Matthew W. Gillman, MD, SM EN Power of Programming March 2014 - - PowerPoint PPT Presentation

Systems Science to Guide Implementation of Whole-of-community Childhood Obesity Interventions

Matthew W. Gillman, MD, SM

EN Power of Programming March 2014

Note: for non-commercial purposes only

Thanks to…

Faculty, Trainees, & Staff

Obesity Prevention Program Department of Population Medicine Harvard Medical School/Harvard Pilgrim Health Care Institute

Questions about Obesity

• Population trends

– What caused/is causing the epidemic? – How can we reverse it?

• Not necessarily the same as what caused it
• Individual (between-person) variability

– Why do some people develop obesity and

• thers not?

– How can we use that information to tailor, and evaluate, prevention and treatment

• What works, for whom, & under what circumstances

• Population trends

– What caused/is causing the epidemic? – How can we reverse it?

• Individual (between-person) variability

– Why do some people develop obesity and

• thers not?

– How can we use that information to tailor, and evaluate, prevention and treatment

Questions about Obesity

• Population trends

– What caused/is causing the epidemic? – How can we reverse it?

• Individual (between-person) variability

– Why do some people develop obesity and

• thers not?

– How can we use that—and other—information to tailor, and evaluate, prevention and treatment

Questions about Obesity

• Population trends

– What caused/is causing the epidemic? – How can we reverse it?

• Individual (between-person) variability

– Why do some people develop obesity and

• thers not?

– How can we use that information to tailor, and evaluate, prevention and treatment

Questions about Obesity

It’s because of what happened to them in utero and in early childhood

• Early (developmental) origins of obesity

– [Motivation, Evidence] – How systems science may help

• Untangle the complex webs of etiology
• Help drive design of prevention

The Childhood Obesity Epidemic

US DHHS, 2001; Hedley et al., 2004; Ogden et al., 2006, 2008

5 10 15

Prevalence of Overweight Year 24-71 months 0-11 months 12-23 months 1980 1985 1990 1995 2000

…in Younger Children Too

Including Infants

Kim et al., Obesity 2006; ~500,000 well child visits in Mass. HMO

a

Standardized for age, race/ethnicity, and HVMA site, using the year 1999-2000 as reference

13.7 8.1 10.6 14.2 11.9 13.3 13.3 13.5 13.8 12.3 13.0 12.1 12.7 12.6 13.1 12.9 9.1 9.8 10.0 9.0 8.1 7.4 9.9 9.8 9.6 9.3 9.8 10.2

4 5 6 7 8 9 10 11 12 13 14 15 16 1980-1982 1983-1984 1985-1986 1987-1988 1989-1990 1991-1992 1993-1994 1995-1996 1997-1998 1999-2000 2001-2002 2003-2004 2005-2006 2007-2008

Year

Prevalence

Overweight (standardized) Obesity (standardized)

Boys

• X. Wen et al. Pediatrics 2012;129:823-831

Downward trend in BMI since 2004 in 0-6-year-olds

Girls

• X. Wen et al. Pediatrics 2012;129:823-831

a

Standardized for age, race/ethnicity, and HVMA site, using the year 1999-2000 as reference

12.3 13.3 11.7 12.6 11.7 6.5 6.3 7.0 8.5 8.4 6.8 11.4 11.1 10.9 11.3 11.9 12.1 11.7 12.3 12.6 5.8 7.4 7.8 7.5 8.0 8.5 8.6 7.3

4 5 6 7 8 9 10 11 12 13 14 15 16 1980-1982 1983-1984 1985-1986 1987-1988 1989-1990 1991-1992 1993-1994 1995-1996 1997-1998 1999-2000 2001-2002 2003-2004 2005-2006 2007-2008

Year Prevalence

Overweight (standardized) Obesity (standardized)

