statistically valid results using flight crews will be a daunting problem.1 In general, very small sample sizes make it impossible to state either quality control or research findings with reasonable confidence intervals or to compare alternatives using tests of statistical significance. The committee recognizes that health-related studies based on observations of space mission crews will, for the foreseeable future, suffer from small sample size. Consequently, inferences based on single missions will have inadequate statistical power unless, in the context of reliability analysis, the problem under study is so prevalent that it is detected in the first few subjects (Virzi, 1992; Lewis, 1994). Methods are available to address this problem, including the pooling of data from multiple studies or missions in the manner of sequential clinical trials (IOM, 2001) and Bayesian sequential trials. The committee proposes that rather than rely on data from a single mission for inference, NASA analyze data pooled from several missions. More specifically, the committee proposes that studies be designed to incorporate as many missions as possible, somewhat in the manner of sequential clinical trials, and also that they incorporate prior information from archival data and ground-based studies to the extent practicable. In a Bayesian framework, a prior uncertainty distribution for extent of bone mass loss as a function of age, gender, and time in space, for example, would be incrementally modified by new information gained from—and incidental to—a series of missions. The goal would be to develop a sequence of posterior distributions about the

1  

By way of example, managers of NASA’s bioastronautics program have tried to determine whether sufficient astronaut subjects will be available to conduct the forward research needed to successfully mitigate the risks imposed by the Design Reference Missions. A workshop was convened in Houston, Texas, in May 2004 to estimate the requirements for human subjects in exploration research. Representatives of the extramural and intramural research communities, flight surgeons, astronauts, bioastronautics management, and ISS payload planners attempted to derive realistic guesses of the minimum number of subjects that would be needed to mitigate each BR risk to an acceptable level. Without a priori calculation of statistical power, they concluded that at a minimum, 1,025 data points would be required to complete the logistical tasks associated with the current BR, with the majority (71%) derived from in-flight experiments. In the best case (i.e., if each ISS crew member participated in six experiments, the experiments did not interact, and they were performed without loss of data), at least 120 astronaut subjects would be required, nearly double the total number of ISS crew members expected for the rest of its useful life. (This assumes a three-person crew rotating every six months; this estimate and the utility of the resultant data would be further affected by changes in the size of the crew or the length of their sortie.)



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement