Skip to content

So, caffeine in seawater stresses marine life, eh?

October 13, 2012

There is a story going around on the Internet about the concentration of caffeine in seawater. It is so prevalent that Googling data-oriented phrases like ‘caffeine concentration tea ng/L’ returns four news articles discussing the paper before any data on the concentration of caffeine in tea – this is a serious subversion of my Internet! The CBC News version of it, titled “Caffeine flushed into Pacific Ocean stresses marine life” was linked to me by a friend, and struck me as fishy. So I thought I’d do a little digging.

The CBC News article has a pretty ominous run of statements in the body text:

“Caffeine is affecting marine life like mussels in much the same way it can affect humans, Granek said.. [sic]
“We found that the mussels ramped up what we call ‘the stress response.’”
She said that response can impact their ability to thrive and reproduce.”

… and yet, that whole run of statements just set my Skeptic Sense buzzing. Most of the studies I’ve read indicate the health effects of caffeine in humans are pretty mild, even up to some pretty high doses. You can detect some negative health effects if you’ve got a huge dataset to play with, but the ratio of noise to signal is huge. So, I asked myself, what data is the CBC article based on? Is dose-response reported well? How do the seawater concentrations stack up against concentrations actually consumed by humans, and known to have health effects in them? Does anyone actually have data on the actual health effects of caffeine on marine organisms, and if so, is there any evidence of it reducing their ability to ‘thrive and reproduce’?

Luckily for me, the CBC article links to a paper, so I found the original research easily. One thing that became immediately apparent was that finding caffeine in seawater is not new. A bunch of other researchers have found it. The introduction even discusses a quite cool suggestion – originally made by other researchers – about using the abundance of caffeine in coastal waters to work out if city plumbing is leaking untreated water because of the very high efficiency of some treatment plants in removing caffeine.

Another thing that is obvious is that this study does not even try to assess the effect of caffeine on marine life – if we’re looking for information on the health effects of caffeine on marine organisms, we’re going to have to look elsewhere. What actually happened in the study is that the researcher drove to 14 different sites on the Oregon coast, which were selected on the basis of things like their population density and proximity to water treatment facilities, and took water samples from oceanic water and (sometimes) nearby streams or estuaries, from which caffeine would presumably flow to the ocean. Sensible precautions were taken against contaminating the samples with caffeine from the experimenters, like wearing gloves while collecting, and not drinking any caffeinated beverages for 24 hours before collecting. The samples were analysed for caffeine concentration back at the lab.


Nine of their 21 samples (14 sites, 21 samples, 7 sites sampled both in the river and the ocean) had measurable caffeine. Some of their field blanks (which should contain no caffeine – caffeine in blanks indicates the degree of contamination in the lab) also contained caffeine – anywhere up to 6.6 ng/L though mainly around 1.5 – 1.9 ng/L. Five coastal ocean sites had detectable caffeine: 9, 18, 18, 30, and 45 ng/L. Four river/estuary sites had detectable caffeine: 11, 67, 91, and 152 ng/L. Caffeine concentration didn’t correlate with population density or water treatment plant proximity, and the two highest caffeine readings came from areas not considered to be at risk of caffeine pollution. A potential confound is mentioned, where the more northerly sites were sampled just after a major storm, which the authors posit may have washed a bunch of caffeine into the ocean, hence the high mean caffeine readings for some of their sites.

There is a graph that purports to show the relationship between river concentrations and coastal ocean concentrations in sites where both were sampled, but they don’t plot any of the datapoints that scored below the reporting limit on the river caffeine axis (no, seriously), and this renders it uninterpretable. And not to be too catty, but someone forgot to check that their axis labels were displaying correctly, so the data presented in the graph is all out of sync with the table which duplicates those data. In any case, the graph of four datapoints claims to show that river concentrations and coastal sea concentrations are negatively correlated – that is, that if there is more caffeine in the river, there is less caffeine in the nearby ocean. How that graph made it through peer review, I will never know.


The crunch point of the study is that there is some detectable caffeine in about half of the sampled rivers and in about one-third of the patches of ocean, but we can’t predict where caffeine concentrations will be high, because we don’t really understand how caffeine pollution happens. It doesn’t look like more people = more caffeine. It possibly looks like more storms = more caffeine, but we can’t even say that to any degree of certainty because no sites were sampled both before and after the storm.

