Check page 39 of the document.

The second column of tables 19 and 20 are what you are looking for. These tables account for occurrences of C19 any time after the first dose.

Table 19's number count is 1178.

Table 20's number count is 1835.

Together they add up to 3013 total children.

Here is an additional sub-stack article that explains what the FDA and CDC did with the data.

Quoting from the article:

of the 375 Sars-Cov-2 infections in the trial, 365 occurred before the third dose. Only 10 occurred after the third dose.

Yes, you are reading that right. The efficacy figure is based on 3 PERCENT OF ALL THE INFECTIONS IN THE TRIAL. (Which is why the confidence intervals are so large.)

I think what this means is that they threw out 3000 case studies that didn't align with their efficacy goal. They narrowed their criteria in such a way as to guarantee that their data was skewed overwhelmingly towards approval of these vaccines as highly effective.

Don't rely on searching. Read the document.

Page 16

4.2. Phase 2/3 4.2.1. Phase 2/3 study design In Phase 2/3, a total of 4,526 participants 6 months through 4 years of age were randomized 2:1 to receive two doses of 3 µg BNT162b2 or placebo, 3 weeks apart.>

After reading through the 9 results tables, I think that the problem is the results were divided by age. Table 6 is for ages 6 - 23 months and table 8 for ages 2 - 4 years. The totals for those add up to the original 4526 and account for participant loss.

Not saying the study or the approval is legit, just this concern about losing 3000 subjects may not be legit.

4,526 is at least part of your answer.