The confidence bands of my survival curve don't extend to the last time point.

Prism only recomputes the confidence bands (error envelope) at the time of each death. So if your last few subjects are censored, Prism will not continue the confidence bands past the last death.

We fixed this problem in Prism 5.02 and 5.0b.

With a bit of cleverness, you can bypass the problem.

1. Go to the results table with the percent survivals.
2. Select the entire table.
3. Copy to the clipboard
4. Make a new XY table formatted the same way.
5. Paste the data.
6. Add error values to the last row, that match the bottom-most error values computed by Prism. Simply copy the last pair of error values computed by Prism, and paste into the last row in the table.
7. Go to the linked graph.
8. Fuss with the formatting to draw a staircase connecting line and plot the error 'bars' using the "-----" format (to create confidence bands).