- Email: [email protected]
Impact of Public Health Interventions on Obesity and Type 2 Diabetes Prevention: A Simulation Study
ARTICLE IN PRESS
RESEARCH ARTICLE
Impact of Public Health Interventions on Obesity and Type 2 Diabetes Prevention: A Simulation Study D1X XRoch A. Nianogo, D2X XMD, MPH, PhD,1,2 D3X XOnyebuchi A. Arah, D4X XMD, MSc, DSc, MPH, PhD1,2,3,4
Introduction: Little is known about what interventions worked or did not work in slowing the obesity epidemic. The long-term comparative effectiveness of environmental and behavioral public health interventions for obesity and type 2 diabetes prevention over an individual’s life course is relatively unexplored. The potential impact and long-term collective effectiveness of environmental and behavioral interventions on obesity and type 2 diabetes throughout the life course was evaluated.
Methods: The Virtual Los Angeles Obesity Model developed in 2016 was used to estimate the incidence and prevalence of obesity and type 2 diabetes under current and hypothetical interventions among 98,000 individuals born in 2009 and followed from birth to age 65 years. Analyses were performed in 2016 and completed in 2018.
Results: The 48-year risk of type 2 diabetes was 0.533 (95% CI=0.446, 0.629) under the natural course, 0.451 (95% CI=0.334, 0.570) under the physical activity intervention, and 0.443 (95% CI=0.389, 0.495) under the fast-food intervention. The 64-year risk of obesity was 0.892 (95% CI=0.879, 0.903) under the natural course, 0.876 (95% CI=0.850, 0.899) under the physical activity intervention, and 0.864 (95% CI=0.856, 0.873) under the fast-food intervention. The other interventions had little or no long-term effects. When all the interventions were applied, the population risk ratios were 0.942 (95% CI=0.914, 0.967) and 0.634 (95% CI=0.484, 0.845) for obesity and type 2 diabetes, respectively. Conclusions: Implementing health interventions continuously throughout the life span and in combination with other interventions could substantially halt the obesity and the type 2 diabetes epidemics. Am J Prev Med 2018;000(000):1−8. © 2018 American Journal of Preventive Medicine. Published by Elsevier Inc. All rights reserved.
INTRODUCTION
O
besity remains a major public health problem in the U.S.1 and worldwide,2 and has been one of the most predominant players in the increase of the incidence of type 2 diabetes mellitus (T2DM).3 Despite ongoing efforts, the rates of obesity have remained persistently high, especially in many disadvantaged groups, but have begun leveling off among children in the U.S.4 However, little is known about what interventions worked or did not work in slowing the obesity epidemic over the course of decades. Further, similar questions of increasing importance remain: which interventions were impactful over the long run and which interventions yielded the greatest impact?
When should these interventions be implemented or to whom should they be given? Answering these questions often entail an evaluation of the interventions that have already been implemented5 and a projection of the From the 1Department of Epidemiology, Fielding School of Public Health, University of California, Los Angeles (UCLA), Los Angeles, California; 2 California Center for Population Research, Los Angeles, California; 3 UCLA Center for Health Policy Research, Los Angeles, California; and 4 Department of Statistics, UCLA College of Letters and Science, Los Angeles, California Address correspondence to: Roch A. Nianogo, MD, MPH, PhD, Department of Epidemiology, Fielding School of Public Health, University of California, Los Angeles, 650 Charles E. Young Drive South, Los Angeles CA 90095. E-mail: [email protected]. 0749-3797/$36.00 https://doi.org/10.1016/j.amepre.2018.07.014
© 2018 American Journal of Preventive Medicine. Published by Elsevier Inc. All rights reserved.
Am J Prev Med 2018;000(000):1−8
1
ARTICLE IN PRESS 2
Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
impact of potential interventions.6 Information about which interventions are most efficacious could assist policymakers in their decision making and help avoid the squandering of resources on ineffective interventions, while allocating more resources on interventions that have the potential to be successful. The long-term evaluation of policy interventions for obesity and T2DM prevention implemented in largescale populations and in real settings can be quite difficult to achieve, and such evaluations can be hard to find.7 Policymakers and researchers have, therefore, relied in the past on other evaluation strategies in the hopes of generalizing results to the larger population. In fact, although RCTs remain the gold standard in evaluating interventions, their evidence is not always generalizable to the population of interest. Moreover, they are typically costly and often cannot include participants for an extended period of time.8 Others have used existing observational studies for their attractive edge (i.e., longer follow-up, less-restrictive eligibility criteria, and lower cost) to evaluate hypothetical interventions in given populations using complex methods.9 Unfortunately, the results of such endeavors, although carefully used, are subject to the misspecification of the model and the presence of uncontrolled confounding, to list a few.9 Finally, some researchers have used computer models, such as the Coronary Heart Disease Policy Model6 or Archimedes,10 to evaluate the impact of public health interventions in large-scale populations. Such models (commonly referred to as microsimulation models) are very promising, but fall short by focusing only on the individual level and not including aspects of the built-environment, a key player in the obesity epidemic.11 Agent-based models (ABMs), on the other hand, are computer simulation models that can incorporate individual characteristics as well as the social and physical environments in assessing intervention impacts.11 To evaluate the potential impact and the long-term effectiveness of environmental and behavioral public health interventions on obesity and T2DM, the Virtual Los Angeles (VILA) Obesity Model, an ABM of obesity and T2DM is used. This study proposes to evaluate interventions implemented in Los Angeles County (LAC), California, as there have been major documented efforts implemented throughout the county to curb the obesity epidemic.12,13
METHODS Study Population The VILA Obesity Model is a stochastic dynamic discrete-time ABM developed in 2016 to study obesity and T2DM in U.S.
individuals followed from birth to age 65 years. Briefly, the model has two subparts: the environment and the individuals (Appendix Figure 1, available online). The characteristics of the environment (i.e., residential makeup, food environment, and physical activity opportunities) were first simulated to represent the residential areas of LAC. Second, the individuals were simulated at birth and spread out in the 235 simulated neighborhoods. Simulated individuals were also characterized by their sociodemographics, dietary behaviors, and physical activity patterns. Each of the 98,230 simulated individuals was “born” in 2009, grew up in their neighborhood, and was followed until age 65 years. During young adulthood (18−24 years) and middle adulthood (40−49 years), the characteristics of the neighborhoods where the agent lived could change to reflect what would happen if the agent actually changed location; this was done by resampling anew from the neighborhood sociodemographics. The VILA Obesity Model was used to describe the trends in obesity and T2DM incidence and prevalence among U.S. individuals born in LAC (Forecasting Obesity and T2DM Incidence and Burden Using an Agent-Based Model. The VILA-Obesity Simulation Model, unpublished observations, 2017). Although a full-fledged ABM would ideally incorporate (1) a set of agents, (2) an environment where the agents live, and (3) a set of agent relationships and underlying conceptual/theoretic decision-making rules outlining how agents behave and interact with each other and with their environment14−16; the VILA model only included agents, their environment, and cross-level agent −environment interactions over time. It can be argued then that VILA could be thought of in the broad sense as an ABM and could be seen as an extension of a microsimulation model,17 especially because it incorporated environmental and contextual exposures (and the resulting cross-level agent−environment interactions).
