Fuzzy estimates of BAC

The following article I wrote was published in the The Network: News From the Front email newsletter on April 18, 2008. Because it does not appear that this newsletter is still being archived on the web, I've placed it here as well:

If you've been following the controversy over the term binge drinking you'd know that a few years back, the National Institute for Alcohol Abuse and Alcoholism (NIAAA) settled on a new definition. They chose to eschew the often-used consumption-based definition of 5 or more (4 for women) drinks in favor of a new criterion. Their definition is "a pattern of drinking alcohol that brings blood alcohol concentration (BAC) to 0.08 gram percent or above. " Unfortunately, they included an example that some have latched on to as the definition: "this pattern corresponds to consuming 5 or more drinks (male), or 4 or more drinks (female), in about 2 hours." But while some have adopted this 2-hour timeframe as the new 5+/4+ definition, doing so misses the point. There are lots of ways to reach 0.08, and NIAAA was offering an example of just one. So if you're interested in using BACs as a criterion for risky drinking, then you'll need a way to convert consumption patterns into resulting BACs. Estimation formulas are the answer. But what's the science behind those estimates? Are the estimates valid?

Two recent studies call into question our ability to use self-reported consumption to accurately estimate BACs for college students. First, Hustad and Karey (2005) reported that when they compared breath-test alcohol concentrations (BrAC) with estimated BACs (eBAC) among college-student partygoers, the correlations were good, but the amount of variance between the estimates and the BrAC was substantial. Estimates were particularly off when consumption was high. Second, Clapp and his colleagues (myself included) similarly found deviations between estimates and BrAC (Clapp, et al, 2006). In fact, we found that only 24% of the cases yielded eBACs that were within ± 0.02 g/dl of the BrAC. And for those 76% of the cases that were off, the margin of error was huge. Worse yet, the error was in part dependent on factors such as party size and consumption amounts.

Where do the formulas go wrong? It's difficult to say with the existing research; however, the problem likely rests in a number of factors. One is that the estimation formulas rely on accurate reporting of consumption in "standard drink" units. In the U.S., that's a drink with 14 grams of alcohol. Even with simple beverages like canned beer, alcohol content can vary by as much as 50% from brand to brand. In other words, 4 cans of one brand can equal 6 cans of another. So it is not surprising that subjects find it difficult to report accurately. Indeed, research has been quite clear that people find reporting standard drinks problematic. Aside from this, others may fail to properly track the number of drinks, or the drinking event's duration; intoxication likely interferes directly with report accuracy. The motivation to fudge not only consumption but also reported weight might also play a role. These factors compound error stemming from estimation formulas that gloss over wide variability within individuals on alcohol absorption and metabolism rates.

What can we do? Well, until better estimation procedures are developed, we need to be more cautious about the statements we make. Our drink-quantity advice needs to reflect the variability between students. And even our program evaluation activities need to treat eBACs humbly. An intervention may result in more accurate reporting, changing resulting eBACs without any change in the underlying drinking behavior or problems.