Combining Data from Clinical Trials on Obesity

August 29 - 30 , 2013
Bethesda, Maryland


The National Heart, Lung, and Blood Institute (NHLBI)’s Division of Cardiovascular Sciences (DCVS) convened a Working Group Meeting titled “Obesity Intervention Taxonomy and Pooled Analysis” on August 29-30, 2013, in Bethesda, Maryland. The NIH Office of Disease Prevention co-sponsored the meeting. The purpose of the Working Group meeting was to discuss:

  • methods for developing intervention taxonomy across diverse obesity interventions within multiple consortia.
  • analytic methods and approaches to combining data across different interventions, populations and settings.
  • methods for testing differences in subgroup responses to intervention (e.g., based on age, gender, race/ethnicity).
  • research recommendations regarding intervention taxonomy and analysis of pooled data from diverse intervention studies.



There are many obesity intervention research consortia in different locations in the United States. Within a consortium, multiple studies, each with different interventions, are being implemented. Some have common outcome measures. Examples of such consortia are the Childhood Obesity Prevention and Treatment Research (COPTR, 4 trials), the Early Adult Reduction of weight through LifestYle intervention (EARLY, 7 trials), the Obesity Related Behavioral Intervention Trials (ORBIT, 7 studies), and the Lifestyle Interventions in Overweight and Obese Pregnant Women (LIFE-Moms, 7 trials).

Except for LIFE-Moms, NHLBI is the lead funding Institute with co-funding from NICHD and OBSSR (COPTR); NIDDK, NCI, OBSSR (ORBIT); and NICHD (EARLY). The Centers for Disease Control and Prevention has also funded the Childhood Obesity Research Demonstration (CORD) projects that have multiple evaluation studies.

These studies all focus on obesity and have common measures such as weight, height, BMI, and psychosocial variables but different intervention strategies at multiple levels of influence (e.g., individual, family, schools). COPTR is testing multi-level intervention approaches to prevent excess weight gain in youth, and to reduce weight in obese and severely obese youth. Targeted age groups are preschoolers (2-5 year olds) and pre-adolescents and adolescents (7-15 year olds) of diverse racial and ethnic groups. EARLY is testing innovative behavioral approaches for weight control in young adults 18-35 years of age at high risk for weight gain. ORBIT is testing methods to translate findings from basic research on human behavior into more effective clinical, community, and population interventions to reduce obesity in a diverse group of subjects. LIFE-Moms is testing behavioral/lifestyle interventions in overweight and obese pregnant women designed to improve weight and metabolic outcomes in both the women and their offspring.

These consortia provide unique opportunities to pool data that would yield larger sample sizes and power to explore unique hypotheses. However, because the studies have different designs and analytic approaches, populations and intervention strategies, there could be analytic challenges in testing the impact of interventions when combining studies even though the studies have common measures. There may be situations in which a common analysis would be useful. For example, study designs that are comparable (e.g., 2-arm experimental), study objectives and populations that are similar, and interventions that share the same objective (e.g., reducing weight). A common analysis could be used to examine effect modification by type of study, gender, and race/ethnicity. A major objective of this working group was to address these issues and to discuss how, when, or why multi-level interventions can be combined in mutually reinforcing ways. Recent research by Belle et al., (2003) and Weiner et al., (2012) provide some suggestions for tackling these issues.


The workshop brought together experts in epidemiology, psychology, biostatistics, intervention methodology, outcomes research, social sciences, population research, health education and behavior, nutrition, nursing, physical activity, and child health. NIH staff attended the meeting.

After brief opening remarks on NIH prevention research priorities and strategic plans, and charge to the working group, two keynote speakers discussed methodologies for combining results across diverse interventions. The REACH (The Resources Enhancing Alzheimer's Caregiver Health) study (Czaja et al., 2003) was discussed as an example of a consortium that decomposed different interventions into similar components using task analysis. Two analytic approaches were used in that study: pre-planned meta-analysis and meta-regression. Discussions ensued on using the Behavior Change Taxonomy (Michie et al., 2013) and the Behavior Change Wheel (Michie et al., 2011) for developing taxonomies across diverse interventions rather than the method used by REACH to code the content of multi-component interventions.


