127,000 dives and plenty of questions – critical comments on the 2026 DAN study

The recent publication of the large DAN analysis attracted considerable attention within the diving medicine community. The study analyzed more than 127,000 dives to identify risk factors for decompression sickness (DCS). Based on the results, the authors even propose an algorithm designed to calculate a diver’s individual DCS risk. Some of the identified factors seem quite plausible: higher decompression stress, longer dive times, or greater depths are, as expected, associated with an increased risk. Other findings, however, raise questions and deserve a more critical examination.

For those who wish to delve deeper into the statistical and methodological aspects of the study, this detailed analysis is recommended. As a physician, I am particularly interested in the plausibility of risk factors. Here, I would like to focus on three findings that exemplify how statistical correlations can appear at first glance to be genuine risk factors, even though confounding factors or other methodological effects are likely at play.

Women face a Higher Risk of DCS – Really?

I have already addressed the observation that women showed a higher risk of decompression sickness (DCS) in statistical modeling in this blog post. Part of this effect could be explained by differences in reporting behavior. Ultimately, the database relies on affected individuals recognizing symptoms, reporting them, and seeking medical evaluation. It is entirely conceivable that women are more likely to take even milder symptoms seriously and report them, while men are more likely to downplay symptoms or simply sit them out. In this case, the database would not necessarily capture more actual DCS events among women, but rather more reported DCS events. Such a reporting bias could explain at least part of the observed gender effect.

Two other results are particularly striking: First, a lower body mass index (BMI) appears to be associated with a higher risk of DCS. Second, divers with more repetitive dives seem to have a lower risk of DCS.

The BMI finding: Are lean divers really at greater risk?

The DAN analysis found a statistical association between a lower BMI and an increased risk of DCS. Put simply, leaner divers appeared to develop decompression sickness more frequently than heavier divers.

But the question immediately arises: Does this even make biological sense? For decades, the opposite was generally discussed. Since nitrogen dissolves well in fatty tissue, being overweight was long considered a potential risk factor. While the evidence for this was never particularly convincing, the idea that a low BMI, of all things, is now supposed to be problematic is surprising.

At this very point, it is worth taking a look at a well-known phenomenon in medicine: the so-called Smoker’s Paradox.

The Smoker's Paradox

In the 1980s and 1990s, several registry studies of patients with acute heart attacks reported that smokers appeared to have a better prognosis than nonsmokers. In some cases, mortality was even lower in nonsmokers. At first glance, one might have concluded that smoking protects against the consequences of a heart attack—a conclusion that is biologically hardly credible.

Upon closer analysis, however, it became clear that smokers and nonsmokers in these studies were by no means comparable. On average, smokers suffered heart attacks significantly earlier in life. They were often ten or more years younger than nonsmokers and therefore had fewer comorbidities such as diabetes, heart failure, or chronic kidney disease. In addition, the groups differed in numerous other characteristics that influence survival after a heart attack.

The seemingly better prognosis was therefore not attributable to smoking itself. Rather, smoking was linked to other characteristics that influenced the outcome. Once the analyses took these differences into account more effectively, the supposed protective effect largely disappeared.

The Smoker’s Paradox is now considered a textbook example of how observational studies can lead to misleading conclusions when the groups being compared differ systematically or when important confounding factors are not fully accounted for.

A similar problem with BMI?

Exactly such a mechanism could also be behind the BMI findings of the DAN analysis. A low BMI may not describe the actual risk, but merely a certain type of diver. If, for example, slender divers undertake more demanding dives because they are younger and fitter, the observed association could in reality be attributable to these activities. BMI would then not be an independent risk factor, but merely a proxy for other influencing factors that were not fully recorded in the database or statistically accounted for. The finding would thus not mean that a low BMI causes DCS, but merely that both characteristics occur more frequently together in certain groups of divers.

Why are repetitive dives supposed to be protective?

Even more remarkable is that divers who make more repetitive dives appear to have a lower risk of DCS. This directly contradicts everything we know about decompression physiology.

Every repetitive dive takes place at a certain level of residual saturation. This is precisely why all decompression models explicitly account for repetitive dives. No one would seriously claim that additional nitrogen loading suddenly has a protective effect.

So how can such a correlation arise? The answer likely lies in a classic statistical fallacy.

A statistical illusion

Let’s imagine two divers. Diver A completes four dives in a single day. Diver B develops DCS after just the first dive. Which of the two will have more repetitive dives in the database at the end of the day? Diver A, of course. Diver B will not make a second, third, or fourth dive after symptoms appear.

This creates a seemingly paradoxical picture: The group with many repeat dives inevitably consists of people who have not developed DCS up to that point. The group with few repeat dives, on the other hand, automatically includes all divers whose dive day was cut short by DCS.

The result is a classic form of survivorship bias. Repeat dives do not protect against DCS. Rather, only those who have not previously experienced DCS can accumulate many repeat dives in the first place.

A well-known problem in large observational studies

Such effects are well known in epidemiology. If time-dependent variables are not modeled correctly, correlations can arise that appear to be the exact opposite of the actual causality. The statistical correlation is then real. However, the interpretation is incorrect.

What can we learn from this?

The DAN study represents an impressive dataset and provides valuable insights into real-world diving conditions. Nevertheless, it also highlights the limitations of large observational databases. Not every statistical correlation automatically reflects a biological mechanism.

If a low BMI suddenly appears as a risk factor, women are said to have a higher risk of DCS, and repetitive dives seem to protect against DCS, one should not be too quick to throw one’s understanding of decompression physiology overboard. Often, the more plausible explanation is the more likely one: The observed correlations reflect differences among divers, reporting bias, or methodological peculiarities of data collection.

The most important finding, therefore, may not be that lean divers are at greater risk or that repetitive dives provide protection. Rather, the most important finding is that correlation and causation are not the same thing.

It is particularly regrettable that the authors barely discuss these obvious methodological issues. Even more problematic is the proposal to develop an algorithm for individual risk calculation based on the identified odds ratios. For example, the model includes each additional repetitive dive with an odds ratio of 0.94. Taken literally, this would mean that the risk of DCS decreases by about 6% with each additional repetitive dive: an impossibility. At this point, at the latest, it becomes clear that it is probably not physiology speaking here, but statistics.