Department of Biostatistics
Neurostatistics Working Group

2015 - 2016

Coordinators: Dr. Rebecca Betensky, and Dr. Catherine Lee

Schedule: Wednesdays, 12:30-1:30 p.m.
SPH2, Room 426 (unless otherwise notified)

Contract All | Expand All
Seminar Description
This working group provides a forum for presentation and discussion of completed, ongoing, or planned statistical analyses of neurological data. Such data include, for example, in vivo human brain images (anatomic, functional and spectroscopic magnetic resonance imaging), gene expression studies of human and non-human animal brain tissue (brightfield and immunofluorescence microscopy, DNA microarrays, laser micro-dissection), in vivo micro-dialysis, clinical trials data for a variety of neurologic diseases, and genetic data from family studies. Non-statistical presentations of neurological, psychiatric and technological background material will also be included. Through this seminar, statisticians will gain exposure to the statistical issues that arise in the broad field of neurology and brain imaging psychiatry and to the diverse ongoing research in this area throughout Harvard and the world. A main goal of the seminar is to stimulate statistical interest in neuroscience and neurology and to develop strategies for collaboration within these fields.

September 16

Jessica Gronsbell
Doctoral Student, Department of Biostatistics, Harvard University

"NIH F31 Fellowship Application Process"
ABSTRACT: The Ruth L. Kirschstein NRSA for Individual Predoctoral Fellows (F31) Award supports promising doctoral candidates who will perform dissertation research and training for a PhD degree in a scientific health-related field relevant to the missions of the NIH during the tenure of the award. In this talk, I will provide an overview of the application process. First, second, and third year students are encouraged to attend.

Catherine Lee, Ph.D.
Doctoral Student, Department of Biostatistics, Harvard University

"Methods for Analyzing Left-truncated Time-to-event Data with Time-varying Biomarkers Measured Only at Study Entry, with Applications to Alzheimer's Disease"
ABSTRACT: Over the past decade, several biomarkers for Alzheimer's disease (AD) have been identified and well validated. Efforts are currently being made to understand how individual biomarker levels change with time and the order of biomarker changes over time relative to each other. Such time-dependent biomarker models would allow for more accurate disease staging and could be used to predict disease progression. However, analyses of such AD biomarker data present two challenges: time-to-event analyses necessitate the selection of a time origin and the Cox hazard regression requires observation of the covariate process at all failure times. The choice of time origin directly affects the treatment of time-varying predictors in the analyses.

We examine the choice of time origin and implication on analysis and provide methods for relating a sigmoidal time-varying predictor of AD to time-to-event in the setting of delayed entry when the time-varying predictor process is only measured at study entry. We provide an analytic derivation of the bias that occurs when an incorrect time origin is used in a Cox model with time-varying predictor or when a time-varying predictor measured at study entry is incorrectly treated as fixed. These analytical results are supported through a simulation study. Finally, our methods are applied to an Alzheimer's dataset. Note: I will be practicing a 15 minute talk that I will be giving at a Alzheimer's conference in mid-September.

September 30

Sy Han Steven Chiou, Ph.D.
Research Fellow, Department of Biostatistics, Harvard University

"Joint Scale-change Models for Recurrent Events and Failure Time"
ABSTRACT: Recurrent event data arise frequently in various fields such as biomedical sciences, public health, engineering, and social sciences. In many instances, the observation of the recurrent event process can be stopped by the occurrence of a correlated failure event and/or dependent censoring. In this talk, I will present a joint scale-change model for the recurrent event process and the failure time that allows the censoring time to be informative about the recurrent event process. In particular, a shared frailty variable is used to model the association between the two types of outcomes. In contrast to the popular Cox-type joint modeling approaches, the regression parameters in the proposed joint scale-change model have marginal interpretations. Moreover, the proposed approach is robust in the sense that no parametric assumption is imposed on the distribution of the unobserved frailty and the strong Poisson-type assumption is not needed for the recurrent event process. The consistency and asymptotic normality of the proposed semiparametric estimators are established under suitable regularity conditions. To estimate the corresponding variances of the estimators, a computationally efficient resampling-based procedure is proposed. Simulation studies and an analysis of hospitalization data from the Danish Psychiatric Central Register illustrate the performance of the proposed method.

