Read on Twitter
Excellent to see new Canadian seroprevalence estimates from @CANBloodServ https://nationalpost.com/news/only-1-5-per-cent-of-canadians-report-covid-19-antibody-immunity-in-second-pandemic-wave-study
I think there are some issues with how these data are interpreted, which make it sound as though sero-reversion (loss of antibody) is a major challenge to using seroprevalence data for surveillance in Canada. I think there are some errors in interpretation which I'd like to flag
The article suggests that the findings (0.88% in October, 1.51% in November) are lower than the number of confirmed infections: "But the number is still low...lower than a hypothetical 2.3 per cent...if antibodies had been detected in all Canadians with confirmed infections."
As we discussed in class at @UofT_dlsph yesterday (CHL5412!): when estimating symptomatic attack rates using seroprevalence data important to use an appropriate comparator for serological and PCR-based estimates.
These estimates for October and November can be assigned to the midpoint of each month...say, Oct 15, and Nov 15. The appropriate PCR comparator would be 2 weeks earlier, b/c it takes 2 weeks to seroconvert...
Cases that are PCR positive in mid-Oct/Nov won't show up in serology data until the end of each month.
Given exponential increase in Canadian COVID-19 cases this fall, mis-specifying the comparator period makes a BIG difference!
Given exponential increase in Canadian COVID-19 cases this fall, mis-specifying the comparator period makes a BIG difference!
To the numbers: that 2.3% estimate for total cases x 37.59 million Canadians would give us 864,570 counted cases, higher than the 757,435 as of today. But that's not the appropriate comparator, as we are almost in February, following a rough period in the epidemic.
On October 1, Canada had 162,777 counted cases (0.43% of our pop) as compared to CBS seroprevalence of 0.88%.
On November 1, Canada had 238,328 counted cases (0.64%) as compared to CBS seroprevalence of 1.51%.
On November 1, Canada had 238,328 counted cases (0.64%) as compared to CBS seroprevalence of 1.51%.
What we see is that case detection fraction (reluctant to call it symptomatic attack rate, as not all symptomatic individuals tested) is 49% in October and 42% in November.
This is around double the 17.1% we reported based on May/June data. Likely reflects expanded testing.
Declining fraction in November probably reflects a rapidly growing epidemic outstripping test capacity. https://www.medrxiv.org/content/10.1101/2020.11.09.20223396v1
Declining fraction in November probably reflects a rapidly growing epidemic outstripping test capacity. https://www.medrxiv.org/content/10.1101/2020.11.09.20223396v1
In summary, there is no indication from the data presented in this article that seroreversion is invalidating the usefulness of serotesting as an important approach to population surveillance for COVID-19. Like any tool, it's only helpful if you use it right!
