TY - CHAP
T1 - [4] Methods for validating deconvolution analysis of pulsatile hormone release
T2 - Luteinizing hormone as a paradigm
AU - Mulligan, Thomas
AU - Johnson, Michael L.
AU - Veldhuis, Johannes D.
N1 - Funding Information:
The authors thank Drs. William F. Crowley for the human GnRH-stimulation series, Iain Clarke for the sheep data, and Ernst Knobil for the monkey data. This work was supported in part by the U.S. Department of Veterans Affairs (T.M.), NIH RCDA 1 KO4 HD00634 from NICHHD (J.D.V.), the National Science Foundation Science and Technology Center for Biological Timing (J.D.V., M.L.J.), and the Diabetes and Endocrine Research Center NIDDK D.K.-38942 (J.D.V., M.L.J.).
PY - 1995
Y1 - 1995
N2 - Deconvolution analysis is an important analytical tool which can provide valuable information regarding pulsatile hormone secretion. However, as with all analytical tools, it must be properly validated for the application under consideration. To maximize sensitivity and specificity, the investigator should obtain representative samples as frequently and as for as long a duration as practical, use a precise and specific assay, and apply a validated algorithm with statistical confidence limits chosen a priori. In addition, the investigator must be aware of the limitations inherent in deconvolution analysis when the amplitude of the secretory pulse is small, the interpulse interval is short, the sampling frequency is low, the amount of experimental uncertainty (noise) is large, and/or the assumptions of the model -xb(3) are violated. We emphasize that no analytical tool can perform in a valid manner under conditions that violate the assumptions that underlie it. For example, a pure burst model of hormone secretion cannot be expected to estimate pulse amplitude and mass accurately if the data to be analyzed consist of predominantly nonpulsatile hormone release. Thus, the model must be shown to be relevant to the data. The data must be collected optimally (e.g., at adequate sampling frequency) and derived by exemplary measurement techniques (high assay precision, sensitivity, and specificity). Finally, optimal analytical parameters should be documented as illustrated here to permit high test performance sensitivity and specificity as well as accurate recovery of estimated half-life and production/secretion rates.
AB - Deconvolution analysis is an important analytical tool which can provide valuable information regarding pulsatile hormone secretion. However, as with all analytical tools, it must be properly validated for the application under consideration. To maximize sensitivity and specificity, the investigator should obtain representative samples as frequently and as for as long a duration as practical, use a precise and specific assay, and apply a validated algorithm with statistical confidence limits chosen a priori. In addition, the investigator must be aware of the limitations inherent in deconvolution analysis when the amplitude of the secretory pulse is small, the interpulse interval is short, the sampling frequency is low, the amount of experimental uncertainty (noise) is large, and/or the assumptions of the model -xb(3) are violated. We emphasize that no analytical tool can perform in a valid manner under conditions that violate the assumptions that underlie it. For example, a pure burst model of hormone secretion cannot be expected to estimate pulse amplitude and mass accurately if the data to be analyzed consist of predominantly nonpulsatile hormone release. Thus, the model must be shown to be relevant to the data. The data must be collected optimally (e.g., at adequate sampling frequency) and derived by exemplary measurement techniques (high assay precision, sensitivity, and specificity). Finally, optimal analytical parameters should be documented as illustrated here to permit high test performance sensitivity and specificity as well as accurate recovery of estimated half-life and production/secretion rates.
UR - http://www.scopus.com/inward/record.url?scp=0008821268&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=0008821268&partnerID=8YFLogxK
U2 - 10.1016/S1043-9471(06)80030-3
DO - 10.1016/S1043-9471(06)80030-3
M3 - Chapter
AN - SCOPUS:0008821268
T3 - Methods in Neurosciences
SP - 109
EP - 129
BT - Methods in Neurosciences
PB - Academic Press Inc
ER -