October 7

Ryan Seals, Sc.D.
Research Fellow, Departments of Biostatistics and Epidemiology, Harvard T.H. Chan School of Public Health

"Amyotrophic Lateral Sclerosis: Trends and Risk Factors"
ABSTRACT: Amyotrophic lateral sclerosis (ALS) is a progressive debilitating disease of the upper and lower motor neurons. There are few well-established risk factors for ALS, and there is conflicting evidence regarding the trends in ALS incidence and mortality over the past several decades. In this talk I will present the results of an age-period-cohort analysis of trends in ALS in Denmark over the past several decades, and the results of two Danish national registry studies: 1) of physical trauma and ALS; and 2) of military occupation and ALS.

October 21

Jessica Gronsbell
Doctoral Student, Department of Biostatistics, Harvard University

"Semi-Supervised Approaches to Efficient Evaluation of Model Prediction Performance"
ABSTRACT: We consider the evaluation of a binary classifier in a semi-supervised setting in which a small or moderate sized `labeled' dataset is accompanied by a large amount of `unlabeled' data. This setting is directly relevant to many practical applications where the outcome is expensive or time-consuming to collect, but information on the predictors is readily available. Such data is increasingly prevalent with the rise of electronically recorded databases such as electronic medical records (EMR). While supervised estimation procedures make use of only labeled data, it is often of interest whether unlabeled data can improve estimation efficiency. In the context of evaluating risk prediction models, we propose semi-supervised (SS) estimators of various prediction performance measures. We validate our proposals via extensive simulation studies as well as a real data analysis of EMR studies of rheumatoid arthritis and multiple sclerosis.

October 28, MGH (50 Staniford St., Suite 560, Rm 560F)

Harvard Catalyst Biostatistics Journal Club led by Joseph Locascio, Ph.D.
Assistant Professor of Neurology, Harvard Medical School

"The Null Hypothesis Significance Test (NHST) Controversy"

Dr. Locascio will discuss "The Null Hypothesis Significance Test (NHST) Controversy" concerning the criticisms that statisticians have had over the years about NHST, especially the recent banning of the use of NHST in the journal Basic and Applied Social Psychology (BASP).

He will include the following articles:

"The Earth is Round (p<.05)" by Jacob Cohen (1994).

"What Scientific Idea is Ready for Retirement?" by Charles Seife (2015), Statistical Significance. In Brockman, J. (Ed.) This Idea Must Die: Scientific Theories That Are Blocking Progress. (pgs. 519-522). New York, NY: Harper Perennial. The Edge Foundation.

"Psychology and statistics, continued... Readers respond to the BASP ban on p-values"

David Trafimow & Michael Marks (2015) Editorial, Basic and Applied Social Psychology, 37:1, 1-2, DOI: 10.1080/01973533.2015.1012991Basic and Applied Social Psychology

Attached is a list of suggested additional readings.

Please contact Letizia Allais for call-in information.

November 4

Yakeel T. Quiroz-Gaviria, Ph.D.
Instructor in Psychology, Department of Psychiatry / Neuropsychologist, Massachusetts Alzheimer's Disease Research Center, Massachusetts General Hospital

"Relationships Between Baseline Biomarkers and Subsequent Cognitive Decline in Cognitively Unimpaired PSEN1 E280A Mutation Carriers from the Colombian Kindred with Autosomal Dominant Alzheimer's Disease"
ABSTRACT: Background: While brain imaging and cerebrospinal fluid (CSF) biomarkers have been used in the early detection and tracking of Alzheimer's disease (AD), their ability to predict subsequent clinical decline remains to be defined. In this study, we compared the ability of baseline PET amyloidβ(Aβ) and CSF measurements to predict subsequent cognitive decline in unimpaired Presenilin1 (PSEN1) E280A mutation carriers from the Colombian autosomal dominant AD (ADAD) kindred, up to almost 25 years before the kindred's estimated median age of 44 at the onset of mild cognitive impairment.

