@ibtnetwork #ibtn2020
Data harmonization and individual patient data meta-analysis: exploring new avenues for evidence summaries and intervention development
Jovana Stojanovic, PhD, MSc, MPharm
Postdoctoral fellow
Department of Health, Kinesiology, and Applied Physiology;
Concordia University, Montreal, Canada
May 28, 2020
Contact : [email protected]
CONTENT
• Introduction
• Overview of the Action study (DecreAsing sedentary behaviour and inCreasing
physical acTIvity fOr healthy ageing)
• Data harmonization: process & practical aspects
• Individual Patient data (IPD) meta-analysis: process & practical aspects
• Conclusions
INTRODUCTION - POPULATION AGEING
most advanced in Europe and in Northern America global population aged 60 an over:
→ 1980 (382 million) → 2017 (962 million )→ 2050 (2.1 billion )
United Nations: World Population Ageing 2017; Chang et al. Lancet Public Health 2019; 4: e159–67
79 % of older individuals will be living in the developing regions
GBD (195 countries 1990 – 2017): age-related disease burden accounts for 51.3% of all burden
• % of older adults meeting current guidelines: 2.4-83%
Guthold et al. Lancet Glob Health 2018; Sun et al. BMC Public Health 2013; Hallal et al. Lancet 2012; Harvey et al. Journal of Aging and Physical Activity, 2015; Palmer et al. Gerontologist, 2019; Copeland JL et al. Br J Sports Med 2017; Hansen et al. Med Sci Sports Exerc. 2012 Feb; Cunningham et al. Scand J Med Sci Sports; 2020.
• ≈ 9.4 hr∕day (Objective)
• ≈ 5.3 hr∕day sitting ≈ 3.3 hr∕day TV watching (Subjective)
↓ overall mortality↓ CVD risk ↓ breast and prostate cancer↓fractures, and recurrent falls↑ functional and cognitive capacity↓ depression↓ dementia and Alzheimer
↓ overall mortality↓ CVD risk ↓ functional and cognitive capacity ↓ autonomy↑ functional and cognitive decline ? depressionLack of longitudinal evidence The importance of context → cognitively engaging SB
PHYSICAL ACTIVITY (PA) SEDENTARY BEHAVIOUR (SB)
HEALTHY AGEING
healthy ageing = active ageing = successful ageing = productive ageing
Rowe and Kahn 1987
Usual ageing vs Successful aging 1997:
low probability of disease and disease-related disability and related risk factors high cognitive and physical functional capacity sustained engagement in social and productive activities
2015: societal-level principles (opportunities for employment, voluntary work, social
activity…)
Adapted from: Lu et al. Gerontologist, 2019;
HEALTHY AGEING
Domains of Healthy Aging: A review of 55 articles published within March 2017.Number of papers measuring each domain of HA
Adapted from: Lu et al. Gerontologist, 2019;
ACTION STUDY OVERVIEW
Through the use of existing international longitudinal studies, we aim to understand whether engaging in PA and SB after the age of 65 might contribute to healthy ageing in later life.
