How did ChAdOx1 nCoV-19 fare against the beta variant
Introduction
First off I am grateful Astrazeneca (AZ) published this trial data in the New England Journal of Medicine (NEJM) on March 16th 2021, data that looks not so great for the company and also provides some tough data for public health officials in general. Also well done to the company for providing detailed information in the supplementary appendix, information that other vaccine manufacturers have been very reluctant to share.
AZ had earlier published interim analysis of their trial using their vaccine ChAdOx1 nCoV-19 with data from UK and Brazil alone in the Lancet (December 8th 2020), the South African trial data was in phase 1/2 stage only at that point. This interim analysis from UK and Brazil was positive — 62.1% efficacy against symptomatic COVID-19 using their standard dose protocol, 55.1% efficacy against positive NAAT swabs.
So what about the South African data?
2026 individuals recruited between June 24th and November 9th 2020.
1912 individuals received first dose of placebo or vaccine
For these individuals, at baseline, 1606 were seronegative , 290 were seropositive, 16 had missing serology.
1806 received second dose of placebo or vaccine
Of whom 1513 had baseline negative serology, 280 had baseline positive serology, 13 had missing baseline serology.
1749 reached 14 days post second dose, 1467 had baseline negative serology, 271 had baseline positive serology and 11 had missing baseline serology.
Efficacy
Poor trend of efficacy, that does not even reach statistical significance — thus we can say no efficacy in the seronegative population.
Effect of natural immunity
Using this table (Table 2) if you subtract cases in the seronegative group from the any group, you can infer that there were 4 cases in the seropositive or missing serology group, of which 1 was in the placebo and 3 in the vaccine. Furthermore, using Table S6(2), you can infer of these 4 cases, 3 were in the seropositive (1 placebo, 2 vaccine) and 1 in the missing serology.
You could use the risk ratio between seropositive and seronegative groups to give you a “vaccine efficacy” of sort and give us an indication of the protection afforded by prior covid infection. 3/271 in seropositive vs 42/1467 and then accounting for total follow up period between the two groups would give you RR of 0.43 = 57% efficacy. But could this be impacted by the effect of the vaccine? Well there are more cases in the vaccine group than the placebo group for the seropositive, so the effect of the vaccine should not have any impact.
And if we only want to look at the effect of natural immunity, let’s look at what happens post dose 1.
The authors actually split the study period into two, pre 31st October 2020 was deemed the point prior to emergence of the beta variant (but this is even before the recruitment window closed!)
Using this data, you can infer for the period 31st October 2020 to 15th January 2021 (the beta transmission period). Only 2 additional cases in seropositive (both in vaccine group) vs 44 aditional in seronegative group. This would give an “efficacy” of 75% (follow up period equal in both groups).
If we looked at NAAT positive swabs only during this time period, the efficacy would increase to 82%.
What about vaccine efficacy in the seropositive group, it is 12.4% for NAAT positive swab, 12.4% for mild-moderate disease and negative efficacy (-100%) for moderate disease, none of these efficacies reach statistical significance, I think it would be fair to say there is no efficacy in vaccinating seropositive patients in this trial.
Lot of data, but it seems clear that the longer the follow up, the more the protection of natural immunity.
So in summary, a holy triad of
- Vaccine not showing efficacy in the trial population
- Strong evidence of protection from previous infection against the beta variant (the variant with the greatest reduction in neutralization in invitro studies)
- No efficacy in giving the vaccine to seropositive individuals either
But why is there such an early excess cases in seropositive patients (data upto 31st October 2020)?
We can calculate mean followup times for seropositive groups vs seronegative groups — 112.9 days vs 125.8 days. The lower follow up time for the seropositive makes sense as the likelihood of the general population being seropositive would increase with time during the active pandemic. It also suggests that many of the seropositive had quite recent infections. So why did so many of the seropositive get “reinfected” soon after recruitment, well maybe they are not reinfections, but just prolonged RNA shedding after a recent infection. Admittedly they had to test negative on the PCR on visit 1 to not be removed from the study population, but given the false negative rates of PCR testing, it is possible that these are still representative of prolonged RNA shedding.
On the day of randomization, 67 individuals tested positive for PCR (and were subsequently removed from the per protocol set), of these 24 (35%) were also seropositive on the same day. By definition these individuals should not have had a previous PCR positive (or previous history of COVID-19) prior to recruitment, yet they were diagnosed PCR positive at the same time of being found seropositive. So it is quite conceivable some early patients are found to be seropositive before their first positive PCR.
Other similar trends
The above effect is also seen in a couple of other trials.
Pfizer Biontech phase 3 trials FDA data
So very soon after recruitment, seropositive patients show a lot of early positive PCRs, but this disappears with more follow up. This is what is found in the study - Association of SARS-CoV-2 Seropositive Antibody Test With Risk of Future Infection — Harvey et al — JAMA Internal Med Feb 2021. The ratio of positive PCR among those with antibodies vs those with a negative antibody test at index was 2.85 0 to 30 days, 0.67 at 31 to 60 days, 0.29 at 61 to 90 days, and 0.10 at more than 90 days. They felt the early excess was simply prolonged RNA shedding and not true reinfections — makes sense.
Similar trend seen in Novavax SA trial, where they published an interim analysis with 60 days follow up, they stated “Notably, during the initial 60 days of follow-up in the placebo group, the preliminary incidence of Covid-19 that was observed among participants who were seronegative at baseline (5.3%; 95% CI, 4.3 to 6.6), which included 33 mild and 47 moderate cases among 1516 participants, was similar to the incidence among seropositive participants (5.2%; 95% CI, 3.6 to 7.2), which included 14 mild and 21 moderate cases among 674 participants (Figure 2C). “ — No protection from previous infection against the South African variant?
Not surprisingly they released a statement later dated Mar 11 — “A previously reported initial analysis from the study through 60 days indicated that prior infection with the original COVID-19 strain might not completely protect against subsequent infection by the variant predominantly circulating in South Africa. However, the complete analysis of the South Africa trial indicates that there may be a late protective effect of prior exposure with the original COVID-19 strain. In placebo recipients, at 90 days the illness rate was 7.9% in baseline seronegative individuals, with a rate of 4.4% in baseline seropositive participants.” So clearly with just 30 more days follow up, there is a significant protective effect from previous infection against the South African variant and surely this should increase further with time.
Still waiting for the full analysis of the South African Novavax trial to be published…
Please provide further data if you do have them and offer corrections if you spot mistakes.