Methods: Thirty-seven cognitively unimpaired mutation carriers and non-carriers, aged 2044 years, were recruited from the Alzheimer's Prevention Initiative (API) Colombia Registry. Baseline cerebral-to-cerebellar florbetapir PET standard uptake value ratios (SUVRs) and CSF Aβ142, total tau and phosphotau181 levels were related to 2-3 year subsequent decline on the API preclinical ADAD composite cognitive test score, previously found to be associated with preclinical progression. The mixed random effect model was used to estimate the relationship between baseline measures and subsequent cognitive decline in the mutation carriers.

Results: In an independent replication, 2-3 year decline on the composite cognitive test score distinguished between carriers and non-carriers (p=0.03). In the carrier group, baseline florbetapir SUVRs and CSF ptau/Aβ142 ratios were associated with subsequent decline on the composite cognitive test score (p=0.008, 0.04, respectively). CSF Aβ142, total tau, and ptau levels alone were not (p= 0.19, 0.43 and 0.88, respectively), even after adjusting for age. Florbetapir SUVRs were slightly but not significantly better than CSF ptau/Aβ142 ratios (p=0.09), and better than CSF Aβ142, total tau, and ptau levels (p= 0.04, 0.04, and 0.06, respectively) in predicting subsequent cognitive decline.

Conclusions: Aβ PET and, to a lesser extent, CSF ptau/Aβ142 measurements may provide prognostic indicators of AD-related cognitive decline in ADAD mutation carriers. Indeed, Aβ PET may be a better prognostic indicator than CSF Aβ and tau levels in the group of mutation carriers assessed up to 25 years before their estimated age at clinical onset. Research is needed to further clarify the prognostic value of these biomarkers in cognitively unimpaired persons at risk for autosomal dominant and late-onset AD.

November 18

Shibani Mukerji, M.D., Ph.D.
Clinical Fellow in Neurology, Massachusetts General Hospital

"Lipid Profiles and APOE4 Allele Impact Midlife Cognitive Decline in HIV+ Men on ART"
ABSTRACT: Dyslipidemia and the ApolipoproteinE4 (APOε4) allele are recognized risk factors for cognitive decline in the general population, but how these risks are modified by HIV serostatus is unclear. This longitudinal study examined relationships between lipid profiles, APOε4 allele, and cognitive decline in a cohort of ART-treated HIV+ men over age 50. In a nested case-control study from the Multicenter AIDS Cohort Study (MACS) public data, 273 HIV+ men (50-65 years, without active heavy drug use, ≥2 visits with neurocognitive testing between 1996-2010, baseline VL<400 copies/ml, and continuous ART use ≥95% visits) were matched by baseline age, race, education, smoking, and alcohol use to 516 HIV- controls using R MatchIt. A composite measure of global cognition was created from 15 cognitive tests in 6 domains. The association between HIV serostatus, time-varying lipid markers (total cholesterol, LDL, HDL, and triglycerides), APOE genotype (n=344), and cognitive decline was examined using multivariable mixed effects models, adjusting for baseline age, CD4 count, IQ, CES-D, and smoking. The median baseline age of participants was 51 years (IQR: 50-54); 81% white, 89% had education >12 years; smoking (p=0.3) and alcohol use (p=0.07) were similar between HIV+ and HIV- men, respectively. HIV+ men had baseline median CD4 count of 514 cells/uL, and 70% were virally suppressed in study (<50 copies/ml with ≥2 blips, blip <400 copies/ml). Higher total cholesterol, LDL, and triglycerides, and lower HDL levels were associated with faster rate of cognitive decline in HIV+ men in mixed effects models (p<0.05). There was no significant relationship between HIV serostatus and baseline cognitive score, suggesting that cognitive decline diverged after study entry. In mixed models adjusted for the same terms, APOε4 allele was associated with accelerated cognitive decline in HIV+ men (p=0.01). In all models, CES-D ≥16 and older baseline age were associated with lower cognitive scores; IQ was associated with higher cognitive scores. Measures of lipid markers were associated with faster rates of cognitive decline among ART-treated HIV+ men over the age of 50. HIV+ APOε4 carriers are at substantial risk for accelerated midlife cognitive decline, and are predicted to decline at younger ages than HIV- carriers. These findings suggest pathways affecting lipid metabolism may accelerate cognitive aging among older HIV+ individuals.

