Comments from scientists on dose spacing of the COVID-19 vaccines.
Prof Sheila Bird, formerly Programme Leader, MRC Biostatistics Unit, University of Cambridge, said:
“Week-on-week comparison of COVID-19 diagnoses is critically important between patient-facing healthcare workers who received their 1st mRNA immunization in December 2020 and received their scheduled 2nd mRNA vaccine dose [A] versus had their 2nd dose postponed to an unspecified date [D]. Is UK set up to deliver answers quickly that patient-facing healthcare workers surely have a right to expect?
“Background: Many hospitals throughout UK received supplies of Pfizer/BioNTech mRNA vaccine early in December 2020 to enable the 1st of two pre-scheduled doses to be administered in December 2020 to patient-facing healthcare workers (HCWs). The 2nd dose was due to be administered 21 days after the 1st. Many 2nd doses due this week were cancelled, as were all those pre-scheduled for administration after 8 January 2020.
“Some hospitals, such as University College London, exercised clinical discretion: their HCWs whose date for their 2nd mRNA vaccine had been assigned in 2020 were reassured that the 2nd dose booking in January 2021 would be honoured.
“Many hospitals, however, deferred to the advice given by UK’s four Chief Medical Officers (CMOs) on 31 December 2020 and cancelled the 2nd dose bookings that had been pre-scheduled for patient-facing HCWs whose first mRNA dose was injected in December 2020.
“Tilting at windmills with 1: 3 tilted randomization ratio? UK’s CMOs have taken an historic, pandemic, public interest decision to deviate from the authorized and trialled dose-schedule for the world’s first mRNA vaccine ever deployed. Dose-interval was 21 days in the Pfizer/BioNTech randomized trial (see NEJM) with the 2nd dose window pre-defined as 19 to 42 days.
“UK has ordered 3m to 10m doses of the Pfizer vaccine. With no allowance for inter-dose mortality, the corresponding maximum number of persons who could be administered their 1st dose would be 1.5 million to 5 million on the understanding that their delayed 2nd dose at 12 weeks would be the Pfizer vaccine. [Any other intention should be explicit before the 1st Pfizer dose is administered.] Since over 1 million Pfizer doses have already been administered (including 2nd doses), there are sufficient ordered doses to enable 800,000 new 1st dose Pfizer recipients to be randomized so that one-quarter receive their 2nd dose on day 22 [schedule 1/22], as originally authorized by Medicines and Healthcare products Regulatory Agency, while three-quarters do not receive their 2nd dose until 12 weeks later, as proposed by CMOs [schedule 1/85]. Comparison between the two randomized groups would give robust evidence about how vaccine-effectiveness alters week-on-week.
“Good, better, but not best for gleaning intelligence: Suppose that UK opts out from the 1:3 tilted randomization proposal, how can hospitals glean evidence informatively about week-on-week vaccine-effectiveness when 2nd mRNA dose is delayed from 3 to 12 weeks? Answer as follows.
“Many UK hospitals have landed in the midst a CMOs-advised, historically controlled, pandemic experiment.
“Each hospital with patient-facing healthcare workers (pfHCWs) who received their 1st Pfizer dose in December 2020 are eligible provided that some pfHCWs received their 2nd Pfizer dose as scheduled (ie 21 days after 1stand so in December 2020 or early in January 2021) [group A] while others had their pre-scheduled 2nd dose in January 2021 cancelled, generally without a re-scheduled deferral-date [group D]. The first question to answer, per hospital, is how many of their pfHCWs fall into groups A versus D for all December 2020 and for four different 1st dose-periods.
“Almost all of those who received their 1st Pfizer dose before 12 December 2020 should have received their 2nd scheduled dose in 2020 or by 3rd January. Many who received their 1st Pfizer dose during 12-18 December will have had their 2nd dose cancelled. Almost all who received their 1st Pfizer dose on or after 19th December 2020 may find themselves in Group D.
“TOTALS first: Until we know the TOTALS for eligible hospitals, and especially their * and ** subtotals, we shall not know just how powerful the above non-randomized, incidental allocation between groups A and D might be. Suppose, however, that 120,000 pfHCWs feature somewhere in my table, of whom one-quarter received their two mRNA vaccines as scheduled [1/22day schedule].
