Introduction
In winter and spring 2021, hope was high that mRNA vaccines were going to be able to bring herd immunity and end the pandemic if enough were vaccinated.
After phase 3 studies of mRNA vaccines showed 95% efficacy in preventing symptomatic COVID-19, the rapid vaccine rollouts raised hope that population vaccination could eventually get the pandemic under control. This hope was reinforced by sharp declines of confirmed COVID-19 cases in Spring 2021 in countries with early vaccine rollouts like Israel, the UK, and USA, the publication of results from many observational studies (~20 of them) in spring 2021 providing real-world validation that the vaccines were preventing infections with estimated population effectivenesss vs. infection of 85-95% , and the publication of secondary transmission studies that suggested breakthrough infections after vaccination had lower rates of transmission than unvaccinated infections.
However, starting in the summer 2021, cases started to uptick as the more transmissible Delta variant emerged and quickly dominated, and studies began to emerge showing that the high level of vaccine protection vs. infection waned at 5-6m after 2nd dose, driven primarily by the natural reduction of circulating antibody (cAbs) levels that were key factors in providing the protection vs. infection. This protection vs. infection further eroded in winter 2022 with the emergence of the Omicron variant, with a large set of mutations leading to immune escape properties driven by much lower levels of effective neutralizing antibodies provided by vaccination (or pre-Omicron infection). Numerous observational studies showed vaccine effectiveness vs. infection was only 35-70% soon after 2nd dose of mRNA vaccine, and waned to negligible levels within 5-6m after 2nd dose (0-40% across 10-15 published studies). Boosting restored protection vs. infection to a certain degree, but this protection also quickly waned within a few months. Numerous other studies have emerged showing protection of natural immunity from previous infection vs. reinfection also wanes to low levels with Omicron and its subvariants.
Given the greatly reduced protection vs. infection from vaccines and previous infection and the continued rapid evolution of SARS-CoV-2 to more transmissible variants with greater immune escape potential, the past hopes of herd immunity by vaccination and/or infection have dimmed. It appears we are stuck dealing with COVID-19 as an endemic disease for the indefinite future.
The news isn't all bad, however. While the immune protection vs. infection is heavily dependent on circulating antibody levels, the protection vs. severe/fatal disease also relies heavily on T-cells and memory B-cells, whose levels do not wane as rapidly as cAbs and are more robust against future variants, even Omicron. We still see substantial protection vs. severe/fatal disease from vaccination, with a dozen or so published population studies demonstrating VE vs. severe disease at 60-90% after 2 vaccine doses, waning slightly to 40-80% after 6m+, but restored to 80-95% after 3rd dose booster. The endurance of protection vs. severe/fatal disease is one of the keys to our ability to treat COVID-19 as an endemic disease, ensuring we do not see the high mortality and morbidity of earlier in the pandemic, or seen in places like Hong Kong coming off extreme zero-COVID policies that left much of their population immune naive to SARS-CoV-2 when experiencing their first real COVID-19 surges.
However, a key question is whether the protection vs. severe Omicron-based disease will wane in a matter of months like it does for infection. One early study first published online by Tartof et al. in the Lancet Respiratory Medicine in April showed that VE vs. COVID hospitalization (defined as hospitalization for acute respiratory infection with concurrent positive COVID-19 test) waned from 85% within 3 months of 3rd dose booster down to 55% >3 months after booster, suggesting a similar waning as vs. infection. Some have extrapolated these numbers to suggest that protection vs. severe disease might wane to negligible levels by 6m or so, requiring frequent boosters to maintain protection vs. severe disease. In this blog post, I will evaluate the current evidence for durability of protection vs. severe disease after 3rd dose boosters using mRNA vaccines, considering 5 published studies during the Omicron era. As I will show, the available studies suggest that after 3rd dose, immune protection vs. severe disease remains relatively durable with little drop off of VE over time, and with drop offs only evident at the longest time points after vaccination, a time period subject to confounding by indication given that these are comprised of individuals in high risk groups prioritized for early boosting.
What does the Tartof study show?
