Please join us for the next installment of the MGH Biostatistics Applied Biostatistics Seminar Series, a new seminar series designed to introduce researchers to intermediate topics that are highly relevant to clinical biostatistical research. Presenters will introduce us to their area of expertise and motivate the use of these methods with concrete clinical examples.

The essential role of early phase clinical trials in oncology: design and interpretation

Thursday May 19, 2022, 12:00-1:00pm

Hybrid event: In-person at 50 Staniford St, Ste 560 & on Zoom (https://partners.zoom.us/j/82088584012 )

Speaker: Alexia Iasonos, PhD, Attending Biostatistician, Director, Clinical Research Development, Department of Epidemiology and Biostatistics, Memorial Sloan Kettering Cancer Center, New York.

Abstract: The specific aims and overall purpose of Phase I clinical trials in oncology have undergone substantial changes over the last several years. The trials originally had a single objective of showing adequate tolerance of an investigational agent or regimen in a heterogeneous patient population. The last several years have seen this single objective evolve into an ambitious set of objectives: to identify the most effective dose of a single agent or combination of agents, the right treatment schedule, and the right patient population in terms of the likelihood of antitumor activity. In this talk, I will provide a broad overview of how the objectives, design, and interpretation of early phase trials have evolved and how the inclusion of dose-expansion cohorts has changed the landscape of early drug development. I will present two recent methodological advancements in this field that aim to correct the bias that can result from errors in toxicity attribution and the bias that can result when the administered dose differs from the assigned dose, which can occur in Phase I trials of CAR T therapy.

This will be a hybrid in-person and virtual event. Dr. Iasonos will be joining us in person for this presentation. Please contact tthaweethai@mgh.harvard.edu with any questions.

Please join us for the next installment of the MGH Biostatistics Applied Biostatistics Seminar Series, a new seminar series designed to introduce researchers to intermediate topics that are highly relevant to clinical biostatistical research. Presenters will join us on every 3^rd Friday of each month to introduce us to their area of expertise and motivate the use of these methods with concrete clinical examples.

Early Detection of Ovarian Cancer through Statistical Modeling of Longitudinal Biomarkers

Friday April 15, 2022, 1:00-2:00pm

https://partners.zoom.us/j/82231967249

Speaker: Steven J. Skates, PhD, Associate Biostatistician at Massachusetts General Hospital and Associate Professor of Medicine at Harvard Medical School.

Abstract: Ovarian cancer (OC) is a prime target for early detection due to a high proportion of cases detected in late-stage disease and a large difference in prognosis between patients detected in early stage compared to late-stage detection. However, the low annual incidence of OC presents a serious medical and statistical challenge – how to detect the cases as early as possible in time without generating too many false positives potentially resulting in surgical interventions. A pilot trial of 1,000 women detected one ovarian cancer using a then recently discovered ovarian cancer serum biomarker CA125 with a single threshold and referral to ultrasound if positive. The CA125 profile over time in the case was striking with an exponential growth in distinct contrast to the relatively stable CA125 profile in all other subjects. Subsequent larger pilot trials showed a similar phenomenon. The statistical challenge was how to use the extra information in the CA125 profile to increase sensitivity in early stage while maintaining the same low and acceptable false positive rate. We developed a calculation for the probability of having undetected OC given a series of CA125 values built on a longitudinal change-point model for OC cases and from a longitudinal flat model for controls. This calculation replaced the single CA125 threshold for screening decisions. We implemented the calculation in five prospective early detection trials in the US and UK, including a 20-year RCT of over 200,000 women. Results from the five trials were mixed, with a negative result in the largest trial but encouraging results in the other trials. We reflect on the current state of the field and describe a parallel biomarker discovery and validation program to identify novel biomarkers and approaches that may improve on the longitudinal serum CA125 approach to early detection of OC.

This will be a virtual event. Please contact tthaweethai@mgh.harvard.edu with any questions.

Upcoming seminars:

Thursday May 19, 2022 (12-1pm): Alexia Iasonos, Attending Biostatistician, Memorial Sloan Kettering Cancer Center

Please join us for the next installment of the MGH Biostatistics Applied Biostatistics Seminar Series, a new seminar series designed to introduce researchers to intermediate topics that are highly relevant to clinical biostatistical research. Presenters will join us on every 3^rd Friday of each month to introduce us to their area of expertise and motivate the use of these methods with concrete clinical examples.

An Introduction to Interval-Censored Failure Time Data, with Applications to HIV Prevention Trials

Friday March 18, 2022, 1:00-2:00pm

https://partners.zoom.us/j/86995026265

Speaker: Kaitlyn Cook, PhD, Postdoctoral Research Fellow, Harvard Medical School and Harvard Pilgrim Healthcare Institute

Abstract: Interval-censored failure time data naturally arise in biomedical and epidemiological studies in which the event of interest is subject to periodic follow-up (as in electronic health records research) or otherwise not directly observable (as in infectious disease prevention studies). Since questionnaires, physical exams, blood cultures, and serological tests are conducted on an intermittent basis, the exact timing of this event is known only up to the interval between the last negative and first positive tests. When the subjects in these studies also belong to existing, non-investigator determined groups—such as hospitals, communities, transmission networks, or insurance networks—the resulting data may be both interval-censored and clustered. The presence of these two features leads to a loss of statistical information and power and presents complications for both estimation and inference. In this talk, I will provide an introduction to interval-censored failure time data and discuss some common methods for analyzing this data in the independent data setting. I will then introduce a novel extension of the proportional hazards framework to the clustered data context, and will illustrate its use by re-analyzing a large-scale cluster-randomized trial of combination HIV prevention methods.

This will be a virtual event. Please contact tthaweethai@mgh.harvard.edu with any questions.