Curious trends in SGA & LGA, U.S. 1990-2005

N = 502,716 low-risk mothers: 37-41 wk, age 25-29 y, white, >13 y educ, married, 1st trim prenatal care, non-smoker, no complications, NSVD, had U/S, GWG 26-35 lb

Donahue et al., Obstet Gynecol 2010; 115:357

Curious trends in SGA & LGA, U.S. 1990-2005

N = 502,716 low-risk mothers: 37-41 wk, age 25-29 y, white, >13 y educ, married, 1st trim prenatal care, non-smoker, no complications, NSVD, had U/S, GWG 26-35 lb

Donahue et al., Obstet Gynecol 2010; 115:357

Message

• The obesity epidemic has spared no

age group, not even our youngest children

• Once present, obesity tenaciously

resists treatment

• Prevention must start early

• Usual etiologic epidemiology

– 1 determinant at a time – Independent of others

• Moving toward systems approach

– More than 1 determinant

Developmental Origins of Obesity

Developmental Origins of Obesity

How Important Can It Be?

Prenatal Infancy Maternal smoking GWG (IOM cat.) Breastfeeding duration Daily sleep

P (Ob) at 7 y N

Inadequate/ Adequate

12+ m 12+ h 0.04 Y

Excessive

<12 m <12 h 0.28

Gillman , Ludwig. New Engl J Med 2013 (5 Dec); 369:2173-2175

Adjusted for maternal BMI, education; HH income; child race/ethnicity

Risk of obesity at age 7-10 y according to combinations of 4 pre/post-natal risk factors

Smoking – – – + – – + – + – + + – + + +

• Gest. weight gain

– + – – – + + + – – – + + + – + Breastfeeding – – + – – + – – + + – + + – + + Sleep – – – – + – – + – + + – + + + +

• Prob. obesity

0.04 0.06 0.07 0.07 0.08 0.10 0.10 0.11 0.11 0.13 0.13 0.16 0.18 0.18 0.20 0.28

• Pred. BMI-z

0.07 0.24 0.22 0.23 0.31 0.39 0.40 0.48 0.38 0.46 0.47 0.55 0.63 0.64 0.62 0.79

• Pred. DXA % fat 23.2

23.0 24.5 24.1 24.4 24.4 24.0 24.2 25.4 25.7 25.3 25.3 25.5 25.2 26.6 26.5 Prevalence in this cohort 6.9% 10.4% 20.3% 0.2% 5.2% 26.6% 0.2% 5.6% 1.1% 7.2% 0.1% 3.5% 9.2% 0.3% 1.5% 1.9%

Combinations of 4 risk factors 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Combinations of 4 risk factors Probability of obesity

Gillman , Ludwig. New Engl J Med 2013 (5 Dec); 369:2173-2175

Smoking – – – + – – + – + – + + – + + +

• Gest. weight gain

– + – – – + + + – – – + + + – + Breastfeeding – – + – – + – – + + – + + – + + Sleep – – – – + – – + – + + – + + + +

• Prob. obesity

0.04 0.06 0.07 0.07 0.08 0.10 0.10 0.11 0.11 0.13 0.13 0.16 0.18 0.18 0.20 0.28

• Pred. BMI-z

0.07 0.24 0.22 0.23 0.31 0.39 0.40 0.48 0.38 0.46 0.47 0.55 0.63 0.64 0.62 0.79

• Pred. DXA % fat 23.2

23.0 24.5 24.1 24.4 24.4 24.0 24.2 25.4 25.7 25.3 25.3 25.5 25.2 26.6 26.5 Prevalence in this cohort 6.9% 10.4% 20.3% 0.2% 5.2% 26.6% 0.2% 5.6% 1.1% 7.2% 0.1% 3.5% 9.2% 0.3% 1.5% 1.9%

Combinations of 4 risk factors 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Combinations of 4 risk factors Probability of obesity

PAR% ~ 20-50% Risk of obesity at age 7-10 y according to combinations of 4 pre/post-natal risk factors

More risk factors

• Prenatal

– Smoking, GWG, GDM

• Perinatal

– C-section, leptin

• Infancy

– Type of feeding, sleep duration, rapid adiposity gain, early intro solids

More risk factors

• Prenatal

– Smoking, GWG, GDM

• Perinatal

– C-section, leptin

• Infancy

– Type of feeding, sleep duration, rapid adiposity gain, early intro solids

• Emerging

– Epigenetics, toxic environment, microbiota….

