AI links caffeinated coffee consumption to reduced heart failure risk
The study draws on the power of machine learning to rank coffee consumption as important as age, blood pressure, heart rate, and weight in links to HF risk, unintentionally becoming the research focus.
“The association between caffeine and heart failure risk reduction was surprising,” states Dr David Kao, senior study author, assistant professor of cardiology at the University of Colorado.
“Coffee and caffeine are often considered by the general population to be 'bad' for the heart because people associate them with palpitations, high blood pressure, etc. The consistent relationship between increasing caffeine consumption and decreasing heart failure risk turns that assumption on its head.”
"However, there is not yet enough clear evidence to recommend increasing coffee consumption to decrease risk of heart disease with the same strength and certainty as stopping smoking, losing weight or exercising."
The team employed machine learning via the American Heart Association's Precision Medicine Platform to examine data from the Framingham Heart Study.
Findings here were compared to data from the Atherosclerosis Risk in Communities Study and the Cardiovascular Health Study to help confirm the team’s conclusions.
The researchers cited advantages to using machine learning that allowed for the assessment of large patient characteristic numbers in a comparatively unbiased manner.
Other benefits include the reduction of false positives and could potentially pick up patterns that otherwise may be missed when using a hypothesis-based approach.
In an accompanying editorial, Amanda Vest, Assistant Professor at Tufts University School of Medicine, praised the broad hypothesis-free approach that used 204 potential measures to construct random models.
“The benefit of this approach is that relationships within the data drive feature selection, allowing the strongest risk factors to rise to prominence even in the setting of collinearity,” she added.
The analysis revealed that in all three studies, individuals that said they drunk one or more cups of caffeinated coffee had a decreased risk of long-term heart failure.
The risk of heart failure over a number of decades decreased by 5-to-12% per cup per day of coffee, compared with no coffee consumption in the Framingham Heart and the Cardiovascular Health studies.
In the Atherosclerosis Risk in Communities Study, heart failure risk remained unchanged between 0 to one cup per day of coffee. However, it was about 30% lower in people who drank at least two cups a day.
Drinking decaffeinated coffee appeared to have an opposite effect on heart failure risk - significantly increasing the risk of heart failure in the Framingham Heart Study.
In the Cardiovascular Health Study however; there was no increase or decrease in risk of heart failure associated with drinking decaffeinated coffee.
Further analysis found caffeine consumption from any source was associated with decreased heart failure risk, and caffeine was at least part of the reason for the apparent benefit from drinking more coffee.
“While unable to prove causality, it is intriguing that these three studies suggest that drinking coffee is associated with a decreased risk of heart failure and that coffee can be part of a healthy dietary pattern if consumed plain, without added sugar and high fat dairy products such as cream," said Penny Kris-Etherton, distinguished professor of nutrition at The Pennsylvania State University.
“Also, it is important to be mindful that caffeine is a stimulant and consuming too much may be problematic -- causing jitteriness and sleep problems."
‘An integral component of research’
Dr Vest emphasised that the study used traditional survival modelling to corroborate the importance of coffee as an HF risk factor in the dataset, although she warned the potential for false discovery remained.
“Artificial intelligence techniques, including machine learning, are becoming an integral component of our research infrastructure and hold particular promise in synthesising the complex, multidimensional, heterogenous relationships believed to explain the biological and sociological pathways connecting dietary choices to subsequent HF,” she remarked.
“Novel features such as geolocation information, participant-driven food and beverage photographs, physical activity sensors and social network analyses offer a future state where creative approaches to cardiovascular nutrition will allow us to ask more ambitious research questions than appeared possible with traditional unidimensional statistics.
“Data fidelity and high-quality model validation will remain essential components to build confidence in these approaches, but the potential benefits of moving from participant-reported dietary recalls towards real-time monitoring of complex dietary patterns could be transformative. “
Source: Circulation: Heart Failure
Published online: doi.org/10.1161/CIRCHEARTFAILURE.119.006799
“Association Between Coffee Intake and Incident Heart Failure Risk, A Machine Learning Analysis of the FHS, the ARIC Study, and the CHS.”
Authors: Laura Stevens et al