Screening for potentially eligible cohorts• PUBMED (171 retrieves)
• Inter-university Consortium for Political and Social Research (77 studies)
Studies Continent Number of studies
Number of individuals
Americans' Changing Lives (ACL); Canadian longitudinal study on Ageing (CLSA); Cardiovascular Health Study (CHS); Charleston Heart Study; Cornell Study of Occupational Retirement; Costa Rican Longevity and Healthy Aging Study (CRELES); Framingham cohort study; Framingham Gen III; Framingham Offspring; Hispanic Established Population for the Epidemiological Study of the Elderly (HEPESE); Honolulu Heart Program (HHP); Long Beach Longitudinal Study; Midlife in the United States (MIDUS) Series; Multi-Ethnic Study of Atherosclerosis (MESA); National Health and Aging Trends Study (NHATS); National Social Life, Health, and Aging Project (NSHAP); Sacramento Area Latino Study on Aging (SALSA Study); Study of Women's Health Across the Nation (SWAN) Series; The Health and Retirement Study (HRS); Women's Health Initiative Study (WHI)
America 20 240K
Chinese Longitudinal Healthy Longevity Survey (CLHLS); Survey of Midlife in Japan (MIDJA); Social Environment and Biomarkers of Aging Study (SEBAS) in Taiwan
Asia 3 99K
The Irish Longitudinal Study on Ageing (TILDA); Swedish Adoption/Twin Study on Aging (SATSA); English Longitudinal study on Ageing (ELSA)
Europe 3 20K
The Survey of Health, Ageing and Retirement in Europe (SHARE); Study on Global AGEing and Adult Health (SAGE) International 2 180K
Australian [Adelaide] Longitudinal Study of Aging Australia 1 2K
Total 29 540-545K
ACTION STUDY OVERVIEW
Data harmonization
Harmonized Dataset
Dataset C
Dataset A
Dataset B
Process of retrospective data harmonization involves poolingheterogeneous data from disparate data sets and transforming them into1 cohesive data set
Rolland et al.. 2015;
Assemble information and
select studies
Process data Estimate quality of
the harmonized dataset
Define variables and evaluate
harmonization potential
Step 2
Step 4Step 3
Step 1
Step O Step 5
Pre-harmonization Core process Post-harmonizationProject Proposal Dissemination
Data harmonization
Fortier et al. International Journal of Epidemiology, 2017;
Data harmonization – simple example
Harmonized variable: Ever alcohol drinking
0=No1=Yes
Did you drink alcohol in the past?
0=No, never
1=Yes, more than 3 servings per
week
2=Yes, less than 3 servings per
week
Do you currently drink alcohol?
0=No
1=Yes, less than 3 servings
per week
2=Yes, more than 3 servings
per week
Do you consume alcohol in the
present time?
0=No, ex-drinker
1=Yes
Did you ever drink alcohol?
0=No
1=Yes
If noIf yes
Study 2
Study 1
Data harmonization exampleHRS SHARE TILDA CRELES
w1 - w13 w1-w2, w4-w7 w1-w2 w1-w3 Waves
aerobics, running,
swimming, bicycling
sports, heavy housework,
job involves physical labor
heavy lifting, digging, aerobics,
fast bicycling
sports, jogging,
dancing, or heavy
work
Activity
every day,
more than 1
time/week,
1 time/week,
1-3 times/month,
never
> than 1 time/week,
1 time/week,
1-3 times/month,
never
Number of days/week
&
Number of hours/day
3 times/week Unit
cathegorical cathegorical continious binary Type
w7-w13 w1-w2, w4-w7 w1-w2 - Waves
gardening, cleaning
the car, walking
gardening, cleaning the
car, walking
carrying light loads, bicycling at a
regular pace, or doubles tennis- Activity
every day,
more than 1,
time/week,
1 time/week,
1-3 times/month,
never
> than 1 time/week,
1 time/week,
1-3 times/month,
never
Number of days/week
&
Number of hours/day
- Unit
cathegorical cathegorical continious - Type
w7-w13 - w1 w1-w3 Waves
laundry, home
repairswalking walking Activity
every day,
more than 1,
time/week,
1 time/week,
1-3 times/month,
never
-
Number of days/week
&
Number of hours/day
number of hours Unit
cathegorical - continious continious Type
Mild
Vigorous
ModeratePhysical Exercise
Measure
HRS SHARE TILDA CRELES
w1 - w13 w1-w2, w4-w7 w1-w2 w1-w3 Waves
aerobics, running,
swimming, bicycling
sports, heavy housework,
job involves physical labor
heavy lifting, digging, aerobics,
fast bicycling
sports, jogging,
dancing, or heavy
work
Activity
every day,
more than 1
time/week,
1 time/week,
1-3 times/month,
never
> than 1 time/week,
