SAVE THE DATE!

The rhetoric used to describe the findings of a second government report on alcohol and health designed to inform federal drinking guidance nudges up against the argument that there is "no safe level" of alcohol consumption. But as is often the case with research used to advocate that position, the data doesn't quite support it as directly as the authors suggest. That is, if the data can be trusted in the first place. A panel of six scientists gathered by the Interagency Coordinating Committee on the Prevention of Underage Drinking (ICCPUD) present in the draft report some highly technical details about some of the sophisticated math they used to arrive at their conclusion that "the risk of dying from alcohol use begins at low levels of average use" and the risk for 7 types of cancer "begins with any alcohol use." However, their full methodology is still unknown. And key aspects of it involved data manipulation and opaque processes that can not be reviewed or replicated based on the document released this week.By the time the draft was released on Tuesday, advocates on either side were ready. Public health advocates with close ties to some of the report's authors cheered its conclusions and urged the Dept of Health & Human Services (HHS) to lower drinking thresholds in the next edition of the US Dietary Guidelines for Americans (DGA). A coalition of industry trade assns continued efforts to undercut the ICCPUD panel's work, claiming its "flawed" process was "rife with bias." Supportive legislators also weighed in to continue allegations about lack of transparency leveled against ICCPUD and HHS for much of 2024.With a public comment period about both this report and the one released by the National Academies of Science, Engineering and Medicine (NASEM) late last year open until at least mid-Feb, plenty more reactions are yet to come. Yet major questions remain about what folks at HHS will do with those comments, who will actually be holding the pen to craft the alcohol guidance and how the agency might change under the incoming administration.Data Disconnects: No Clear Risk at Low Consumption Levels; Low Confidence? Among the most core findings and conclusions of the ICCPUD draft report estimates the "risk of dying from alcohol use" at different drinking levels. That risk is 1 in 1000 at over 7 drinks per week and 1 in 100 at over 9 for both male and female drinkers, it claims. Elsewhere, the report details the relative risk curves (or RRs) for a variety of outcomes, emphasizing inflated risks for certain cancers, infectious diseases and injuries at even 1-3 drinks per week. Ref 1Graphs accompanying those results do not show the RRs for non-drinkers as a reference point, thereby eliminating the possibility of a J-curve. However, they do show that the risks for females and males consuming less than 7 and 6.5 drinks per week, respectively, are negative. That implies a net-benefit at low levels of consumption. And in fact, detailed tables in the appendix confirm that those better outcomes are driven by lower risks for cardiovascular diseases.Furthermore, tables detailing the "lifetime risk of an alcohol attributable death" for different age groups emphasize that the risks are not clear until about 14 drinks per week for either sex. Authors estimated that 1.7 out of 1000 males consuming 7 drinks per week will die from an alcohol-attributable cause. But the range of numbers that they're 95% certain the definitive result falls within (called a "confidence interval") is very broad. It ranges from almost -29 to over 29. (Again, negative numbers suggest a net-benefit.) Among females, the risk is just 0.15 per 1000 at 7 drinks a week with a confidence interval almost as broad (-19 to about 22).Generally speaking, such wide ranges mean there isn't a lot of certainty. And if the confidence interval includes zero, there could be no net effect at all. Only at 14 drinks per week do both ends of the range land above 0 for either sex, with the estimated risk at around 40 per 1000. Put another way, authors acknowledge that the same 1 in 1000 and 1 in 100 lifetime risk levels could be experienced from consuming as little as 1 drink per week or as many as 14, depending on sex. Despite these wide ranges, co-author and ICCPUD panel member Tim Naimi confidently told the press this week that "what a lot of folks may have previously considered 'moderate' drinking is actually moderately risky."Methodology: Opacity and Manipulation But can folks outside of the panel &mdash; whether drafters of coming drinking guidance or the American public &mdash; be equally confident in the panel's conclusions or their data? The draft report purports to explain the researchers' "methods." The key effort "to estimate alcohol-specific risk relationships comprised two steps," they wrote. But both steps involve opaque processes that are not fully explained as well as adjustments to data inputs by the panel and other unnamed participants.To determine the all-important relative risk (RR) curves for each condition related to alcohol consumption, ICCPUD's group