Measures In this study, the terms behavioral interventions in particular, and intervention in general, were used to represent a hypothetical or counterfactual intervention in which the modeler hypothetically alters the level of the behavior or exposure status. The underlying assumption is that individuals who are exposed to an intervention intended at a behavior change, for instance, are going to engage in the intended behavior with a 100% uptake. In other words, the impact of such “hypothetical” interventions would represent the potential maximum effect of the intervention. All interventions were designed on the basis of specific behavior status. Generally, an intervention consisted of simulating and altering the current behavior status to become the desired level of that particular behavior. For instance, the breastfeeding intervention was implemented in the first year of life and included altering the breastfeeding exposure status of simulated individuals to become “breastfed exclusively for ≥ 6 months,” if not already so. Likewise, the sugar-sweetened beverage (SSB) intervention was implemented throughout the life course at eight possible timepoints and included altering the SSB consumption exposure status of simulated individuals to become “drink zero glasses of soda or other sugary drinks,” if not already so (Figure 1 and Appendix Table 1, available online). www.ajpmonline.org
ARTICLE IN PRESS Nianogo and Arah / Am J Prev Med 2018;000(000):1−8 In addition to behavioral interventions, environmental interventions were also explored. The neighborhood park access intervention was implemented at three possible timepoints: birth, young adulthood, and middle adulthood and consisted in altering the neighborhood park access exposure status to become “neighborhood with a high access to parks,” if not already so. Access to parks in neighborhoods was defined as the percentage of the population living within a quarter-mile buffer and was based on California aggregated data obtained from Wolch et al.18 All interventions were evaluated singly and in combination with one another. The first set of combined interventions, which included the five behavioral interventions (i.e., referred to as combined behavioral interventions), was composed of four dietary interventions (i.e., breastfeeding promotion, reduction of SSB consumption, reduction of fast-food consumption, and increase of fresh fruit and vegetable consumption) and one physical activity intervention (i.e., increase of the level of physical activity). The second set of combined interventions, which included both behavioral and environmental interventions (i.e., referred to as combined mixed interventions), was composed of two individual-level dietary interventions (i.e., breastfeeding promotion,
3
and reduction of SSB consumption) and two environmental physical activity−related interventions (i.e., increasing access to parks and recreations and designing pedestrian-friendly communities; Appendix Table 1, available online). Furthermore, interventions were implemented throughout the individual life course, during childhood, young adulthood, and middle adulthood. Specifically, this study defined and projected the cumulative incidence of obesity and T2DM under a natural course (i.e., no intervention, status quo), an optimistic scenario (i. e., idealistic scenario), and a pessimistic scenario (i.e., worst-off scenario; Figure 1). Childhood obesity was defined using the WHO guidelines on the basis of BMI z-scores calculated using Centers for Disease Control and Prevention−provided SAS codes for ages 2−17 years.19 A child with a BMI z-score ≥ 2 was classified as obese.20 Adult obesity was also defined using WHO guidelines for ages 18−65 years. An individual with a BMI ≥ 30 was classified as obese.21 To calculate measures of incidence, only the first time that an individual was diagnosed as being obese was considered (i.e., first occurrence of obesity among
Figure 1. Hypothetical intervention regimens implemented throughout an individual life span. Note: Throughout the life course (i.e., eight discrete time-steps from age 2 to 65 years) interventions were implemented in childhood, in young adulthood, in middle adulthood, and at all relevant time-points (i.e., optimistic or idealistic scenario) and compared to the natural course (i.e., status quo). For reference a pessimistic scenario is also implemented (i.e., worse-off scenario). In the natural course (i.e., no intervention), the individual behavior status remained unchanged throughout follow-up. In the optimistic scenario, the exposure status of the individual at time=3, 5, and 7 was simulated and altered to become “the desirable level of that particular behavior” and would remain unchanged at the other time points at time t=1, 2, 4, 6, and 8. In the pessimistic scenario, however, the exposure status of individuals was simulated and altered to become “Less desirable level of the particular behavior” a time t=1, 2, 4, 6, and 8 and unchanged at time 3, 5, and 7.
& 2018
ARTICLE IN PRESS Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
4
at-risk individuals, that is individuals who were not obese in the previous time-step). Incident T2DM between ages 18 and 65 years was the outcome of interest. The following covariates were considered: age (continuous), sex (binary), race (binary), SES (binary), marital status (binary), and family history of T2DM (binary).
Statistical Analysis Assumptions about the underlying data generating mechanisms of obesity and T2DM and the relationships between covariates, exposures, mediators, and outcomes were presented using a directed diagram (Appendix Figure 2, available online). Monte Carlo simulations and the g-computation algorithm of Robins22 were used to predict the potential cumulative incidences of obesity and T2DM under various hypothetical scenarios. Briefly, this was done by first projecting the potential outcomes under an intervention scenario and then projecting the potential outcomes under a no-intervention scenario (i.e., generally the natural course or status quo) and then taking the contrast between these two potential outcomes. To generate the mean prevalences, incidences, and population risk ratios (RRs) and 95% CI, 100 replications of the simulation model were generated by Monte Carlo sampling of the parameters from a normal distribution using parameter inputs outlined in Appendix Table 2 (available online). All analyses were performed in SAS, version 9.4.
RESULTS Table 1 and Appendix Figure 3 (available online) describe the simulated end of follow-up cumulative
incidence of obesity and T2DM under various hypothetical interventions. The 64-year risk of obesity (from 2 to 65 years) and the 48-year risk of T2DM (from 18 to 65 years) at the end of follow-up under the natural course were 0.892 (95% CI=0.879, 0.903) and 0.533 (95% CI=0.446, 0.629), respectively. The 48-year risks of T2DM under the breastfeeding, the neighborhood walkability, and the neighborhood park access interventions were 0.533 (95% CI=0.446, 0.631), 0.522 (95% CI=0.425, 0.620), and 0.525 (95% CI=0.421, 0.631) respectively. The 64-year risks of obesity under the breastfeeding, the neighborhood walkability, and the neighborhood park access interventions were 0.892 (95% CI=0.880, 0.903), 0.890 (95% CI=0.875, 0.903), and 0.890 (95% CI=0.875, 0.902), respectively. Figure 2 and Appendix Figure 4 (available online) present the cumulative incidence of incident T2DM and obesity over time under the combined behavioral interventions. The curves of the cumulative incidence of obesity under the optimistic scenario departed slightly and were lower than that of the natural course curve for children, adults, and for the population as a whole. The cumulative incidence curve of T2DM under the optimistic scenario was much lower than that of the natural course throughout follow-up. The cumulative incidence of obesity and T2DM was consistently high among the non-white segment of the population throughout follow-up. The cumulative incidence for all other
Table 1. Simulated Cumulative Incidence of Obesity and Type 2 Diabetes Under Hypothetical Interventions in VILA (n=98,230) Interventions 00-Natural course (no intervention) 01-Eliminate SSB consumption 02-Exclusively breastfeed for ≥ 6 months 03-Increase neighborhood walkability 04-Increase neighborhood access to parks 05-Engage in moderate to vigorous physical activity 06-Consume ≥ 5 fresh fruit and vegetable/day 07-Eliminate fast-food consumption 08-Combined mixed interventions (all) 09-Combined behavioral interventions (all) 10-Combined mixed interventions (childhood) 11-Combined behavioral interventions (childhood) 12-Combined mixed interventions (young adulthood) 13-Combined behavioral interventions (young adulthood) 14-Combined mixed interventions (adulthood) 15-Combined behavioral interventions (adulthood)
64-year risk of obesity
48-year risk of type 2 diabetes
0.892 (0.879, 0.903) 0.890 (0.878, 0.900) 0.892 (0.880, 0.903) 0.890 (0.875, 0.903) 0.890 (0.875, 0.902) 0.876 (0.850, 0.899) 0.888 (0.875, 0.900) 0.864 (0.856, 0.873) 0.887 (0.870, 0.900) 0.839 (0.813, 0.867) 0.890 (0.877, 0.902) 0.873 (0.851, 0.891) 0.889 (0.875, 0.901) 0.872 (0.852, 0.888) 0.892 (0.878, 0.904) 0.880 (0.860, 0.895)
0.533 (0.446, 0.629) 0.515 (0.406, 0.608) 0.533 (0.446, 0.631) 0.522 (0.425, 0.620) 0.525 (0.421, 0.631) 0.451 (0.334, 0.570) 0.522 (0.428, 0.620) 0.443 (0.389, 0.495) 0.496 (0.379, 0.590) 0.338 (0.261, 0.419) 0.529 (0.440, 0.622) 0.481 (0.395, 0.589) 0.517 (0.414, 0.616) 0.462 (0.380, 0.545) 0.544 (0.422, 0.674) 0.452 (0.375, 0.523)
Note: Data presented as M (95% CI). All interventions were implemented under the optimistic scenario. The combined mixed interventions include the SSB consumption, breastfeeding, neighborhood walkability, and the neighborhood access to park interventions. The combined behavioral interventions include the SSB consumption, breastfeeding, physical activity, fruit and vegetable, and fast-food interventions. The mean and CI were obtained from 100 repeated replications of the model. SSB, sugar-sweetened beverage; VILA, Virtual Los Angeles Obesity Model.
www.ajpmonline.org
ARTICLE IN PRESS Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
interventions evaluated singly or in combination with one another are presented in Appendix Figures 5−9 (available online). Figure 3 and Appendix Figure 10 (available online) present the population impact of various interventions on T2DM and obesity, respectively. The two single most effective interventions on T2DM prevention were the fast-food intervention (population RR=0.831, 95% CI=0.712, 0.984), followed by the physical activity intervention with RR=0.847 (95% CI=0.730, 0.951). Eliminating fast-food consumption had a small effect on obesity prevention (RR=0.969, 95% CI=0.960, 0.979). Combining all behavioral interventions yielded the greatest effect for both T2DM (RR=0.634, 95% CI=0.484, 0.845) and obesity (RR=0.942, 95% CI=0.914, 0.967).
