Will Drinking Coffee Prolong Your Life?

Press reports have highlighted a study published in the New England Journal of Medicine on May 17 on the relationship between coffee drinking and mortality (Freedman et al, Association of Coffee Drinking with Total and Cause-Specific Mortality. N Eng J Med 2012;366:1891-1904).  This study involved an analysis of data from the NIH-AARP Diet and Health Study.  Between 1995 and 1996 over 600,000 AARP members completed a comprehensive questionnaire assessing diet and lifestyle.  In this study responses to that questionnaire were correlated with follow-up data – most specifically death and cause of death – in order to address the association between coffee drinking and mortality.

The results were quite interesting.  Overall men or women who reported drinking four or more cups of coffee per day, after taking into account age, had a rate of death which was statistically greater that those who drank no coffee.  For example, men who drank 6 or more cups of coffee per day, after taking into account age, had a rate of death 60% greater than those who drank no coffee.

However, coffee drinking was found to be associated with other behaviors which in turn have been associated with increased mortality such as cigarette smoking, high level of alcohol consumption and red meat consumption.  Coffee drinkers also tended to have a lower level of education, exercised less, and had a lower intake of fruits, vegetables and white meat. 

The investigators statistically adjusted the results to account for such other factors.  This type of statistical adjustment attempts to correct for these other factors so as to be able to measure the specific effect of coffee drinking alone on mortality.  After this statistical adjustment the results were quite different.  Men who drank one or more cups of coffee per day, and women who drank two or more cups of coffee per day, had an adjusted rate of death which was statistically lower that those who drank no coffee.  For example, men who drank 6 or more cups of coffee per day had an adjusted rate of death 10% lower than those who drank no coffee.  Women who drank 6 or more cups of coffee per day had an adjusted rate of death 15% lower than those who drank no coffee.

Interestingly, the reduction in the adjusted rate of death was found to persist over a wide range of causes of death.  For example, in both men and women who drank coffee there was a significant reduction in adjusted rates of death due to heart disease, respiratory disease, stroke, injuries and accidents, and diabetes.  In men there was also a significantly lower adjusted rate of death due to infection.  The one exception to the pattern was that men who drank coffee had a higher adjusted rate of death due to cancer.

In summary, after taking other factors into consideration, these investigators found that adjusted overall death rates were statistically significantly reduced in men and women who drink coffee; this reduction applied to a wide range of specific causes of death.

Does this mean that drinking coffee will extend your life?  No, not necessarily.  As the investigators point out this type of study cannot differentiate between correlation and causation.

Of course, one possible explanation of the results is that coffee drinking actually does prolong one’s life.  Coffee contains thousands of compounds that could affect various metabolic processes in a beneficial (or non-beneficial) manner. 

Conversely, it could be that individuals with physiological or behavioral characteristics associated with a longer life span, but which were not specifically adjusted for in this study, are more likely to be coffee drinkers. 

Or perhaps people, who are more likely to die, avoid drinking coffee.  Maybe coffee is more likely to make such individuals nauseated.  This explanation is perhaps unlikely in this study because coffee seemed to reduce death due to such a wide range of causes and many of the deaths occurred many years after the individuals reported their coffee drinking habits.

Furthermore, while the authors used standard statistical methods to adjust the death rates to compensate for other factors, the statistical methods assume a specific mathematical form for the interaction between the various factors.  This mathematical model might not be applicable and thus the statistical adjustment might produce incorrect results.

Finally, for a study to be considered completely prospective all decisions regarding the data analysis have to be made ahead of time in the complete absence of any knowledge – direct or indirect - of the data collected in the study.  If knowledge of such data informed any decisions regarding how the data were to be analyzed, the study would be considered a retrospective analysis which is far less reliable.  For example, in this study approximately one third of the individuals who provided completed questionnaires were excluded from the analysis for various reasons.  From a statistical perspective, the decisions on exactly which groups to exclude needed to have been made with absolutely no knowledge of the study data.  The selection of which possibly confounding factors were to be adjusted for would also have to be made similarly with no knowledge – direct or indirect – of the study data.  While there is no evidence in this study to suggest that any decisions were at all informed by prior knowledge of study data, it is often very difficult in the process of analyzing data to prevent such knowledge from, even subconsciously, affecting seemingly minor decisions regarding data analysis which can end up affecting results.

This study does provide very intriguing results.  However, as the investigators state this study by design could at best demonstrate correlation and not causation.  Furthermore, since the beneficial effect of coffee drinking on survival emerged only after a set of statistical adjustments were applied, one has to question whether the reported effect of coffee drinking on survival could be an artifact of the statistical adjustment process itself.

The study does suggest a hypothesis that coffee drinking reduces mortality.  This hypothesis could be tested in a prospective randomized interventional trial in which individuals are randomly assigned to drink or not drink coffee over a prolonged period.  This undoubtedly would be a difficult trial to conduct, but only such a trial could reliably determine whether coffee drinking actually reduces mortality.  Such prospective randomized interventional clinical trials are the gold standard in medicine for testing hypotheses regarding new treatments and new diagnostic methods. It might be worthwhile for Starbucks to fund such a study.

In the meantime, I will continue drinking coffee because it tastes good and is not fattening.