Tartof, et al. initially published results from a study done using health records from the Kaiser Permanente Health System (KPHS) in Southern California, a large health system with 4.7 million members who are representative of the broader Southern California population.
The study considered all KPHS patients 18 years or older who were admitted to the hospital or emergency department between December 1, 2021 and February 6, 2022 with a diagnosis of acute respiratory infection based on ICD-10 codes and with concurrent PCR test for SARS-CoV-2. Vaccination status was either documented in the medical record, or captured by the California Immunization Registry, to which all vaccinations are required to be reported within 24 hours.
A test-negative case-control design was used with cases defined as individuals with positive PCR tests, and controls those with negative PCR tests. Logistic regression models were used to compute the odds ratio of cases to controls for those fully vaccinated (with 2 doses) or boosted (with 3 doses), while adjusting for age, sex, race/ethnicity, BMI, comorbidities, previous influenza and pneumococcal vaccination status, and documentation of previous SARS-CoV-2 infection. Analyses were done separately for two outcomes: (1) hospitalizations and (2) ER visits, and also stratified by variant (Delta vs. Omicron) and time since 2nd or 3rd dose. Their key results showed that VE vs. hospitalization with Omicron-based disease after 3rd dose was 85% within 3m of booster, but waned to 55% >3m after booster, and similarly the VE vs. ER visit waned from 77% to 53% once >3m after booster.
Based on these results, their conclusions were that the protection vs. hospitalization or ER admission waned strongly by 3m after 3rd dose, suggesting a 3 dose course of the original Wuhan-based vaccine is not sufficient to provide long term protection vs. severe disease, requiring additional boosters or updated vaccines to maintain protection vs. severe disease.
However, since the boosters were originally approved for only high risk groups on Sept 22 2021, the >3m boosted group as of Feb 6, 2022, the end date of the study, was highly enriched for high risk individuals, including immunocompromised, who were prioritized for early boosting. Thus, they updated their analysis with follow up through Mar 1 2022 once available to include a broader demographic in the >3m past 3rd dose group. This was published as a correspondence in the Lancet Respiratory Medicine alongside the original paper above.
Here are the updated results vs. Omicron-based disease with 6 weeks of additional followup.
We see that the overall VE vs. hospitalization/ER admission was 83%/76% <3m after booster, and still 79%/74% >3m after booster, showing very little waning at all.
In addition to providing updated data with the extra 6wk of follow up, they also split results out by immunocompromised status, which is important given it is known that immunocompromised have much weaker responses to vaccination or infection.
For immunocompetent individuals, they found little evidence of waning vs. severe disease, with 86%/78% VE vs. COVID-19 hospitalization/ER admission <3m after boost, and 86%/76% >3m after boost.
For immunocompromised, the VE was much lower, 55%/52% <3m and dropping to 25%/47% >3m after boosting.
These results suggested that in this study, a 3 dose regimen maintained strong protection vs. severe disease, even with Omicron, in the time frames considered in the study, especially for immunocompentent adults. It seems that the waning of protection vs. severe disease after 3m in the original published paper were artifacts of confounding by indication, given that those in the >3m group were strongly dominated by those prioritized for early boosting, who tended to have an inherently higher risk of severe disease than those not prioritized for early boosting. This type of confounding has been common in the pandemic given the prioritization of older and higher risk groups for early vaccination, especially in studies looking at waning of vaccine effectiveness over time. It is one of the many types of confounding that can potentially cause confusion, and need to be carefully considered and accounted for.
The study did covariate adjustment for various confounders including age, comorbidites, BMI, and other key factors, but often covariate adjustment is not sufficient for removing all confounding, which apparently was the case here.
This is just one study in one area (southern California in the USA). There are several other studies that have also looked at VE of 3rd dose mRNA boosters vs. severe COVID-19 disease in the Omicron era. I will now briefly summarize those. Nyberg et al. Lancet Study, England
Nyberg, et al. published a study in the Lancet that looked at potential waning of VE vs. hospitalization in an English cohort study.
This study looked at laboratory-confirmed COVID-19 cases among residents in England between Nov 29 2021 and Jan 9 2022, linked to national medical records data including vaccination status, hospital attendance and admission, and mortality.
