DOSING DECISIONS: COULD COVID VACCINES HAVE WORKED BETTER? - Nature
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Feature
DOSING DECISIONS:
COULD COVID VACCINES
HAVE WORKED BETTER?
Past experience and best guesses won the day in the mad
rush to beat back the pandemic. But dose-modelling tools
might have made a difference. By Elie Dolgin
DHIRAJ SINGH/BLOOMBERG VIA GETTY
Speed to approval and manufacturing capabilities factored in to dosing decisions for COVID-19 vaccines.
22 | Nature | Vol 604 | 7 April 2022
©
2
0
2
2
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
W
hen Moderna joined the hunt optimizing immune responses, they have been Moderna’s vaccine programme was no
for a coronavirus vaccine in developing mathematical and computational exception. Researchers involved in the ear-
early 2020, the company had models over the past several years to inform liest mouse studies administered two-shot
only limited clinical experi- dose decision-making for vaccine trials. Not regimens, with doses of up to 20 µg each2. But,
ence with its technology. everyone is convinced the models are ready according to Graham, they made little attempt
Scientists had tested the for prime time; many aren’t even aware that to quantitatively map the immune responses
company’s messenger RNA the platforms exist. But those who embrace observed in mice that received different doses
(mRNA)-based vaccines the technology say that, if companies had to anticipated outcomes in people: the plan all
against a few viruses, such as avian influenza simply capitalized on all the tools at their dis- along was to anchor human trials around the
and Zika, in humans. They found that the high- posal, COVID-19 vaccines might be doing an 100-µg dose that worked best for Moderna’s
est dose levels — upwards of 300 micrograms even better job at containing viral spread and bird-flu vaccine candidate.
— often triggered undesirable side effects. The limiting collateral damage. “We missed a huge Company executives defend the approach
lowest doses (around 10 μg) did not always opportunity,” Evans says. because of the time and data constraints.
elicit a sufficient immune response. “You make the best decision you can,” says
There seemed to be a happy medium: in a two- Leap of faith Jacqueline Miller, head of infectious diseases
dose vaccine for another respiratory virus with Pharmaceutical companies have long used at Moderna, “but that is informed by some
pandemic potential1, a new strain of bird flu, the computational modelling strategies to fine- of the previous programmes that had small
sweet spot was around 100 µg. So, it made intu- tune drug dosing, but such techniques have phase I data with other vaccine antigens.”
itive sense for Moderna, based in Cambridge, rarely been applied to vaccine development. Business considerations factor in as well.
Massachusetts, and its collaborators at the Past experience and animal testing typically Pfizer in New York City and BioNTech in Mainz,
US National Institute of Allergy and Infectious guide dose selection for experimental vaccines Germany, opted for a shorter gap between
Diseases (NIAID) in Bethesda, Maryland, to try — and things were no different for those against doses for their mRNA jab, in part to help them
something similar to tackle SARS-CoV-2. COVID-19. The result was a range of doses for the beat Moderna in the race to marketing author-
Within days of confirming that the products (see ‘Dosing decisions’). ization, whereas Johnson & Johnson in New
coronavirus vaccine it had developed offered Vaccine developers chose amounts that Brunswick, New Jersey, initially advanced a one-
protection in mice, the company started worked in other disease contexts, but immune dose regimen to differentiate the company’s
human trials, testing doses of 100 μg to see responses can vary widely from one pathogen COVID-19 vaccine from others in development.
whether its intuition was right. It also trialled to the next. Animal studies add a degree of con- There were also public-health arguments
25- and 250-μg doses in case it wasn’t. fidence in a vaccine’s success — but the immune for these dosing decisions, such as getting
“The whole point was to go quickly,” says system of a mouse or monkey is not the same people vaccinated quickly. And the choices
Barney Graham, former deputy director of the as that of a human, and scientists don’t fully made during the sprint to provide a vaccine
NIAID’s Vaccine Research Center, who oversaw understand how to scale doses across species. had real-world consequences.
