Calculating Baseline Risk For Non-Drinkers From Population Lifetime Risk Estimates: Communicating Alcohol Guidelines

Harriet RUMGAY, Cancer Research UK, United Kingdom

1 Analysis and Evaluation Team, Cancer Research UK, London, United Kingdom

Purpose: National guideline limits for the consumption of alcohol should be communicated using robust calculations of risk of disease at different levels of consumption. Using population lifetime risk figures as the baseline for these calculations overestimates the risk for alcohol drinkers, because population lifetime risk includes all levels of exposure, from non-drinkers to heavy drinkers. We present a methodology to calculate a reasonable baseline (lifetime risk for non-drinkers), using the recently revised UK alcohol consumption guidelines as an example, and present the differences in absolute risk for alcohol drinkers when calculated using the general population, or non-drinkers, as the baseline.

Methods: Risk figures based on two baselines were compared. The first uses UK 2012 lifetime risk estimates for all cancer types classified by IARC as caused by alcoholic beverages: breast (female), colorectal, laryngeal, liver, oesophageal, oral cavity and pharyngeal cancers. The comparator was a non-drinkers baseline calculated by applying the population distribution of a range of drinking levels to the relative risks (RRs) for those drinking levels, and combining them to find the lifetime risk for non-drinkers which, when used as the baseline to calculate lifetime risk for different levels of drinking, gives a total population lifetime risk as near as possible to the general population baseline.

Results: The two different baselines generated markedly different lifetime risk estimates for all cancer types. The largest differences were found in cancer types where the relative risk with alcohol consumption is higher, and in populations where higher alcohol intake levels are more common.

Conclusions: Reasonable baseline lifetime risk estimates can be produced using population distribution of risk exposure and relative risk at each exposure level. This methodology can be applied to other risk factors, such as tobacco or overweight, where exposure prevalence and RRs are known.

Funding source: Cancer Research UK.