Twenty-four people put on a wristband, drank three beers each in a lab, and blew into a breathalyzer again and again while the band logged the alcohol coming out through their skin. The question was simple. Does the wrist agree with the breath?

Mostly, yes. But the band also picked up something about who was wearing it, and that part interests me more.

I build an app that estimates blood alcohol, so I keep an eye on wearables that try to measure it directly. The newest consumer one is the BACtrack Skyn. It reads TAC (transdermal alcohol concentration), the alcohol that diffuses out through your skin. The appeal is obvious. No tube to blow into, no drinks to log, just a continuous, objective record. So how close does it get to breath?

What they did

The team recruited 24 adults. Average age was 34.4, and 55% were men. About a third (32%) were living with HIV. Alcohol research in people with HIV needs tools that work for them too, so the researchers wanted to know whether the sensor behaves the same way in both groups.

Across 40 lab sessions, everyone followed the same protocol: three 4.5% beers. They wore the Skyn the whole time and gave repeated breath samples, measured as BrAC (breath alcohol concentration).

Then they compared the two signals three ways:

  • peak vs. peak: the highest wrist reading against the highest breath reading
  • total vs. peak: the AUC (area under the curve), meaning all the alcohol the band registered over the session, against the breath peak
  • the full time series, point by point

They also checked which personal traits moved the wrist readings once the breath peak was accounted for: age, race and ethnicity, HIV status, BMI (body mass index), and AUDIT score (a standard questionnaire about drinking habits).

What they found

The wrist and the breath moved together. How tightly depended on the comparison.

  • Peak readings correlated modestly across the whole group: r = 0.36 to 0.39.
  • Point by point over the session, the match was much stronger: r = 0.64, with no difference by HIV status.
  • Peak-to-peak, both HIV groups looked alike (around 0.5). The total-exposure comparison split: 0.57 in people without HIV and 0.33 in people with HIV, and that weaker link wasn't statistically significant.

The modest peak numbers have a dull explanation, and the authors flag it themselves. Everyone drank exactly the same amount, so breath peaks landed in a narrow range. When the reference barely varies, even a decent instrument can't show a strong correlation with it. A full session has far more spread to work with: rising, topping out, coming down. That's my guess for why it scores higher.

Now the part I keep coming back to. Even with the breath peak held constant, several traits shifted what the wrist reported:

  • Age. Older participants had slightly higher peak wrist readings.
  • HIV. People with HIV had higher peak readings and a larger total area. The effect on total area was the biggest in the study.
  • AUDIT. Higher scores went with higher peak and total readings.
  • Ethnicity. Participants who weren't non-Hispanic White had lower peak readings.
  • BMI. Slightly smaller total area at higher BMI.

The paper doesn't explain why any of these traits change the skin signal, and I'm not going to invent a mechanism. With 24 people, some of these effects may shrink in a larger sample. The authors say as much.

What it means

First, the curve beats the peak. In this study the whole-session shape tracked breath alcohol far better than any single top reading. If you wear one of these, trust the line over the night more than the highest number on it.

Second: same drinks, different readings. Everyone had three identical beers, and the band still read them differently depending on age, health status, and drinking history. Strictly, that's a finding about the sensor. It still rhymes with something I've written about before. Alcohol peaks can vary more than 2.5x between people, and they arrive at different times.

And objective data is getting closer, slowly. Self-reported drinking is shaky, which is why researchers keep looking for ways to stop relying on memory. A wristband is one route. But a lab session with one fixed dose is a long way from a Friday night, and the authors want bigger, more diverse samples before anyone leans hard on these numbers.

This is also why I don't think of BAC as one number that fits everyone. Two people with the same three beers aren't in the same place. AlcoBalance tunes its estimate to you, not to averages, so on a given night you can see where you are relative to your own peak and decide about the next drink while you're still steering the evening.

As for the Skyn: it follows breath alcohol reasonably well across a session, less well at the peak. And who's wearing it changes the readout.

Source: Alcohol, Clinical and Experimental Research, DOI