivalent vaccines” to be used as the “Autumn 2022”boost: a time when BA.1 had almost disappeared and BA.5 was still dominant, but with emerging more resistant BA.5 subvariants as well as BA.2 recombinants (e.g. XBB).

  1. Europe and UK decided to take on the “BA.1 bivalent vaccine”, because there was more robust clinical data on this vaccine
  2. US and Israel decided to go for BA.5, although only few animal and human data were available.

However, in both cases, the WT was included, which seems not logical, because it could worsen “imprinting”.


    1. First real world” results: rather weak induction of neutralizing Ab by bivalent against recent subvariants (Ep 301)

Ep 301-1:  Hitoshi Kawasuji medRxiv 18 Nov: Induction of neutralizing Ab by the wild-type/Omicron BA.1 bivalent vaccine as the second booster dose against Omicron BA.2 and BA.5



Pseudotype virus-based neutralizing activity against Omicron BA.2 and BA.5 in the WT (n = 168) and the WT+BA.1 vaccinated group (n = 23). The numbers at the top indicate the median inhibition by diluted serum in each group.  Hence: results are not expressed as neutralizing titer, but as % inhibition. 

Therefore, bivalent (WT + BA.1) vaccine induces more neutralization than monovalent (WT) against both BA.2 and BA.5, but, overall both mono- and bivalent vaccines induce less neutralization against BA.5 < BA.2 << WT virus  


Ep 301-2: : Davis Gartner bioRxiv 1 Nov 2022: Bivalent BA.5 mRNA booster enhances neutralization against new omicron subvariants BA.2.75 and BQ.1.1



Ep 301-3: Chaitanya Kurhade Nat Med 6 Dec: Induction of neut Ab by BA.5 bivalent Pfizer or Moderna mRNA booster without or with previous infection.


BA.5 Bivalent vaccine increased the titer against subvariant more than monovalant, but it is still 10 X lower than against WT virus.

Prior infection has an additional effect = hybrid immunity.


    1. Background on Ab and T cell responses as potential “correlates of protection”

Ep 307-1: Peter Gilbert NEJM 15 Dec 2022

Panel A shows the vaccine efficacy observed among participants in five randomized, controlled trials of Covid-19 vaccines  who tested negative for SARS-CoV-2 at baseline, according to postvaccination neutralizing antibody titers of  participants.

Vaccine efficacy was defined as the percentage reduction in the average risk of Covid-19 among vaccine recipients, as compared with the risk among placebo recipients, and was estimated with the use of a marginalized Cox proportional-hazards model.

The data are from trials of four vaccines: COVE for mRNA-1273, ENSEMBLE for Ad26.COV2.S (U.S. study sites only), PREVENT-19 for NVX-CoV2373, AZD1222 for ChAdOx1nCoV-19, and COV002 also for ChAdOx1 nCoV-19.




Clearly, neutralizing antibody titers as correlate of protection (CoP) can only be used at population level, not to assess an individual risk on infection, because susceptibility to symptomatic infection is also dependent on other individual variables (age, immune genetics, co-morbidities, load of exposure etc).


The great advantage is that this CoP could replace large scale clinical trials in order to:

  • Compare lot to lot variability and expiration time of a given vaccine
  • “Immunobridging”: comparing activity of a new vaccine types with more established ones
  • Evaluate immune response when other vaccines (e.g. Flu) are co-administered with  COVID vaccines.


Limitation: This CoP was identified in trials involving people

  •  who had not previously been infected with SARS-CoV-2
  • received intramuscular, spike only vaccines
  • and were then exposed to pre-delta viruses.


Some open questions are if  this CoP will remain valid

  • for new viral variants (i.e. Omicron subvariants),
  • for new populations including previously infected people,
  • for new vaccine classes,  
  • and for various aspects of Covid-19 disease (e.g., symptom types, durations, and severities).


Another confounding variable (not mentioned by Gilbert) is the fact that neutralization essays are not really standardized:  various set-ups are used (e.g. live virus,  Spike recombinant pseudovirus, binding inhibition to ACE-2…) and various isolates: although they may broadly belong to the same variant, the different isolates may have some distinct mutations, e.g. additional mutations in Spike may result in different susceptibility to the same neutralizing Ab, whereas, in live virus, different mutations in non-Spike (e.g. polymerase or nucleoprotein) may lead to different replicative capacity, which indirectly also influences susceptibility to neutralization.


