On March 26,National Cancer Institute Senior Investigator Hormuzd Katki spoke at Swarthmore as part of the 2023-2024 Math/Statistics Colloquium Series. In his talk, titled “The Role of Statisticians and Quantitative Thinking in Cancer Prevention and Screening,” Katki discussed developing advancements in the field of biostatistics.
Katki has had a winding path through statistics to becoming a biostatistical researcher. He majored in statistics in college and then got a master’s degree in the subject. Then, he left academia and worked at the Bureau of Labor Statistics (BLS).
“I wanted to see if statistics would be useful in real life. I took a job at the BLS, where I worked on the survey where they calculate unemployment rates for the U.S. And that showed me just how important statistics was and so I was very excited about that.”
Finally, his career took him to biostatistics at the National Cancer Institute (NCI), a part of the National Institutes of Health (NIH), where he got his Ph.D and received tenure.
In his talk, Katki explained that screening is one of the best ways to prevent death from cancer.
“There are lots of things that go into it: who should be eligible for screening, what tests you should get, when you should come back for your next screening, and when you should exit from lifetime screening. All those questions can be addressed directly with data where we try to answer that in a way that is specific to the person who’s actually considering screening.”
To understand the scale and complexity of screening, Katki gave the example of cervical cancer screening.
“For cervical cancer screening in the US, we spend around $5 billion a year. That is an extremely effective program. It is estimated to have reduced cervical cervical cancer death by about 85%. So it’s worth it.”
Katki said his work on cervical cancer screening was some of the most exciting he had done.
“One of the best things that I’ve ever been able to have the privilege of being a part of was the process to revise cervical cancer screening guidelines. Rather than just having the traditional pap smear, they incorporate testing for human papillomavirus HPV, which causes nearly every cervical cancer,” he said.
Katki then went on to discuss different aspects of screening. One is a new blood test that could potentially test for many types of cancers at once. By screening these uncommonly tested organ sites all at once, researchers can catch more cases.
“Possibly the most exciting new frontier right now in cancer research is the potential for multi-cancer early detection tests that can, within a single blood sample, try to find tumor DNA in the blood that could come from any organ,” Katki said.
The mechanism for this broad screening ability is through a protein transcription process involving methylation.
“There is methylation of DNA that turns genes on and off. I think it was that if a methyl group was off [a DNA fragment], then the corresponding protein does transcribe, then makes a protein, but it turns out that those patterns are really specific to organs,” Katki explained.
Using this knowledge, this essentially means that it is possible to identify the origin of the cancerous DNA. Katki also discussed the use of risk prediction models and how they might change in the future.
“People at highest risk for cancer should typically also have the highest benefit from screening, defined as reduction in risk of death due to the procedure.”
However, this may be a flawed approach. Katki explains that they have been considering alternatives, such as models for life gains in expectancy. Because older people with very high risk for cancer are typically more likely to die from other comorbid conditions even if their cancer is caught, resources could be focused on younger people who could live for many more years.
“A modeling study showed that if you use a risk calculator to base lung cancer screening on, you actually don’t increase the life years gained in that population. So if we choose people instead based on life gains, it could be an attractive alternative. ”
However, this model has seen some pushback when it has been put into use in transplant lists.
“Immediately, people who were in line screamed out ‘This is age discrimination. We can’t have this.’ There was so much blowback that in the end, they didn’t use the system, which is unfortunate. ”
Katki also talked about an AI project he had worked on recommending lung cancer screening either for the next year or in two years, based on the results of their past year’s screening. In order to recommend them for screening in two years, there should be low risk of cancer in the next year.
“If we send these people for a two-year interval, we want them to have a very low risk that cancer would appear in one year. If we [miss] [the cancer], we lose one year of lead time to hopefully cure it,” he explained.
The AI was significantly better than the model currently in use, and at lower confidence levels, was also better than a statistical algorithm Katki tested.
Finally, Katki stressed how much he loves the larger-scale applications of his job.
“One thing is how fulfilling this research can be when your work comes into practice. There’s a big future right now for being able to do this,” he explained. “Biostatistics is fundamentally the information science of medical research and you need that good grounding if you want to go further into medical research.”
