David's New Book

Wednesday, July 20, 2011

Parachutes: Early vs. Late Endpoints

George Drusano, Paul Ambrose and Brad Spellberg - Special Guest Bloggers.


“It’s not the fall that kills you, it’s the sudden stop.”--Chuck Yeager

A recent news story described a skydiver who jumped out of an airplane at 15,000 feet http://www.grindtv.com/outdoor/blog/28817/skydivers+15000-foot+plunge+to+earth+an+amazing+survival+story/.  He had 2 parachutes (primary and backup).  Both failed to deploy.  Nevertheless, after landing in some bushes he walked away with 2 broken bones and some bruises.
Smith and Pell previously called into question the efficacy of parachutes in a thoughtful analysis of the existing literature at the time.1  Their conclusions were shocking.  It turned out, no one had ever conducted randomized, placebo controlled trials of parachutes.  Furthermore, the current news story is only the latest of many describing people who have jumped from “lethal” heights without parachute support but survived with only minor injuries.1  Such scientific descriptions have called into doubt one hundred years of “anecdotal” successful experience with parachutes.
Because of the absence of historical evidence of sensitivity to device effect (HESDE) for parachutes versus placebo/background therapy, the conduct of non-inferiority trials of parachutes must be reconsidered.  While parachutes do not seem to improve end-of-therapy survival based on available published literature, they do appear to provide early benefit to the jumper.  Analyses from the late bronze age2-3 showed that, compared to cliff jumpers using background therapy (e.g., flapping arms real fast), early after the jump those with functioning parachutes had superior mood scores and improved physiology (including lower heart rate, lower blood pressure, and cleaner underwear).
Thus, the primary endpoint for parachute non-inferiority trials should not occur at end of therapy (i.e., at impact) since there is no HESDE for that endpoint.  Rather, based on these relevant and timely bronze-aged studies, the primary endpoint should occur at 48 to 72 seconds after jumping, since that’s when screaming and bodily functions of parachuted vs. control cliff-jumpers differed the most.2,3  Specifically, the primary endpoint should include a composite of: 1) loudness of scream; 2) proportion of words screamed with 4 letters in them; and 3) change in weight of underwear in flight.  As for all endpoints and irrespective of data, the non-inferiority margin for such a scientifically rigorous endpoint should be 10%, because we said so.
Other, less scientifically rigorous critics have felt that it was important in some vague, abstract way for the jumper to be physiologically normal when the jump was completed, not just happier and cleaner early during the jump.  Thus, we are willing to concede that less important, secondary endpoints that are not used to define success of therapy could be analyzed at end of therapy.  Such exploratory endpoints could include the volume of the “thud” or bone snapping that occur at the time of ground-contact, or the magnitude of post ground-contact shrieks.  Unresolved are what type of performance characteristics such endpoints would have.  For example, can physician observers accurately describe the intensity of such thudding, snapping, or shrieks simply by listening, or does a specific device need to be employed?  Should such measurements be made in decibels, killidecibels, or another measure?  How many inches of extrusion from the surface of the skin must a compound fracture extend for it to be considered meaningful to jumpers post ground-contact?
Furthermore, since the distance splintered bone rises above the skin and the volume of thudding, crackling, and shrieks are just biomarkers and not symptoms, they may have no relevance to the jumpers.  Rather, it may be necessary to use a validated Patient Reported Outcome (PRO) instrument so jumpers can directly document the symptoms they are feeling post ground-contact.  Presumably voice recognition software could be used to capture the PRO results, since jumpers with less effective parachutes may fracture key bones in their limbs post ground-contact, and thus be unable to write on PRO forms.  Alternatively, various sorts of rudimentary scratch marks and other primitive symbols that are feasible to generate with fractured hand bones may convey the patient’s symptoms accurately on a well designed PRO.
A final, critically important point.  The jumper in the current news story had 2 parachutes, a primary and a backup.  Such a jumper should be ineligible to enroll in future non-inferiority studies because the effect of the second parachute could potentially mask the lack of efficacy of the first parachute.  Not even a single additional parachute should be allowed for jumpers in future trials.  Sadly, this means that the authors will not be eligible to enroll (BS has indicated he will jump only with a minimum of 5 parachutes in place, for example).  However, we are confident that those truly dedicate to regulatory science will lead by example by enrolling in such studies and volunteering to wear only 1 parachute.
We believe that the science of designing and analyzing parachute non-inferiority clinical trials has been greatly advanced in recent years.  While initially skeptical that non-inferiority trials could be justified based on available science, we now believe that early endpoints focusing on patient reported symptoms provide a “scientifically rigorous,” “compromise” pathway forward.

References
1.     Gordon Smith and Jill Pell.  2003.  Parachute use to prevent death and major trauma related to gravitational challenge: systematic review of randomised controlled trials.  Br Med J.  327:1459-61.

2.     Hippocrates,* George Drusano,* Paul Ambrose, and Brad Spellberg.  947 B.C.  The impact of parachutes versus traditional therapy on how cliff jumpers “feel, function, and survive”: a Bayesian analysis of early versus late endpoints.  Journal of Ha, We Made You Look!
*these authors contributed equally to this work

3.     Asklepios, Paul Ambrose, George Drusano, and Brad Spellberg.  1106 B.C.  Really, after reading reference 2 you still went on to 3?  The Punked Journal.