They estimated relative risk of hospital attendance or admission within 14 days, or death within 28 days after positive PCR test, using proportional hazards regression, stratifying by test date, 10-year age band, ethnicity, residential region, and vaccination status, and covariate-adjusted for sex, SES status, previous infection status, and year within 10-year age band. Their primary goal was to estimate reduction of risk of hospitalization, and of death, from previous infection, 1st dose vaccination, 2nd dose vaccination, or booster vaccination, with results also presented based on time since 1st., 2nd or 3rd vaccine dose to look at waning. Results were presented separately for Astrazeneca and the mRNA vaccines (Pfizer or Moderna).
VE estimates are obtained by 100% x (1-HR), where HR is the estimated hazard ratio from the proportional hazards model. Here are the HR results for the various outcomes:
Here, we are interested in the results on potential waning of effectiveness after 3rd dose, which is split out based on time since 3rd dose (<2wk, 2-7wk, 8-11wk, >12wk). For mRNA vaccines, they found the HR for 3rd dose vs. hospital admission was 0.23 at <2wk, 0.26 for 2-7wk, 0.22 for 8-11wk and 0.19 for >12wk, which corresponds to estimates of VE of 77%, 74%, 78%, and 81%, respectively. This shows no evidence of waning of protection vs. hospital admission through >4m after 3rd dose, or for the other outcomes.
Note that this paper looked at infections through Jan 9 2022, and the booster campaign in the UK started on Sept 16 2021 and ramped up slowly. Out of the 31 million given 3rd dose by Dec 31, 2021, only 1m were given by Oct 1, only ~8m given by Nov1, and only 19m by Dec 1. Thus, the >12wk group was dominated by those prioritized for early boosting. In this study, it does not appear there was much of a confounding by indication effect.
Incidentally, they also estimated protection provided by past infection during the Omicron era, and found an HR of 0.55, corresponding to 45% effectiveness of previous infection vs. hospitalization in the unvaccinated. Among vaccinated, the HR of previous infection was 0.96, meaning that previous infection only provided an additional 4% reduction of risk of hospitalization for those who are vaccinated.
Gram et al. study in Denmark
Gram, et al. posted a preprint to medRXiv on Apr 20 2022 presenting results of VE vs. infection and COVID-19 related hospitalization during Alpha, Delta, and Omicron eras in a nationwide Danish cohort study.
This study was based on a data from a nationwide registry in Denmark with the 3rd dose Omicron study based on residents 18yr+ between Dec 21 2022 and Jan 31 2022. Risk of infection or COVID-19 based hospitalization, defined as any new admission lasting >12hr and occurring between 2 days before and 14 days after a sample producing a positive PCR test for SARS-CoV-2.
They used proportional hazard regression models to assess the HR of COVID-19 hospitalization after 3 doses of vaccine while adjusting for age group, sex, geographical region, and calendar time to adjust for temporal and seasonal effects, with VE estimated as 1-HR. VE was stratified by time since 3rd dose, using 14-30d, 31-60d, 61-90d, 91-120d, and >120d, and by age (18-59yr, 60yr+)
Table 5 contains their key results:
In the Omicron period, we see that for the 60yr+ cohort, the VE vs. hospitalization went from 94.4% for 14-30d to 93.8% for 30-60d to 91.7% for 61-90d, to 84.5% for 91-120d and 77.3% for >120d. These results suggest some potential waning of protection here in the 91d-120d and >120d groups, but there may be some confounding by indication within these groups given that followup went through Jan 31, 2022, and only 4.8% of the Danish population had been boosted 90d before (Oct 29 2021) and only 1.1% of the Danish population had been boosted 120d before (Sept 30 2021). In this data set, among the boosted 60yr+, 33% were 14-30d, 38% were 31-60d, 17.5% were 61-90d, and only 7.7% and 3.3% were 91-120d and >120d, respectively. Thus, the 91d-120d and >120d >60yr cohorts were comprised of the small percentage of the total cohort; those initially prioritized for early boosting.