December 2 (With Harvard Catalyst Journal Club)

Robert Glynn, SD
Professor of Biostatistics at Harvard T.H. Chan School of Public Health, and Professor of Medicine at Harvard Medical School

"Matching and Trimming with Propensity Scores"
Treatment Effects in the Presence of Unmeasured Confounding: Dealing With Observations in the Tails of the Propensity Score Distribution - A Simulation Study by Til Stürmer, Kenneth J. Rothman, Jerry Avorn, and Robert J. Glynn
Optimal full matching for survival outcomes: a method that merits more widespread use by Peter Austin and Elizabeth Stuart

December 9

Bernard Hanseeuw, M.D., Ph.D.
Research Fellow in Neurology, Massachusetts General Hospital

"Evidence that striatal amyloidosis is a marker of progression across the spectrum of Alzheimer's disease"
ABSTRACT: Background: Autopsy studies indicate that amyloidosis starts in neocortex and extends into striatum. Because amyloid (Aβ) is a risk factor for memory decline and Alzheimer's disease (AD), this research investigates in-vivo the association between Aβ and disease progression in patients with symptomatic AD and cognitively normal (CN) older adults participating in the Harvard Aging Brain study.

Methods: Baseline Aβ PET imaging (C11-PiB) and longitudinal clinical dementia rating (CDR-SB, median follow-up: four years) were analyzed in 52 patients with symptomatic AD (ages=50-85) and 284 CN older adults (ages=63-90). Memory was assessed annually in CN elderly and composite z-scores were derived from three challenging tests. The association between longitudinal PiB accumulation, memory decline, and tau deposition was investigated in a subset of CN older adults with multiple PiB observations and tau PET imaging (T807) at follow-up. Striatal (caudate and putamen) and neocortical PiB aggregates were compared, using cerebellar gray as reference region. We used linear mixed-effect models with random intercepts, covarying age, sex, E4 genotype, and time. Education was also adjusted for in the models predicting clinical impairment and memory decline.

Results: R2 between baseline striatal- and cortical was 0.76. In the spectrum ranging from normal aging to AD (n=334): At baseline, striatal (T=8.6, p=3*e-16) and cortical (T=8.8, p=8*e-17) PiB both related to clinical impairment (CDR-SB). Adjusting for cortical PiB (T=2.6, p=.009), striatal PiB still related to impairment (T=2.0, p=.047). Striatal (T=9.9, p=2*e-22) and cortical (T=8.9, p=2*e-18) PiB both predicted clinical decline (CDR-SB). Modeled as simultaneous predictors, baseline striatal PiB (T=4.3, p=2*e-5) better predicted increasing clinical impairment over time than baseline cortical PiB (T=0.6, p=.562).

In CN older adults (n=284): Both baseline striatal (T=-4.9, p=e-6) and cortical (T=-5.3, p=e-7) PiB predicted memory decline, but cortical PiB (T=-2.4, p=.018) was a better predictor than striatal PiB (T=-0.5, p=.597).

In CN older adults with longitudinal PIB (n=137): memory decline was associated with striatal (T=-3.4, p=7*e-4) and cortical (T=-3.3, p=.001) longitudinal PiB accumulation. Adjusting for baseline cortical PiB, longitudinal striatal PiB still related to memory decline (T=-1.9, p=.054).