DISCUSSION The purpose of this study was to evaluate the long-term effectiveness of public health interventions on obesity and T2DM. This study complements the previous study based on the same model that forecasted the future prevalence and incidence of obesity and T2DM in LAC. Interventions, such as breastfeeding for 6 months or longer, increasing fresh fruit and vegetable consumption, increasing neighborhood walkability or neighborhood access to parks, when implemented singly, were not effective in reducing the cumulative incidence of obesity and T2DM. Engaging in moderate to vigorous physical activity and eliminating fast-food consumption appeared to be effective in reducing the excess risk
5
in obesity and T2DM incidence. In addition, combining interventions with one another throughout the life span showed the greatest impact, especially when combining all behavioral interventions together. Furthermore, for a given effective intervention, the impact seemed greater in reducing T2DM risk than in reducing obesity risk. Lastly, to have an impact, most interventions needed to be implemented at all possible timepoints (i.e., optimistic scenario). Interestingly, interventions implemented in childhood appeared more effective in reducing obesity risk than interventions implemented in young adulthood or middle adulthood; and interventions implemented in middle adulthood appeared more effective in reducing T2DM risk than interventions implemented in young adulthood and childhood. These results highlight many important insights worth mentioning. First, some interventions are more effective than others. Second, some periods in the life course appear to be more critical than others in the prevention of obesity or T2DM. Third, to be effective, most interventions have to be implemented continuously (virtually at every stage of life) and have to be implemented together. Fourth, the modest impacts of the interventions evaluated here testify to the persistence of obesity and T2DM and to the difficulty to curb these epidemics, which might explain why there has been only a slight leveling off of childhood obesity after many years of prevention efforts. The findings provide supporting evidence for the beneficial effects of engaging in physical activity, as is already established in the literature.23 Contrary to expectations, eliminating SSB throughout the life course
Figure 2. Cumulative incidence of type 2 diabetes under a combination of interventions. Note: The interventions include the sugar-sweetened beverage (SSB), breastfeeding (EBF), moderate to vigorous physical activity (MVPA), fruit and vegetable (FFV), and fast-food (FFD) interventions. (A) Type 2 diabetes cumulative incidence among adults aged 18−65 years; (B) type 2 diabetes cumulative incidence among adults aged 18−65 years by race and ethnicity. The mean cumulative incidences were obtained from 100 repeated replications of the model.
& 2018
ARTICLE IN PRESS 6
Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
Figure 3. Population impact (in terms of risk ratios) of various interventions on the cumulative incidence of type 2 diabetes. Note: The combined mixed interventions are the interventions that include the sugar-sweetened beverage consumption, breastfeeding, neighborhood walkability, and the neighborhood access to park interventions. The combined behavioral interventions include the sugar-sweetened beverage consumption, breastfeeding, physical activity, fruit and vegetable, and fast-food interventions. The mean and CI were obtained from 100 repeated replications of the model.
did not reduce the risk of obesity. Nevertheless, even though obesity rates did not decline as a result of this intervention, any weight stabilization could be considered beneficial, especially in a context where weight may be generally rising.24 In this study, breastfeeding for 6 months or longer had no overall effect over the life course of an individual. Although there is some evidence as to the benefits of breastfeeding in childhood obesity prevention,9 its role has been debated in the literature.25 It may be that breastfeeding results in a short-term effect on obesity rates that dissipates over time; this is consistent with
what was found in post-hoc analyses (Appendix Figure 11, available online). Contrary to expectations,26 living in a highly walkable neighborhood or neighborhood with high access to parks did not appear to reduce obesity and T2DM incidence in this study. Other studies have also reported inconsistent or no effect of the built-environment on BMI.27 The absence of an effect of the built-environment on obesity and T2DM in this study could be because of the presence of other barriers to engaging in physical activity in the neighborhood (e.g., safety) or other aspects of the food environment that could potentially outweigh the www.ajpmonline.org
ARTICLE IN PRESS Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
importance of the neighborhood physical activity characteristics. Furthermore, policies, systems, and environments (PSE), such as having a walkable neighborhood, may affect population health indirectly and slowly, if at all.28,29 The modest findings in this study are also in line with recent simulation studies that evaluated the potential health impacts of implementing PSE interventions in LAC. Their simulation, albeit a system dynamics simulation, showed that the PSE changes, if sustained, have the potential to reduce the burden of obesity in the county.30 Other studies involving microsimulation models, which tested various hypothetical scenarios in California, also show promising decline in obesity and T2DM prevalence.31 Studies like these and the present can help policymakers direct their efforts to interventions that have the potential to yield the greatest impact and alert them to the resulting slow decline in obesity and T2DM, despite major prevention efforts. This study suggests that some interventions appear to be more effective than others. Such knowledge can be useful to policymakers, as it can guide them on where to allocate more resources in the fight against obesity and T2DM. For instance, this study suggests that investing in policies that promote physical activity and discourage fast-food consumption would yield a greater decrease in obesity and T2DM incidence than those that discourage SSB consumption or those that promote fresh fruit and vegetable consumption. Nevertheless, to have an impact on the obesity and T2DM epidemic, most interventions have to be implemented continuously over the individual life course and in synergy and in combination with one another.
Limitations This study is not without limitations. First, the application of the findings may be limited in that computer models remain a simplification and abstraction of the real world and, as such, may not capture other important aspects that can influence how one becomes obese or develops T2DM. Second, some variables, such as income and ethnicity, were not set conditionally on one another in the VILA cohort; this can limit the generalizability of the results to the actual LAC population. Third, the synthetic cohort used in this study is inherently a closed population admitting no new individuals after the start of follow-up, and not allowing any process, such as death, to remove individuals from the population. Fourth, in this iteration of the model, only a few combinations of interventions were tested, and focused only on obesity and T2DM as outcomes. Future research should test other combinations of interventions, allow for dynamic or open populations, target social networks, and evaluate other outcomes besides obesity and T2DM and do so over the entire lifetime.
& 2018
7
CONCLUSIONS This study reinforces the notion that implementing health interventions continuously throughout the life span and in combination with other interventions could substantially halt the obesity and the T2DM epidemics. Further, allocating more resources to interventions that promote physical activity and discourage fast-food consumption would likely yield greater reduction in obesity and T2DM incidences. The slow decline in obesity and T2DM may be the result of the modest impact of health interventions and their isolated implementation. Lastly, this study illustrates the usefulness of using computer simulation models in assisting policy making in obesity and diabetes prevention.
ACKNOWLEDGMENTS RAN participated in the study conception, design, and analysis and wrote the first draft of the article. OAA supervised and participated in the study conception, design, and analysis and reviewed and revised the manuscript. All authors provided critical input and insights into the development and writing of the article and approved the final manuscript as submitted. RAN wishes to thank his doctoral dissertation committee for providing constructive feedback that helped improved this article. RAN was supported by a Burroughs Wellcome Fellowship and the Dissertation Year Fellowship from the University of California Los Angeles (UCLA). OAA was partly supported by grant R01HD072296-01A1 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development. In addition, RAN benefited from facilities and resources provided by the California Center for Population Research at UCLA (CCPR), which receives core support (R24-HD041022) from the Eunice Kennedy Shriver National Institute of Child Health and Human Development. The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. An abstract based on a previous version of this manuscript was presented at AcademyHealth’s Annual Research Meeting, New Orleans, Louisiana, June 25−27, 2017. No financial disclosures were reported by the authors of this paper.
