PREVALENCE AND SIGNIFICANCE OF SPINAL DISC ABNORMALITIES IN AN ASYMPTOMATIC ACCELERATION SUBJECT PANEL

Background: A protocol to allow for human centrifuge exposures up to 12 Gz (12 times gravity) required a screening spinal MRI. MRI-derived spinal disc abnormalities were observed in three of the first four asy mptomatic volunteer subjects. The protocol was interrupted and a secon d study was initiated to determine the possible cause and effect relat ionship between the disc findings and previous +Gz exposure. Methods: A T1 or T2 weighted sagittal MRI of the entire spine was accomplished on each of 22 asymptomatic male acceleration panel members, and a simi lar, age-matched control panel of 19 asymptomatic male subjects with n o history of previous acceleration exposure. The MRIs from all 41 subj ects were read at 2 diagnostic facilities by 9 radiologists. The evalu ating radiologists were aware asymptomatic centrifuge subjects were be ing evaluated but were unaware a control group was included. Results: Initial results from any one reader revealed spinal disc abnormalities (bulging, degeneration or herniated nucleus pulposus-HNP) in 91% of t he centrifuge panel and 79% of the control group, a non-significant di fference. Within-reader and between-reader variability was very high. Comparison of 1st vs. 2nd reading of the same data by one radiologist demonstrated a 28% agreement and a 72% disagreement on observed abnorm alities. Comparison of the same MRIs read by two different radiologist s revealed a 23% agreement and a 77% disagreement, pointing out the am biguity of the data and subjectiveness of the interpretation. Two addi tional neuroradiologists agreed to independently read all 41 MRIs afte r establishing unique reading criteria. There remained a non-significa nt difference between the two subject groups, whereas reader disagreem ent was still high (56%). Conclusions: No significant difference was f ound between the two subject groups. The power of the test was low bec ause of the small sample size. Our confidence in the interpretation is low because of the high degree of between reader and within-reader va riability.