This is binomial distribution. You should start with assuming that 40 % and 15 % (if the drug is not effective, the prob. should still be 15 %) respectively are actually the real probabilities, and from there create 90 % single sided confidence intervals. If the observed cases (the counted numbers, 7 and 15 resp.) are outside of the intervals, you can consider your initial assumption to be false. 1) and 2) are solved separately of course.