The problem here is that we have a condition ("jumping out of aeroplanes") where, without an intervention ("give the patient a parachute") the consequences are almost certainly fatal (yes, a handful of people have survived falls from altitude without parachutes: but a fatal outcome is sufficiently certain that defenestration has historically been used as a means of execution).

It's a major ethics no-no to expose healthy patients to potentially-fatal environments. It's also a major ethics no-no to withhold known life-saving treatment in the name of continuing a trial. So we can't contrive a proper randomized double-blind placebo-controlled test of the life-saving efficacy of parachutes, where we push people out of planes repeatedly to see how many survive without a parachute vs. with a parachute.

We can't test parachutes any more than we can contrive an evidence-based test of the efficacy of being-rescued-from-burning-buildings-by-firefighters, by arranging for a city's fire department to randomly not rescue half the victims of house fires, so that we can subsequently compare their survival rate with the survival rate of those who were rescued.

Basically, evidence-based assessment of medical treatments breaks down when it hits the edge condition defined by the patients automatically dying if not treated. At best we can compare different types of parachute, or different types of emergency tratment, and see which has a lower associated mortality rate. But we can't, ethically, compare treatment/no-treatment if there's a high likelihood of no-treatment resulting in death.