Therefore, extrapolating the correlations found between protection and neutralizing Ab in early clinical trials to present day would need validation, but since there is continuous evolution both in the virus and in the host, it is really a “moving target” situation.      


In addition, from earlier results (Khoury Nat Med vol 27 p. 205–12112021 https://www.nature.com/articles/s41591-021-01377-8 )  it is evident that vaccine efficacy against severe disease is always more pronounced than against any  infection.




It is generally accepted that protection against severe infection is a more important endpoint from an individual as well a public health point of view, but that it is rather determined by T cell immunity.  See Barouch NEJM 15 Sept 2022 https://www.nejm.org/doi/10.1056/NEJMra2206573?url_ver=Z39.88-2003&rfr_id=ori:rid:crossref.org&rfr_dat=cr_pub%20%200pubmed



Although there are many assays to evaluate T cell immunity showing some relation with protection against severe infection,  no single “gold standard T cell correlate of protection” that is immediately applicable in either natural infection or vaccination studies.   



Par 2 Recent results, in line with expectations on induction of neutralizing Ab, hybrid immunity and protection against hospitalization by booster vaccination


Ep 307-2: Tauzin medRxiv Nov 2022 Effect of hybrid immunity in 3 X vaccinated subjects on neutralization of Omicron subvariants in Canada


Previously infected = before vaccination




  • BA.4/5 and BQ.1.1 are much less neutralized compared to the D614G and the other Omicron subvariants.
  • The titers were lowest in infection-naïve individuals, but those who have been infected before or after vaccination present better humoral responses


Thus confirming that hybrid immunity generates better humoral responses also against these BA/5 subvariants.


Ep 307-3: Tauzin Dec 2022: follow-up on fourth dose and BQ.1.1 neutralization in Canada: 

  1. Confirmation of hybrid effect for 4th-dose vaccinated individuals who have been recently infected: they recognize and neutralize better the BQ.1.1, than donors who have never been infected or have an older infection.





  1. Not different between Pfizer or Moderna monovalent or bivalent = remarkable, but low numbers





Ep 307-4: Ngoc Tan medRxiv 19 Dec: Effect of bivalent BA.1 mRNA on Neut titers and T cells in  Netherlands


  1. Bivalent booster vaccination with mRNA-1273.214 (Moderna bivalent) or BNT162b2 OMI BA.1 (Pfizer bivalent) resulted in a rapid recall of humoral and cellular immune responses independent of the initial priming regimen (either Ad26 Janssen or mRNA).


Neutralizing Ab: as expected lower titers against BA.1 and BA.5 than against ancestral virus



T cell IFN-gamma production upon Spike peptide stimulation


  1. No preferential boosting of variant-specific responses was observed:  T-cells cross-reacted with Omicron BA.1 and BA.5



Note: Participants with a known history of prior SARS-CoV-2 infection were eligible, unless the infection occurred less than three months before the start of the study (based on self-reporting). Those with breakthrough infections between study visits were excluded.  No separate analysis in “hybrid immunity” was made.


Ep 307-5: Sara Carazo medRxiv 27 Dec 2022:  Prior infection- and/or vaccine-induced protection against Omicron BA.1, BA.2 and BA.4/BA.5-related hospitalisations in older adults: a test-negative case-control study in Quebec, Canada.


Set-up: test-negative case-control study among symptomatic elderly adults tested for SARS-CoV-2.




Previous infection (with or without vaccination) protects strongly against hospitalisation




Sorry for the following Table, but it is important  (PI = previous infection)




  • “Infection-induced immmity” (without vaccination and “hybrid immunity” (vaccination + previous infection) is always highly protective against hospitalization, with no clear influence of omicron subvariant.
  • Vaccination alone provides also good protection and (as could be expected):
    • Increases with increasing number of doses
    • Decreases with more “immune escaping” variant: protection better for BA.1 > BA.2 > BA.5
    • Decreases with time since last vaccination.


Ep 307-6: Eric Topol Ground Truth 11 Jan: provides a number of arguments that the BA.5 bivalent vaccine outperforms in inducing neutralizing antibodies against novel subvariants and clinical effectiveness


  1. BA.5 bivalent vaccine induces higher neutralization titers than vaccine based on ancestral Wuhan


Importantly, the superior induction of neut is not only against BA.5, but also against XBB (which is the precursor of XBB.1.5 that now is potentially inducing a new wave in US


  1.  Ep 307-7: A retrospective population study in Israel, comparing hospitalization and death in 65+ subjects who either received a BA.5 bivalent booster or not shows a clear protective effect.



