Y'all blend together. Can't quite keep you straight when I keep hearing the same argument over and over.
Since you apparently haven't had the data versus anecdote things explained directly to you:
What matters here is the rate. How many people per X (for example 100,000) have a certain reaction or outcome that is "harmful" or dangerous, especially with demographic controls (age, sex, other stuff).
For the worst outcomes (like death), all cases are likely to be reported. We can then compare the deaths (say within 2 weeks of injection) per 100,000 people injected and by demographic group compare to the regular death rate for the same demographic group people who *weren't* injected every two weeks. (These statistics already exists, so the "deaths within 2 weeks of vaccination" group can be compared.) Then see if the difference is statistically significant.
For less drastic issues, it may require a survey made by the vaccinating clinic in a follow up: "Have you had any unusual medical symptoms since your vaccinations? Did you experience X? Y? Z?" If need be surveys of non-vaccinated people can be used as a control.
The various anecdotes that have been mentioned in this thread (and a couple others) and have often been rapidly dismissed are not necessarily invalid nor are they "excluded" from data sets, they just need to be collected in a systematic fashion to be data with adequate controls to prevent selection biases, etc. Data isn't just a collection of measurements, but a carefully selected set of measurements such that it is clear what is and is not in the data.