The main goal of this project is to develop and apply methods using finite-mixture statistical models to provide a patient-specific estimate of the probability that the patient has responded to treatment. The estimated probability of response will provide an empirically grounded and conceptually sound statistical technique for monitoring success of treatment for osteoporosis. The methods also can be used for basic research on patient factors associated with responsiveness to treatment. Since phenotypes defined in terms of response to treatment can be related to genetic markers, these methods are applicable to the emerging field of pharmacogenomics. This project will extend our recently published methodological work [1], in which we focused on methods and application to the use of a patient's pre to post treatment change in total hip bone mineral density (BMD) for judging whether or not the patient has responded to treatment with alendronate. We calibrated the procedure with data from FIT [2], a randomized placebo-controlled trial, which evaluated alendronate for treatment of osteoporosis. We found that clinicians may be overly ready to conclude that treatment with alendronate has failed. For example, when a treated subject has no increase in total hip BMD from baseline, there is only a small probability that a treated patient has actually failed to respond to treatment. Although different from current clinical opinion, this conclusion results from proper consideration of background changes in placebo-treated control subjects as a backdrop for judging the changes in treated subjects. To make the previously published methods more versatile, the proposed developments will: (1) deal with many of the different configurations of patient monitoring data that are typically encountered in clinical practice, and (2) can take patient-specific characteristics like age, race, BMI, baseline BMD, genetic markers, etc. into consideration. Application of the methods to FIT, MORE [3] and other data will provide new substantive results that will: (1) contribute to useful clinical guidelines for judging how well an individual patient is responding to osteoporosis treatment, and (2) provide basic empirical evidence about responsiveness to treatment. We expect that the results from this project will be of both methodological and clinical interest so articles appropriate for both biostatistical and substantive journals will be prepared to present the findings.
| Lowe, R M; Genin, A; Orgun, N et al. (2014) IL-15 prolongs CD154 expression on human CD4 T cells via STAT5 binding to the CD154 transcriptional promoter. Genes Immun 15:137-44 |
| Lin, Haiqun; McCulloch, Charles E; Rosenheck, Robert A (2004) Latent pattern mixture models for informative intermittent missing data in longitudinal studies. Biometrics 60:295-305 |