SUPPLEMENTAL MATERIAL Supplemental materials associated with this article can be found in the online version at https://doi.org/10.1016/j. amepre.2018.07.014.
REFERENCES 1. Ogden CL, Carroll MD, Kit BK, Flegal KM. Prevalence of childhood and adult obesity in the United States, 2011−2012. JAMA. 2014;311 (8):806–814. https://doi.org/10.1001/jama.2014.732. 2. WHO. Obesity and Overweight. Fact Sheet No. 311. Geneva: WHO, 2014.
ARTICLE IN PRESS 8
Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
3. Felber J-P, Golay A. Pathways from obesity to diabetes. Int J Obes Relat Metab Disord. 2002;26(suppl 2):S39–S45. https://doi.org/ 10.1038/sj.ijo.0802126. 4. Skinner AC, Skelton JA. Prevalence and trends in obesity and severe obesity among children in the United States, 1999−2012. JAMA Pediatr. 2014;168(6):561–566. https://doi.org/10.1001/jamapediatrics.2014.21. 5. Diabetes Prevention Program Research Group. Long-term effects of lifestyle intervention or metformin on diabetes development and microvascular complications over 15-year follow-up: the Diabetes Prevention Program Outcomes Study. Lancet Diabetes Endocrinol. 3 (11):866–875. https://doi.org/10.1016/S2213-8587(15)00291-0. 6. Bibbins-Domingo K, Chertow GM, Coxson PG, et al. Projected effect of dietary salt reductions on future cardiovascular disease. N Engl J Med. 2010;362(7):590–599. https://doi.org/10.1056/NEJMoa0907355. 7. Brownson RC, Haire-Joshu D, Luke DA. Shaping the context of health: a review of environmental and policy approaches in the prevention of chronic diseases. Annu Rev Public Health. 2006;27(1):341–370. https://doi.org/10.1146/annurev.publhealth.27.021405.102137. 8. Mccurley JL, Crawford MA, Gallo LC. Prevention of type 2 diabetes in U.S. Hispanic youth: a systematic review of lifestyle interventions. Am J Prev Med. 2017;53(4):519–532. https://doi.org/10.1016/j.amepre. 2017.05.020. 9. Nianogo RA, Wang MC, Wang A, et al. Projecting the impact of hypothetical early life interventions on adiposity in children living in low-income households. Pediatr Obes. 2017;12(5):398–405. https:// doi.org/10.1111/ijpo.12157. 10. Eddy DM, Schlessinger L. Validation of the Archimedes diabetes model. Diabetes Care. 2003;26(11):3102–3110. https://doi.org/ 10.2337/diacare.26.11.3102. 11. Nianogo RA, Arah OA. Agent-based modeling of noncommunicable diseases: a systematic review. Am J Public Health. 2015;105(3): e20–e31. https://doi.org/10.2105/AJPH.2014.302426. 12. Los Angeles County Department of Public Health Office of Health Assessment and Epidemiology. Obesity and Related Mortality in Los Angeles County: A Cities and Communities Health Report. Los Angeles, CA: Los Angeles County Department of Public Health Office of Health Assessment and Epidemiology; 2011. http://publichealth.lacounty.gov/ha/reports/habriefs/2007/obese_cities/obesity_ 2011fs.PDF. 13. Los Angeles County Department of Public Health. Project RENEW: Fact Sheet. Los Angeles, CA: Los Angeles County Department of Public Health; 2010. http://publichealth.lacounty.gov/chronic/. 14. North MJ, Macal CM. Managing Business Complexity: Discovering Strategic Solutions with Agent-Based Modeling and Simulation. New York, NY: Oxford University Press; 2007. https://doi.org/10.1093/ acprof:oso/9780195172119.001.0001. 15. Macal CM, North MJ. Tutorial on agent-based modelling and simulation. J Simul. 2010;4(3):151–162. https://doi.org/10.1057/ jos.2010.3. 16. Nianogo RA, Arah OA. Agent-based modeling of noncommunicable diseases: a systematic review. Am J Public Health. 2015;105(3):e20– e31. https://doi.org/10.2105/AJPH.2014.302426.
17. Basu S. Modeling Public Health and Healthcare Systems, Vol. 1. Oxford, UK: Oxford University Press; 2017. https://doi.org/10.1093/ med/9780190667924.001.0001. 18. Wolch J, Wilson J, Fehrenbach J. Parks and park funding in Los Angeles: an equity-mapping analysis. Urban Geogr. 2005;26(1):4–35. https://doi.org/10.2747/0272-3638.26.1.4. 19. CDC. A SAS Program for the 2000 CDC Growth Charts (ages 0 to < 20 years). www.cdc.gov/nccdphp/dnpao/growthcharts/resources/sas.htm. Published 2016. Accessed September 16, 2016. 20. WHO. WHO BMI-for-age growth charts. www.who.int/growthref/ who2007_bmi_for_age/en/. Published 2015. Accessed September 29, 2016. 21. WHO. BMI classification 2016. http://apps.who.int/bmi/index.jsp? introPage=intro_3.html. Published 2016. Accessed October 6, 2016. 22. Robins JM. A new approach to causal inference in mortality studies with sustained exposure periods—application to control of the healthy worker survivor effect. Math Model. 1986;7(9−12):1393–1512. https:// doi.org/10.1016/0270-0255(86)90088-6. 23. Aune D, Norat T, Leitzmann M, Tonstad S, Vatten LJ. Physical activity and the risk of type 2 diabetes: a systematic review and doseresponse meta-analysis. Eur J Epidemiol. 2015;30(7):529–542. https:// doi.org/10.1007/s10654-015-0056-z. 24. Zheng M, Allman-Farinelli M, Heitmann BL, Rangan A. Substitution of sugar-sweetened beverages with other beverage alternatives: a review of long-term health outcomes. J Acad Nutr Diet. 2015;115 (5):767–779. https://doi.org/10.1016/j.jand.2015.01.006. 25. Casazza K, Fontaine KR, Astrup A, et al. Myths, presumptions, and facts about obesity. N Engl J Med. 2013;368(5):446–454. https://doi. org/10.1056/NEJMsa1208051. 26. Sallis JF, Saelens BE, Frank LD, et al. Neighborhood built environment and income: examining multiple health outcomes. Soc Sci Med. 2009;68(7):1285–1293. https://doi.org/10.1016/j.socscimed. 2009.01.017. 27. Papas MA, Alberg AJ, Ewing R, Helzlsouer KJ, Gary TL, Klassen AC. The built environment and obesity. Epidemiol Rev. 2007;29(27):129– 143. https://doi.org/10.1093/epirev/mxm009. 28. Sterman JD. Learning from evidence in a complex world. Am J Public Health. 2006;96(3):505–514. https://doi.org/10.2105/AJPH.2005. 066043. 29. Orr MG, Kaplan GA, Galea S. Neighbourhood food, physical activity, and educational environments and black/white disparities in obesity: a complex systems simulation analysis. J Epidemiol Community Health. 2016;70(9):862–867. https://doi.org/10.1136/ jech-2015-205621. 30. Kuo T, Robles B, Trogdon JG, Ferencik R, Simon PA, Fielding JE. Framing the local context and estimating the health impact of CPPW obesity prevention strategies in Los Angeles County, 2010−2012. J Public Heal Manag Pract. 2016;22(4):360–369. https://doi.org/ 10.1097/PHH.0000000000000334. 31. Shi L, van Meijgaard J, Fielding J. Forecasting diabetes prevalence in California: a microsimulation. Prev Chronic Dis. 2011;8(4):A80. www. cdc.gov/pcd/issues/2011/jul/10_0177.htm. Accessed October 4, 2018.
www.ajpmonline.org
RESEARCH ARTICLE
Impact of Public Health Interventions on Obesity and Type 2 Diabetes Prevention: A Simulation Study D1X XRoch A. Nianogo, D2X XMD, MPH, PhD,1,2 D3X XOnyebuchi A. Arah, D4X XMD, MSc, DSc, MPH, PhD1,2,3,4
Introduction: Little is known about what interventions worked or did not work in slowing the obesity epidemic. The long-term comparative effectiveness of environmental and behavioral public health interventions for obesity and type 2 diabetes prevention over an individual’s life course is relatively unexplored. The potential impact and long-term collective effectiveness of environmental and behavioral interventions on obesity and type 2 diabetes throughout the life course was evaluated.