1 time/week,
1-3 times/month,
never
Number of days/week
&
Number of hours/day
3 times/week Unit
cathegorical cathegorical continious binary Type
Measure
Physical Exercise Vigorous
Harmonized Vigorous physical exercise
≤ than 1 time/weekNever ≥ than 1 time/week Everyday
WHAT IS IPD meta-analysis
• Instead of extracting aggregate data from the published literature, IPD meta-analysis allows the original participant level data to be requested and re-analysed
• ‘gold-standard’ in evidence summaries
Stewart et al.. JAMA.2015;
Data harmonization and IPD meta-analysis in the era of open data
*Adapted from: Nevitt SJ et al. BMJ 2017; Ioannidis. Milbank Q. 2016;
Number of IPD- MAs published from 1994 to 2014
Number of SRs and MAs published from 1994 to 2014
Study 1 Study 2 Study 3 Study N
Standard fixed or random effects MA
Study specific
estimates
IPD meta-analysis methods: 2-STAGE APPROACH
Study 1 Study 2 Study 3 Study N
Participant level
Studylevel
IPD meta-analysis methods: 1 STAGE APPROACH Multilevel model
Inconsistently defined exposures and outcomes
Complex data types, such as long-term outcomes, and time-dependent data
Subgroup analysis and exploration of interactions
IPD vs. Aggregate data (AD) meta-analysis
• A review of 39 studies that compared IPD and AD meta-analysis of randomized trials
190 comparisons
80% agreement in statistical significance
! 23 comparisons disagreed in direction of effect
20% disagreement in statistical significance
! 15% IPD-MA significant & AD-MA non-significant! 5% IPD-MA non-significant & AD-MA significant
144 comparisons for the main effect
46 comparisons for effect modifiers
Tudur et al. Cochrane Database of Systematic Reviews 2016.
• https://g2aging.org/
• https://www.maelstrom-research.org/
• https://biolincc.nhlbi.nih.gov/home/
• https://www.icpsr.umich.edu/web/pages/ICPSR/index.html
• https://vivli.org/
• https://yoda.yale.edu/
• https://clinicalstudydatarequest.com/Default.aspx
Harmonization:
Individual patient data access:
Data harmonization and IPD meta-analysis in the era of open data
Opportunities
• Include unpublished, and/or poorly reported data• Include studies with lots of missing data• Standardize/harmonize outcomes and exposures• Suitable for complex data types• ↑ data quality• ↑ generalizability of results• ↑ statistical power• Refined subgroup and sensitivity analysis, explore interactions• Ability to focus on rare outcomes• Encouraging more efficient secondary usage of existing data• Collaboration
Challenges
• Time & resource-consuming
• Multi-disciplinary teams (data management and statistical expertise required)
• Heterogeneity of the collected variables
• Understanding the meaning of specific variables
• Ethical, legal, and consent-related restrictions (especially for international research):
• Data sharing agreements
• Confidentiality & protection agreements
• Availability (selection) bias
Performing an IPD meta-analysis: practical considerations
1. Research objectives: - Outcomes &exposures- Subgroups or covariates- Data type and modelling
2. Data availability Adopt a systematic approach: search for all relevant published and
unpublished studies → contact study authors Convenience: include research groups from collaborative initiatives Beware of availability bias: sensitivity with AD and IPD dataPRISMA-IPD checklist
3. Available resources
Stewart et al. JAMA.2015;
ACTION study → consideration for PA/SB intervention development in the ageing population
development of evidence-based behavioral goals → improve health outcomes
include environmental, organizational and societal perspective → WHO definition of an
age-friendly community
effective communication and messaging
World Health Organization. (2007).
@ibtnetwork #ibtn2020
ACKNOWLEDGEMENTS
ACTION study investigators: International Research team lead by Drs. Bacon and Lavoie Simon L. Bacon, PhD
ProfessorConcordia University, Canada
MBMC Co-director
Kim L. Lavoie, PhDProfessorUQAM, Canada
MBMC Co-director
Funding bodies
MBMC Team
@ibtnetwork #ibtn2020
References: Healthy ageing
• Cunningham et al. Consequences of Physical Inactivity in Older Adults: A Systematic Review of Reviews and Meta-Analyses. Scand J Med SciSports 2020 May;30(5):816-827.
