Training Program in Neurostatistics and Neuroepidemiology



Current Fellows (2012-2013)

Ritesh Ramchandani is a third year student (dissertation advisors: Dianne Finkelstein/David Schoenfeld) in the Biostatistics doctoral program. Currently he is working on two projects. One is a rank test for right-censored data using intermediate states as auxiliary information: they are modifying rank tests for censored data to take into account auxiliary information on intermediate states that subjects may pass through before failure. Using this additional information can improve power that is lost due to censoring. The method uses existing methods for modelling state-space processes and has been implemented using R. Some preliminary analyses and simulations have shown that power can be improved over existing rank tests without sacrificing type 1 error. Further simulations need to be completed to assess the method for type 1 error and power in a variety of settings. In addition, potential bias arising from unequal censoring distributions,and censoring distributions that depend on state occupied is being explored.

The second project is on chronic myeloid leukemia: A well-established mathematical theory on drug resistance, when combined with the analysis of clinical data on chronic myeloid leukemia patients treated with TKIs, provides evidence that point mutations responsible for resistance to targeted therapies are likely pre-dating the start of the treatment. Thus, a key factor in the possible development of TKI resistance is given by the size of the leukemic burden found in a patient at the time of clinical detection of the disease. They show how the statistical analysis of a group of CML patients treated at the M.D. Anderson Cancer Center confirms this prediction. Moreover, estimates for the reduction in the occurrence of relapses if the therapy were to be started at some earlier given times are provided.

Tomasetti C, Ramchandani R, Jabbour E, Quintas-Cardama A, Kantarjian H, Parmigiani G, Cortes J. (2012) The average baseline BCR-ABL levels are significantly higher in patients with resistance to dasatinib as first-line treatment for early chronic phase chronic myeloid leukemia. Submitted to Leukemia and Lymphoma.

David Swanson is a fourth year student (dissertation advisors: Christoph Lange/Rebecca Betensky) in the Biostatistics doctoral program. He has completed all his coursework and is working on his thesis. Currently he is working on the following three projects:

  1. Gene-based testing project: We have shown theoretically and through simulation weaknesses of currently implemented gene-based tests. We propose two alternative gene-based tests, one of which only makes use of summary statistics from already-conducted analyses, thereby making access to the original data unnecessary. The other alternative test reduces the dimension of the data, which, under certain situations, results in power gains. This latter test can be thought of as a score test alternative to a previously-proposed dimension-reduction gene test approach.
  2. Survival bias project: We provide a framework for thinking about prevalence-incidence bias, sometimes present in epidemiology studies, and obtain formulas for the observed, biased odds ratio. We describe situations and give examples of when the bias may be a source of concern in studies. Lastly, we propose three different hypothesis tests for detection of prevalence-incidence bias, all of which make minimal use of external data sources.
  3. Research participation / Dependent censoring project: We show through simulation how variation in payment structure may affect the bias of the Kaplan-Meier survival curve estimated in clinical trial time-to-event analyses. An inverse-probability weighted Kaplan-Meier estimator is advocated for when dependent censoring may be present in the data. This estimator can be used as a means of sensitivity analysis to underlying assumptions of the standard survival curve estimator.
Swanson D, Blacker D, AlChawa T, Ludwig K, Mangold E, Lange C. "Gene-based association tests using dimension reduction methods and marker correlation structure", Biostatistics, submitted for publication.

Swanson D, Betensky RA. "Testing for prevalence-incidence bias in case-control studies", Biometrika, to be submitted.

Swanson D, Betensky RA. "Research participant compensation: a matter of inference as well as ethics", Annals of Internal Medicine, to be submitted.

Amin S, Yip W, Minvielle S, Li Y, Hanlon B, Swanson D, Shah P, Moreau P, van der Holt B, van Duin M, Broyl A, Magrangeas F, Sonneveld P, Anderson K, Li C, Avet-Loiseau H, Munshi N. "Gene expression profile alone Is inadequate in predicting response In multiple myeloma", Journal of Clinical Oncology, accepted for publication.

Folefac Atem is a new postdoctoral research fellow who will be working on collaborative research under the direction of Dr. Rebecca Betensky and subject matter mentors. His projects include:

  1. Statistical methodology on analyzing censored covariates: Neurologists are often faced with the problem of randomly censored covariates, for instance predicting offspring dementia status using parent status (subject to censoring). The simplest and most straightforward approach for dealing with such data is to remove variables with censored observations or delete all censored observations. The former leads to model misspecification while the latter leads to overestimation of standard error due to a loss in power. We proposed a two stage approach; the first phase involves estimating the survival probability for each individual and in the second phase we used these survival probabilities to impute the censored covariate using multiple imputations. This approach produces unbiased estimates with a great coverage probability.
  2. Variable selection in very high dimensional problems: For this project, we will try to find genes that are highly correlated with brain dementia in animals.

Nicte Mejia is also a new postdoctoral research fellow who is working on collaborative research under the direction of Dr. Rebecca Betensky and Dr. Lee H. Schwamm and subject matter mentors. Her current projects include the following:

  1. Racial/Ethnic Disparities in Parkinson Disease Patient Care in the United States: 1996-2009 Medical Expenditure Panel Survey: In collaboration with Benjamin Cook, Ph.D. We are data mining the Medical Expenditure Panel Survey to learn about possible racial and social disparities in Parkinson Disease Care in the United States. Initial hypotheses have been tested and an abstract submitted to the American Academy of Neurology was accepted for 3/2013 Platform Presentation. Manuscript in preparation.
  2. Health Literacy and Parkinson disease care: This pilot study in which I mentored a second year medical student evaluates health literacy among a diverse group of Parkinson disease patients seen at Cambridge Hospital. Data obtained, currently pursuing data analysis and manuscript preparation.
Wahlster S, Berkowitz A, Zepeda R, Sagar V, Lyons J, Klein J, Mejia NI (2013). Neurology and Global Health: Initiatives led by Residents and Staff from the Partners Neurology Program. American Academy of Neurology 65th Annual Meeting, San Diego, CA.