Methods: The Virtual Los Angeles Obesity Model developed in 2016 was used to estimate the incidence and prevalence of obesity and type 2 diabetes under current and hypothetical interventions among 98,000 individuals born in 2009 and followed from birth to age 65 years. Analyses were performed in 2016 and completed in 2018.
Results: The 48-year risk of type 2 diabetes was 0.533 (95% CI=0.446, 0.629) under the natural course, 0.451 (95% CI=0.334, 0.570) under the physical activity intervention, and 0.443 (95% CI=0.389, 0.495) under the fast-food intervention. The 64-year risk of obesity was 0.892 (95% CI=0.879, 0.903) under the natural course, 0.876 (95% CI=0.850, 0.899) under the physical activity intervention, and 0.864 (95% CI=0.856, 0.873) under the fast-food intervention. The other interventions had little or no long-term effects. When all the interventions were applied, the population risk ratios were 0.942 (95% CI=0.914, 0.967) and 0.634 (95% CI=0.484, 0.845) for obesity and type 2 diabetes, respectively. Conclusions: Implementing health interventions continuously throughout the life span and in combination with other interventions could substantially halt the obesity and the type 2 diabetes epidemics. Am J Prev Med 2018;000(000):1−8. © 2018 American Journal of Preventive Medicine. Published by Elsevier Inc. All rights reserved.
INTRODUCTION
O
besity remains a major public health problem in the U.S.1 and worldwide,2 and has been one of the most predominant players in the increase of the incidence of type 2 diabetes mellitus (T2DM).3 Despite ongoing efforts, the rates of obesity have remained persistently high, especially in many disadvantaged groups, but have begun leveling off among children in the U.S.4 However, little is known about what interventions worked or did not work in slowing the obesity epidemic over the course of decades. Further, similar questions of increasing importance remain: which interventions were impactful over the long run and which interventions yielded the greatest impact?
When should these interventions be implemented or to whom should they be given? Answering these questions often entail an evaluation of the interventions that have already been implemented5 and a projection of the From the 1Department of Epidemiology, Fielding School of Public Health, University of California, Los Angeles (UCLA), Los Angeles, California; 2 California Center for Population Research, Los Angeles, California; 3 UCLA Center for Health Policy Research, Los Angeles, California; and 4 Department of Statistics, UCLA College of Letters and Science, Los Angeles, California Address correspondence to: Roch A. Nianogo, MD, MPH, PhD, Department of Epidemiology, Fielding School of Public Health, University of California, Los Angeles, 650 Charles E. Young Drive South, Los Angeles CA 90095. E-mail: [email protected]. 0749-3797/$36.00 https://doi.org/10.1016/j.amepre.2018.07.014
© 2018 American Journal of Preventive Medicine. Published by Elsevier Inc. All rights reserved.
Am J Prev Med 2018;000(000):1−8
1
ARTICLE IN PRESS 2
Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
impact of potential interventions.6 Information about which interventions are most efficacious could assist policymakers in their decision making and help avoid the squandering of resources on ineffective interventions, while allocating more resources on interventions that have the potential to be successful. The long-term evaluation of policy interventions for obesity and T2DM prevention implemented in largescale populations and in real settings can be quite difficult to achieve, and such evaluations can be hard to find.7 Policymakers and researchers have, therefore, relied in the past on other evaluation strategies in the hopes of generalizing results to the larger population. In fact, although RCTs remain the gold standard in evaluating interventions, their evidence is not always generalizable to the population of interest. Moreover, they are typically costly and often cannot include participants for an extended period of time.8 Others have used existing observational studies for their attractive edge (i.e., longer follow-up, less-restrictive eligibility criteria, and lower cost) to evaluate hypothetical interventions in given populations using complex methods.9 Unfortunately, the results of such endeavors, although carefully used, are subject to the misspecification of the model and the presence of uncontrolled confounding, to list a few.9 Finally, some researchers have used computer models, such as the Coronary Heart Disease Policy Model6 or Archimedes,10 to evaluate the impact of public health interventions in large-scale populations. Such models (commonly referred to as microsimulation models) are very promising, but fall short by focusing only on the individual level and not including aspects of the built-environment, a key player in the obesity epidemic.11 Agent-based models (ABMs), on the other hand, are computer simulation models that can incorporate individual characteristics as well as the social and physical environments in assessing intervention impacts.11 To evaluate the potential impact and the long-term effectiveness of environmental and behavioral public health interventions on obesity and T2DM, the Virtual Los Angeles (VILA) Obesity Model, an ABM of obesity and T2DM is used. This study proposes to evaluate interventions implemented in Los Angeles County (LAC), California, as there have been major documented efforts implemented throughout the county to curb the obesity epidemic.12,13
METHODS Study Population The VILA Obesity Model is a stochastic dynamic discrete-time ABM developed in 2016 to study obesity and T2DM in U.S.
individuals followed from birth to age 65 years. Briefly, the model has two subparts: the environment and the individuals (Appendix Figure 1, available online). The characteristics of the environment (i.e., residential makeup, food environment, and physical activity opportunities) were first simulated to represent the residential areas of LAC. Second, the individuals were simulated at birth and spread out in the 235 simulated neighborhoods. Simulated individuals were also characterized by their sociodemographics, dietary behaviors, and physical activity patterns. Each of the 98,230 simulated individuals was “born” in 2009, grew up in their neighborhood, and was followed until age 65 years. During young adulthood (18−24 years) and middle adulthood (40−49 years), the characteristics of the neighborhoods where the agent lived could change to reflect what would happen if the agent actually changed location; this was done by resampling anew from the neighborhood sociodemographics. The VILA Obesity Model was used to describe the trends in obesity and T2DM incidence and prevalence among U.S. individuals born in LAC (Forecasting Obesity and T2DM Incidence and Burden Using an Agent-Based Model. The VILA-Obesity Simulation Model, unpublished observations, 2017). Although a full-fledged ABM would ideally incorporate (1) a set of agents, (2) an environment where the agents live, and (3) a set of agent relationships and underlying conceptual/theoretic decision-making rules outlining how agents behave and interact with each other and with their environment14−16; the VILA model only included agents, their environment, and cross-level agent −environment interactions over time. It can be argued then that VILA could be thought of in the broad sense as an ABM and could be seen as an extension of a microsimulation model,17 especially because it incorporated environmental and contextual exposures (and the resulting cross-level agent−environment interactions).