the vaccine’s early development. “That’s just So, most vaccine companies simply made what Moderna’s 100-μg shot proved to offer
how it’s done. It’s not a real precise process.” Jeff Barrett, a quantitative pharmacologist at the greater protection against infection, disease
Other efforts to develop COVID-19 vaccines Critical Path Institute, a non-profit organization and hospitalization than the one from Pfizer–
followed a similar playbook. The University of in Tuscon, Arizona, describes as “a leap of faith” BioNTech, which used only 30 μg of mRNA per
Oxford, UK, in partnership with AstraZeneca from animal models to human testing. injection. By one estimate, recipients of the
in Cambridge, UK, started testing its vaccine Pfizer–BioNTech vaccine had a 58% greater
DOSING DECISIONS
— made from an engineered adenovirus — at risk of infection from the Delta variant of SARS-
a dose of 50 billion viral particles. It chose CoV-2 than those who got Moderna’s shot3.
Companies chose different dosing regimens for
the dose in large part because that amount their vaccines, often for different reasons. (The trade-off is that Moderna’s product came
was used in previous trials of the same vac- with more frequent post-vaccine reactions.)
cine platform for other pathogens, including Vaccine type Company Dose* Differences in formulation and adminis-
the coronavirus responsible for Middle East tration schedules could be at play, says John
mRNA Pfizer–BioNTech 30 µg
respiratory syndrome. Moore, an immunologist at Weill Cornell
This approach to dose selection produced Medicine in New York City. But he, like many
several safe and effective COVID-19 vaccines Moderna 100 µg researchers, points to dose size as the most
in record time — which has helped to save mil- probable explanation for differences in effi-
Adenovirus AstraZeneca– 50 billion
lions of lives around the world — but it did not vectored Oxford viral particles cacy and tolerability. Moderna’s own head-to-
necessarily take full advantage of the vaccines’ head trials of 50- and 100-μg dosing schemes
50 billion
pandemic-altering potential, scientists say. Johnson & Johnson support this conclusion4.
viral particles
Some blame companies’ inexact, educated Perhaps a more consequential decision was
Gamaleya 100 billion
guesses for a high rate of adverse events con- (Sputnik V) viral particles the one that executives at Pfizer and BioNTech
nected to many shots, the diminished efficacy made for their child-sized COVID-19 vaccine.
Inactivated
of others and several high-profile trial failures, Sinopharm 4 µg In a small trial involving a few dozen children
virus
including initial attempts to develop a shot for under 5 years old, the companies found that a
young children. Sinovac Biotech 3 µg pair of 3-μg shots was sufficient to prompt an
“It still feels like people just threw a lot of antibody response comparable to that in teen-
spaghetti at the wall and saw what stuck,” Bharat Biotech 6 µg agers and young adults who had received two
says Thomas Evans, chief scientific officer of full doses. What’s more, the mini-dose didn’t
Protein**
Vaccitech in Oxford, which devised the adeno- Novavax 5 µg trigger the severe fevers observed among chil-
virus-mediated gene-delivery system found in dren given a 10-μg shot, so Pfizer and BioNTech
the Oxford–AstraZeneca vaccine. Biological E 25 µg moved ahead with the smallest dose.
A growing number of scientists think that But in a later-stage trial involving thousands
*Approved adult dose for each shot
the industry can do better. With an eye towards **Antigen dose, not including adjuvant of infants and young children, protection
Nature | Vol 604 | 7 April 2022 | 23
©
2
0
2
2
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.Feature
infectious-disease modelling been better
positioned to obtain those kinds of quantita-
tive insight early in the pandemic response,
“it could have changed the way that we devel-
oped vaccines”, Rhodes says. CanSino chief
scientific officer Tao Zhu stands by the compa-
ny’s dosing decisions: the LSHTM analysis “is a
good model”, he says, but it doesn’t account for
logistical challenges in administering higher
doses and later-stage trial data that shaped
the company’s final call on which dose to use.