  1. Ep 307-8: MMWR report, essentially confirming the protection of additional BA.5 booster against hospitalization in 65+  (both as compared to unvaccinated and to monovalent only)





  1. Ep 307-9: Tenforde MMWR Dec 2022 finds a lower, but still substantial VE of bivalent vaccine in adults > 18

VE of a biva­lent booster dose (after 2, 3, or 4 monovalent doses) against hospitalization for COVID-19–

  • 57% (95% CI = 41%–69%) compared with no vaccination
  • 45% (95% CI = 25%–60%) compared with receipt of last monovalent doses, with last dose ≥11 months earlier




(No sub-analysis of > 65 years old provided) 


Because of waning of monovalent vaccine-conferred immunity, relative effectiveness of bivalent vaccines was higher with increased time since the previous monovalent dose.


  1. Ep 307-10: Overview from CDC of adult hospitalizations according to vaccination status in all US until Nov



Clearly, the bivalent vaccine reduced hospitalizations in Nov, when a new wave by BQ.1.1 (BA.5 subvariant) hits.

From all these data, Eric Topol concludes that his own skepticism against the BA.5 bivalent vaccine (because it had not been properly tested in animals or humans) was unfounded.



Par 3 Some unexpected results


3.1 Result from Qatar


Par 307-11: Chemaitelly medRxiv 1 Nov 2022 shows a remarkable result in a Qatar cohort, followed from Dec 2021 -Sept 2022:


A primary-series vaccination enhanced immune protection against omicron reinfection,

but a  booster vaccination compromised protection against omicron reinfection.   


Set-up: Matched, retrospective, observational cohort.  Data retrieved from the integrated, nationwide, digital-health information platform. Most infections are diagnosed not because of symptoms, but because of routine testing. The majority of primary omicron infections was the BA.2 subvariant.





Interestingly, there were no cases of severe disease in any of the cohorts.


Importantly also, all these cohorts consisted of  “young adults” without any comorbidity




Discussion:  in fact both results are remarkable:


  1. The protection against omicron BA.2 by a primary vaccination, that preceded the observation for almost one year is higher than expected. It is not clear to me how sure we are that some individuals had an unnoticed omicron BA.1 infection?  In fact the authors admit undocumented reinfections may have occurred
  2. The increase of infection rate after a first (wild-type) booster is considered a consequence of “immune imprinting” and these observation are late events :
  • follow-up commenced >4 months after the booster, at a time when booster effectiveness is expected to be marginal.
  • There is no question that the booster dose reduced infection incidence right after its administration, based on evidence from this same population


Their conclusion: These findings do not undermine the utility of booster vaccination in the short-term, but may point to potentially significant public health complexities requiring fine-tuning of booster vaccination to those who can best benefit from it, such as those most clinically vulnerable to severe COVID-19.


Remark by Geert Molenberghs:

Chemaitelly et al. do adjust for age, using age bands of 10 years. That's fine, but two caveats. Due to the very strong dependence on age, even narrower bands are indicated, or age as a continuously variable, which is then modeled carefully. Furthermore, in addition to confounding, age also causes effect modification, which means that the effect of the exposure (here: previous vaccinations) changes from age group to age group. It would actually be better to determine the effect separately for each age band and then, for example, to make a graph of this, or possibly to average it, weighted for the population fraction per age band. Also note that a number of HRs in the third panel of Table 2 (page 23) do not differ significantly from 1. The correction for testing frequency is interesting, but complex. There are feedback loops between testing and infection/other outcomes of course. Thus, testing is a co-endpoint rather than a covariate. Here a more complex structural model is needed


3.2. Three papers by  Nabin Shrestha following a cohort of over 50,000 employees of Cleveland Clinic (OH, US)


General remarks:

  • Presumably symptomatic infections: Most positive tests would have been done to evaluate suspicious symptoms or as part of quarantine and return-to-work testing of employees exposed to patients with COVID-19. A small proportion would have been tests done as part of preoperative or preprocedural screening.
  • No data on disease severity or hospitalization.


Ep 307-12: CID April 2022 on effect of previous infection and/or primary vaccination (2 doses) 


  • Both previous infection and vaccination provide substantial protection against COVID-19.
  • Vaccination of previously infected individuals does not provide additional protection against COVID-19 for several months, but after that provides significant protection at least against symptomatic COVID-19.


Ep 307-13: CID April 2022 on booster effect


  • Risk of COVID-19 increased with time since proximate SARS-CoV-2 exposure,
  • Boosting protected those >6 months since prior infection or vaccination.


