Impact of Different Estimation Methods on Obesity-Attributable Mortality Levels and Trends: The Case of The Netherlands

Abstract

The available methodologies to estimate the obesity-attributable mortality fraction (OAMF) affect the levels found and hamper the construction of time series. Our aim was to assess the impact of using different techniques to estimate the levels and the trends in obesity-attributable mortality for The Netherlands between 1981 to 2013. Using Body Mass Index (BMI), all-cause and cause-specific mortality data, and worldwide and European relative risks (RRs), we estimated OAMFs using three all-cause approaches (partially adjusted, weighted sum, and the two combined) and one cause-of-death approach (Comparative Risk Assessment; CRA). We adjusted the CRA approach to purely capture obesity (BMI ≥ 30 kg/m2). The different approaches led to a range of estimates. The weighted sum method using worldwide RRs generated the lowest (0.9%) while the adjusted CRA approach using 2013 RRs generated the highest estimate (1.5%). Using European-specific RRs instead of worldwide RRs resulted in higher estimates. Most of the approaches revealed an increasing OAMF over the period 1981 to 2013 especially from 1993 onwards except for the adjusted CRA approach among women. Estimates of OAMF levels and trends differed depending on the method applied. Given the limited available data, we recommend using the weighted-sum method to compare obesity-attributable mortality across European countries over time.

The combined impact of smoking, obesity and alcohol on life-expectancy trends in Europe

“…The techniques employed to estimate smoking-, obesity- and alcohol-attributable mortality were selected based on a careful assessment of different estimation techniques 48–50 and data availability. However, these techniques provide estimates only and should be interpreted as such.…”

Impact of obesity on life expectancy among different European countries: secondary analysis of population-level data over the 1975–2012 period

“…Based on the available data, only a fairly simple method—although commonly applied namely the weighted sum method could be applied27 to estimate the OAMFs. The application of a more advanced methodology27 could have affected the OAMFs and thus the PGLE levels, but less the trends 31. The lack of information on the uncertainty of the RRs we used limited us in estimating CIs for the OAMFs and PGLEs.…”

Methodological guidelines for the estimation of attributable mortality using a prevalence-based method: the STREAMS-P tool

“…Last but not the least, when attributing mortality to a risk factor assessing the entire spectrum of disease for which there is evidence of causal association, a causal-linked approach must be the norm [9]. In the case of tobacco, alcohol, or obesity attributable mortality, the all-cause approach is not ruled out [7,28] and is sometimes even recommended [15]. However, including pathologies not causally linked to a risk factor, an all-causes approach overestimates the figures of smoking attributed mortality [8,29,30] or produces slight differences for alcohol related mortality [31].…”