Participants acknowledged that the recommendations below may require substantial resource (time, personnel) investment. However, resource needs could be reduced by implementing the recommendations in a sample of individuals or a suitable subset of the studies.
Consortia willing to develop intervention taxonomy and pool data should consider the following:

Taxonomy-related recommendations

  • Decompose and code content of each intervention utilizing established theory or taxonomy. Examples of taxonomies include behavior change techniques (BCTs) and more extensive taxonomies that address other aspects of studies, such as populations studied, mode of intervention administration, training, measures used, timing of measures, intervention adaptability, and interventionist characteristics.
  • If established theory or taxonomy is amended, there should be appropriate scientific rigor to justify the change(s), e.g., calculate inter-rater reliability.
  • Determine the intervention components and dose intended to be delivered (according to protocol) per each intervention component (e.g., BCT).
  • Determine the intervention components and dose actually delivered per each intervention component (e.g., BCT).
  • Determine the intervention components (e.g., BCTs) and dose actually received by participants.

Analysis methodology related recommendations

The methodology should take into account relevant theory and be driven by the research question(s). Issues to consider include variable selection and interactions.

  • Pooling results across studies, within or among consortia, must account for heterogeneity among studies.
  • In general, pooling is used for exploratory analyses to help identify intervention components (e.g., BCTs) that may work better than others and to identify subsets of participants across studies in which particular components work better. As such, these analyses are not to replace the standard analysis plan for each study. If the analyses were not specified in advance, analyses of pooled data are viewed as ‘post-hoc’ exploratory analyses. If such analyses are to be performed, split-sample or cross-validation techniques should be employed.
  • Pooling may also be performed to test hypotheses which were specified a priori.
  • Approaches to consider:
    • Traditional meta-analysis, with or without meta-regression. Including only study-level covariates in meta regression limits the number of observations. Using participant level covariates in meta regression is recommended to the extent possible.
    • Ignoring randomization but utilizing intervention components and, potentially, other study or intervention level data rather than indicator(s) of intervention. Participant level covariates should be included with this approach.
  • Methodologies to consider
    • Linear mixed-effects models (multilevel analysis)
    • Non-linear models and approaches (e.g., classification/regression trees/forests) Multi-group structural equation modeling
    • Latent class models

Workshop Co-Chairs

Shrikant Bangdiwala, PhD
University of North Carolina, Chapel Hill.

Steven Belle, PhD
University of Pittsburgh

Workshop Participants

Jennifer Beaumont, MS
Northwestern University

Alok Bhargava, PhD
University of Maryland

Dave Cella, PhD
Northwestern University

Rebecca Gersnoviez Clifton, PhD
Assistant Research Professor of Epidemiology and Biostatistics

Jennifer Foltz, MD, MPH
Centers for Disease Prevention

Debra Haire-Joshu, PhD
Washington University

Leslie Lytle, PhD
University of Carolina at Chapel Hill

Simon J. Marshall, PhD
University of California, San Diego

Paras Mehta, PhD
University of Houston

Susan Michie, PhD
Sciences University College London

Shirley Moore, RN, PhD, FAAN
Case Western Reserve University

Dan O’Connor, PhD
University of Houston

Thomas Robinson, MD, MPH
Stanford UniversitySchool of Medicine

Nancy E. Sherwood, PhD
University of Minnesota and Health Partners Institute for Education and Research

Evan Sommer, MS
Vanderbilt University

June Stevens, PhD
University of North Carolina

Deborah Tate, PhD
University of North Carolina, Chapel Hill

Thomas N. Templin, PhD
Wayne State University

Elizabeth Thom, PhD
George Washington University

Dianne Ward, EdD
University of North Carolina at Chapel Hill


Sonia Arteaga, PhD
Division of Cardiovascular Sciences, NHLBI

Denise Bonds, MD
Division of Cardiovascular Sciences, NHLBI

Susan Czajkowski, PhD
Division of Cardiovascular Sciences, NHLBI

Layla Esposito, PhD
Child Development & Behavioral Branch, NICHD

Mary Evans, PhD
Division of Digestive Diseases & Nutrition, NICHD

Larry Fine, MD, Dr PH
Division of Cardiovascular Sciences, NHLBI

Mary Horlick, MD, NIDDK
Division of Digestive Diseases & Nutrition, NIDDK

Catherine (Cay) Loria, PhD, MS, MA, FAHA
Division of Cardiovascular Sciences, NHLBI

