Unlocking Alzheimer's through statistical methods
W. P. Carey School Professor Jeffrey Wilson dives into statistics to uncover insights into treatment options for Alzheimer’s, other neurodegenerative diseases
Jeffrey Wilson, professor of statistics and biostatistics in the economics department at Arizona State University's W. P. Carey School of Business, a former chair of the editorial board of the American Journal of Public Health and a director of the biostatistics core in the NIH Center for Alzheimer’s at ASU, joined Lori La Bey with Alzheimer’s Speaks Radio to discuss his book, "Fundamental Statistical Methods for Analysis of Alzheimer’s and Other Neurodegenerative Diseases."
Brittany Dugger, co-author of the book and assistant professor of pathology and laboratory medicine at the University of California, Davis, and co-leader of the Neuropathology Core within the University of California, Davis' Alzheimer’s Disease Center, joined Wilson for the interview.
Their work dives into statistics that help uncover insights into treatment options for Alzheimer’s and other neurodegenerative diseases and focuses on uncovering how data analytics can aid in unlocking the secrets of this devastating disease.
Editor's note: Questions and answers have been edited for length and clarity.
Question: Why do people fight the disease differently than others?
Dugger: So much of our life is due to serendipitous circumstances. I have been very blessed to have opportunities to work with such great people on this disease, including Drs. Wilson and Irimata, which has led to our certain approach to fighting the disease through statistical analyses. Both my grandmothers were diagnosed with the disease but had very different behaviors. If they hadn’t had dementia, I might not have known about its impact so early in my career.
Q: Have you been touched by dementia?
Wilson: During my time at Iowa State University, I had a mentor named George Jackson. He was very influential in my life. About 10 years ago, I went to see him in a Florida Alzheimer’s home and could hardly recognize him. The visits taught me how difficult it is dealing with the disease and I realized soon that I was more useful in providing family support. After his funeral, I realized his impact and am grateful for the many years our paths crossed, and I was glad he was able to inspire my life to such an extent.
Q: Why was there a need to write a book consisting of analyzing statistics and Alzheimer’s data?
Dugger: This goes back to when Dr. Wilson and I met. I was working on a dementia project and trying to understand topics such as cardiovascular risk factors. Many results were showing factors such as obesity, hypertension and diabetes relate to dementia — but what is the driving force, as cardiovascular risk factors are very intertwined? As a scientist, I needed a collaborator with the skill sets to help answer these questions. I was put in contact with Dr. Wilson about 10 years ago. During that time, we both realized how necessary a book was on the topic of statistics and Alzheimer’s.
Q: How does data tie into Alzheimer’s?
Dugger: This is a disease that gains momentum as it goes on. Most persons who get dementia end up with the disease later in life. Over time, people have an accumulation of event choices and/or exposures. If someone has cardiovascular risk factors, they may have diabetes as well as obesity and hypertension. If symptoms are intertwined, it’s called co-linearity. Also, it’s necessary to know that people can have multiple types of dementia. A patient can have Alzheimer’s disease and vascular dementia (brain damage caused by multiple strokes), too. All of these items are data, and analyzing these types of data aid in understanding the disease, leading to better prognosis, treatments and diagnoses.
Q: How do you explain this in terms of statistics?
Wilson: In statistics, we have basic principles that we use to determine the results. Researchers measure one response variable at a time. These observations are usually treated as coming from a mechanism that is unrelated to each other. However, consider a patient; each provides several different outcomes. To use a separate outcome analysis is to ignore any interrelationship. Our book pays attention to the usual neglectful perspective of correlation, as it shows instances of the shortcomings when it is ignored. It also helps readers to understand various simultaneous responses regarding the disease.
Q: Can you explain how Alzheimer’s is different than other neurodegenerative diseases?