Upcoming seminars:

Friday April 15, 2022 (1-2pm): Steven J. Skates, PhD, Associate Investigator at MGH Biostatistics and Associate Professor of Medicine at Harvard Medical School. Topic: Longitudinal Biomarkers for Early Detection of Cancer

Tanayott Thaweethai, PhD will lead a journal club with Harvard Catalyst on Wed 3/2 from 1-2pm.

Title: Missing Data Challenges in Electronic Health Records-Based Studies of COVID and Long COVID

Abstract: Since the beginning of the COVID-19 pandemic, researchers have repeatedly turned to electronic health records (EHR) to rapidly answer complex questions about the short and long-term consequences of SARS-CoV-2 infection. However, because EHR are not collected for research purposes, observational studies using EHR are subject to various challenges and biases, including bias due to missing data. Standard missing data methods generally fail to address the complex nature of EHR data, particularly the interplay of numerous decisions by patients, physicians, and insurers that collectively determine whether “complete” data is observed. Tanayott Thaweethai, PhD, Massachusetts General Hospital, will discuss some statistical methods for handling bias due to missing data in the EHR setting, and conclude with an introduction to a semi-supervised learning technique for handling the “positive unlabeled” problem of phenotyping individuals based on the presence or absence of clinical codes.

The talk will occur on Wednesday March 2, 2022, 1-2pm.

Register here.

Please join us for the next installment of the MGH Biostatistics Applied Biostatistics Seminar Series, a new seminar series designed to introduce researchers to intermediate topics that are highly relevant to clinical biostatistical research. Presenters will join us on every 3^rd Friday of each month to introduce us to their area of expertise and motivate the use of these methods with concrete clinical examples.

Using Neural Networks to Predict Breast Cancer Risk

Friday February 18, 2022, 1:00-2:00pm

https://partners.zoom.us/j/89183981161

Speaker: Zoe Guan, PhD, Postdoctoral Research Fellow, Memorial Sloan Kettering Cancer Center

Abstract: Neural networks are a flexible machine learning method inspired by the structure of the brain. They are able to capture complex nonlinear relationships between inputs and outputs and have achieved remarkable accuracy in many prediction and classification problems, including natural language processing, image recognition, and prediction of health outcomes. Recently, there has been growing interest in applying neural networks to clinical problems, especially with the acceleration of large-scale data collection efforts in healthcare. In this talk, I will give an overview of two types of neural networks, fully-connected and convolutional, and discuss an application of these approaches to breast cancer risk prediction using family history data.

This will be a virtual event. Please contact tthaweethai@mgh.harvard.edu with any questions.

Upcoming seminars:

Friday March 18, 2022 (1-2pm): Kaitlyn Cook, PhD, Postdoctoral Research Fellow, Harvard Medical School and Harvard Pilgrim Healthcare Institute. Topic: Interval-censored data and HIV prevention trials

Brian Healy, PhD will be teaching the course “Basic Biostatistics for Clinical Research” between Friday January 7 and Friday February 4, 2022. The course is sponsored by the MGH Division of Clinical Research and MGH Biostatistics.

This course will provide clinical researchers with a solid foundation in biostatistical concepts. Intended for those with minimal statistical experience, these five lectures will serve as an introduction to biostatistical issues in clinical investigation and will prepare students for more advanced courses on clinical trial design and biostatistics offered through the DCR’s Education Unit.

The course is open to learners at MGB. Click here to register.

Please join us for the inaugural seminar of the MGH Biostatistics Applied Biostatistics Seminar Series, a new seminar series designed to introduce researchers to intermediate topics that are highly relevant to clinical biostatistical research. Presenters will join us on every 3^rd Friday of each month to introduce us to their area of expertise and motivate the use of these methods with concrete clinical examples.

Semi-competing risks: modeling and joint prediction of dependent non-terminal and terminal events

Friday November 19, 2021, 1:00-2:00pm

Hybrid event: In-person and on Zoom (see details below)

MGH Biostatistics Conference Room, 50 Staniford St Ste 560

or https://partners.zoom.us/j/86551137602

Speaker: Harrison T. Reeder, PhD Candidate, Department of Biostatistics, Harvard TH Chan School of Public Health

Abstract: Semi-competing risks refers to the survival analysis setting where the occurrence of a non-terminal event is subject to whether a terminal event has occurred, but not vice versa. Semi-competing risks arise across a broad range of clinical contexts, but are not always recognized as such, leading researchers to pursue analyses that ignore the dependence between events, or focus solely on a single or composite outcome. In particular, unlike standard competing risks, semi-competing risks provide an opportunity to learn about the joint risk of the two events, enabling individualized risk prediction of patients’ entire outcome trajectories across time. In this talk we will build on familiar survival analysis tools to introduce the frailty-based illness-death model for semi-competing risks. This framework captures the complex interplay between risk factors and the joint outcomes, and aligns with the needs of clinical decision makers. We illustrate this with recent work on joint prediction in two clinical settings: preeclampsia and delivery during pregnancy, and shock and death among heart failure patients receiving implantable cardioverter defibrillators.

This will be a hybrid in-person/virtual event. Harrison will be joining us in-person and MGH employees who would like to attend in-person are welcome to do so. The option to attend virtually will be made available to all. Due to COVID protocols, non-MGH employees are not permitted to attend in-person at this time. Please contact tthaweethai@mgh.harvard.edu with any questions.

Upcoming seminars:

Kaitlyn Cook, PhD, Postdoctoral Research Fellow, Harvard Medical School and Harvard Pilgrim Healthcare Institute. Topic: Interval-censored data and HIV prevention trials

Zoe Guan, PhD. Postdoctoral Research Fellow, Memorial Sloan Kettering Cancer Center. Topic: Neural networks and prediction of hereditary cancers