In CN with T807 (n=100): Longitudinal PiB accumulation (in both the striatum and the cortex) was associated with higher levels of entorhinal and inferior temporal T807 at follow-up (p<3*e-4). Adjusting for baseline cortical PiB, longitudinal striatal PiB related to entorhinal (T=2.3, p=.023) and inferior temporal (T=1.8, p=.080) T807.

Conclusion: This data provides evidence that striatal Aβ accumulation relates to disease progression across the AD spectrum, and that the association is stronger than that of cortical Aβ, suggesting a temporal relationship more proximal to clinical decline. Additional time-points are required to confirm the sequential order of Aβ accumulation in neocortical and subcortical regions. These findings should be replicated using F18-PET tracers, before suggesting striatal Aβ as potential endpoint in ongoing anti-amyloid preventive clinical trials.

January 13 (11:00 am - 12:00 pm with Harvard Catalyst Journal Club)

Harvard Catalyst Biostatistics Journal Club led by Linda Valeri, Ph.D.
Assistant Biostatistician at McLean Hospital, and Instructor in Psychiatry at Harvard Medical School

"The Impact of Measurement Error in Causal Mediation Analysis: When Common 'Beliefs' Fail"

Dr. Valeri will discuss the following articles:

"The role of measurement error and misclassification in mediation analysis." by VanderWeele, T.J., Valeri, L., Ogburn, E.L. (2012).

"DNA methylation mediates the effect of maternal smoking during pregnancy on birthweight of the offspring" by Küpers, L. K., et al. (2015).

Please contact Letizia Allais for call-in information.

January 20

Scott Plotkin, M.D., Ph.D.
Director, Neurofibromatosis Clinic
Director, MGH/DFCI/BWH Neuro-Oncology Fellowship Program
Associate Professor of Neurology, Harvard Medical School

"Individualized Endpoint Analysis for Use in Clinical Trials of Neurofibromatosis"
ABSTRACT: None Given

January 27

Melissa Sands
Doctoral Student, Department of Government, Harvard University

"How Human Subjects Research Rules Mislead You and Your University, and What to Do About it"
ABSTRACT: Universities require faculty and students planning research involving human subjects to pass formal certification tests and then submit research plans for prior approval. Those who diligently take the tests may better understand certain important legal requirements but, at the same time, are often misled into thinking they can apply these rules to their own work which, in fact, they are not permitted to do. They will also be missing many other legal requirements not mentioned in their training but which govern their behaviors. Finally, the training leaves them likely to completely misunderstand the essentially political situation they find themselves in. The resulting risks to their universities, collaborators, and careers may be catastrophic, in addition to contributing to the more common ordinary frustrations of researchers with the system. To avoid these problems, faculty and students conducting research about and for the public need to understand that they are public figures, to whom different rules apply, ones that political scientists have long studied. University administrators (and faculty in their part-time roles as administrators) need to reorient their perspectives as well. University research compliance bureaucracies have grown, in well-meaning but sometimes unproductive ways that are not required by federal laws or guidelines. We offer advice to faculty and students for how to deal with the system as it exists now, and suggestions for changes in university research compliance bureaucracies, that should benefit faculty, students, staff, university budgets, and our research subjects.

February 10 (Harvard Catalyst Biostatistics Symposium)

11:00 am - 5 pm, Yawkey Conference Center, Dana Farber Cancer Institute

"Statistical Issues in the Study & Design of Cardiovascular Disease"
ABSTRACT: See flyer attached here.

