Preprint on immune response to mixed dose scheduling of COVID vaccines from the Com-COV study

Preprint on immune response to mixed dose scheduling of COVID vaccines from the Com-COV study

This article was published on
June 28, 2021

This explainer is more than 90 days old. Some of the information might be out of date or no longer relevant. Browse our homepage for up to date content or request information about a specific topic from our team of scientists.

This article has been translated from its original language. Please reach out if you have any feedback on the translation.

A preprint, an unpublished non-peer reviewed study, from the Com-COV trials, looks at the immune response to mixed dose scheduling of the Oxford-AstraZeneca and Pfizer-BioNTech COVID-19 vaccines.

A preprint, an unpublished non-peer reviewed study, from the Com-COV trials, looks at the immune response to mixed dose scheduling of the Oxford-AstraZeneca and Pfizer-BioNTech COVID-19 vaccines.


Safety and immunogenicity report from the Com-COV study – A single-blind randomised non- inferiority trial comparing heterologous and homologous prime-boost schedules with an adenoviral vectored and mRNA COVID-19 vaccine

Not peer-reviewed
This work has not been scrutinised by independent experts, or the story does not contain research data to review (for example an opinion piece). If you are reporting on research that has yet to go through peer-review (eg. conference abstracts and preprints) be aware that the findings can change during the peer review process
This work was reviewed and scrutinised by relevant independent experts.

What our experts say

Context and background


Media briefing

Media Release

Expert Comments: 

Prof Kevin McConway

The reporting of this study in the preprint is, unavoidably, rather complicated. That’s because effectively the researchers carried out two so-called ‘non-inferiority’ trials, and there are some special considerations in reporting the results from such trials. One has to be careful not to make too much of findings that weren’t planned in the registered trial protocol, and that makes some of the wording rather impenetrable to the uninitiated. But I believe the analysis and the reporting in the preprint is appropriate. The main comparisons are, first, between one of the standard UK dose schedules, where people have two doses of the Oxford AZ vaccine, and a schedule where they have AZ first followed by the Pfizer-BioNTech vaccine (and below I’ll refer to these as AZ/AZ and AZ/Pf), and second, between the other common standard dose schedule in the UK, two doses of Pfizer-BioNTech and a schedule where participants have Pfizer first followed by AZ (and I’ll call those Pf/Pf and Pf/AZ). The researchers didn’t formally set out to make other comparisons, such as between AZ/AZ and Pf/AZ, presumably because they were thinking in terms of what to give for the second dose to someone who had already had a first dose. If someone has already had one dose of AZ, they can’t then go onto a schedule of Pf/AZ. But, despite that kind of comparison not being made formally, one can use data in the preprint to make the comparisons anyway. You’ve just got to be a bit cautious in interpreting the results.

The researchers not only found that AZ/Pf didn’t have an inferior immune response (in terms of antibody level) than AZ/AZ – they found it had a demonstrably superior response. The same didn’t happen with their other primary comparison, though. They couldn’t show that Pf/AZ was not inferior to Pf/Pf, and in fact their results indicate that the antibody levels were, on average, worse on Pf/AZ than on Pf/Pf. There are important reasons why this result doesn’t mean than the Oxford-AZ vaccine is worse overall than the Pfizer vaccine, though – I’ll explain why below. Within their pre-planned analysis, the researchers didn’t compare (for instance) AZ/Pf with Pf/Pf in terms of antibody levels, though such comparisons can be made informally. In the press release, it’s (correctly) said that the highest antibody response was with the standard Pf/Pf schedule, though that isn’t too much higher than the antibody response from AZ/Pf – the difference between them is small enough to be explained just by chance variability. The antibody response with Pf/AZ is lower than with AZ/Pf, and enough lower that it’s unlikely to be due to chance – and the antibody response with AZ/AZ is lower still. But on another measure of immune response, the T cell response, the highest level was with AZ/Pf, and indeed both of the mixed dose schedules produced a higher T-cell response than did either of the schedules using the same vaccine both times. So the picture is complicated.

The press release, and the Interpretation sections in the preprint report, both emphasise that both of the mixed dose schedules produced stronger immune responses than did one of the existing standard unmixed schedules, two doses of AZ. I agree that this is a very important finding. It’s known from the original clinical trials and from several studies of real-world effectiveness that two doses of AZ are highly effective in raising an immune response and, more importantly, in preventing severe Covid-19. So if the mixed schedules do even better in terms of immune response than two doses of AZ, and since no issues of safety arose in relation to the mixed dose schedules, this study hasn’t shown any reason to avoid these mixed dose schedules.

