Transcript Presentation Slides

Allostatic load:
conceptualisation and measurement
in the English Longitudinal Study of
Ageing
Sanna Read and Emily Grundy


http://pathways.lshtm.ac.uk
[email protected]
@PathwaysNCRM
Allostatic load
• a multisystem dysregulation state resulting from
accumulated physiological ‘wear and tear’
• Allostasis = a process whereby organism maintains
physiological stability by adapting itself to
environmental demands
- > health is a state of responsiveness and optimal
predictive fluctuation to adapt to the demands of
the environment -> dynamic biological process
interacting with context
Allostatic load
Environmental stressors
(work, home, neighbourhood)
Major life events
Trauma,
abuse
Perceived stress
Brain’s evaluation of threat ->
activates
Sympathetic-adrenal- medullary
(SAM) axis -> catecholamines
Hypothalamic-pituitary-adrenal
(HPA) axis - > glucocorticoids
Individual
differences
(genes,
development,
experience)
Behavioural
responses
(fight or flight, healthrelated behaviour –
smoking, alcohol use,
diet, exercise)
Physiological responses
Allostasis
Adaptation
Allostatic load
Disease
Adapted from McEwen, 1998
Allostasis
Adaptation
Multiple mediators of adaptation:
1) Primary effects: stress hormones (e.g. epinephrine, norepinephrine
and cortisol), anti-inflammatory cytokines (e.g. Interleukin-6)
2) Secondary outcomes: metabolic (e.g. insulin, glucose, total cholesterol,
triglycerides, visceral fat depositing), cardiovascular (e.g. systolic and
diastolic blood pressure) and immune system (e.g. C-reactive protein,
fibrinogen).
3) Tertiary outcomes: poor health, disease, death
Mediators interconnected and reciprocal, non-linear effects on many organ
systems in body - > should be measured as multisystem concept,
challenging to develop measures
Allostatic load accumulates throughout the life -> study processes in
longitudinal settings
Measures of allostatic load
Measure
Description
Group allostatic load index
the number of biomarkers falling within a high risk
percentile (e.g. upper or lower 25th percentile)
based on the sample distribution of biomarkers
values
Z-score allostatic load index
Summary measure of individual’s obtained zscores for each biomarker based on the sample
distribution of biomarker values.
A number of other methods also used for calculating composite measures:
bootstrapping, canonical correlations, recursive partitioning, grade of
membership, k-means cluster analysis, genetic programming, multivariate
distance.
Examples of biomarkers used
in measuring allostatic load
Type
Biomarker
Neuroendocrine
Epinepherine, norepinephrerine, dopamine,
cortisol, dehydroepiandrosterone (DHEAS),
aldosterone
Immune
Interleukin-6, tumor necrosis factor-alpha, creactive protein (CRP), insulin-like growth factor-1
(IGF-1)
Metabolic
HDL and LDL cholesterol, triglycerides,
glucosylated hemoglobin, glucose insulin,
albumin, creatinine, homocysteine
Cardiovascular and
respiratory
Systolic blood pressure, diastolic blood pressure,
peak expiratory flow, heart rate/pulse
Anthropometric
Waist-to-hip ratio, body mass index (BMI)
Factors associated with
allostatic load in previous studies
Socioeconomics:
education, income,
occupational status,
downward mobility,
homelessness
Individual: type
A/hostility, locus of
control, a polymorphism
of ACE gene
Neigbourhoods:
crowding, noise, lack
of housing,
rural/urban
Ethnicity: Nonwhites (U.S.)
Allostatic load
Social
networks:
emotional
support, social
position
Spirituality: religious
attendance, sense of
meaning/purpose
Family: attachment,
violence, single
parent, separation,
care-giving,
demands/criticism,
spouse
Work: control,
demands,
decisions, career
instability, effortreward imbalance
Sample
• English Longitudinal Study of Ageing (ELSA) waves 1-6 (2002-2012)
• men and women (n = 11223*) aged 50+ in 2002
• Measures:
– Biomarkers available in waves 2, 4 and 6
– Health and functioning: self reported health, limitating long-term
illness, walking speed
– Fertility history: number of children, birth before age 20 (women) or
age 23 (men), birth after age 34 (women) and 39 (men), coresidence
with child
– Background and intermediate factors: age, marital history, childhood
health, qualification, net wealth quintile, smoking, physical activity,
social support and social strain
* Core members who where interviewed in-person in 2002 (wave 1).