• More than 1 determinant
• Interacting with each other

Developmental Origins of Obesity

• More than 1 determinant
• Interacting with each other
• Over time (age)

– Life course approach

Developmental Origins of Obesity

• More than 1 determinant
• Interacting with each other
• Over time (age)
• At multiple levels of influence

– Different influences at different developmental periods

Developmental Origins of Obesity

• More than 1 determinant
• Interacting with each other
• Over time (age)
• At multiple levels of influence

***********************************************

• Dynamic

– Feedback loops

Developmental Origins of Obesity

Maternal and child inter-generational vicious cycles

• More than 1 determinant
• Interacting with each other
• Over time (age)
• At multiple levels of influence

***********************************************

• Dynamic

– Feedback loops

• That may operate in different directions at different

stages of the life course

Developmental Origins of Obesity

Hormone feedback loops in older children and adults Tend to impede weight loss

Cord blood leptin predicts slower WFL gain in 1st 6 mo, lower 3 & 7-y BMI, but ...3-y leptin predicts higher 7-y BMI

 early sensitive period of leptin action, then tolerance Boeke et al, Obesity 2013;21:1430-7

Body Composition Whole Body Total Energy Expenditure Thermic Effect of Feeding Adaptive Thermogenesis Physical Activity Energy Expenditure Resting Metabolic Rate Daily Average Lipolysis Rate Ketone Oxidation Rate Daily Average Ketogenesis Rate Daily Average Ketone Excretion Rate Daily Average Glycogenolysis Rate Glycerol 3-Phosphate Production Rate Gluconeogenesis From Amino Acids De Novo Lipogenesis Rate Macronutrient Oxidation Rates Respiratory Gas Exchange Nutrient Balance Parameter Constraints Carbohydrate Perturbation Constraint Protein Perturbation Constraint Physical Inactivity Constraint Model Parameter Values

Predicting metabolic adaptation, body weight change, and energy intake

KD Hall

• More than 1 determinant
• Interacting with each other
• Over time (age)
• At multiple levels of influence

***********************************************

• Dynamic

– Feedback loops

• At multiple levels of influence

Developmental Origins of Obesity

Health

• utcomes

d

Pre- and peri- natal factors Micro Macro

Time Axis Hierarchical Axis

Weight gain

Energy in (diet) Energy out (physical activity)

Health Behaviors Genes Appetite Metabolism Mood HPA axis

Built environments

(e.g., connectivity, walkability)

Commercial messaging

(e.g., TV ads to kids)

Psychosocial hazards

(e.g., crime)

Local food environment

(e.g., presence of fast food)

Area deprivation

(e.g., poverty)

Cultural norms

(e.g., body image)

Laws, regulations, policies (e.g., farm subsidies)

Social, built, natural environment

http://kim.foresight.gov.uk/Obesity/Obesity.html

Yikes!

Complex System

• “…one whose properties are not fully

explained by an understanding of its component parts”

• Whole is greater than sum of parts

Gallagher & Appenzeller, quoted in Luke and Stamatakis, Annu Rev Public Health 2012; 33:357-76

Complex System

• Elements

– Large number – Heterogeneous, within and between – Interact with each other

• Interactions produce emergent properties
• Effects

– Persist over time – Adapt to changing circumstances

Luke and Stamatakis, Annu Rev Public Health 2012; 33:357-76

Complex System

Need to understand drivers

• Leverage for most powerful and efficient

effects on health outcomes

– Seemingly unimportant elements with large downstream effects? – Combinations of elements? – Unforeseen adverse effects (unintended consequences)?