Since the AZ/AZ schedule comes out worse than the mixed schedules (and than Pf/Pf) on the measures used in this study, why doesn’t that mean that the Oxford-AZ vaccine isn’t as good as the Pfizer vaccine? It’s because of limitations in what this new research could show, and because antibody response isn’t the only aspect of immunity. I’ve already mentioned that the highest level of T cell response was from the AZ/Pf schedule, not from Pf/Pf. Also, as the researchers point out, the interval between the doses in these results was only 4 weeks, and the recommended interval between two AZ doses is considerably longer than that, 8 to 12 weeks. They also point to evidence that previous studies have shown that a longer interval between doses produces a better antibody response with AZ. So perhaps the non-standard 4 week interval between doses was just not long enough for the AZ vaccine to have its best effect. This study also used an interval of 12 weeks between doses for other participants – those results aren’t available yet. They are promised soon, and may (or may not) show different patterns of findings.

There are other limitations too. The data analysed so far only involve antibody levels measured 28 days after the second vaccine dose. It’s known that antibody levels take some time to build up after vaccination. So they will change over time, and indeed the participants will be followed up to examine their antibody levels for a year after they joined the study. Those longer-term results might (or might not) make the comparison results different. If the build-up of antibodies is slower with AZ than with the Pfizer vaccine, that could possibly be another reason why AZ appears to do worse than Pfizer in this study, on some measures – though I’m no immunologist and can’t comment on how likely that is. But, anyway, whether that is the case will be known when further follow up results become available. Also, the participants in this trial were aged 50-70 and results might (or might not) be different in younger age groups.

Perhaps most importantly, though, the research hasn’t investigated how these vaccine schedules compare on actually preventing infection, or preventing serious disease. They’ve only looked at measures of the immune response. I’m no immunologist, so can’t really comment, but they do point out that there is evidence from other studies that good immune responses, in terms of what they measured, do correlate well with actual efficacy against severe disease. However, the study can’t on its own establish how effective the schedules are against severe disease.

Prof Deborah Dunn-Walters

This new pre-print, currently under review at The Lancet, provides encouraging data regarding mixed vaccine schedules involving the AstraZeneca/Oxford and the Pfizer/BioNTech COVID-19 vaccines. As part of the Com-COV study, researchers evaluated the “mix and match” combinations, with a gap of four weeks between the first and second dose, reporting high concentrations of antibodies against the SARS-CoV2 spike IgG protein.

This is welcome news as it shows that all possible schedules involving the AstraZeneca/Oxford and the Pfizer/BioNTech generate a strong immune response against COVID-19. While we look forward to more data on doses given 12 weeks apart, mixing these two vaccines not only appears to be safe but can give a higher immune response than the standard dosing regimens where first and second doses of the vaccine are the same.

While the findings of this study might allow for some flexibility and optimisation of the vaccination programme in the future, we continue to strongly encourage people to accept the offer of both doses of the vaccine within the current programme to ensure maximum protection from COVID-19.

Prof Peter Openshaw

This beautifully designed study has already reported on the side effects of the ‘mix and match’ approach, and now reports the immunology results of the one-month dose interval. The most interesting aspect is that the antibody responses were best with the two-dose Pfizer-BioNTech schedule, whereas the best T cell responses were in those who had the AstraZeneca followed by Pfizer vaccines. Which of these is more protective in the long run remains to be determined, but in the meantime the study is reassuring that using a mixed vaccine approach is not only safe but can potentially give immune responses that are as good or better than those induced by single vaccine regimens. Being able to boost with a different vaccine would add flexibility to the vaccine programme. The next set of data from the 12 week dose interval is keenly awaited.

Prof Paul Hunter

This latest paper is an important contribution to the debate about whether it is better to give identical vaccines for first or second boost (homologous prime boost) or to give vaccines from different manufacturers for the second dose (heterologous prime boost). Given that we have already given second doses to over 60% of the adult population in the UK, the findings of this study are unlikely to affect strategy for the first and second doses but this could have major impact on strategy for a booster campaign in the autumn. The is evidence from before the appearance of COVID that using different vaccines for other infectious diseases can give better immunity So it was not too much to expect that this would be the case for COVID.