Selected biomarkers to measure allostatic
load in ELSA
Neuroendocrine
Immune
Cardiovascular
Respiratory
Metabolic
Body fat
DHEASa
(dehydroepia
ndrostorone
sulphate)
C-reactive
protein
Systolic blood
pressure
Peak
expiratory
flow
Total blood
cholesterol/
HDL
cholesterol
ratio
Waist-to-hip
ratio
Fibrinogen
Diastolic blood
pressure
IGF-1b
(insulin-like
growth
hormone)
a
only in wave 4
b only in waves 4 and 6
Triglycerides
Glycated HgB
Availability of valid measures in ELSA
Measure
% valid measure crosssectionally
% valid measure longitudinally among
those who participated in wave 1
Wave 2
Wave 4
Wave 2
Wave 4
Blood pressure
70
72
58
46
Waist-to-hip ratio
78
76
65
48
Lung function
75
70
62
44
Blood measures*
63
58
52
37
* CRP, Fibrinogen, cholesterol, triglycerides, glycated HgB, IGF-1, DHEAS
Allostatic load scores in ELSA
• Group allostatic load index: biomakers indicating
high risk (25th percentile) and mean calculated for
each five subsystems, range 0 - 5
Subsystem
Upper 25th percentile
Lower 25th percentile
Cardiovascular
Systolic blood pressure
Diastolic blood pressure
Immune
Fibrinogen
C-reactive protein
Metabolic
Triglycerides
Glycated HgB
Total/HDL cholesterol ratio
Body fat
Respiratory
Waist-to-hip ratio
Peak expiratory flow
Allostatic load scores in ELSA
Challanges in creating composite scores:
• Extreme values
• Medication
• Fasting
• Age and ageing
• Non-linearity and skewness
• Missing values
Extreme values in CRP
Medication and AL score
Prescribed medication was taken into account so that an individual was given
the value 1 (indicating health risk):
for diastolic and systolic blood pressure if they used blood pressure lowering
medication
for fibrinogen if they used anticoagulants
for triglycerides and HDL cholesterol ratio if they used lipid lowering
medication
for glycosylated haemoglobin if they used diabetes medication
for peak expiratory flow if they used lung function medication
Moreover, because the literature suggests that diabetic, cholesterol and blood
pressure lowering medication reduced the values of C-reactive protein between
25-30%, the values in the second highest 25 percentile were given value 1 to
indicate health risk.
Fasting
Respondent was asked to fast (%),
wave 6
1.3
22.5
76.2
Yes
No, advised respondent not safe to fast
Did not contact respondent prior to visit
Fasting was controlled
using the time when
last eaten (varies
between the waves
how it was asked).
Age, ageing and AL score
• AL score calculated in two age groups: under 65 and 65+
• Age (continuous) adjusted in the final models
• Cut-offs in very old age? Change over time?
Papers on biochemical values in old-old population (Swedish twins aged 82+):
Nilsson, S., Read, S., & Berg, S. (2009). Heritabilities for fifteen routine biochemical values: Findings in 215 Swedish twin
pairs 82 years of age or older. Scandinavian Journal of Clinical Laboratory Investigation, 69, 562 – 569.
Nilsson, S., Takkinen, S., Tryding, N., Evrin, P.-E., Berg, S., McClearn, G., & Johansson, B. (2003). Association of
biochemical values with morbidity in the elderly: a population-based Swedish study of persons aged 82 or more years.
Scandinavian Journal of Clinical Laboratory Investigation, 63, 457-466.
Nilsson, S.E., Evrin, P.E., Tryding, N., Berg, S., McClearn, G., & Johansson, B. (2003). Biochemical values in persons older
than 82 years of age: report from a population-based study of twins. Scandinavian Journal of Clinical Laboratory
Investigation, 63, 1-14.
