This will depend on the distribution of the outcome, the test you will perform, and the hypothesis being tested. It's not a straightforward answer.

What I recommend is for you to intuit your way by considering a hypothetical observed result congruent with what you hope to see and performing the hypothesis test. Does the test based on the hypothetical data reject the research null hypothesis? If not, you may want to consider a bigger sample size for the actual study. You may want to repeat this exercise with different hypothetical data sets that are less optimistic, and with some that are congruent with the research null hypothesis.

This isn't a formal power or sample size calculation, but it gives you a more hands-on feel of what you need.

The program G*power is freely available and is designed to answer questions such as this via a "power analysis."

For a study comparing outcomes of the same individuals at different times (paired or repeated measures design), you would typically use a paired t-test or a related statistical test. The sample size determination in this context is based on the paired differences between the two-time points.

The following steps outline the approach to determine the sample size for such a study:

Keep in mind that the above formula assumes a normal distribution of the differences and that the paired differences are independent and identically distributed. Also, consider potential non-response or missing data and adjust your sample size accordingly.

Finally, since you're dealing with medical data involving deceased individuals (postmortem findings), ethical considerations are crucial. Make sure to adhere to all ethical guidelines and obtain necessary permissions when conducting such research.

Overall, it's a good practice to consult with a biostatistician or a statistical expert to ensure