One boost was independently associated with lower risk of COVID-19

  • among those vaccinated but not previously infected (hazard ratio [HR], 0.43; [CI], 0.41– 0.46)
  • as well as those previously infected (HR, .66; 95% CI, .58–.76).


Two (wild type) boosts compared to one: higher risk of COVID (HR, 1.54; 95% CI, 1.21–1.97) in previously infected

there is no advantage to administering more than 1 dose of WT vaccine to previously infected persons.


  1. The 2 dose vaccinated, but not boosted and not previously infected have higher incidence of COVID-19.




  1. Effect of time since exposure (either previous infection or 2 dose vaccine) on incidence of (new) infection



  • In previously infected subjects, protection seems to wane after 6 months
  • In previously vaccinated (not infected), waning seems to start earlier (> 3 months), but those vaccinated 9-12 months have an equally low incidence as those vaccinated < 3 months ago (!?). 


The latter counter-intuitive finding is difficult to explain, but the authors suggest on page 2175 (highlighted text) that it could be due to a “boost” from an unrecognized Delta infection in those who were vaccinated 9-12 months earlier, while those who were vaccinated more recently were protected against Delta (thus not boosted) and therefore more susceptible to omicron. .  


  1. Effect of first booster on previously infected subjects



As we know from Fig 2, protection wanes by progressive time since infection (nonboosted conditions in different colors).  A booster always reduces the incidence of new infection (compare boosted with non-boosted conditions in the same color).


Summary multivariate analysis




Both in vaccinated not infected (Table 3) and in previously infected (Table 4):

  • The longer time since previous exposure (either vaccine or infection), the higher chances on (re)infection (middle part of Tables)
  • Boosting resulted in lower risk (upper line) if performed > 6 months after primary exposure (lower lines).
  • As expected, patient contact was associated with slightly higher infection chance, but, remarkably, higher age or male sex were associated with slightly lower chances.   


  1. Effect of number of doses on those who were previously infected



Among previously infected individuals:

  • 1 dose had a significantly lower risk of COVID-19 than those who received no vaccine,
  • 2 doses                no lower risk than for those who received no vaccine

higher risk than those who received 1 dose  

  • 3 doses                lower risk than those who received no vaccine

higher risk than those who received 1 dose.


In a multivariate analysis, no vaccine or 2 doses were worse than 1 (= reference in Table 5), while 3 doses not different from 1 dose.




Authors Conclusions:

  1. Immunity from prior COVID-19 infection provides substantial protection against the Omicron variant for at least 6 months even in the absence of a vaccine.

BUT: little to be gained by vaccinating those who are within 6 months of SARS-CoV-2 infection.

  1. Among individuals with waning immunity, boosting with a COVID-19 vaccine not designed for the Omicron variant protects against Omicron variant infection in both vaccinated and previously infected individuals.

BUT: no advantage to administering more than 1 dose of vaccine to previously infected persons.


Limitation: elderly, children, and the immunocompromised were not adequately represented and caution should be exercised in extrapolating these findings to those populations.

Ep 307-14:  medRxiv Dec 2022: Effectiveness of BA.5 bivalent vaccine


Among 51011 working-aged Cleveland Clinic employees, the bivalent COVID-19 vaccine booster was 30% effective in preventing infection, while the virus strains dominant in the community were represented in the vaccine.




Essentially this Table on time-dependent risk of infection shows 3 main observations:

  1. The bivalent vaccine has a significant but moderate protective effect (30 %)  = expected, but less than one could have hoped.
  2. With regard to the protective effect of a prior infection: BA.4/5 is much more protective than previous variants = expected, since BA.4/5 is more related to the present subvariants and the time is shorter, hence more specific and no waning of antibodies. (illustrated in Fig 1)




  1. The higher the number of previous vaccine doses, the higher the chances on (re-) infection.  Obviously, this is highly counter-intuitive. (illustrated in Fig 2)




The authors are also puzzled by this counter-intuitive positive relation between number of doses and risk on (re)infection).  They state

  • A simplistic explanation might be that those who received more doses were more likely to be individuals at higher risk of COVID-19. But this is not very likely.
  • In fact: those who received fewer than 3 doses (>45% of individuals in the study) were not those ineligible to receive the vaccine, but those who chose not to follow the CDC’s recommendations. Thus NOT less but MORE  likely to have exhibited higher risk taking behavior: hence more infections expected in those with fewer doses.

