Biomarker classification derived from finite growth mixture modeling with a time-varying covariate: an example with phosphorus and glomerular filtration rate |
| |
Authors: | Sterling McPherson Celestina Barbosa-Leiker |
| |
Institution: | Washington State University, College of Nursing, PO Box 1495, SNRS 314C, Spokane, WA 99210, USA |
| |
Abstract: | Finite growth mixture modeling may prove extremely useful for identifying initial pharmacotherapeutic targets for clinical intervention purposes in chronic kidney disease. The primary goal of this research is to demonstrate and describe the process of identifying a longitudinal classification scheme to guide timing and dose of treatment in future randomized clinical trials. After discussing the statistical architecture, we describe the model selection and fit criteria in detail before choosing and selecting our final 4-class solution (BIC = 1612.577, BLRT of p < .001). The first class (highly elevated group) had an average starting point of 3.969?mg/dl of phosphorus at Visit 1, and increased 0.143 every two years until Visit 4. The second, elevated class had an average starting point of 3.460?mg/dl of phosphorus at Visit 1, and increased 0.101 every two years until Visit 4. The normative class had an average starting point of 3.019?mg/dl of phosphorus at Visit 1, and increased 0.099 every two years until Visit 4. Lastly, the low class had an average starting point of 2.525?mg/dl of phosphorus at Visit 1, and increased 0.158 every two years until Visit 4. We hope that this example will spur future applications in biomedical sciences in order to refine therapeutic targets and/or construct long-term risk categories. |
| |
Keywords: | finite growth mixture modeling time-varying covariate chronic kidney disease biomarker classification latent growth curve modeling |
|
|