Measures In this study, the terms behavioral interventions in particular, and intervention in general, were used to represent a hypothetical or counterfactual intervention in which the modeler hypothetically alters the level of the behavior or exposure status. The underlying assumption is that individuals who are exposed to an intervention intended at a behavior change, for instance, are going to engage in the intended behavior with a 100% uptake. In other words, the impact of such “hypothetical” interventions would represent the potential maximum effect of the intervention. All interventions were designed on the basis of specific behavior status. Generally, an intervention consisted of simulating and altering the current behavior status to become the desired level of that particular behavior. For instance, the breastfeeding intervention was implemented in the first year of life and included altering the breastfeeding exposure status of simulated individuals to become “breastfed exclusively for ≥ 6 months,” if not already so. Likewise, the sugar-sweetened beverage (SSB) intervention was implemented throughout the life course at eight possible timepoints and included altering the SSB consumption exposure status of simulated individuals to become “drink zero glasses of soda or other sugary drinks,” if not already so (Figure 1 and Appendix Table 1, available online). www.ajpmonline.org
ARTICLE IN PRESS Nianogo and Arah / Am J Prev Med 2018;000(000):1−8 In addition to behavioral interventions, environmental interventions were also explored. The neighborhood park access intervention was implemented at three possible timepoints: birth, young adulthood, and middle adulthood and consisted in altering the neighborhood park access exposure status to become “neighborhood with a high access to parks,” if not already so. Access to parks in neighborhoods was defined as the percentage of the population living within a quarter-mile buffer and was based on California aggregated data obtained from Wolch et al.18 All interventions were evaluated singly and in combination with one another. The first set of combined interventions, which included the five behavioral interventions (i.e., referred to as combined behavioral interventions), was composed of four dietary interventions (i.e., breastfeeding promotion, reduction of SSB consumption, reduction of fast-food consumption, and increase of fresh fruit and vegetable consumption) and one physical activity intervention (i.e., increase of the level of physical activity). The second set of combined interventions, which included both behavioral and environmental interventions (i.e., referred to as combined mixed interventions), was composed of two individual-level dietary interventions (i.e., breastfeeding promotion,
3
and reduction of SSB consumption) and two environmental physical activity−related interventions (i.e., increasing access to parks and recreations and designing pedestrian-friendly communities; Appendix Table 1, available online). Furthermore, interventions were implemented throughout the individual life course, during childhood, young adulthood, and middle adulthood. Specifically, this study defined and projected the cumulative incidence of obesity and T2DM under a natural course (i.e., no intervention, status quo), an optimistic scenario (i. e., idealistic scenario), and a pessimistic scenario (i.e., worst-off scenario; Figure 1). Childhood obesity was defined using the WHO guidelines on the basis of BMI z-scores calculated using Centers for Disease Control and Prevention−provided SAS codes for ages 2−17 years.19 A child with a BMI z-score ≥ 2 was classified as obese.20 Adult obesity was also defined using WHO guidelines for ages 18−65 years. An individual with a BMI ≥ 30 was classified as obese.21 To calculate measures of incidence, only the first time that an individual was diagnosed as being obese was considered (i.e., first occurrence of obesity among
Figure 1. Hypothetical intervention regimens implemented throughout an individual life span. Note: Throughout the life course (i.e., eight discrete time-steps from age 2 to 65 years) interventions were implemented in childhood, in young adulthood, in middle adulthood, and at all relevant time-points (i.e., optimistic or idealistic scenario) and compared to the natural course (i.e., status quo). For reference a pessimistic scenario is also implemented (i.e., worse-off scenario). In the natural course (i.e., no intervention), the individual behavior status remained unchanged throughout follow-up. In the optimistic scenario, the exposure status of the individual at time=3, 5, and 7 was simulated and altered to become “the desirable level of that particular behavior” and would remain unchanged at the other time points at time t=1, 2, 4, 6, and 8. In the pessimistic scenario, however, the exposure status of individuals was simulated and altered to become “Less desirable level of the particular behavior” a time t=1, 2, 4, 6, and 8 and unchanged at time 3, 5, and 7.
& 2018
ARTICLE IN PRESS Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
4
at-risk individuals, that is individuals who were not obese in the previous time-step). Incident T2DM between ages 18 and 65 years was the outcome of interest. The following covariates were considered: age (continuous), sex (binary), race (binary), SES (binary), marital status (binary), and family history of T2DM (binary).
Statistical Analysis Assumptions about the underlying data generating mechanisms of obesity and T2DM and the relationships between covariates, exposures, mediators, and outcomes were presented using a directed diagram (Appendix Figure 2, available online). Monte Carlo simulations and the g-computation algorithm of Robins22 were used to predict the potential cumulative incidences of obesity and T2DM under various hypothetical scenarios. Briefly, this was done by first projecting the potential outcomes under an intervention scenario and then projecting the potential outcomes under a no-intervention scenario (i.e., generally the natural course or status quo) and then taking the contrast between these two potential outcomes. To generate the mean prevalences, incidences, and population risk ratios (RRs) and 95% CI, 100 replications of the simulation model were generated by Monte Carlo sampling of the parameters from a normal distribution using parameter inputs outlined in Appendix Table 2 (available online). All analyses were performed in SAS, version 9.4.
RESULTS Table 1 and Appendix Figure 3 (available online) describe the simulated end of follow-up cumulative
incidence of obesity and T2DM under various hypothetical interventions. The 64-year risk of obesity (from 2 to 65 years) and the 48-year risk of T2DM (from 18 to 65 years) at the end of follow-up under the natural course were 0.892 (95% CI=0.879, 0.903) and 0.533 (95% CI=0.446, 0.629), respectively. The 48-year risks of T2DM under the breastfeeding, the neighborhood walkability, and the neighborhood park access interventions were 0.533 (95% CI=0.446, 0.631), 0.522 (95% CI=0.425, 0.620), and 0.525 (95% CI=0.421, 0.631) respectively. The 64-year risks of obesity under the breastfeeding, the neighborhood walkability, and the neighborhood park access interventions were 0.892 (95% CI=0.880, 0.903), 0.890 (95% CI=0.875, 0.903), and 0.890 (95% CI=0.875, 0.902), respectively. Figure 2 and Appendix Figure 4 (available online) present the cumulative incidence of incident T2DM and obesity over time under the combined behavioral interventions. The curves of the cumulative incidence of obesity under the optimistic scenario departed slightly and were lower than that of the natural course curve for children, adults, and for the population as a whole. The cumulative incidence curve of T2DM under the optimistic scenario was much lower than that of the natural course throughout follow-up. The cumulative incidence of obesity and T2DM was consistently high among the non-white segment of the population throughout follow-up. The cumulative incidence for all other
Table 1. Simulated Cumulative Incidence of Obesity and Type 2 Diabetes Under Hypothetical Interventions in VILA (n=98,230) Interventions 00-Natural course (no intervention) 01-Eliminate SSB consumption 02-Exclusively breastfeed for ≥ 6 months 03-Increase neighborhood walkability 04-Increase neighborhood access to parks 05-Engage in moderate to vigorous physical activity 06-Consume ≥ 5 fresh fruit and vegetable/day 07-Eliminate fast-food consumption 08-Combined mixed interventions (all) 09-Combined behavioral interventions (all) 10-Combined mixed interventions (childhood) 11-Combined behavioral interventions (childhood) 12-Combined mixed interventions (young adulthood) 13-Combined behavioral interventions (young adulthood) 14-Combined mixed interventions (adulthood) 15-Combined behavioral interventions (adulthood)
64-year risk of obesity
48-year risk of type 2 diabetes
0.892 (0.879, 0.903) 0.890 (0.878, 0.900) 0.892 (0.880, 0.903) 0.890 (0.875, 0.903) 0.890 (0.875, 0.902) 0.876 (0.850, 0.899) 0.888 (0.875, 0.900) 0.864 (0.856, 0.873) 0.887 (0.870, 0.900) 0.839 (0.813, 0.867) 0.890 (0.877, 0.902) 0.873 (0.851, 0.891) 0.889 (0.875, 0.901) 0.872 (0.852, 0.888) 0.892 (0.878, 0.904) 0.880 (0.860, 0.895)
0.533 (0.446, 0.629) 0.515 (0.406, 0.608) 0.533 (0.446, 0.631) 0.522 (0.425, 0.620) 0.525 (0.421, 0.631) 0.451 (0.334, 0.570) 0.522 (0.428, 0.620) 0.443 (0.389, 0.495) 0.496 (0.379, 0.590) 0.338 (0.261, 0.419) 0.529 (0.440, 0.622) 0.481 (0.395, 0.589) 0.517 (0.414, 0.616) 0.462 (0.380, 0.545) 0.544 (0.422, 0.674) 0.452 (0.375, 0.523)
Note: Data presented as M (95% CI). All interventions were implemented under the optimistic scenario. The combined mixed interventions include the SSB consumption, breastfeeding, neighborhood walkability, and the neighborhood access to park interventions. The combined behavioral interventions include the SSB consumption, breastfeeding, physical activity, fruit and vegetable, and fast-food interventions. The mean and CI were obtained from 100 repeated replications of the model. SSB, sugar-sweetened beverage; VILA, Virtual Los Angeles Obesity Model.
www.ajpmonline.org
ARTICLE IN PRESS Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
interventions evaluated singly or in combination with one another are presented in Appendix Figures 5−9 (available online). Figure 3 and Appendix Figure 10 (available online) present the population impact of various interventions on T2DM and obesity, respectively. The two single most effective interventions on T2DM prevention were the fast-food intervention (population RR=0.831, 95% CI=0.712, 0.984), followed by the physical activity intervention with RR=0.847 (95% CI=0.730, 0.951). Eliminating fast-food consumption had a small effect on obesity prevention (RR=0.969, 95% CI=0.960, 0.979). Combining all behavioral interventions yielded the greatest effect for both T2DM (RR=0.634, 95% CI=0.484, 0.845) and obesity (RR=0.942, 95% CI=0.914, 0.967).