“Next step: The next step is to know, by swab-date, about each and every COVID-19 diagnosis that occurs for any pfHCW who features in the above table. Record-linkage and hospitals’ absentee-list should enable us to do this promptly. The swab-date for each recorded COVID-19 diagnosis is used to determine which week, post-1st Pfizer dose, COVID occurred in: for example, days 1-7 = week 1; days 43-49 = week 7*. Currently, we are in week 4 or week 5 after 1st Pfizer dose for those pfHCWs who received their first Pfizer dose during 8-11 December 2020.
“Good, not best: This proposal is less good than 1:3 tilted randomization but has merit because, even in lockdown, the exposure risk of hospitals’ pfHCWs is substantially occupational and because we owe it to pfHCWs to find out as quickly as possible how vaccine effectiveness in and after week 4 differs between those who, having received their 1st Pfizer dose in December 2020, have also received their 2nd [A group] versus those whose 2nd scheduled dose has been deferred [D group].”
* [days 1-7 = week; days 8-14 = week 2; . . .; days 22-28 = week 4; days 29-35 = week 5; days 36-42 = week 6; days 43-49 = week 7; days 50-56 = week 8; days 57-63 = week 9; days 64-70 = week 10]
Dr Stephen Griffin, Associate Professor in the School of Medicine, University of Leeds, said:
“The recommendation of the JCVI to delay second vaccine doses across the board has caused controversy and confusion. The motivation is understandable given the predicament we find ourselves in with the epidemic once again spiralling out of control. However, vaccination cannot be the sole strategy to curtail the spread of SARS-CoV-2 throughout the UK, new variant or not, and we once again face the scenario where harsh restrictions will be necessary to re-establish control. These difficulties were both predicted and eminently avoidable had advice been heeded; whilst the new variant compounds these issues, this also could arguably have been avoided if an appropriate suppression and quarantine strategy had been followed.
“Nevertheless, the availability of effective vaccines significantly alters the picture for the better compared to the situation in spring 2020. The two vaccines currently approved in the UK are both safe and effective, although they are by no means the same, with each offering distinct advantages. The data supporting the Pfizer vaccine robustly supports a defined protocol with two doses given 21 days apart for the vast majority of the patients involved in the phase III trial. The data for the AZ/Oxford vaccine is more nuanced and broader, resulting from changes to dosing and scheduling experienced during the trial period. Each vaccine therefore brings different benefits, with Pfizer being highly effective within a somewhat inflexible framework, whereas the greater breadth of conditions within the AZ trial revealed the apparent increase in efficacy where doses were spaced further apart.
“Thus, given the differences in both the nature of these vaccines and the way in which trials were delivered, it is a concern that aspects deemed favourable for one vaccine should now also be applied to the other. Such changes would entail using the Pfizer vaccine “off-label”, and the company has understandably distanced itself from this notion; the UK is notably the only country currently considering this path. Given the tremendous effort and investment required to mount clinical trials on this scale, it feels unwise to alter the Pfizer protocol – put simply, if it ain’t broke, don’t fix it…
“The Pfizer vaccine has been judged by JCVI to confer 90% protection following the first injection, yet there is no data to support how long this might last and what may or may not happen if the second dose is delayed. This is also difficult to reconcile with data showing that robust antibody responses are coincident with the second Pfizer injection at day 21 – without data on appropriate patients, we should not merely assume that these will be the same if second doses are delayed.
“Given these differences between the vaccines (there are others relating to the immunogens themselves that I won’t explore here), it is also notable that SARS-CoV-2 vaccination strategies have two obvious objectives – first to rapidly provide protection to those most vulnerable to disease (including those at risk of acute exposure such as healthcare workers and teachers), and second to provide widespread population coverage to eventually suppress and ideally eliminate SARS-CoV-2 from the UK. I would suggest strongly that these aims are in fact satisfied by the Pfizer and AZ vaccines, respectively.
“I would therefore urge a stratified approach both to the vaccines, as well as, critically, vaccinees. We should remember that shielders and the elderly might not only respond differently to vaccines, but have also undergone the most profound life-altering issues as a result of the pandemic, living with the heightened threat of severe disease since March 2020. The shorter, dogmatic, yet reliable protocol of the Pfizer vaccine therefore suits this population despite the issues concerning the cold chain, plus many of this group have already received their first dose. Similarly, the timescales involved and greater availability of the AZ vaccine lend themselves to the second objective, concomitantly providing time for restrictions to reduce the global burden of disease. Thus, as current guidance advises against the mixing of vaccines within individuals, should we not apply the same logic to vaccination strategies and populations as well?