Model makers
The LSHTM researchers, along with Evans and
others, have been at the forefront of efforts to
create a mathematical framework for inform-
ing vaccine dose decision-making6. In 2015,
ALEXANDRE SCHNEIDER/GETTY
they convened the world’s first workshop
dedicated to the topic. (Only a few dozen
people attended.) In the years since, they have
honed their modelling techniques with an eye
to streamlining the process of determining
vaccine doses for early-stage trials.
For any vaccine candidate, the modelling
Dosing decisions might have slowed approval for a COVID vaccine in young children. framework starts with data. The researchers
feed immune-response results from animal
came up short. Two- to four-year-olds failed risks missing an optimal window. And dose– experiments into their equations to produce
to produce enough antibodies, and a third response dynamics often differ widely by age. a predicted dose–response curve. They then
booster might be necessary to develop ade- A child is not simply a small adult when it comes scale that dose–response relationship to
quate immune protection in those children. to vaccines — more so than for other medicines. humans using clinical data from a more limited
Meanwhile, Moderna announced last month Then, there’s the question of what to number of doses, often from historical work
that a 25-μg version of its shot provided the measure when calculating vaccine-mediated on similar vaccines. In this way, they come up
same level of immune protection against protection. Antibodies or immune cells? Rates with expected ‘best’ doses for testing in human
COVID-19 in children under six, as did a full of infection or of disease and death? trials — and they can further refine the model’s
100-μg dose in young adults. Moderna is now “A definition of optimal dose may vary predictions as more data become available
moving forwards with global regulatory sub- depending on which of these factors you (see ‘Immune modelling’).
missions in every age bracket. care about,” says John Benest, a mathematical As a proof of principle, the researchers fit
biologist at the London School of Hygiene & their model with mouse and human data on
Tricky business Tropical Medicine (LSHTM). the response of T cells — a type of immune cell
With the benefit of hindsight, most scientists Together with his supervisor Richard White — to an experimental tuberculosis vaccine. The
now think that Pfizer and BioNTech chose and former group member Sophie Rhodes, mathematics then predicted that lower doses
too low a dose for children under five. But it’s would offer the best immune response7. Inde-
hard to fault the thinking of drug executives, “It’s a shame. It could pendent clinical studies run in parallel to the
who were trying to minimize side effects, group’s modelling project bore this out.
says Karim Azer, who has previously worked
have changed the Jennifer Linderman, a systems biologist
on tuberculosis-vaccine modelling at the Bill way that we develop at the University of Michigan in Ann Arbor,
& Melinda Gates Medical Research Institute vaccines.” says such approaches could be useful for
in Cambridge, Massachusetts. “The dose– guiding dose decisions in the future. “We’re
response relationship with vaccines can be in a position now where, going forward, we
very tricky,” he says. Benest took published data from an early clin- can be much more intentional about vaccine
There are several complicating factors. With ical study sponsored by CanSino Biologics in design,” she says. She and Denise Kirschner, a
conventional pharmaceuticals, greater drug Tinjian, China, maker of a one-shot COVID-19 computational immunologist at the Univer-
concentrations usually yield more potent vaccine based on a viral vector. The research- sity of Michigan Medical School, developed
effects, at least up to a certain level. This isn’t the ers modelled5 dosing schemes, prioritizing HostSim, a model that incorporates lung biol-
case with vaccines, because higher doses can population-level immunity in one scenario, indi- ogy alongside simulations of what happens in
sometimes produce less favourable responses. vidual immunogenicity and safety in another the blood and lymph nodes8.