We have seen some initial results from this trial reporting an increase in mild and short live adverse reactions in people who had different vaccines for first and second dose Though looking at that paper the risk of adverse reactions after a second different injection, the risk was not that much different from the risks of adverse reactions in the first dose, just that adverse reactions were generally lower after second rather than after first doses.

This recent paper looks at the impact of the dosing schedule on the immune response. This was not a phase 3 trial where the researchers  looked at the incidence of infections in the volunteers after inoculation. It should also be stated that the prime and booster doses were given just 4 weeks apart not 12 weeks as has been done in the UK. The researchers investigated both Antibody and T cell responses.

“People who had an AstraZeneca followed by a Pfizer injection had rather higher indicators of protection, across all measures, than people who had both AstraZeneca Injections. By contrast there was little difference in response between people who had two Pfizer injections compared to those who had a Pfizer followed an AstraZeneca injection and if anything two Pfizer doses were somewhat less immunogenic than a mixed Pfizer/ AstraZeneca schedule. That the AstraZeneca Vaccine is less immunogenic and protective than the Pfizer vaccine is now well known

We will have to wait till the analyses of this study with 12 week gaps between doses are completed but given what we know already it is likely that the schedules with one or two AstraZeneca shots would perform somewhat better than in this study. Also in subsequent analyses both the Moderna and Novavax vaccines are being included and these results will be interesting.

As to the implications for future vaccine strategy in the UK. The big question at present is whether or not we will be being offered booster vaccines in the Autumn. With the evidence available from this and other sources, I suspect that will be likely for those most at risk from the virus, either due to age or being clinically vulnerable. From this study, people who have had a first course of AstraZeneca should probably be offered the Pfizer vaccine (or possibly Moderna or Novavax dependent on the future trial data from this study) in the autumn rather than a repeat AstraZeneca. People who had a Pfizer first course may not need an autumn booster but if they do then it probably does not matter much which vaccine they are offered.

Dr Peter English

When the first Covid-19 vaccines became available, there was a lot of discussion about how we could use them most effectively. I have commented on this previously.2 3

One of the facts that is familiar to vaccinologists, but perhaps not so widely understood, is that heterologous boosting – priming with one vaccine, and then using a different vaccine for the booster dose – is commonly more effective than homologous boosting, using the same vaccine.4 5

Previous studies have looked at antibody responses to heterologous boosters for Covid-19 vaccines. For example.6-8

This is another paper confirming that, at least with a four-week interval (and as the authors comment, and was supported by another UK paper published today,9 a longer prime-boost provides a stronger booster response), mixing Pfizer-BioNTech and Oxford-AstraZeneca vaccines produces at least as strong a response as do heterologous schedules. (It obviously takes longer to accumulate data with a longer prime-boost schedule, and further data from this study, looking at this, will be published in due course.)

The fact that the antibody responses are marginally lower (although the T-cell responses remain higher) using a mix of the vector (AstraZeneca) and mRNA (Pfizer-BionTech) vaccines than using homologous mRNA vaccine is a curious wrinkle, but not too surprising since antibody responses to the mRNA vaccine are higher than to the vector vaccine.

This paper’s findings, therefore, are no surprise to vaccinologists. They confirm what we had predicted. But it remains very important to check that our predictions were correct, and this paper does just that.


  1. Liu X, Shaw RH, Stuart AS, Greenland M, Dinesh T, Provstgaard-Morys S, et al. Safety and immunogenicity report from the Com-COV study 1 – A single-blind randomised non inferiority trial comparing heterologous and homologous prime-boost schedules with an adenoviral vectored and mRNA COVID-19 vaccine (UNDER EMBARGO UNTIL MONDAY, 28 JUNE, 16:00 BST). 2021; Updated; Accessed.
  2. English PMB. The UK’s vaccine approach isn’t ‘anti-science’. The Spectator 2021; Updated 04 Jan 2021; Accessed: 2021 (04 Jan): (
  3. English PMB. The UK approach to COVID-19 vaccination: why was it so different? Drugs in Context 2021;10:2021-4-5. (
  4. English P, Boseley S. Covid-19: why mix and match vaccines? – podcast. The Guardian: Science Weekly podcast 2021; Updated 16 Feb 2021; Accessed: 2021 (16 Feb): (
  5. English PMB. Can you use a different vaccine for the booster dose? Peter English’s random musings [blog]. Selected Covid-19 vaccine Q&As 2021; Updated 29 Jan 2021; Accessed: 2021 (19 Feb): (
  6. Instituto de Salud Carlos III. The combined use of AstraZeneca and Pfizer vaccines against SARS-CoV-2 offers a powerful immune response [google translate]. Instituto de Salud Carlos III press release 2021; Updated 18 May 2021; Accessed: 2021 (19 May): (
  7. Duarte-Salles T, Prieto-Alhambra D. Heterologous vaccine regimens against COVID-19. Lancet. (
  8. Spencer AJ, McKay PF, Belij-Rammerstorfer S, Ulaszewska M, Bissett CD, Hu K, et al. Heterologous vaccination regimens with self-amplifying RNA and adenoviral COVID vaccines induce robust immune responses in mice. Nature Communications 2021;12(1):2893  PMID: 34001897. (
  9. Cunningham AM, Flaxman A, Marchevsky N, Jenkin D, Aboagye J, Aley PK, et al. Tolerability and immunogenicity after a late second dose or a third dose of ChAdOx1 nCoV-19 (AZD1222) (Manuscript draft- embargoed until 28 Jun 2021): Oxford Vaccine Group, 2021 (28 Jun 2021).

