A Break in the Obesity Epidemic? Explained by Biases Or Misinterpretation of the Data?
Total Page:16
File Type:pdf, Size:1020Kb
International Journal of Obesity (2015) 39, 189–198 © 2015 Macmillan Publishers Limited All rights reserved 0307-0565/15 www.nature.com/ijo REVIEW A break in the obesity epidemic? Explained by biases or misinterpretation of the data? TLS Visscher1,2,3, BL Heitmann4,5,6, A Rissanen3,7, M Lahti-Koski3,8 and L Lissner3,9 Recent epidemiologic papers are presenting prevalence data suggesting breaks and decreases in obesity rates. However, before concluding that the obesity epidemic is not increasing anymore, the validity of the presented data should be discussed more thoroughly. We had a closer look into the literature presented in recent reviews to address the major potential biases and distortions, and to develop insights about how to interpret the presented suggestions for a potential break in the obesity epidemic. Decreasing participation rates, the use of reported rather than measured data and small sample sizes, or lack of representativeness, did not seem to explain presented breaks in the obesity epidemic. Further, available evidence does not suggest that stabilization of obesity rates is seen in higher socioeconomic groups only, or that urbanization could explain a potential break in the obesity epidemic. However, follow-ups of short duration may, in part, explain the apparent break or decrease in the obesity epidemic. On the other hand, a single focus on body mass index (BMI) ⩾ 25 or ⩾ 30 kg m−2 is likely to mask a real increase in the obesity epidemic. And, in both children and adults, trends in waist circumferences were generally suggesting an increase, and were stronger than those reported for trends in BMI. Studies concluding that there is a recent break in the obesity epidemic need to be interpreted with caution. Reported studies presenting a break were mostly of short duration. Further, focusing on trends in waist circumference rather than BMI leads to a less optimistic conclusion: the public health problem of obesity is still increasing. International Journal of Obesity (2015) 39, 189–198; doi:10.1038/ijo.2014.98 INTRODUCTION salads in fast-food restaurants. Some go even beyond the It took some time, but public health policy makers have now ‘traditional’ energy-intake and energy-expenditure explanations developed a common worry regarding the obesity epidemic, when explaining a potential levelling off in obesity levels, by based on strong epidemiologic studies showing increases in body pointing at a similarity with infectious diseases epidemics that are 13,14 mass index (BMI) throughout the world, during the past three or also followed by levelling off. four decades.1 During recent years, however, several reports have However, although these claims and theories are appealing, suggested a potential break or even decrease in the obesity encouraging and interesting, it may be too early for optimism. epidemic.2–7 By a ‘break in the obesity epidemic’ others meant Recent review studies on the trends in obesity make clear that 2,4,7 4 fi there is large heterogeneity with respect to the quality and stabilization, or plateauing, de ned as no changes occurring 2,5,15–17 in the prevalence of obesity over the period studied, as a ‘levelling comparability of available data. There may be biases and off’ where an increase was followed by stabilization, or as distortions in the data that could explain these trends. Before 7 fi concluding that a plateauing is taking place, the validity of the reversing de ned as decreasing rates. 16,17 Suddenly, based on those recent studies, optimistic views are presented data should be discussed more thoroughly. And, in heard that we may be entering a new era, in which the obesity addition to biases and distortions in the data itself, there may be epidemic might come to an end. biases in what we are looking at and how we are interpreting There is reason for an optimistic interpretation of recent studies the data. suggesting a halt in the obesity epidemic. Decreases and breaks in Potential biases that may affect the study of time trends in the obesity epidemic are being presented in studies from across obesity are varying participation rates, self-reported data on body weight and height, sample sizes and representativeness of data, the globe, and they represent a wide variety of subgroups. Policy and variation in socioeconomic background and difference in makers attribute potential decreases in obesity to their policy urban and rural data. efforts, and city councils and related interdisciplinary consortia point to successes coming from large, integrated community- based prevention and management programs.8–12 The food Decreasing participation rates industry is happy to claim their part of the success by making Non-participation is generally higher in the obese than in the reference to increased sales of low-calorie products, and offering general population.18–20 A Finnish study suggested a difference in 1Research Centre for the Prevention of Overweight (Zwolle), Windesheim University of Applied Sciences and VU University, Zwolle, The Netherlands; 2Institute of Health Sciences, VU University, Amsterdam, The Netherlands; 3Prevention and Public Health Taskforce, European Association for the Study of Obesity, London, UK; 4Institute of Preventive Medicine, Bispebjerg and Frederiksberg Hospitals, The Capital Region, Copenhagen, Denmark; 5National Institute of Public Health, University of Southern Denmark, Copenhagen, Denmark; 6The Boden Institute of Obesity, Nutrition, Exercise and Eating Disorders, University of Sydney, Sydney, New South Wales, Australia; 7Helsinki University Central Hospital, Helsinki, Finland; 8Finnish Heart Association, Helsinki, Finland and 9Department of Public Health and Community Medicine, University of Gothenburg, Gothenburg, Sweden. Correspondence: Dr TLS Visscher, Research Centre for the Prevention of Overweight (Zwolle), Windesheim University of Applied Science and VU University, PO Box 10090, 8000 GB Zwolle, The Netherlands. E-mail: [email protected] Received 16 July 2013; revised 14 March 2014; accepted 1 April 2014; accepted article preview online 9 June 2014; advance online publication, 22 July 2014 Potential break or decreases in the obesity epidemic TLS Visscher et al 190 response rates between different educational levels, and also that studies of long duration that show increases in obesity, do at the this so-called selective participation had increased over the past same time also show short-term decreases.38–43 Therefore, it is 25 years.21 Such selective non-participation is an unwanted important to study whether apparent breaks and levelling-offs can phenomenon in trend studies, as it may lead to severe bias. be seen from both studies with a short and long duration, or from Therefore, it is important to study whether participation rates are studies with a short duration only. decreasing over time, which may mask an increase in the obesity trends. Distribution of BMI Levels of BMI have generally been seen to increase mainly in the Self-reported versus measured body weight and height upper centiles.6,33,41,44–46 In other words, increases in obesity have Obese subjects have a higher tendency to under-report their body been stronger than increases in overweight, and those who have weight than normal weight subjects.22 As a result, use of self- become obese, have become even more obese. A focus on mean reported body weight and height may lead to increasingly biased values of BMI or percentage of overweight may therefore mask obesity rates in case obesity rates are increasing. It is therefore increases in obesity. And, especially relevant when the prevalence important to study whether presented breaks in the obesity of obesity is already high, a focus on obesity alone may mask epidemic may be due to the use of self-reported data. increases in more extreme categories. Thus, it is important to study whether studies suggesting a break or levelling off have Sample size and representativeness studied different percentiles or different obesity categories. Over the past few decades, trend studies required optimal methodology including large numbers of participants and BMI and waist circumference representativeness to get published. Since the increase in obesity BMI is the most frequently used measure to estimate fat levels in to date is old news, it is a possibility that increases in obesity are individuals and to study trends in populations. BMI, however, is a being published only in those cases when the methodology is difficult measure to interpret, as the ‘mass’ reflects both fat mass very strong. At the same time, small studies lacking optimal and lean mass.47 Therefore, at least theoretically, decreasing levels methodology are more likely to be published when showing of BMI do not necessarily reflect decreases in fat mass. In case lean unexpected changes in obesity rates such as stabilizing or mass levels are decreasing, for instance, owing to decreased levels decreasing rates.23,24 It is therefore important to study whether of activity, fat mass could increase, while body mass remains evidence for a break in the obesity epidemic is mainly coming stable. In adults, waist circumference is reported to be a better 48–50 from studies with small and non-representative sample sizes. predictor of coronary heart disease and diabetes than is BMI. Also, concerns are being raised whether the BMI is the most optimal measure for fatness in children. Although BMI and waist Trends across categories of socioeconomic background circumference seem to be equally useful in predicting cardiovas- Obesity is more prevalent in lower than in higher socioeconomic 51 25–34 cular risk in children, BMI may not be the best marker of success categories. It is therefore important to study whether breaks in interventions. Doak et al.,52 for instance, showed that some or levelling-offs in obesity are seen across all socioeconomic interventions not having an impact on BMI levels did have an fi categories or in speci c categories only. impact on skinfold thickness.