That’s because repeat exposure to vaccine and factoring in cost containment in a third. In Although initially focused on tuberculosis,
antigens can cause certain arms of the immune every situation, the optimal dose — predicted on Kirschner notes that, with the appropriate data
system to secrete enough pro-inflammatory the basis of data from CanSino’s first-in-human inputs, her team’s tool could guide vaccine
signalling molecules to trigger a phenome- trial — was more than double the amount now development for any pathogen that infects
non known as immune exhaustion, leading to approved for use in China and elsewhere. the airways. “We can use our model for flu. We
impaired protection. “It’s a shame,” says Rhodes, now a staff can use our model for COVID. We can use it for
Timing is also important: a long interval scientist at Certara, a drug-development lots of things,” she says.
between shots might coax out more pro- consultancy headquartered in Princeton, In simulated trials, at least, such modelling
tective antibodies, but one that’s too long New Jersey. Had she and other specialists in approaches allow vaccine developers to vet
24 | Nature | Vol 604 | 7 April 2022
©
2
0
2
2
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.many more doses than would otherwise be
feasible, even in the largest of clinical studies.
IMMUNE MODELLING
Scientists have been developing mathematical
scheduling regimens of other mRNA shots,
and determined that a dose between the 30 μg
“You can test a wider range of scenarios in silico,” models to optimize vaccine doses before used by Pfizer–BioNTech and the 100 μg used
they are given to humans.
says Luca Marchetti, a computer scientist at the by Moderna might provide the ideal balance
Microsoft Research–University of Trento Centre 1. Animal testing of immunogenicity and tolerability. They
Scientists administer a wide range of
for Computational and Systems Biology in Rov- doses in small animal models. planned to press forwards with an initial trial
ereto, Italy, who last year developed a model to evaluating up to 60-μg doses of mRNA.
support mRNA-vaccine development9. But when the company, in partnership with
Certara, simulated immune responses to the
Pandemic response vaccine in virtual participants, they found that
Before the pandemic, few companies wanted older individuals failed to mount robust anti-
2. Initial modelling
to invest in this kind of vaccine-dose model- Researchers identify doses correlating to the
body responses at the highest planned dose.
ling. Around five years ago, Evans and Kent largest and smallest immune responses in animals. “This result triggered an internal discussion on
Kester, then head of translational sciences They use modelling to estimate the relationship the phase I study design,” says Daiichi’s Ryoko
between dose and response.
at Sanofi Pasteur (the vaccines division of Sawamura, who leads a modelling team at the
Paris-based Sanofi), tried to create a research Dose 1 2 3 4 company. Ultimately, her firm added a 100-μg
Response (at time T)
consortium focused on tool development in dose to its first-in-human trial protocol.
this area, but they failed to get buy-in from AstraZeneca scientists are now using
Response
industry or regulatory authorities. “No one Certara’s model to simulate scenarios not
was really interested in pursuing it,” says captured by earlier clinical studies of the
Kester, now vice-president of translational company’s vaccine. They are interrogating
medicine at IAVI, a vaccine non-profit organi- 1 2 3 4 immune responses in populations that were
Time Dose
zation based in New York City. under-represented in trials — particular ethnic
Because of COVID-19, more vaccine man- groups, for example, and immunocompro-
3. Translation to humans
ufacturers are now experimenting with dose Data from the animal model are scaled up to mised people — to predict who might benefit
modelling, and regulatory agencies are mon- predict a theoretical dose–response relationship from non-standard dosing. And they are look-
for humans, which helps scientists to choose
itoring the science closely. ing at long-term immunity trends to inform
initial doses for testing.
“This is an important area,” says Marco optimal timing of booster-dose regimens.
