I always wonder at how results get simplified so much. Â And, how the authors sleep at night.
From that article
Things looked good until I saw that the group that had heterologous doses had an average of 85 days between the two doses, while the Pfizer+Pfizer group had an average of 29 days. Â The article indicates both sets of first doses were started in January. Â The incidence of COVID was counted for cases occurring 14 days after the second dose.
There seem to be at least two things wrong with that particular study:
 1) More days at risk.  The  higher incidence group was counted from (Jan 15 (as good as I can estimate from the article) 1st dose; Feb 13, 2nd dose; Feb 27 begins counting)  Feb 27 to Aug 15 (227-58 = 169 days), while the lower incidence group was counted from (Jan 15, 1st dose; Apr 10, 2nd dose; Apr 24 begins counting)  Apr 24 to Aug 15 (227-114 = 113 days).  So if risk was equal from Feb 27 to Apr. 24 as it was Apr 24 to Aug 15, the number of cases in the higher incidence group should have been decreased by about a third to compare with those in the lower incidence group.
 2) Unequal exposure.  Using the WHO's COVID Dashboard, I find that France had 38.7M cases on 2/27/21(day 58), 50.6M cases on 4/24/21 (day 113), 62.8M on 8/15/21 (day 227).  The higher incidence group was at risk of being counted during a period when the country had 24.1M cases in 169 days.  The lower incidence group was at risk of being counted during a period when the country had 12.2M cases in 113 days.  The population of France is about 65.27M at mid-year. Â
Using the entire population of France is not ideal as a control group, but it is the best I can do to try to remove those issues.
There were 10,609 in the higher incidence group. Â During their at-risk period, one would expect 24.1/65.27 of the group (ie, 3,917) to get sick. Â Only 85 got sick, for an Pfizer+Pfizer effectiveness of 97.83% during the evaluated period.
There were 2,512 in the lower incidence group. Â During their at-risk period, one would expect 12.2/65.27 of the group (ie, 470) to get sick. Â Only 10 got sick, for an Az+Pfizer effectiveness of 97.87% during the evaluated period.
The effectiveness of a Pfizer booster seems virtually indistinguishable in the two groups. Â I do not believe this study's results support anything more than a Pfizer booster worked equally well in both cases.
From that article
Quote:The team showed that SARS-CoV-2 infection rates in the mix-and-match [AstraZeneca + Pfizer] group were half that of the group that received two doses of Pfizer.From the referenced study:
Quote:individuals who received each vaccination regimen (which started in January 2021
in both groups)...
Histograms show the infection rate among groups of healthcare workers who were vaccinated with the homologous BNT/BNT combination (n=81/10609) within the recommended 4-week timeframe between the two doses, or with the BNT boost after receiving the first ChAd dose (n=10/2512) approximately 12 weeks before.
Data show the infection that occurred after the 14-days post-boost period, up to the end of recording (08/15/2021).Â
Things looked good until I saw that the group that had heterologous doses had an average of 85 days between the two doses, while the Pfizer+Pfizer group had an average of 29 days. Â The article indicates both sets of first doses were started in January. Â The incidence of COVID was counted for cases occurring 14 days after the second dose.
There seem to be at least two things wrong with that particular study:
 1) More days at risk.  The  higher incidence group was counted from (Jan 15 (as good as I can estimate from the article) 1st dose; Feb 13, 2nd dose; Feb 27 begins counting)  Feb 27 to Aug 15 (227-58 = 169 days), while the lower incidence group was counted from (Jan 15, 1st dose; Apr 10, 2nd dose; Apr 24 begins counting)  Apr 24 to Aug 15 (227-114 = 113 days).  So if risk was equal from Feb 27 to Apr. 24 as it was Apr 24 to Aug 15, the number of cases in the higher incidence group should have been decreased by about a third to compare with those in the lower incidence group.
 2) Unequal exposure.  Using the WHO's COVID Dashboard, I find that France had 38.7M cases on 2/27/21(day 58), 50.6M cases on 4/24/21 (day 113), 62.8M on 8/15/21 (day 227).  The higher incidence group was at risk of being counted during a period when the country had 24.1M cases in 169 days.  The lower incidence group was at risk of being counted during a period when the country had 12.2M cases in 113 days.  The population of France is about 65.27M at mid-year. Â
Using the entire population of France is not ideal as a control group, but it is the best I can do to try to remove those issues.
There were 10,609 in the higher incidence group. Â During their at-risk period, one would expect 24.1/65.27 of the group (ie, 3,917) to get sick. Â Only 85 got sick, for an Pfizer+Pfizer effectiveness of 97.83% during the evaluated period.
There were 2,512 in the lower incidence group. Â During their at-risk period, one would expect 12.2/65.27 of the group (ie, 470) to get sick. Â Only 10 got sick, for an Az+Pfizer effectiveness of 97.87% during the evaluated period.
The effectiveness of a Pfizer booster seems virtually indistinguishable in the two groups. Â I do not believe this study's results support anything more than a Pfizer booster worked equally well in both cases.