• Chang et al. Measuring population ageing: an analysis of the Global Burden of Disease Study 2017. Lancet Public Health 2019; 4: e159–67• Lu et al. Domains and Measurements of Healthy Aging in Epidemiological Studies: A Review. The Gerontologist, Volume 59, Issue 4, August
2019, Pages e294–e310, • Palmer et al. What Do Older People Do When Sitting and Why? Implications for Decreasing Sedentary Behavior. Gerontologist 2019 Jul
16;59(4):686-697. • Guthold et al. Worldwide Trends in Insufficient Physical Activity From 2001 to 2016: A Pooled Analysis of 358 Population-Based Surveys With
1·9 Million Participants. Lancet Glob Health 2018 Oct;6(10):e1077-e1086. • Copeland et al. Sedentary Time in Older Adults: A Critical Review of Measurement, Associations with Health, and Interventions. Br J Sports
Med 2017 Nov; 51(21):1539. • United Nations, Department of Economic and Social Affairs, Population Division (2017). World Population Ageing 2017 • Harvey et al. How Sedentary Are Older People? A Systematic Review of the Amount of Sedentary Behavior. J Aging Phys Act 2015 Jul;
23(3):471-87. • Sun et al. Physical Activity in Older People: A Systematic Review. BMC Public Health 2013 May 6;13: 449. • Hallal et al. Global Physical Activity Levels: Surveillance Progress, Pitfalls, and Prospects. Lancet 2012 Jul 21; 380(9838):247-57. • Hansen et al. Accelerometer-determined Physical Activity in Adults and Older People. Med Sci Sports Exerc 2012 Feb; 44(2):266-72. • World Health Organization. (2007). Global age-friendly cities: A guide. Geneva, Switzerland.
@ibtnetwork #ibtn2020
References: IPD meta-analysis and data harmonization
• Tierney JF et al. Comparison of aggregate and individual participant data approaches to meta-analysis of randomised trials: An observational study. PLoSMed 17(1): e1003019, 2020.
• Fortier et al. Maelstrom Research guidelines for rigorous retrospective data harmonization. International Journal of Epidemiology, 2017, 103–115 • Nevitt et al. Exploring changes over time and characteristics associated with data retrieval across individual participant data meta-analyses: systematic
review. BMJ 2017; 357: j1390. • Ioannidis. The Mass Production of Redundant, Misleading, and Conflicted Systematic Reviews and Meta-analyses. The Milbank Quarterly, Vol. 94, No. 3,
2016 (pp. 485-514)• Smith et al. Individual participant data meta-analyses compared with meta-analyses based on aggregate data. Cochrane Database of Systematic Reviews
2016, Issue 9. Art. No.: MR000007. • Debray et al. Get real in individual participant data (IPD) meta‐analysis: a review of the methodology. Res Synth Methods. 2015 Dec; 6(4): 293–309. • Rolland et al. Toward Rigorous Data Harmonization in Cancer Epidemiology Research: One Approach. Am J Epidemiol. 2015; 182(12):1033–1038• Stewart et al. Preferred Reporting Items for a Systematic Review and Meta-analysis of Individual Participant Data the PRISMA-IPD Statement. JAMA.2015;
313(16):1-1665.• Huang Y et al. Distribution and Epidemiological Characteristics of Published Individual Patient Data Meta Analyses. PLoS ONE 9(6): e100151, 2014. • Thomas et al. Systematic review of methods for individual patient data meta- analysis with binary outcomes. BMC Medical Research Methodology 2014,
14:79• Ahmed et al. Assessment of publication bias, selection bias, and unavailable data in meta-analyses using individual participant data: a database survey. BMJ
2011; 344:d7762• Riley et al. Meta-analysis of individual participant data: rationale, conduct, and reporting. BMJ 2010; 340:c221• Stewart and Tierney. To IPD or not to IPD? Advantages and Disadvantages of Systematic Reviews Using Individual Patient Data. Evaluation & the health
professions, vol. 25 no. 1, march 2002 76-97.