Dugger: Both my grandmothers had dementia, and their presentations varied. This was due to them having perhaps different underlying causes or neurodegenerative diseases. For instance, when a person has a stomach ache and high fever, what is the underlying cause? With my grandmother on my mom’s side, she had hallucinations, thinking people were coming out of the TV. She would have personality changes as well. On the opposite side was my dad’s mom, who had more classic memory problems associated with dementia. Every Sunday she would go to church and sit in the same spot but couldn’t remember who I was.
Dementia in the clinical sense is when a person has cognitive impairment as well as a decrease in activities of daily living. And with neurodegenerative disease, these are the underlying mechanisms of this outward clinical symptom of dementia. One type of neurodegenerative disease is Alzheimer’s disease. Other types are out there, too, including dementia with Lewy bodies and frontotemporal dementia. These neurodegenerative diseases can only be definitively diagnosed after death.
We work with data from what's called the National Alzheimer's Coordinating Center. It compiles data from over 30 sites from around the country. They're called Alzheimer’s Disease Research Centers. And a lot of them are in major cities. We have one at the University of California, Davis. There's one in Arizona where they have a consortium that includes Banner Sun Health Research Institute and Arizona State University.
Q: What would you say are some of the challenges with explaining your work to potential readers?
Wilson: For the reader, we build on what is commonly known. So, the book builds on what readers understand in basic approaches. For instance, if you have one predictor such as age and whether or not someone has dementia, you would want to look at the logistic regression model with two variables. But if you have baseline knowledge, our work would give you opportunities to incorporate past information along with several outcomes. The book tries to simplify things while maintaining the accuracy of the analysis. Simplicity and accuracy are the aims of the book. We focus a great deal on correlated observations. By correlated observations, we mean the mechanism by which the observations are obtained are related.
Dugger: Statistics consists of a lot of jargon and has a unique vocabulary, so this can be a hurdle for many. The book tries to break down these hurdles by providing motivating examples. Researchers do this frequently, where we discuss how certain variables will affect Alzheimer’s patients. For example, we study obesity and dive into whether weight is different between people who had Alzheimer’s and people who didn’t. In the book, we walk through these factors and, in a step-by-step manner, we give the workflows for the use of the correct statistical model with certain programs.
Q: How does this book differ from other statistical books?
Wilson: This book has 12 chapters and talks about different statistical methods such as survival analysis and hierarchical models or Bayesian statistics: when different things happen with the same patient, a doctor would be looking at these different measures. That’s because, in the doctor's mind, they’re related. We know that all these things together are correlated. Because we know they’re correlated, we know that you can’t treat medical diagnoses separately. They’re not independent observations.
Q: How is this book organized?
Dugger: The book is organized to serve as a statistical guide for those who wish to analyze data focused on Alzheimer’s disease. Statistical techniques are presented, addressing questions regarding data analysis. It begins with simple introductions and builds to more complex models. In the book, we present different models of analysis, and this is important as one can get different results depending on what model they use. For example, this occurred in a recent clinical trial with an Alzheimer’s drug where the initial statistical model was composed, and they ended up stopping the drug trial based on the results. Then, after re-analyzing the data with a different model and getting different results, they restarted the trial. It matters how you look at the data and how you’re modeling things. Our understanding of how we can improve is growing, especially since there currently is no cure for Alzheimer’s disease.
One last concept to touch upon is the notion that most data published on Alzheimer’s disease is based on wealthier white people. We know this disease affects people from various cultures, socioeconomic statuses and life experiences. To succeed in research, we must pursue an inclusive process. Many people are struggling, and we hope this book and the way it’s organized will aid many people in being able to analyze data.
Wilson: Also, this book includes basic statistical programs that people are more likely to use in conjunction with Alzheimer’s and dementia, and some case studies are highlighted. Let me thank Johns Hopkins University Press for publishing this book on our behalf. Also thanks to the other co-author (Dr. Katherine Irimata), who was instrumental in our effort.
To order a copy of “Fundamental Statistical Methods for Analysis of Alzheimer's and Other Neurodegenerative Diseases," visit Johns Hopkins University Press. Use the code HTWN.