David M. Murray, Ph.D.
Office of Disease Prevention, NIH

Victoria Pemberton, RNC, MS, CCRC
Division of Cardiovascular Sciences, NHLBI

Charlotte Pratt, PhD, RD, FAHA
Division of Cardiovascular Sciences, NHLBI

William (Bill) Riley, PhD
Division of Cancer Control and Population Sciences, NCI

Caroline Signore, MD, MPH
Division of Extramural Research, NICHD

Denise Simons-Morton, MD, PhD
Office of Disease Prevention, NIH

Erica L. Spotts, Ph.D
Office of Behavioral and Social Sciences Research, NIH

Sue Yanovski, MD
Division of Digestive Diseases & Nutrition, NIDDK



  • Czaja SJ, Schulz R, Lee CC, Belle SH; REACH Investigators. A methodology for describing and decomposing complex psychosocial and behavioral interventions. Psychol Aging, 2003;18(3):385-95.
  • Davidson KW, Goldstein M, Kaplan RM, Kaufmann PG, Knatterud GL, Orleans CT,Spring B, Trudeau KJ, Whitlock EP. Evidence-based behavioral medicine: what is it and how do we achieve it? Ann Behav Med, 2003 Dec;26(3):161-71.
  • Michie S, Ashford S, Sniehotta FF, Dombrowski SU, Bishop A, French DP. A refined taxonomy of behaviour change techniques to help people change their physical activity and healthy eating behaviours: the CALO-RE taxonomy. Psychol Health, 2011;26(11):1479-98. doi: 10.1080/08870446.2010.540664.
  • Michie S, Richardson M, Johnston M, Abraham C, Francis J, Hardeman W, Eccles MP, Cane J, Wood CE. The Behavior Change Technique Taxonomy (v1) Hierarchically Clustered Techniques: Building an International Consensus for the Reporting of Behavior Change Interventions. Ann Behav Med, DOI 10.1007/s12160-013-9486-6, 2013.
  • Michie S, van Stralen MM, West R. The behaviour change wheel: A new method for characterising and designing behaviour change interventions. Implementation Science, 2011;6:42.
  • Schulz R, Czaja SJ, McKay JR, Ory MG, Belle SH. Intervention taxonomy (ITAX): describing essential features of interventions. Am J Health Behav, 2010;34(6):811-21.
  • Weiner BJ, Lewis MA, Clauser SB, Stitzenberg KB. In Search of Synergy: Strategies for Combining Interventions at Multiple Levels. J Natl Cancer Inst Monogr, 2012;44:34 -41.

Analytical Methods

  • Bangdiwala SI, Villaveces A, Garrettson M, Ringwalt C. Statistical methods for designing and assessing the effectiveness of community-based interventions with small numbers. Int J Injury Control & Safety Promotion, 2012; 19(3):242-248.
  • Bangdiwala SI. The multilevel diagram. Int J Injury Control & Safety Promotion, 2012; 19(4):388-390.
  • Belle SH, Czaja SJ, Schulz R, Zhang S, Burgio LD, Gitlin LN, Jones R, Mendelsohn AB, and Ory MG. Using a New Taxonomy to Combine the Uncombinable: Integrating Results Across Diverse Interventions. Psychology and Aging, 2003;18(3):396-405.
  • Breiman L, Friedman JH, Olshen RA, Stone CJ. Classification and regression trees. Chapman & Hall/CRC. 1984.
  • Broeze et al., (2009). Individual patient data meta-analysis of diagnostic and prognostic studies in obstetrics, gynaecology and reproductive medicine. BMC Medical Research Methodology 2009, 9:22 doi:10.1186/1471-2288-9-22.
  • Goldstein H. Multilevel statistical models. Halsted Press. 1995.
  • Strobl C, Malley J, Tutz G. An introduction ot recursive partitioning: rationale, application and characteristics of classification and regression trees, bagging and random forests. Psychological Methods, 2009, 14(4):323-348.
  • Sutton AJ, Abrams KR, Jones DDR, Sheldon TA, Song F. Methods for meta-analysis in medical research. John Wiley & Sons, Lt

Last Updated: December 2013