Dr Andrew Garrett

The Oxford-led paper describes the results of the previously announced randomized clinical trial that explores the impact of combining the Pfizer and AstraZeneca vaccines on the immune response. This is an important study since understanding the impact of mixing vaccines and modifying the dosing interval have important scientific and public policy implications. Since this a randomized study with a clear methodology and appropriate blinding of patients, it represents an important next step in understanding vaccine optimization.  In this respect the paper will generate global interest and have global implications.

The paper compares a heterologous two-dose schedule with its corresponding homogenous two-dose schedule, based on switching the second jab. The paper describes the results of the cohort of 463 participants who were randomized to receive one of the four vaccine combinations that followed a 4-week dosing interval.  The study was conducted in adults aged 50 years or above, including those with well-controlled comorbidities, and it uses a non-inferiority method whereby the aim is to show that a heterologous combination (e.g. AZ then Pfizer) does not produce a poorer immune response that the corresponding homogeneous schedule (e.g. AZ, then AZ), within a well-defined tolerance limit (the so-called non-inferiority margin).  The analysis method compares the ratio of the geometric means, where a ratio of 1 would imply an identical immune response when comparing the two regimens. Non-inferiority is defined as a ratio no worse than 0.63 for anti-spike IgG.

In more simple terms, the study represents a factorial design that explores the four combinations of AZ and Pfizer in terms of the two dose effect on the immune response post-second dose.  What is clear from the data presented is that the inclusion of at least one dose of Pfizer vaccine induces a stronger immune response than AZ alone.  The highest geometric mean for anti-spike IgG is for the Pfizer/Pfizer combination, followed by AZ/Pfizer and Pfizer/AZ, with the AZ/AZ combination somewhat lower.

It leads the authors to conclude that AZ/Pfizer is not only non-inferior to AZ/AZ, but actually superior to it, since the lower confidence limit is not only greater than 0.63, but is also greater than 1. It also reports that the Pfizer/AZ combination is not non-inferior to the Pfizer/Pfizer combination, but arguably this is a tougher test because the Pfizer/Pfizer combination has the highest immune response of all four combinations. When something is not non-inferior, it essentially is stating that it cannot be ruled out that it is worse, and the authors go on to show that the Pfizer/AZ combination does produce a lower immune response than Pfizer/Pfizer.

Since this is a randomized study with appropriate blinding, it provides strong evidence of causality. It provides strong evidence that AZ/Pfizer 4 week dosing provides a better immune response than AZ/AZ. It also demonstrates that Pfizer/Pfizer produces a stronger immune response than Pfizer/AZ.  Overall the strongest immune response is with Pfizer/Pfizer while the weakest is with AZ/AZ.

Since this is essentially a factorial design, it would have been helpful to have seen an analysis that estimated the main effects and their interaction, whilst also presenting the pairwise comparisons – for instance Pfizer/AZ versus AZ/AZ, not just against Pfizer/Pfizer.  This statistical approach would have been more informative in terms of vaccine optimization.  It would also have been helpful for interpretation for the authors to have presented the two-sided confidence limits for the per protocol analysis set (with an upper bound), as was done for the modified intent-to-treat.

It is not possible however at this stage to determine whether the increased immune response would translate into a stronger clinical response in terms on reduced infection, hospitalization or mortality. Furthermore results from the separate 12-week dosing interval cohort will be available in due course, and a different pattern of immune response might emerge, given the recommended AZ dosing interval is longer than 4 weeks.


No items found.