DISCUSSION The purpose of this study was to evaluate the long-term effectiveness of public health interventions on obesity and T2DM. This study complements the previous study based on the same model that forecasted the future prevalence and incidence of obesity and T2DM in LAC. Interventions, such as breastfeeding for 6 months or longer, increasing fresh fruit and vegetable consumption, increasing neighborhood walkability or neighborhood access to parks, when implemented singly, were not effective in reducing the cumulative incidence of obesity and T2DM. Engaging in moderate to vigorous physical activity and eliminating fast-food consumption appeared to be effective in reducing the excess risk
5
in obesity and T2DM incidence. In addition, combining interventions with one another throughout the life span showed the greatest impact, especially when combining all behavioral interventions together. Furthermore, for a given effective intervention, the impact seemed greater in reducing T2DM risk than in reducing obesity risk. Lastly, to have an impact, most interventions needed to be implemented at all possible timepoints (i.e., optimistic scenario). Interestingly, interventions implemented in childhood appeared more effective in reducing obesity risk than interventions implemented in young adulthood or middle adulthood; and interventions implemented in middle adulthood appeared more effective in reducing T2DM risk than interventions implemented in young adulthood and childhood. These results highlight many important insights worth mentioning. First, some interventions are more effective than others. Second, some periods in the life course appear to be more critical than others in the prevention of obesity or T2DM. Third, to be effective, most interventions have to be implemented continuously (virtually at every stage of life) and have to be implemented together. Fourth, the modest impacts of the interventions evaluated here testify to the persistence of obesity and T2DM and to the difficulty to curb these epidemics, which might explain why there has been only a slight leveling off of childhood obesity after many years of prevention efforts. The findings provide supporting evidence for the beneficial effects of engaging in physical activity, as is already established in the literature.23 Contrary to expectations, eliminating SSB throughout the life course
Figure 2. Cumulative incidence of type 2 diabetes under a combination of interventions. Note: The interventions include the sugar-sweetened beverage (SSB), breastfeeding (EBF), moderate to vigorous physical activity (MVPA), fruit and vegetable (FFV), and fast-food (FFD) interventions. (A) Type 2 diabetes cumulative incidence among adults aged 18−65 years; (B) type 2 diabetes cumulative incidence among adults aged 18−65 years by race and ethnicity. The mean cumulative incidences were obtained from 100 repeated replications of the model.
& 2018
ARTICLE IN PRESS 6
Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
Figure 3. Population impact (in terms of risk ratios) of various interventions on the cumulative incidence of type 2 diabetes. Note: The combined mixed interventions are the interventions that include the sugar-sweetened beverage consumption, breastfeeding, neighborhood walkability, and the neighborhood access to park interventions. The combined behavioral interventions include the sugar-sweetened beverage consumption, breastfeeding, physical activity, fruit and vegetable, and fast-food interventions. The mean and CI were obtained from 100 repeated replications of the model.
did not reduce the risk of obesity. Nevertheless, even though obesity rates did not decline as a result of this intervention, any weight stabilization could be considered beneficial, especially in a context where weight may be generally rising.24 In this study, breastfeeding for 6 months or longer had no overall effect over the life course of an individual. Although there is some evidence as to the benefits of breastfeeding in childhood obesity prevention,9 its role has been debated in the literature.25 It may be that breastfeeding results in a short-term effect on obesity rates that dissipates over time; this is consistent with
what was found in post-hoc analyses (Appendix Figure 11, available online). Contrary to expectations,26 living in a highly walkable neighborhood or neighborhood with high access to parks did not appear to reduce obesity and T2DM incidence in this study. Other studies have also reported inconsistent or no effect of the built-environment on BMI.27 The absence of an effect of the built-environment on obesity and T2DM in this study could be because of the presence of other barriers to engaging in physical activity in the neighborhood (e.g., safety) or other aspects of the food environment that could potentially outweigh the www.ajpmonline.org
ARTICLE IN PRESS Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
importance of the neighborhood physical activity characteristics. Furthermore, policies, systems, and environments (PSE), such as having a walkable neighborhood, may affect population health indirectly and slowly, if at all.28,29 The modest findings in this study are also in line with recent simulation studies that evaluated the potential health impacts of implementing PSE interventions in LAC. Their simulation, albeit a system dynamics simulation, showed that the PSE changes, if sustained, have the potential to reduce the burden of obesity in the county.30 Other studies involving microsimulation models, which tested various hypothetical scenarios in California, also show promising decline in obesity and T2DM prevalence.31 Studies like these and the present can help policymakers direct their efforts to interventions that have the potential to yield the greatest impact and alert them to the resulting slow decline in obesity and T2DM, despite major prevention efforts. This study suggests that some interventions appear to be more effective than others. Such knowledge can be useful to policymakers, as it can guide them on where to allocate more resources in the fight against obesity and T2DM. For instance, this study suggests that investing in policies that promote physical activity and discourage fast-food consumption would yield a greater decrease in obesity and T2DM incidence than those that discourage SSB consumption or those that promote fresh fruit and vegetable consumption. Nevertheless, to have an impact on the obesity and T2DM epidemic, most interventions have to be implemented continuously over the individual life course and in synergy and in combination with one another.
Limitations This study is not without limitations. First, the application of the findings may be limited in that computer models remain a simplification and abstraction of the real world and, as such, may not capture other important aspects that can influence how one becomes obese or develops T2DM. Second, some variables, such as income and ethnicity, were not set conditionally on one another in the VILA cohort; this can limit the generalizability of the results to the actual LAC population. Third, the synthetic cohort used in this study is inherently a closed population admitting no new individuals after the start of follow-up, and not allowing any process, such as death, to remove individuals from the population. Fourth, in this iteration of the model, only a few combinations of interventions were tested, and focused only on obesity and T2DM as outcomes. Future research should test other combinations of interventions, allow for dynamic or open populations, target social networks, and evaluate other outcomes besides obesity and T2DM and do so over the entire lifetime.
& 2018
7
CONCLUSIONS This study reinforces the notion that implementing health interventions continuously throughout the life span and in combination with other interventions could substantially halt the obesity and the T2DM epidemics. Further, allocating more resources to interventions that promote physical activity and discourage fast-food consumption would likely yield greater reduction in obesity and T2DM incidences. The slow decline in obesity and T2DM may be the result of the modest impact of health interventions and their isolated implementation. Lastly, this study illustrates the usefulness of using computer simulation models in assisting policy making in obesity and diabetes prevention.
ACKNOWLEDGMENTS RAN participated in the study conception, design, and analysis and wrote the first draft of the article. OAA supervised and participated in the study conception, design, and analysis and reviewed and revised the manuscript. All authors provided critical input and insights into the development and writing of the article and approved the final manuscript as submitted. RAN wishes to thank his doctoral dissertation committee for providing constructive feedback that helped improved this article. RAN was supported by a Burroughs Wellcome Fellowship and the Dissertation Year Fellowship from the University of California Los Angeles (UCLA). OAA was partly supported by grant R01HD072296-01A1 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development. In addition, RAN benefited from facilities and resources provided by the California Center for Population Research at UCLA (CCPR), which receives core support (R24-HD041022) from the Eunice Kennedy Shriver National Institute of Child Health and Human Development. The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. An abstract based on a previous version of this manuscript was presented at AcademyHealth’s Annual Research Meeting, New Orleans, Louisiana, June 25−27, 2017. No financial disclosures were reported by the authors of this paper.