A side note, and some background on caffeine concentrations: caffeine concentrations observed in Oregon water, when caffeine was observed at all, were low. The highest concentration, in a river after a storm, was 152 ng/L. The highest cited concentration of caffeine in seawater anywhere in the world was 5,000 ng/L – I checked the cited paper, and it appears that this number was from a sample taken right next to a wastewater treatment plant outflow pipe: elsewhere in that paper, values for wastewater treatment plant outflow are given as 15,200 ng/L. For comparison, black tea has a caffeine concentration of between 120,000,000 and 360,000,000 ng/L of caffeine: the highest caffeine concentration found anywhere in the Oregon study was about around one millionth the concentration of a weak cup of black tea. Even the concentrations of caffeine in wastewater treatment plant outflow are four orders of magnitude lower than those in a weak cup of black tea. In order for mussels to be harmed by these concentrations of caffeine, they’d have to be extremely caffeine-sensitive compared to humans.

I’ve mentioned that this paper did not deliver any assessment of the impacts of caffeine on marine organisms. In light of that, why are people talking about effects of caffeine pollution on marine organisms? Why is the second author quoted as saying that mussels show a stress response to caffeine if the study doesn’t even mention mussels?

After doing a search on the authors’ names, I found that the lab that produced the Oregon research also has some work that looks at mussels and their responses to caffeine. The mussels paper looks at a gene called Hsp70, which is traditionally associated with stress responses, and claims that it is upregulated by caffeine. They looked at the effects of four different concentrations of caffeine: 0, 50, 200 and 500 ng/L, for three different durations: 10, 20, and 30 days, in two different tissue types: gill and mantle. The 50 ng/L is already quite high in the context of what was found in Oregon, and the other two exceed the highest value found there in what were thought to be exceptional, transient circumstances (immediately after a storm) – so the relevance of these doses in the real world, especially for extended periods, is pretty debatable. In any case, what the paper actually does is badly mess up its statistical analysis, and then extrapolate. I read it and facepalmed.


The experiment tested for effects of caffeine concentration and time, and was run on four datapoints per treatment, using ANOVA and Tukey Post-hoc Tests. What the results show is a remarkably consistent rate of Hsp70 expression in the no-caffeine control group, and huge variance in gene expression in all the treatment groups for gill lamellae, but no response at all in the mantle tissue. The researchers seem unaware that ANOVA and Tukey Post-hoc Tests both assume equal variances, especially where sample sizes are small. They certainly don’t let it get in the way of them making directional interpretations of the data! I’ve taken the plots from that paper so you can look at them if you want – I think this counts as Fair Use (click to see both plots).



The last point I’d like to make is that it’s a pretty big jump to go from ‘upregulation of a gene’ to ‘stress’. I’d think that the sensible experiment to run if you were looking for ‘stress’ would be to assess body condition or fecundity (say, egg number or mass) of marine organisms in response to caffeine. As it is, changes in the regulation of this gene could easily be the metabolic equivalent of me opening the window if the weather gets hot – a response, sure, but not what I would call evidence of ‘stress’.

So, to summarise what is known:

  1. In coastal waters around Oregon, the most common amount of caffeine is ‘below reporting threshold’

  2. Where caffeine is found around Oregon, levels are typically low – in the tens of nanograms per litre. It has been seen at up to 5,000 ng/L elsewhere (apparently right next to a wastewater treatment plant outflow), but between zero and tens of nanograms per litre looks typical.

  3. In a study investigating the response of mussels’ Hsp70 gene expression rates to caffeine, no consistent results were observed, but the authors reported a link anyway, seemingly on the basis of not understanding the assumptions of the statistical tests they used, or the nature of small sample sizes and large variance.

  4. In that same study, the experimental design used extended-duration high doses of caffeine. It is questionable whether concentrations of caffeine that high are ever seen in the real world, other than briefly after storms.

  5. The media loves to hype stories about caffeine. It’s a popular chemical! But seriously, you shouldn’t trust stories about it.


From → Science

Leave a Comment

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: