Retrospective Data Analysis: You Never Know
By Rob Wright, Chief Editor, Life Science Leader magazine
Steve Jobs has referred to computers as being a wonderful tool and a "bicycle for the mind." The possibilities for computers and their application seem limitless and will play a key role in drug development and discovery. Here's an example. A recent news release demonstrates how computers can benefit researchers and improve patient safety. Previously, some ECG data was ignored as it was considered to be noise. University of Michigan (U of M) researchers decided to do some data mining of existing ECG information from 4,557 heart attack patients. By applying some machine-learning methods, researchers hoped to determine if a specific biomarker could be identified for increased risk of death after certain cases of acute coronary syndrome (ACS).
I have always said if you torture the data long enough, it will confess. Using computational methods, researchers were able to identify three previously undiscovered biomarkers for higher risk of cardiovascular death. All three biomarkers created patterns found in the previously gathered clinical trial ECG data. According to Zeeshan Syed, assistant professor of electrical engineering and computer science at U of M, the main point of the study was to make better use of data already being collected. Syed states, "There's prognostic information buried in the noise, and it's almost invisible because of the sheer volume of the data. The sophisticated computational techniques allow us to home in on truly abnormal ECG signals. These patients with unstable hearts are at a greater risk of dying. Identifying them would allow physicians to initiate more aggressive treatment."
Retrospective data analysis is not a new technique. Unfortunately, few outside of academia take the time to conduct retrospective data analysis. In industry, the focus is on innovation, as well as killing nonviable drugs quickly. In so doing, companies can focus their finances and efforts on those drugs with the greatest potential for success and eventual commercialization. With the advent of newer and more sophisticated technologies, one has to wonder if some of the drugs which have been killed should be given a second look. Has the cure for cancer been shelved because the necessary technology did not exist to tease out key analytics?
Having completed the retrospective data analysis, researchers at U of M are interested in testing the biomarkers in hospitalized patients. Syed believes the biomarkers could be used alongside existing methods for categorizing high-risk patients with ACS. I will go one step further the biomarkers could play a key role in drug development and discovery. If you are a company sitting on a mountain of data and don't know what to do with it, here's a tip, share it with an academic institution. University researchers have the time, the resources, as well as the incentive (publishing an article) to conduct a thorough retrospective data analysis. If they uncover a key finding, you can always buy them a new building. Help a graduate student get published, find a new cure and stimulate the economy sounds like a win-win to me. So blow the dust off some data you have lying around and share it. You never know what you might find.