SUPPLEMENTAL MATERIAL Supplemental materials associated with this article can be found in the online version at https://doi.org/10.1016/j. amepre.2018.07.014.
REFERENCES 1. Ogden CL, Carroll MD, Kit BK, Flegal KM. Prevalence of childhood and adult obesity in the United States, 2011−2012. JAMA. 2014;311 (8):806–814. https://doi.org/10.1001/jama.2014.732. 2. WHO. Obesity and Overweight. Fact Sheet No. 311. Geneva: WHO, 2014.
ARTICLE IN PRESS 8
Nianogo and Arah / Am J Prev Med 2018;000(000):1−8
3. Felber J-P, Golay A. Pathways from obesity to diabetes. Int J Obes Relat Metab Disord. 2002;26(suppl 2):S39–S45. https://doi.org/ 10.1038/sj.ijo.0802126. 4. Skinner AC, Skelton JA. Prevalence and trends in obesity and severe obesity among children in the United States, 1999−2012. JAMA Pediatr. 2014;168(6):561–566. https://doi.org/10.1001/jamapediatrics.2014.21. 5. Diabetes Prevention Program Research Group. Long-term effects of lifestyle intervention or metformin on diabetes development and microvascular complications over 15-year follow-up: the Diabetes Prevention Program Outcomes Study. Lancet Diabetes Endocrinol. 3 (11):866–875. https://doi.org/10.1016/S2213-8587(15)00291-0. 6. Bibbins-Domingo K, Chertow GM, Coxson PG, et al. Projected effect of dietary salt reductions on future cardiovascular disease. N Engl J Med. 2010;362(7):590–599. https://doi.org/10.1056/NEJMoa0907355. 7. Brownson RC, Haire-Joshu D, Luke DA. Shaping the context of health: a review of environmental and policy approaches in the prevention of chronic diseases. Annu Rev Public Health. 2006;27(1):341–370. https://doi.org/10.1146/annurev.publhealth.27.021405.102137. 8. Mccurley JL, Crawford MA, Gallo LC. Prevention of type 2 diabetes in U.S. Hispanic youth: a systematic review of lifestyle interventions. Am J Prev Med. 2017;53(4):519–532. https://doi.org/10.1016/j.amepre. 2017.05.020. 9. Nianogo RA, Wang MC, Wang A, et al. Projecting the impact of hypothetical early life interventions on adiposity in children living in low-income households. Pediatr Obes. 2017;12(5):398–405. https:// doi.org/10.1111/ijpo.12157. 10. Eddy DM, Schlessinger L. Validation of the Archimedes diabetes model. Diabetes Care. 2003;26(11):3102–3110. https://doi.org/ 10.2337/diacare.26.11.3102. 11. Nianogo RA, Arah OA. Agent-based modeling of noncommunicable diseases: a systematic review. Am J Public Health. 2015;105(3): e20–e31. https://doi.org/10.2105/AJPH.2014.302426. 12. Los Angeles County Department of Public Health Office of Health Assessment and Epidemiology. Obesity and Related Mortality in Los Angeles County: A Cities and Communities Health Report. Los Angeles, CA: Los Angeles County Department of Public Health Office of Health Assessment and Epidemiology; 2011. http://publichealth.lacounty.gov/ha/reports/habriefs/2007/obese_cities/obesity_ 2011fs.PDF. 13. Los Angeles County Department of Public Health. Project RENEW: Fact Sheet. Los Angeles, CA: Los Angeles County Department of Public Health; 2010. http://publichealth.lacounty.gov/chronic/. 14. North MJ, Macal CM. Managing Business Complexity: Discovering Strategic Solutions with Agent-Based Modeling and Simulation. New York, NY: Oxford University Press; 2007. https://doi.org/10.1093/ acprof:oso/9780195172119.001.0001. 15. Macal CM, North MJ. Tutorial on agent-based modelling and simulation. J Simul. 2010;4(3):151–162. https://doi.org/10.1057/ jos.2010.3. 16. Nianogo RA, Arah OA. Agent-based modeling of noncommunicable diseases: a systematic review. Am J Public Health. 2015;105(3):e20– e31. https://doi.org/10.2105/AJPH.2014.302426.
17. Basu S. Modeling Public Health and Healthcare Systems, Vol. 1. Oxford, UK: Oxford University Press; 2017. https://doi.org/10.1093/ med/9780190667924.001.0001. 18. Wolch J, Wilson J, Fehrenbach J. Parks and park funding in Los Angeles: an equity-mapping analysis. Urban Geogr. 2005;26(1):4–35. https://doi.org/10.2747/0272-3638.26.1.4. 19. CDC. A SAS Program for the 2000 CDC Growth Charts (ages 0 to < 20 years). www.cdc.gov/nccdphp/dnpao/growthcharts/resources/sas.htm. Published 2016. Accessed September 16, 2016. 20. WHO. WHO BMI-for-age growth charts. www.who.int/growthref/ who2007_bmi_for_age/en/. Published 2015. Accessed September 29, 2016. 21. WHO. BMI classification 2016. http://apps.who.int/bmi/index.jsp? introPage=intro_3.html. Published 2016. Accessed October 6, 2016. 22. Robins JM. A new approach to causal inference in mortality studies with sustained exposure periods—application to control of the healthy worker survivor effect. Math Model. 1986;7(9−12):1393–1512. https:// doi.org/10.1016/0270-0255(86)90088-6. 23. Aune D, Norat T, Leitzmann M, Tonstad S, Vatten LJ. Physical activity and the risk of type 2 diabetes: a systematic review and doseresponse meta-analysis. Eur J Epidemiol. 2015;30(7):529–542. https:// doi.org/10.1007/s10654-015-0056-z. 24. Zheng M, Allman-Farinelli M, Heitmann BL, Rangan A. Substitution of sugar-sweetened beverages with other beverage alternatives: a review of long-term health outcomes. J Acad Nutr Diet. 2015;115 (5):767–779. https://doi.org/10.1016/j.jand.2015.01.006. 25. Casazza K, Fontaine KR, Astrup A, et al. Myths, presumptions, and facts about obesity. N Engl J Med. 2013;368(5):446–454. https://doi. org/10.1056/NEJMsa1208051. 26. Sallis JF, Saelens BE, Frank LD, et al. Neighborhood built environment and income: examining multiple health outcomes. Soc Sci Med. 2009;68(7):1285–1293. https://doi.org/10.1016/j.socscimed. 2009.01.017. 27. Papas MA, Alberg AJ, Ewing R, Helzlsouer KJ, Gary TL, Klassen AC. The built environment and obesity. Epidemiol Rev. 2007;29(27):129– 143. https://doi.org/10.1093/epirev/mxm009. 28. Sterman JD. Learning from evidence in a complex world. Am J Public Health. 2006;96(3):505–514. https://doi.org/10.2105/AJPH.2005. 066043. 29. Orr MG, Kaplan GA, Galea S. Neighbourhood food, physical activity, and educational environments and black/white disparities in obesity: a complex systems simulation analysis. J Epidemiol Community Health. 2016;70(9):862–867. https://doi.org/10.1136/ jech-2015-205621. 30. Kuo T, Robles B, Trogdon JG, Ferencik R, Simon PA, Fielding JE. Framing the local context and estimating the health impact of CPPW obesity prevention strategies in Los Angeles County, 2010−2012. J Public Heal Manag Pract. 2016;22(4):360–369. https://doi.org/ 10.1097/PHH.0000000000000334. 31. Shi L, van Meijgaard J, Fielding J. Forecasting diabetes prevalence in California: a microsimulation. Prev Chronic Dis. 2011;8(4):A80. www. cdc.gov/pcd/issues/2011/jul/10_0177.htm. Accessed October 4, 2018.
www.ajpmonline.org