NCCIH Research Blog

Recently, I noticed news articles about a study on chamomile consumption and its potential effect on mortality. Researchers from the University of Texas Medical Branch, Galveston, analyzed data from a population-based study of older Mexican Americans in Southwestern states and reviewed 7-year all-cause and cause-specific mortality. They found a 29 percent decreased risk of death (all-cause mortality) among chamomile users compared with non-users, and concluded, for women at least, that this difference was statistically significant. It’s good news that using this herb is associated with longer life. Nevertheless, let’s remember that, even with statistical corrections, correlations don’t prove causality.

The authors reported that the decreased risk was statistically significant for women after adjusting for age, smoking, chronic conditions, and other known confounding factors. This is the right approach to the data. But it is not enough. Statistically correcting for other factors only captures the impact of measured differences. The women who use chamomile may differ in many ways from those who do not. For example, the use of herbal tea may be a marker of a “healthier” lifestyle. In their paper, the researchers note that “other unmeasured factors, such as frequency and duration of chamomile, level of physical activity, and quality of diet, which were not measured in the survey, could influence the results.”

This study illustrates the challenges of observational (non-experimental) research. This type of research helps us find patterns and signals and can raise new and interesting hypotheses. But let’s remember that there may be a big gap between correlation and causation.

Reference