“Lastly, rapid rollout of the Pfizer vaccine within shielding and elderly populations may also minimise the opportunity for the virus to evolve vaccine resistance within immunocompromised individuals, or in those experiencing only partial vaccine protection before numbers are driven down by enhanced restrictions. This should be a pivotal consideration for the mid- to longer term as we combat the UK epidemic.”
Comment on the need for randomised trials to evaluate the strategy of delaying second dose of vaccine:
Prof Sheila Bird, formerly Programme Leader, MRC Biostatistics Unit, CAMBRIDGE, said:
“Often, decisions to safeguard public health have to be taken in the face of uncertainty. Importantly, however, when robust evidence can be got to evaluate those decisions, we should ensure that robust, randomized evaluation takes place from the outset. Hence, UK’s historic decision to prolong the inter-dose interval for the Pfizer/BioNTech mRNA vaccine from 21 days to nearly 12 weeks should be properly evaluated in the interest of both public health and scientific method.
“Robust randomized evaluation of policy-decision can be done relatively simply: randomize one quarter of those prioritized to be offered their 1st dose of BNT162b2 to receive the 2nd dose after 21 days versus three-quarters to receive their 2nd dose at around 12 weeks. Assume 160,000 1st doses of BNT162b2 are to be administered per week, then after 5 weeks some 200,000 citizens would have been randomized to receive the 1/22 day schedule versus 600,000 assigned to the 1/85 day schedule. This like-with-like powerful comparison will reveal how vaccine effectiveness actually compares during the weeks 4 to 9 after randomization, with answers (in terms of COVID-19 cases and hospitalizations) available by the week 12 after randomization began.
“An historic pandemic decision has been made to vary the dose-interval for the first mRNA vaccine authorized for use in UK. Let the UK’s second decision be randomized evaluation of the first.”
Notes to editors
Longer version of Sheila’s thoughts:
MHRA’s Public Assessment Report Authorisations for Temporary Supply (under regulation 174 of the Human Medicines Regulation); recommendations by Joint Committee for Vaccination and Immunization.
Background: MHRA’s Public Assessment Report Authorization for the COVID-19 Vaccine AstraZeneca for active immunization of individuals >= 18 years old for the prevention of coronavirus disease 2019 (COVID-19) states: “The vaccination course consists of two separate doses of 0.5 ml each. The second dose should be administered between 4 and 12 weeks after the first dose (see section 5.1).” Heterogeneity of dosing intervals across trials of the COVID-19 Vaccine AstraZeneca was more happenstance than by design and comparisons between-intervals did not have randomization as their underpinning. Nonetheless, MHRA granted the same laxity in dose-intervals as the trials had presented.
The MHRA’s corresponding text (last updated 31 December 2020) for the COVID-19 mRNA Vaccine BNT 162b2 reads: “BNT162b2 is a vaccine indicated for active immunization of individuals 16 years of age and older to prevent COVID-19 caused by the SARS-CoV-2 virus . . . Following dilution with saline, BNT162b2 is given to you by an authorised practitioner as an intramuscular injection into the muscle at the top of the upper arm (deltoid muscle). You should receive two doses (each 0.3 mL) given 21 days apart.”
Public interest announcement: The Joint Committee for Vaccination and Immunization (JCVI) has, however, announced a major change from MHRA’s original authorization for temporary supply of BNT162b2. Instead of the second dose being received 21 days after the first, the dosing interval was extended to be from 3 weeks up to 12 weeks. And, in the interest of public health, the UK’s four Chief Medical Officers intimated on 31 December 2020 that UK’s inter-dose interval, including for BNT162b2, was likely to be maximal: 12 weeks.
The Pfizer/BioNTech mRNA vaccine is the first mRNA vaccine ever authorized in UK/USA. It was trialled in an exemplary and well-controlled manner so that, in particular, the scheduled interval of 21 days between doses was strictly adhered to. This was properly reflected in MHRA’s authorization for temporary supply and may contribute to BNT162b2’s higher vaccine efficacy – as the time interval for a mRNA booster injection may be critical for getting the best immune response. Moderna’s mRNA vaccine, authorised in USA, has a 28-day dose-interval. However, around 2,000 persons randomized in the Moderna randomized trials had received only one dose by the time of interim analysis: they have a key role in JCVI’s appraisal (see 2. Below).
More detailed explanation about JCVI’s rationale was needed: On 31 December 2020, a short statement (with Public Health England’s ANNEX A) was issued by JCVI. In respect of both vaccines, the Committee advised: “initially prioritising delivery of the first vaccine dose as this is highly likely to have a greater public health impact in the short term and reduce the number of preventable deaths from COVID-19.”
