Foreword: I know this is a long post. I tried to slim it down, but it’s really hard to do that without leaving out important details. This post was prompted by my daughter telling me that I need to stop using a French press for making my coffee. If you want to jump to the point, scroll down to the last paragraph. If you’d like to see a little more in-depth analysis of the sensationalist headlines, keep reading.
We all know how academic work is often misinterpreted and how lazy, sensationalist journalism can be misleading. There’s also some culpability on the part of the authors in this one. Let’s look at their “aim” and their conclusions in the abstract section of the paper.
Aim: The aim of this study was to investigate whether the coffee brewing method is associated with any death and cardiovascular mortality, beyond the contribution from major cardiovascular risk factors.
Methods: (skipped for now)
Conclusion: Unfiltered brew was associated with higher mortality than filtered brew, and filtered brew was associated with lower mortality than no coffee consumption.
The authors clearly went for the click-bait conclusion by leaving out a couple of key pieces of information. Before we get to that, let’s talk about their methods. The study is based on a 20-year time period, using participants in a Norwegian cardiovascular survey. The length of the study is good, but having participants all of one nationality is not so good, especially in such a small population (< 5 million) who are very homogeneous (77% of Norwegian birth). Surveys are always suspect because people are really, really bad at self-reporting.
As an example, let’s look at one of the central questions of this study: “How many cups of coffee do you usually drink daily?” (taken directly from their questionnaire). This simple question is really quite complicated. The obvious problem is how big of a cup are they asking about. A “cup” of coffee isn’t a standardized measure. In the U.S., a cup is 240ml, in the metric world, a cup is 250ml. That says nothing about what people call a cup. My preferred coffee “cup” actually holds 20 oz, 2.5 times the U.S. measured cup and 2.4 times a metric cup. Yet, if you asked me about it without clarifying the actual amount, I would probably say I have one cup of coffee a day.
Within the paper, they list a whole host of confounding variables, only some of which they could control in the statistical analysis.
The following variables were included in the analysis as potential confounders: age, number of cigarettes/d, total cholesterol, triglycerides, systolic blood pressure, body mass index, education, and year of examination. Educational length comprised nine categories: <7 years, 7–9 years, ....17–18 years, and > 18 years; and physical activity was >=1 hour vigorous physical activity per week or walking around, cycling, or other activity for at least four hours a week.
They didn’t control for all of the confounding variables: “We adjusted the mean values and percentages across the coffee categories for age and sex.”
This alone doesn’t invalidate their study! This is a normal and widely accepted statistical procedure which helps reduce variability within a test group. The authors used this procedure on age and sex, but did not do that for the other potential confounding factors.
Alright, there’s a lot to unpack here and I don’t want to lose you without getting to the point. Let’s look at two tables within the paper. The first, Table 2, shows “Hazard Ratios” for three specific types and “any” type of death. The Hazard Ratio (HR) is a measure of how many people died of the specific death types compared to “any” death. That means the “any” death HR is 1. This is the reference category and shows up in the table as “Ref” instead of a number. Any number less than 1 means that fewer people in that category died as compared to the control.
The only category that has a HR ratio that is not lower than the reference is unfiltered coffee drinkers who died of Ischemic Heart Disease (IHD). Even so, they have the same HR as non-coffee drinkers who died from anything except Cardiovascular Disease (CVD) or Stroke.
This, of course, is odd. Why would unfiltered coffee drinkers have the same HR ratio as non-coffee drinkers? The logical thing to do is try to unpack the numbers and figure out why! So, that’s what they did. I present Table 3:
To create this table, the authors aggregated the participants into two groups, those 60 years old and older, and those younger than 60. This was their age at the beginning of the study period, 20 years ago. Note that only men, aged 60 and older, who drank either unfiltered coffee or both filtered and unfiltered, died more frequently of CVD. Those HR numbers are important - they mean that these two groups of coffee drinkers died either 5% or 19% more frequently than people 60 years or older died of any other cause.
To make a really long story short, if you’re a man who is 60 years or older, perhaps you should give up drinking unfiltered (french brewed or percolated) coffee. It might, all other confounding factors ignored, lower your risk of dying from cardivascular disease.