Luke and Stamatakis, Annu Rev Public Health 2012; 33:357-76

System Dynamics “top down” Stocks & flows Agent-based Modeling “bottom up” Actors & rules Network Analysis Nodes & ties among them

Systems Science

• Systems science approaches have the

potential to

– Identify the most potent early drivers of the development of obesity and its complications

Systems Science

• Systems science approaches have the

potential to

– Identify the most potent early drivers of the development of obesity and its complications – Use for developing (and evaluating) multi- setting, mutli-component life course interventions

• Address “what works, for whom, and under what

circumstances”

Implementation of Interventions

How can systems science help?

• Like etiology, implementation is complex

– Informed by SNPs, methylated CpG sites, 16s speciation, dopamine reward pathways, insulin resistance… – But primarily involves stakeholders, interactions, processes

Stakeholders, Processes

But too simple—lacks interactions, feedback, etc

Implementation of Interventions

How can systems science help?

• Stakeholders, interactions, processes

– Whole of community interventions try to change them all – Need to study mechanisms (same as ‘omics) to understand “what works, for whom, and under what circumstances?”

R01 Funded by NIH (NHLBI, OBSSR)

2013-2018

Systems Science to Drive Whole-of-community Childhood Obesity Prevention Interventions

Investigator Institution Role Expertise Gillman Harvard Univ PI Obesity etiology and prevention Hammond Brookings Inst PI Agent-based modeling Economos Tufts Univ Co-I CPBR, obesity whole community interventions Hovmand Washington Univ Co-I Participatory group model building Allender Deakin Univ Co-I Systems intervention approaches Swinburn Deakin Univ, Univ Auckland Co-I Community/policy approaches to obesity

Systems Science to Drive Whole-of-community Childhood Obesity Prevention Interventions

Logic:

• Start with 2 completed interventions
• Relevant literature
• Build initial computational model

Systems Science to Drive Whole-of-community Childhood Obesity Prevention Interventions

Logic:

• Start with 2 completed interventions
• Relevant literature
• Build initial computational model
• Refine models with ongoing intervention

Victorian Trial

• Cluster RCT
• 12 intervention, 12 control communities
• Funded by Victoria state and Australia

federal government ($160m)

• Consortium of state government,

academia, NGO evaluation unit

Systems Science to Drive Whole-of-community Childhood Obesity Prevention Interventions

Logic:

• Start with 2 completed interventions
• Relevant literature
• Build initial computational model
• Refine models with ongoing intervention
• Use to design new intervention

Departure from:

Linear thinking Multiple causation Independent levels of influence

Systems Approach to Obesity Prevention: “Whole of community”

Courtesy Christina Economos

Creating a Causal Loop Map: The Shape Up Somerville Experience

• Develop understanding of whole system

– Describes key dynamics of social change within Somerville over time (10 y) – Based on CBPR – Informed by qualitative individual and group interviews with key SUS stakeholders and researchers – Illustrated through integration of complex, reciprocal, interdependent, and interactive relationships among individuals and their environments

• Highlights importance of context

X

Economos et al., submitted

X

X

SUS Model

• Based on deep qualitative knowledge
• But

– Retrospective – Not quantitative – School age, not younger – Replicable? Refinable? Predictive?