February 11

Joel Salinas, M.D., M.B.A.
Clinical Fellow | Behavioral Neurology & Neuropsychiatry
Research Fellow | Schwamm Marriott Clinical Care Research Fellowship
Clinical Fellow in Neurology | Harvard Medical School
Massachusetts General Hospital

"Social and Behavioral Determinants of Neurologic Outcomes"
ABSTRACT: The effectiveness of amyloid-directed therapy for treatment of dementia or antidepressants in treating the psychiatric sequelae of neurologic disease appear limited, thus underscoring the importance of considering a more prominent role for disease prevention and non-pharmacological interventions. Accumulating evidence supports behavioral and psychosocial factors critically influencing various aspects of psychiatric and cardiovascular disease outcomes. While the burden of stroke, and to a lesser degree dementia, can be partially attributed to well-established biological risk factors, the risk and response to brain injury may also be modulated by differences in lifestyle, social influences, and cognitive reserve. We investigate this challenging hypothesis through large epidemiologic cohorts by examining the associations between social and behavioral factors with patient-centered neurologic disease outcomes, such as poststroke depression, as well as the interplay between their influence on neurologic disease risk and potential biological mechanisms underlying observed relationships.

February 17 (With Harvard Catalyst Journal Club)

Henry Feldman, Ph.D.
Associate Professor of Pediatrics, Harvard Medical School, and Principal Biostatistician for the Biostatistics Core in the BCH Clinical Research Program

"Covariate-Adjusted Survival Curves"
Zhang et al. A SAS macro for estimation of direct adjusted survival curves based on a stratified Cox regression model. Computer Methods and Programs in Biomedicine , Volume 88 , Issue 2 , 95 - 101.

Ghali WA, Quan H, Brant R, et al. Comparison of 2 Methods for Calculating Adjusted Survival Curves From Proportional Hazards Models. JAMA. 2001;286(12):1494-1497. doi:10.1001/jama.286.12.1494.

Please contact Letizia Allais for call-in information.

February 24

Jing Qian, Ph.D.
Assistant Professor, Department of Biostatistics, University of Massachusetts - Amherst

"Talk Title TBD"
ABSTRACT: None Given

March 2

Sy Han Steven Chiou, Ph.D.
Research Fellow, Department of Biostatistics, Harvard T.H. Chan School of Public Health

"Talk Title TBD"
ABSTRACT: None Given

March 9

Speaker TBA

"Talk Title TBD"
ABSTRACT: None Given

March 23

Daniel Nevo, Ph.D.
Postdoctoral Fellow, Departments of Biostatistics and Epidemiology, Harvard T.H. Chan School of Public Health

"The Competing Risks Model with Missing Cause of Failure and Auxiliary Case Covariates: An Application to Cancer Subtype Analysis"
ABSTRACT: A competing risks model is often used when in addition to the event time, its cause is also observed. However, in some studies, the event cause is unknown for some of the cases. The probability of a missing cause is typically assumed to be independent of the cause given the time of the event and covariates measured before the event occurred. When further assuming a proportional hazard model for the cause-specific hazard, an estimating equations approach derived by combining two separate likelihood functions (for the same data) can be used. In practice, however, the underlying missing-at-random assumption does not necessarily hold. Here we are motivated by colorectal cancer subtype analysis, where some tumors are more likely to result in a missing tumor tissue, given other tumor characteristics such as tumor location and grade. We develop a method to conduct valid analysis when additional auxiliary variables are measured for cases only. We consider a weaker missing-at-random assumption, when the missing pattern depends on the observed quantities, that include the auxiliary covariates. Overlooking these covariates will potentially result in biased estimates. We use an informative likelihood approach that will yield consistent estimates even when the underlying model for missing cause of failure is misspecified.

March 30

Speaker TBA

"Talk Title TBD"
ABSTRACT: None Given

April 6

Speaker TBA

"Talk Title TBD"
ABSTRACT: None Given

April 13

Speaker TBA

"Talk Title TBD"
ABSTRACT: None Given

April 20

Speaker TBA

"Talk Title TBD"
ABSTRACT: None Given

April 27

Speaker TBA

"Talk Title TBD"
ABSTRACT: None Given

May 4

Speaker TBA

"Talk Title TBD"
ABSTRACT: None Given

May 11

Speaker TBA

"Talk Title TBD"
ABSTRACT: None Given

Back to SPH Biostatistics Maintained by the Biostatistics Webmaster
Last Update: February 2, 2016