Dose 1 2 3 4
Cavaleri, head of biological health threats Few seasoned vaccine developers are con-
Response (at time T)
and vaccines strategy at the European Med- verts to the approach. “There are too many
Response
icines Agency in Amsterdam. “The more we variables to model in a way,” says Emilio Emini,
can refine these techniques, the more we will chief executive of the Gates Medical Research
be prepared in the future.” Institute and a former vaccine research head at
Last June, the US Food and Drug Administra- Pfizer and Merck. “At the moment, there are no
1 2 3 4
tion convened a workshop at which research- Time Dose clear prospective models that exist that allow
ers discussed best practices for modelling one to make that initial prediction — other than
vaccine dose–response relationships. White 4. First-in-human extrapolating as best as one can,” he adds.
and Rhodes spoke, as did Andrzej Kierzek and Scientists feed data from early human testing But such modelling tactics are catching
into the human-response model to further
Piet van der Graaf, modellers at Certara who, refine dose decisions. on. The FDA says that, at the end of 2021,
before the pandemic, had created a tool for it received its first submission of a vaccine
running virtual trial simulations of antibody product created using modelling to optimize
drugs and biological therapies. dose–response relationships.
ADAPTED FROM REF. 6
The tool had helped pharmaceutical compa- Although Moderna’s Miller and other indus-
nies to predict unwanted immune reactions. try executives say it’s too soon to begin pro-
But when COVID-19 hit, the Certara research- 1 2 3 4 spectively selecting vaccine doses for human
ers realized that the same model could be trials, they might change their minds if and
used to forecast desired immune responses before efficacy results were even known for when the tools get validated and prove their
from vaccines. As a first test, they plugged in the first wave of COVID-19 vaccines, Kierzek worth. “As we gain more experience,” Miller
the amino-acid sequence corresponding to and van der Graaf had already concluded says, “we’ll get there.”
the coronavirus’s spike protein, the bit used that longer dosing intervals than those being
by most COVID-19 vaccinese. As van der Graaf evaluated would yield improved antibody Elie Dolgin is a science journalist in
recalls: “We got a surprising, meaningful result.” responses10. Data from the United Kingdom, Somerville, Massachusetts.
The immune responses predicted by the where extended dosing schedules were rou-
model “seemed to be plausible”, he says. And tine during the vaccine roll-out, later con- 1. Feldman, R. A. et al. Vaccine 37, 3326–3334 (2019).
2. Corbett, K. S. et al. Nature 586, 567–571 (2020).
as companies such as Moderna and Pfizer– firmed that advantage. 3. Dickerman, B. A. et al. N. Engl. J. Med. 386, 105–115 (2022).
BioNTech began to publish more human and 4. Chu, L. et al. Vaccine 39, 2791–2799 (2021).
mouse data, the Certara scientists would Dose decisions 5. Benest, J., Rhodes, S., Quaife., M., Evans, T. G. & White, R.
G. Vaccines 9, 78 (2021).
incorporate those results into their simulation Daiichi Sankyo was one of the first drug com- 6. Rhodes, S. J., Knight, G. M., Kirschner, D. E., White, R. G. &
workflow. They added response dynamics for panies to incorporate the Certara platform Evans, T. G. J. Theor. Biol. 465, 51–55 (2019).
7. Rhodes, S. J. et al. NPJ Vaccines 3, 36 (2018).
T cells and B cells, which produce antibodies, into its vaccine-development programme.
8. Joslyn, L. R., Linderman, J. J. & Kirschner, D. E. J. Theor.
into the mix, along with plug-in modules to The Tokyo-based firm began testing its mRNA Biol. 539, 111042 (2022).
account for different vaccine technologies and vaccine in humans in March 2021 — a slow start 9. Selvaggio, G. et al. CPT Pharmacometrics Syst.
Pharmacol. 10, 1448–1451 (2021).
routes of administration. that gave trial organizers the opportunity to
10. Giorgi, M., Desikan, R., van der Graaf, P. H. & Kierzek, A.
Over time, their model — dubbed the Vac- learn from the experiences of other companies. M. CPT Pharmacometrics Syst. Pharmacol. 10, 1130–1133
cine Simulator — grew in sophistication. And Scientists at Daiichi looked at the dosing and (2021).
Nature | Vol 604 | 7 April 2022 | 25
©
2
0
2
2
S
p
r
i
n
g
e
r
N
a
t
u
r
e
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.You can also read