They refer to Chemaitelly etc  but provide no explanation and conclude on the safe side: The effect of multiple COVID-19 vaccine doses on future risk of COVID-19 needs further study.


However: according to the authors (p. 11):  There were too few severe illnesses for the study to be able to determine if the vaccine decreased severity of illness. Our study of healthcare personnel included no children and few elderly subjects, and the majority would not have been immunocompromised.

Therefore, all these results refer mainly to effects on mild/moderate symptomatic infection in a relatively healthy aldult population.


Remark by Geert Molenbergs

A very important risk factor, especially for the more serious manifestations of an infection (severe covid, hosp, ICU, mort) is increasing age.  We get such strong confounding that this is a "textbook situation" for Simpson's paradox.


See Ep 307-15 for explanation on Simpson paradox (or also: confounding and effect modification)

As far as I could see from a quick reading, Shrestha, et al. do not correct for age. They do mention age when describing baseline characteristics. In the light of my previous comments, I would therefore question this analysis.

Nevertheless I agree that Simpson's paradox play less of a role when the study population is limited to younger people on the one hand, or older people on the other. I would not dare to say that confounding/effect modification do not play a role, so in the interpretation of the results it is good to be cautious, but there is indeed a good chance that these effects do not undermine the study results qualitatively, but shift qualitatively.





  1. Bivalent vaccines and/or previous infection (especially with more recent omicron variants) increase neutralizing antibody levels to newer omicron subvariants, but the level remains much lower than against wildtype virus.
  2. There is a clear protective effect against severe disease (hospitalization) in older adults (> 65 years).
  3. The effect of additional boosters on symptomatic (not severe) infection in younger immune-competent adults is complex and puzzling, but it seems that there is little to gain from additional doses, especially in subjects who have been recently (< 6 month) infected with a related variant.


How to explain?


The discrepancy between weak or no protection against infection (in younger adults) and clear protection against hospitalization (in elderly) could be due to divergent humoral and cellular immune responses:

  • Antibody responses are too weak to protect against infection with immune-evasive subvariants: a booster rather “deviated” Ab responses from the new Spike epitopes (from mutations) in  presently circulating subvariants as a consequence of “imprinting”.  
  • T cell responses are more effectively boosted, because the T cell epitopes in Spike differ from the Ab epitopes and are more conserved (hence no problem of imprinting). Since T cells recognize infected cells, they have a limited (if any) role in prevention of infection, but they could block unlimited replication, hence preventing severe disease.  


Clearly, however, some of the results by Chemaitelly and Shrestha (with all caution we expressed) suggest that additional doses of mRNA vaccines (both monovalent and bivalent) do in fact enhance  the risk on infection.  Two possible explanations:

  • Epidemiological: besides already discussed age, risk behavior could have a role: people, who faithfully receive multiple doses could feel so safe that they take more risk. This has not been investigated.
  • Immunological: the long feared “enhancing antibodies”(Enh Ab) .  Some elements to consider:
    • During the development of COVID vaccines, many warnings on Enh Ab were launched.  In  fact, deep analysis of B cell repertoire after infection or vaccination showed that a minority of Ab could have enhancing properties, but in the clinical reality, they were clearly offset by neutralizing Ab (Neut Ab), as there was never any clinical evidence of enhancement upon a second/third contact with the virus (after primo-infection or vaccination).  
    • Enhancement is very well known in Dengue: it is dependent on Ab with weak affinity that “smuggle” the virus into myeloid cells expressing receptors for IgG (FcγR), i.e. monocytes/macrophages/dendritic cells, which are well-known target cells for Dengue virus.  Whether SARS-CoV-2 is able to infect these cell types remains controversial, but they are certainly not a major target.
    • While measuring Neut Ab is relatively straightforward (with the caveats mentioned), there is no consensus on a “correlate of enhancement” in the literature.


In conclusion, it is not impossible that

    • Imprinting could disturb the previous beneficial balance between Neut and Enh Ab AND
    • New subvariants more easily infect FcR (+) myeloid cells  

But I cannot find proof for either element of that hypothesis in recent literature.  


Moreover, it is very obvious that we need more careful studies, especially in vulnerable population, presenting with age- or disease- related immune deficiency.


As discussed before, novel types of vaccines should be developed to induce broader immune responses against Spike and other proteins, by preference covering a range of Sarbeco viruses (other human and potentially zoonotic SARS-related viruses) and eliciting more robust mucosal and T cell mediated responses.  


I’m curious to read your reactions


Best wishes,