Explanation in respect of JCVI’s advice on dose-interval extension for Pfizer/BioNTech mRNA vaccine was critical: it being the first mRNA vaccine ever authorized in UK/USA and having been authorised, as trialled, for 2nd dose at 21 days.
1. The JCVI summary states: “models suggest that initially vaccinating a greater number of people with a single dose will prevent more deaths and hospitalizations than vaccinating a smaller number of people with two doses.” However, neither models nor time-varying assumptions about mRNA vaccine effectiveness during days 23 to 84 are detailed in ANNEX A, nor referenced.
2. JVCI’s considerations include that the Pfizer/BioNTech mRNA vaccine efficacy between dose 1 and dose 2 was 52% (95% CI: 29% to 69%); but, allowing 14 days for immune response to develop, 89% (95% CI: 52% to 97%) between days 15 and 21; and 91% (95% CI: 74% to 97%) between days 15 and 28 (before booster dose takes effect). Lacking data beyond 28 days on BNT162b2, JCVI remarks: “Similar findings were seen with the Moderna mRNA vaccine out to 108 days after the first dose (Annex A)”. However, ANNEX A cautions to the contrary: “These participants had a median follow-up time of 28 days (range: 1 to 108 days). The small, non-random sample and short median follow-up time** limits the interpretation of these results. There appears to be some protection against COVID-19 disease following one dose; however, these data do not provide sufficient information about longer term protection beyond 28 days after a single dose.”
3. JVCI’s considerations include: “With most vaccines an extended interval between the prime and booster doses leads to a better immune response to the booster dose”. However, as no previous mRNA vaccine has been authorized, this statement cannot be read or interpreted as “With most mRNA vaccines an extended interval . . .”
4. JVCI’s considerations include: “The rate of vaccine delivery in the UK is currently limited by vaccine supply rather than by workforce capacity.” This is key as the alternative to a 12-week dose-interval for the mRNA vaccine would be to increase the volunteer workforce capacity. Evidence on supply issues was not adduced; and has been challenged.
** ie at the time of interim analysis of Moderna randomized controlled trial.
Dr Andrew Garrett, Executive VP, Scientific Operations, ICON Clinical Research, said;
“It is important to distinguish controlled experimentation from uncontrolled experimentation. Controlled data come from randomized clinical trials and these provide the basis for determining cause and effect by comparing different groups with respect to an outcome, like infection or hospitalization. Randomization ensures that an unbiased comparison can made between the groups. Uncontrolled experimentation is the introduction of an untested measure for one group (or for everyone) and it is very difficult to determine if any subsequent change observed is a result of the measure or due to other things that are changing at the same time.
“Currently experimentally controlled data are available for a single dose vaccine regimen in the very short-term only – that is, prior to the planned second dose. These controlled experimental data do support one dose vaccine efficacy for that very limited time period. Single dose data currently available beyond that point generally reflect non-adherence to the intended clinical trial protocol, and as a result can be unreliable – making interpretation much harder. However policy makers are faced with difficult decisions and have to balance the effectiveness and wider impact of NPIs with those of PIs, like vaccines. In adopting a policy of delayed second dose boosters, policy makers are having to use their judgment based on previous experiences rather than having access to controlled experimental data. It is similar for the NPIs, where cause and effect is also difficult to determine. On the whole, we have to accept that policy makers are having to make judgments in the face of a worsening situation on the ground. However once such policy is enacted, every effort should be made to source controlled experimental data to inform future judgments and decisions. It is important therefore to consider which experimental methods could be used to generate data that could be used to help answer some of these questions.
“One approach would be to use a platform trial design to study approved vaccines head-to-head, including dose modifications during vaccine deployment. This could be undertaken at specific sites during roll-out, subject to participant consent. Like the UK’s RECOVERY trial of COVID-19 treatments, the platform design could also be used to evaluate new candidates – including ones that may not have the commercial resources to progress independently. This would undoubtedly be an efficient use of resources, address concerns about the continuing use of placebo in vaccine clinical trials, and provide much needed direct comparisons of different vaccines.
“Experimental data could also be generated through vaccine deployment using a randomized stepped-wedge type design. Care homes are the obvious unit, and since roll out will take a number of weeks, it would be straightforward to select care homes at random within a region and roll-out through time, using the lagged time to vaccination as the control. Other units could, at a later date, include university halls of residence, or schools (subject to vaccine approval for children). Vaccination of individuals within a “unit” would of course remain voluntary. This type of approach could provide valuable data on safety and within-unit transmission.