For the 18-59yr group, the VE vs. hospitalization went from 89.8% 14-30d after 3rd dose to 87.4% at 31-60d, 80.5% for 61-90d, 72.8% for 91-120d, and only 33.3% for >120d.
Again, this suggests potential waning, but again there is potential for strong confounding by indication. Considering that among the 18-59yr boosted population, 53% were 14-30d, 38% were 31-60d, 5.6% 61-90d, 3.8% 91-120d, and 0.4% >120d, the 61-90d, 91-120d, and >120d groups for 18-59yr were all very small select groups comprised of the subset of younger adults at high risk and prioritized for much earlier boosting, and it is possible the apparent decline from 87-90% VE down to 80%, 73%, and 33% in the 61-90d, 91-120d, and >120d is at least partially and perhaps mostly, due to the confounding by indication.
Stowe et al. study in England.
Stowe et al. posted a medRXiv preprint on April 1 2022 presenting results from a test-negative case control study in the UK looking at effectiveness of COVID-19 vaccines vs. Omicron and Delta hospitalizations.
This study took COVID-19 testing data in England from hospital and community based testing and matched with vaccination histories from the National Immigration Management System (NIMS) and demographic characteristics.
The part of the study we are interested in here involves two groups:
1. Individuals admitted to the hospital within 14d of a SARS-CoV-2 PCR test (injuries excluded), with an ICD-10 code indicating acute respiratory illness, and staying in the hospital at least 2 days.
2. Subset of these individuals requiring oxygen use, ventilation support, and/or ICU admission.
Data from individuals receiving tests through Feb 23 2022 were considered.
A test-negative case-control design was used to relate case (positive PCR test)/control (negative PCR status) status with vaccination status, defined based on vaccination dose and time since dose, with our interest here in boosted individuals with time points 7-13d, 14-34d, 35-69d, 70-104d, and 105d+ after receiving 3rd doses. Their results present VE vs. COVID-19 hospitalization (group 1 above) or critical COVID-19 (group 2 above) for these time points, stratified by age (18-64yr, 65yr+) and variant. Analysis was done using a logistic regression where VE=1- odds of vaccination in cases/odds of vaccination in controls, stratified by vaccine type and variant, while using covariate-adjustment for 5yr age bands, sex, SES, ethnic group, come care resident status, geographic region, HCW status, clinical risk group status, immunocompromised status, and previous infection status.
Here we focus on the Omicron results.
For the 65yr+ age group,
VE vs. COVID-19 hospitalization was 84.7% for 7-13d, 91.3% for 14-34d, 89.5% for 35-69d, 88.1% for 70-104d, and 86.4% for 105d+
VE vs. critical COVID-19 disease was 94.7% for 7-13d, 95.8% for 14-34d, 92.8% for 35-69d, 92.5% for 70-104d, and 86.8% for 105d+
For the 18-64yr age group,
VE vs. COVID-19 hospitalization was 90.9% for 7-13d, 88.6% for 14-34d, 85.8% for 35-69d, 80.2% for 70-104d, and 67.4% for 105d+
VE vs. critical COVID-19 disease was 90.7% for 0-13d, 97.1% for 14-34d, 94.3% for 35-69d, 89.9% for 70-104d, and 75.9% for 105d+
These data show very little waning of protection vs. severe disease throughout the time period of this study, which went through Feb 23 2022, and with the latest groups >3.5m post boost.
The final group would have been boosted before November 10 2021, a point at which <17% of the English population had been boosted.
There may be some mild confounding by indication, especially in the younger age group given that very few would have been indicated to receive boosters during the early time periods. While there was some covariate adjustment for relevant confounders, it may not have been enough to remove the potential effect.
Baum et al. study of elderly in Finland
Baum et al. published a medRXiv preprint on Mar 13 2022 looking at vaccine effectiveness vs. severe COVID-19 in elderly in Finland after the emergence of Omicron.
This study involved a nationwide, register-based cohort of all residents age 70yr+ in Finland between Dec 27 2020 and Feb 19 2022 including 897,932 individuals. Their interest was in assessing the protection of Pfizer vaccination vs. COVID-19 related hospitalization or COVID-19 related ICU stay during the Omicron era between Jan 1 2022 and Feb 19 2022.
