Updated: Oct 7, 2021
A paper published October 4, 2021 in The Lancet based on >3.4m USA patients found Pfizer vaccine effectiveness (VE) vs. infection decreased from 88% 1m after vaccination to 47% after 5m, but that VE vs. hospitalization remained strong at 93% through 6m.
This post will briefly summarize key results.
This paper followed >3.4m patients >12yr old in the Kaiser Permanente Southern California (KPSC) system between December 2020 and July 2021 using a retrospective cohort design.
All patients needed to have >1yr of previous data for inclusion to establish comorbidities that were used in the modeling.
Their primary analysis computed relative risk of PCR+ infection, comparing unvaccinated with fully vaccinated individuals for each calendar day using Cox regression models with vaccination status including as a time-varying covariate. This approach adjusts for any time confounding since it effectively compares the vaccinated and unvaccinated cohorts for each calendar day and aggregates information across days to arrive at the VE estimate, which is computed as:
VE = 100 x (1 - HR),
where HR is the ratio of hazard of infection for fully vaccinated and unvaccinated cohorts.
To adjust for potential confounding, in the model they adjusted for age, sex, race, previous SARS-CoV-2 infection, SES, previous health-care utilization, & various co-morbidities including BMI, acute myocardial infarction, congestive heart failure, cerebrovascular disease, peripheral vascular disease, organ transplant, diabetes, cancer, renal disease, COPD, hypertension, the Charlson comorbidity index, as well as flu vaccine in the preceding year and pneumococcal vaccine in the preceding 5 years.
It is of particular importance that they adjusted for previous SARS-CoV-2 infection, given substantial data showing that previously infected individuals still tend to have a strong level of immune protection even if unvaccinated, and if vaccinated have an even stronger level of protection than vaccinated individuals without previous infection. Studies that do not adjust for previous infection status have the potential for major bias in estimating VE.
Their results validate what has been seen in Israel, UK and elsewhere: that Pfizer VE vs. infection wanes over time, down near 50%, consistency across all age groups 16yr+
While waning, the VE vs. infection does not go down anywhere near zero, so there is still protection vs. infection. It is just that it is not as strong as it was closer to vaccination time.
This has been consistently found to be the case in papers that model VE using rigorous models trying to adjust for potential confounders, and counters the narrative that after 6m the Pfizer vaccine no longer protects vs. infection but only reduces severity of disease.
They sequenced the positive samples to evaluate the variants. We can see the shift in variants in this region, dominated by Epsilon in the winter, Alpha in the spring, and Delta in the summer.
There is some concern of confounding with variant type since Delta emerged more recently the longest after vaccination.
But looked at waning VE separately by variant, and found that the waning was not an artifact of Delta, and that the waning effect is seen across all variants.
They also analyzed Pfizer vaccine effectiveness vs. hospitalization, with cases defined as hospitalizations with positive PCR test between 14 days before admission to 3 days after admission, using a similar model.
There is no sign of any waning VE vs. hospitalized infections in any age group (see plot above for legend for which color line corresponds to which age group), remaining near 90% through 5m post vaccination. Taken together, this paper validates other data showing:
(1) VE vs. infection wanes down to ~50% after 6m
(2) VE vs. serious disease remains >90% after 6m.
These results match what was found by a study of waning VE done in Israel last month that also rigorously modeled various key confounding factors .
Note: It is disclosed that Pfizer sponsored this study and was involved in the interpretation. Thus, I encourage critical assessment of the design, analysis, and interpretation.
But from my initial reading, the study was well designed, rigorously analyzed using appropriate statistical analysis methods that account for known confounders as best as possible, and the conclusions they draw are appropriately supported by the data, with appropriate accounting for uncertainty. Thus, it appears to be a well-done study that adds to our accruing knowledge of waning vaccine effectiveness for Pfizer vaccines.
Appendix 10/7/21 -- NEJM Study on Waning Immunity from Qatar:
Another paper just came out in NEJM with further corroborating evidence for the waning of vaccine effectiveness vs. infection from a large matched case/control (vaccinated/not) test negative design study out of Qatar.
The study similarly found a sharp reduction in VE vs. infection at 5-6m after vaccination. The magnitude was greater than the USA or Israeli studies showed. This might be related to the fact that Qatar does extensive asymptomatic testing, testing every person in the country every 2 weeks regardless of symptoms, so the vast majority of detected breakthrough cases are symptomatic. We know the vaccines protect less strongly against asymptomatic than symptomatic disease, especially over time as the waning effect is much stronger in asymptomatic disease.
Also, the Qatari population demographics is very different than most other countries -- nearly all younger people and from all over the world -- with only 1-2% over age 65. These dynamics can strongly affect the risk of exposure, especially with certain travel and other privileges only available to vaccinated individuals. The Qatari study looks very well designed, and provides an important complementary perspective on vaccine effectiveness over time in a different population demographic and in a setting with extensive asymptomatic testing.
On the positive side, as we have seen in other studies, even with this sharp decline in VE vs. infection at 5-6m, the VE vs. serious/hospitalized/fatal disease remains much more durable over time, remaining near 90% even at these time points.