ABM = Agent-based modeling GMB = Participatory group model building

Systems Science to Drive Whole-of-community Childhood Obesity Prevention Interventions

Participatory Group Model Building

• Involves participants and
• ther stakeholders in iterative

process of developing system dynamics (and, now, agent-based) models

– Problem conceptualization – Formulation – Policy analysis – Implementation

• Reasons for using GMB

– Sharing of insights – Developing consensus – Design for implementation

Hovmand

Levels of System Insights

There is a system The components of a system How the components are related through feedback How people might think about a system Where one could intervene What is transformation What is the generic structure What are the implications of accumulations and nonlinear relationships What systems can generate the dynamic behavior Where are the leverage points When do boundary conditions determine behavior Why do things happen Deep system insights Surface system insights Graphical models or maps Simulation models Depth Informal Formal Modeling System pictures or diagrams

Hovmand

Levels of System Insights

There is a system The components of a system How the components are related through feedback How people might think about a system Where one could intervene What is transformation What is the generic structure What are the implications of accumulations and nonlinear relationships What systems can generate the dynamic behavior Where are the leverage points When do boundary conditions determine behavior Why do things happen Deep system insights Surface system insights Graphical models or maps Simulation models Depth Informal Formal Modeling System pictures or diagrams

October 5, 2012

There is a system The components of a system How the components are related through feedback How people might think about a system Where one could intervene What is transformation What is the generic structure What are the implications of accumulations and nonlinear relationships What systems can generate the dynamic behavior Where are the leverage points When do boundary conditions determine behavior Why do things happen Deep system insights Surface system insights Graphical models or maps Simulation models Depth Informal Formal Modeling System pictures or diagrams X X X x

Levels of System Insights

There is a system The components of a system How the components are related through feedback How people might think about a system Where one could intervene What is transformation What is the generic structure What are the implications of accumulations and nonlinear relationships What systems can generate the dynamic behavior Where are the leverage points When do boundary conditions determine behavior Why do things happen Deep system insights Surface system insights Simulation models Depth Informal Formal Modeling System pictures or diagrams

Hovmand

Conclusions

• Systems science approaches have the

potential to

– Identify the most potent early drivers of the development of obesity and its complications – Use for developing multi-setting, mutli- component life course interventions

• Address what works, for whom, and under what

circumstances

Questions

• In general…What do system science

approaches offer over and above …

• More specifically

– Initial steps often qualitative – Inputs are often quantitative – Pros and cons of ABM, SD for different

• circumstances. Other approaches?

Questions

• In general…What do system science

approaches offer over and above …

• More specifically

– Initial steps often qualitative – Inputs are often quantitative – Pros and cons of ABM, SD – When do we have enough data?

• Right balance of too little/too much
• Face, construct, criterion validity?

Questions

• In general…What do system science

approaches offer over and above …

• More specifically

– Initial steps often qualitative – Inputs are often quantitative – Pros and cons of ABM, SD – When do we have enough data? – Generalizability (as usual)

Agent-based Model

Finucane et al., Lancet 2011; 377: 557–67

Worldwide increases in adult obesity 1980-2008

Global Burden of Disease Study Lancet 12/21/12

…much of it due to decrease in early childhood mortality …and resulting increase in life expectancy

Developmental Origins

• f Health and Disease
• DOHaD emphasizes

prenatal period and early childhood as important periods for development

• f chronic disease

throughout life

Developmental Origins of Obesity

• Together pre- and post-natal risk factors

predict a substantial fraction of childhood

• besity
• Prevention interventions starting in

pregnancy and infancy have potential to

– Reduce these risk factors – Thus obesity-related disorders – And interrupt intergenerational vicious cycles

Developmental Origins Research

• In animal models, perinatal

programming of adult health outcomes well known

• Programming

– Perturbation at a critical period of development causes alterations with lifelong, sometimes irreversible consequences

Zygote Adult Neonate

45 Divisions 55 Divisions

Godfrey et al., Trends Endocrinol Metab 2010 ; 21:199-205

Why Early Intervention Makes Sense

Conclusions

• Together pre- and post-natal risk factors

predict a substantial fraction of childhood

• besity
• Prevention interventions starting in

pregnancy and infancy have potential to

– Reduce these risk factors – Thus obesity-related disorders – And interrupt intergenerational vicious cycles

Luke and Stamatakis, Annu Rev Public Health 2012; 33:357-76.