“Given so much remains unknown, and many questions remain unanswered, it is imperative that international and national organizations mobilize to create an experimental framework in 2021 to generate empirical evidence.
Prof Stephen Evans, Professor of Pharmacoepidemiology, London School of Hygiene & Tropical Medicine, said:
“This is not a simple problem. The idea that there is a definite clear-cut answer is not true because respected scientists differ in their views.
“In an ideal world, decisions about treatments would only be made within the exact parameters of the trials which have been conducted. In the real world, this is never so.
“It is safest and most cautious to use the vaccines in the exact conditions reflecting the trials. However this will never be possible in the real world and the question is how much moving outside the exact conditions is acceptable. Pharmaceutical companies are constrained by law and must only make claims that are reflected in the authorisation documents set out by regulatory authorities. Clinicians and health authorities have to take the circumstances they are dealing with into account.
“We have a crisis situation in the UK with a new variant spreading rapidly, and as has become clear to everyone during 2020, delays cost lives.
“The trials did not compare different dose spacing or compare one versus two doses, so we simply do not know what is ‘optimal’. So, the information directly from the trials is lacking. We have to utilise what we know from science generally. We know that vaccinating only half of a vulnerable population will lead to a notable increase in cases of COVID, with all which that entails including deaths. When resources of doses and people to vaccinate are limited, then vaccinating more people with potentially less efficacy is demonstrably better than a fuller efficacy in only half.
“It is clearly on stronger ground to use the vaccine in exactly the same way as it was used in the trials, but it is simply not true to say that there is evidence that using the vaccines in a different way will have dramatically reduced efficacy. We have some evidence that the efficacy is quite good, and there are no reasons to believe it will show a sudden decline between three and twelve weeks.
“We must take into account that in the current UK context there will be many more cases of disease and therefore more deaths by vaccinating fewer people.
“The US is a different environment and sticking to the exact form of dosing used in the trials is more cautious but is almost undoubtedly at a high cost in terms of extra illness and deaths. The US may prefer the certainty of efficacy in half the relevant population leaving half unprotected, for that is the choice in the situation of limited doses and other resources. They may also not have a situation of limits on doses. Giving what may be a higher protection to the highest risk groups could be a better strategy in some circumstances, but it may not be – it depends on the context.
“The key question is how many extra cases (and possibly also deaths) arise through the various different plausible assumptions? Does following the licensed dosing interval lead to fewer cases?
“It is vital to realise that vaccine efficacy (VE) is not a value that applies to each individual – it is an average across individuals, some of whom are essentially fully protected, and some of whom have little protection. A booster will increase the proportion with full protection. Getting the second dose later will probably result in better longer-term protection.
“There is no doubt that the UK’s current advice is focussed on public health. The notion that getting one dose results only in protection for 21 days or one month is not biologically plausible.
“What is definitely true is that the UK is uniquely placed to follow-up all those vaccinated and to see if the policy is justified based on the cases of disease among those vaccinated. Data on this will become available within about a month, and policy will no doubt be changed if the data indicate it should be.”
All our previous output on this subject can be seen at this weblink:
Dr Stephen Griffin: “No conflicts of interest.”
Prof Sheila Bird: “SMB is a member of the Royal Statistical Society’s COVID-19 Taskforce and chaired the RSS Working Party on Performance Monitoring in the Public Services. SMB has family-members and friends who are patient-facing healthcare workers who received their 1st Pfizer dose in 2020 but whose 2nd pre-scheduled Pfizer dose has been cancelled. SMB served on UK’s Medicines Commission in the 1990s; and as first statistician on NICE’s Appraisal Committee (1999-2005).”
Dr Andrew Garrett: “I am employed by ICON which is a Contract Research Organization . ICON provides pharmaceutical services to the pharmaceutical and biotechnology industries. ICON conducts clinical trials on behalf of Sponsors, including vaccine trials. I am a member of the UK Statistical Authority’s (UKSA) Research Accreditation Panel.”
Prof Stephen Evans: “No conflicts of interest. I am funded (one day per week) by LSHTM. They get funding from various companies, including Astra Zeneca and GSK but I am not funded by them, I have no involvement in obtaining funding from them and I am not an investigator on any grants obtained from them. I am the statistician to the ‘meta-Data Safety and Monitoring Board’ for CEPI. I am paid for my attendance at those meetings and will be paid expenses for travel if that occurs. I am a participant in the Oxford/Astra Zeneca trial, but am still “blind” to the vaccine received.”