COVID-19 hospitalization was defined as primary COVID-19 diagnosis, acute respiratory tract infections, or severe complications of lower respiratory tract infections along with a positive SARS-CoV-2 PCR test between 14d before and 7d after admission, and COVID-19 related ICU stay was marked by treating physician as due to COVID-19 and with a positive SARS-CoV-2 PCR test.
Their goal was to assess effect of vaccination on risk of COVID-19 hospitalization or ICU stay, stratifying by vaccine type (Pfizer, Moderna, AstraZeneca) vaccine dose, and time since latest vaccine dose (<14d, 15-60d, and >60d), while using covariates to adjusting for potential confounders age, sex, region, long-term care facility residence, influenza vaccination in 2019-2020, and number of nights hospitalized in 2015-2019, and comorbidity status. Our focus here is on VE after boosting. For Pfizer vaccinated,
VE vs. COVID-19 related hospitalization was 91% (95% confidence interval 87%-93%) 0-13d after 3rd dose, 96% (95%-97%) 14-60d after 3rd dose, and 92% (89%-94%) >60d after 3rd dose.
VE vs. COVID-19 related ICU stay was 99% (91%-100%) at 0-13d after 3rd dose, 97% (94%-99%) 14-60d, and 90% (76%-96%) >60d after 3rd dose.
For Moderna vaccinated,
VE vs. COVID-19 related hospitalization was 91% (77%-96%) at 0-13d, 98% (95-99%) at 14-60d, and 93% (82-98%) at >60d after 3rd dose.
VE vs. COVID-19 related ICU stays was 100% for 0-13d, 13-60d, and >60d post 3rd dose vaccination.
So the VE vs. COVID-19 related hospitalization or ICU stay remains strong throughout this study, with 92-93% VE vs. hospitalization and 90-100% VE vs. ICU stays at 60d+.
Followup in this study was through Feb 19, 2022.
Finland was slower to roll out boosters than the UK, with only 3.6% boosted by Nov 21 2021, 14.1% boosted by Dec 19 2021, and 49.2% boosted by Feb 19 2022, so again the later group (60d+) is highly select and comprised of those prioritized for early boosting.
Conclusion
So from all of these studies, we see that at the latest time points sampled after 3rd dose, VE vs. severe disease was relatively durable.
There is some drop off in the latest time points in some of the studies, but this is potentially attenuated by confounding by indication given that in each study these groups are comprised of a highly select group of high risk individuals prioritized for early boosting, like the initial Tartof analysis.
Each of these studies have their own potential critiques (some having weaker definitions of Covid-associated hospitalization than others having strong definitions, some excluding d0-d13 after vaccination while others including them, some adjusting for previous infection or comorbidities while others not doing so), but together they paint a picture of relatively durable protection vs. severe disease at least a few months after boosting.
Of course, we need more data with later followup to better characterize any waning after 3rd dose at later time points, 4m, 5m, 6m+ after boosting, and including the general population, not just the select groups chosen for early boosting. These data won't be available until this summer, and so studies on them may not be available until the fall.
These studies may provide evidence of more substantial waning of protection of 3 doses vs. severe disease, so we need to evaluate them as they are published. But from the published studies to date the protection vs. severe disease after boosting appears quite durable.
Additionally, we need to see whether there is further waning of protection vs. some of the more recent Omicron sub variants, including BA.2.12.1, BA.4, and BA.5, that appear to be potentially more transmissible, more virulent, and /or have greater immune escape than the original BA.1 Omicron sub variant. Studies in the fall should have data on severe reinfection rate with these variants. We have overwhelming evidence that with immune escape variant Omicron, VE vs. infection quickly wanes within several months of vaccine doses or boosting (and also after previous infection). But these studies suggest that the protection vs. severe disease may hold up much better over time, especially after 3rd dose boosting. These emerging data need to be assessed to provide information about which adult age groups should receive additional boosters and when, and whether boosting with the current vaccine formulations is useful or whether they should focus on reformulated boosters targeting more recent sub variants.