Clinical versus Statistical Significance in Medical Imaging

• Activity Number: 467
• Type: Contributed
• Date/Time: Wednesday, August 5, 2009 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Thomas F. Heston*+ and Richard L. Wahl
• Companies: Johns Hopkins University and Johns Hopkins University
• Keywords: medical imaging ; health care ; clinical significance ; statistical significance ; medicine ; imaging

Abstract

OBJECTIVE: How often do statistically significant results achieve clinical significance? METHODS: Medical imaging research studies with a statistically significant difference were evaluated. The crossover point (CP), defined as the point below which a patient would more likely than not to belong to group 1, and above which would more likely than not belong to group 2 was determined. CP = (SD1*X2+SD2*X1)/(SD1+SD2), X = mean and SD = standard deviation. RESULTS: Of the 49 statistically significant results evaluated, the average CP was X1 + 0.559 SD1 = X2 - 0.559 SD2. In 88%, this CP was < 1 SD from the mean. CONCLUSION: In medical imaging research, statistical significance may not always be clinically applicable. A research determination that two patient samples are statistically different does not mean that current imaging technology can correctly categorize the individual patient.