of six researchers started by doing a systematic review of systematic reviews, whittling down the number of papers analyzed from almost 7300 to just 56. As little as 2-4 papers were analyzed for certain health outcomes. The group then handed off those results to five additional "panels of experts," each chosen for authoring the most studies on the topic at hand in the past decade. But the report does not disclose the names of those panelists. Nor does it share a copy of the "questionnaire" each panelist completed or how their responses were combined to determine the ultimate RRs included in the report. Importantly, almost all of the report's findings flow directly from the black boxes of these expert panels.The resulting RRs were then applied to estimates about alcohol consumption among the US population. But not before the core set of six researchers also adjusted that data in an attempt to correct for underreporting. Average consumption by different age groups were adjusted up; national per capita consumption data based on sales was adjusted down; estimates for unrecorded alcohol were added; data from studies based on self-reported consumption were changed assuming participants under-reported their drinking by 10%.Limitations: Garbage In, Then "Corrected" For? The above manipulation of sources used by these researchers to measure the health risks associated with drinking speaks to a great irony in the report. The closing section on the report's "limitations" reminds that it is "informed primarily by observational studies." But the limitations to that research "are considerable," including the well-worn territory of former-drinker biases skewing results. Authors then cite problems with cohort studies, which also underpin their analysis. Data adjustments attempt to correct for these perceived problems. In other words, the panel fed admittedly problematic research into an opaque set of processes too complicated to explain to pull out a set of results that purportedly support a set of favored policies.Citations: Heavily Self-Referential & Late Additions Over a quarter of the report's citations (35 of 129, by our count) point back to papers co-authored by one of the six panelists. That doesn't include another couple dozen citations to reports by various federal agencies or global orgs, to which these researchers may have also contributed. Among those citations: the Surgeon General's advisory on alcohol and cancer released exactly 11 days before this draft report was published, raising questions about the extent to which it was peer-reviewed prior to that.Additional Questions: All in the Timing; Patterns After All Any number of other decisions made by ICCPUD's panelists could (and probably will) be questioned. We'll focus on just two more. In its introduction, the report used 2023 federal survey data, but then uses changes in per capita consumption between 1999 and 2021 to suggest that alcohol consumption is a growing problem. Recall that consumption declined sharply in 2023 with at least as large of a drop seen in 2024. To calculate the burden of disease attributable to alcohol, authors used data from 2022, when deaths were still inflated because of the pandemic. They also cite research from 10-15 years ago to argue that adult consumption impacts teen drinking. But teen drinking has declined by 40-60% in that time, depending on the age group.Finally, recall that in their draft protocol issued last spring, these panelists maintained that they would not look at research that attempted to account for different patterns of drinking, sticking instead to analyzing studies that used weekly averages. Yet the report repeatedly insists that "drinking patterns shape risk" and "even infrequent high per-occasion drinking may eliminate the lower levels of risk for" certain outcomes like ischemic stroke. Authors emphasize the increased risk of patterns like binge drinking and point out that using weekly averages as a measure of consumption can conflate regular low-volume drinkers with infrequent heavy drinkers. They do not, however, attempt to parse how that conflation might skew data to make, say, 7 drinks per week appear more risky than consuming 1 drink per day.Public statements and responses to the report only added more grist for the mill during an incredibly rich period of discussion about alcohol's impact on health. Most headlines continue to focus on the Surgeon General's advisory, but these reports and the announcement late this week that the TTB plans to require mandatory Alcohol Facts labeling only fuel the fire more. Those statements and responses, public discussion and press, ongoing TTB rulemakings and more will all receive attention in subsequent issues.Cheers,References1 Shield, K, et al, "Draft Report: Scientific Findings of the Alcohol Intake & Health Study for Public Comment," Interagency Coordinating Committee on the Prevention of Underage Drinking, Jan 2025.

Archives Search

Keyword

Publication Name

Publication Date

Year

Volume

Issue #

ICCPUD Report Emphasizes Risk, But Data Disconnects, Questions Abound

Publishing Info