How under-reporting undermines credibility of alcohol-cancer studies

The connection between alcohol consumption and certain types of cancer, notably breast cancer, has been increasingly reported with ever larger studies, but they should all be taken with a correspondingly expanding grain of salt.  One reason is underreporting of actual alcohol intake, which universally occurs because population studies rely on self-reporting of amount and type of alcohol consumed. Self-reporting is plagued by recall bias, so no matter how large the study it may only magnify underlying statistical distortions.

I recently heard a presentation on this from Nicole Tran MD, PhD, who is taking up the reins of epidemiologic research at Kaiser Permanente from Arthur Klatsky, a renowned figure in the field. Kaiser has a massive database of lifestyle factors and health outcomes going back decades, and this has yielded significant insights into alcohol and health. When taking a fresh look at cancer risk, they noticed a departure from the classic J-shaped curve that we know describes wine and other health issues: even light-to-moderate drinkers appeared to be at increased risk for certain cancers. Was this real or anomalous?

Everyone under-reports, but not equally

Epidemiologists have long been aware of underreporting, and typically adjust for underestimates of approximately 40-50% of actual alcohol consumption from population surveys. Researchers usually apply this to survey data uniformly such that they match measures of consumption based on records of overall alcohol sales or taxes. But if everyone equally underreported, the curve would look the same but shift to the right; in other words, if healthy light drinkers are really moderate, and moderate really not-so-moderate, that would be good news for oenophiles. There is evidence however that underreporting is anything but uniform. For example, a large recent Australian study[i]found substantial variations based on age, gender, and episodic drinking (i.e.,bingeing.)

Klatsky and Tran [ii]took an approach not possible in most population surveys. By cross-referencing the records from baseline medical evaluations of the more than 127,000 patients in their database with medical events relating specifically to alcohol (e.g., diagnosis of cirrhosis or hospital admissions for acute alcohol intoxication) they were able to more accurately classify them.  Based on this information, subjects were divided into 3 groups: suspected of underreporting, not suspected of underreporting, and unsure underreporting status.

Taken as a whole, the Kaiser data looked like everyone else’s: A drink a day increases risk of alcohol-related cancer by 10%, yielding a linear relationship. But looking at the subgroups changed the relationship dramatically; it turned out that a sizable proportion of the light-to-moderate drinkers were way off the mark in their estimates. This skewed the results beyond the expected underreporting fudge factor. Klatsky and Tran concluded that “the apparent increased risk of cancer among light-moderate drinkers may be substantially due to underreporting of intake.”

For breast cancer specifically, the generally accepted numbers are equally dubious. Results from various studies are still inconsistent and reconciling these with known mechanisms of carcinogenesis remains challenging.  A recent review[iii]by a former director of the NIH’s Institute on Alcohol Abuse & Alcohol is noted that “the current state-of-knowledge about alcohol and breast cancer association is ambiguous and confusing to both a woman and her physician.”Recognizing how underreporting distorts the picture makes things a bit less ambiguous, if a lot more complicated.

[i] Livingston M, Callinan SJ.  Underreporting in alcohol surveys: whose drinking is underestimated? Stud Alcohol Drugs. 2015 Jan;76(1):158-64.

[ii] Klatsky AL, Udaltsova N, Li Y, Baer D, Nicole Tran H. Moderate alcohol intake and cancer: the role of underreporting.  Cancer Causes Control. 2014 Jun;25(6):693-9.

[iii] Zakhari S, Hoek JB. Alcohol and breast cancer: reconciling epidemiological and molecular data. Adv Exp Med Biol. 2015;815:7-39.

One Comment Add yours

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 )

Google photo

You are commenting using your Google 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 )

Connecting to %s