Cut-offs for AL indicators in ELSA W2
Non-linearity
Allostatic load measures in 2004 predicting ADL problems in 2006 in men in ELSA
25
ADL problem %
20
15
10
5
0
1 Lowest
25%
2
3
4 Highest
25%
Skewness
Allostatic load score distribution in men age 60+ in wave 2
Missing values
• Values missing because of medical
contraindication or refusal to participate in nurse
examination. Taking into account the use of
medication made it possible to recover some of
missing information
• Information on at least four out of five
subsystems had to be available to calculate the
allostatic load score
• Full maximum likelihood (use of incomplete data)
in the analysis
Examples of the use of allostatic load
in studying the pathways to health in
older age
Read, S. & Grundy, E. (2014). Allostatic load and health in the older population of
England: A crossed-lagged analysis. Psychosomatic Medicine, 76, 490-496.
Grundy, E & Read, S. (2015). Pathways from fertility history to later life health: Results
from analyses of the English Longitudinal Study of Ageing. Demographic Research, 32,
107−146.
Read, S. & Grundy, E. (2012). Allostatic load - a challenge to measure multisystem
physiological dysregulation. National Centre for Research Methods Working Paper,
04/12.
Allostatic load and health:
study direction of sequences
• Disablement process and accumulation of allostatic load
assume a causal path between the factors.
• An effective method to detect direction of sequences of
effects in longitudinal settings is to apply cross-lagged
models.
Aim
• To investigate the reciprocal association between
allostatic load, self-rated health and walking speed
as a measure of functional limitation.
– allostatic load would predict functional limitation
– the association between self-rated health and allostatic
load may be reciprocal or self-rated health may even
precede allostatic load.
Cross-lagged model
Time 1
Time 2
Allostatic load
Allostatic load
Self-rated health
Self-rated health
Functional
limitation
Functional
limitation
http://pathways.lshtm.ac.uk
Cross-lagged model
Time 1
Time 2
Allostatic load
Allostatic load
Self-rated health
Self-rated health
Functional
limitation
Functional
limitation
http://pathways.lshtm.ac.uk
Cross-lagged model
Time 1
Time 2
Allostatic load
Allostatic load
Self-rated health
Self-rated health
Functional
limitation
Functional
limitation
http://pathways.lshtm.ac.uk
Results: Cross-lagged model
Wave 2
Wave 4
0.54 (0.016)
Allostatic
load
Allostatic
load
-0.09 (0.020)
-0.04 (0.015)
0.51 (0.015)
Self-rated
health
Self-rated
health
0.11 (0.014)
-0.07 (0.013)
0.05 (0.015)
Walking
speed
0.42 (0.016)
Walking
speed
Adjusted for age, gender, education, marital status, wealth, smoking, physical
activity, and social support.
Results: Cross-lagged model
Wave 2
Wave 4
0.54 (0.016)
Allostatic
load
Allostatic
load
-0.09 (0.020)
-0.04 (0.015)
0.51 (0.015)
Self-rated
health
Self-rated
health
0.11 (0.014)
-0.07 (0.013)
0.05 (0.015)
Walking
speed
0.42 (0.016)
Walking
speed
Adjusted for age, gender, education, marital status, wealth, smoking, physical
activity, and social support.
Results: Cross-lagged model
Wave 2
Wave 4
0.54 (0.016)
Allostatic
load
Allostatic
load
-0.09 (0.020)
-0.04 (0.015)
0.51 (0.015)
Self-rated
health
Self-rated
health
0.11 (0.014)
-0.07 (0.013)
0.05 (0.015)
Walking
speed
0.42 (0.016)
Walking
speed
Adjusted for age, gender, education, marital status, wealth, smoking, physical
activity, and social support.
Conclusions & Discussion
• Allostatic load predicts functional limitation
→ allostatic load may be a useful early objective indicator of health
problems. The drawbacks of using it is that it is a complex composite
measure which involves invasive data collection methods and therefore
subject to refusal and drop-out. No standardized way of measuring it.
• The association between self-rated health and allostatic load
and functional limitations were reciprocal, although the
strength of the associations suggested that self-rated health
may be an earlier indicator of health problems
→ The role of self-rate health in the disablement process seem to be less
clear: it predicts better functioning, but it is also an outcome of good
functioning. Self-rated health is simple and quick to use with high
response rates. The limitations are its subjective content and variation
from one population to another.
Conclusions & Discussion
• As hypothesised, allostatic load predicts later functional
limitations. In the future, it is important to include earlier
indicators of chronic stress (neuroendocrine and
inflammatory markers) and study longer time spans from
middle adulthood to old age to detect the accumulation of
stress.
The model to be tested
Is the association between parenthood history and health
mediated by wealth, health-related behaviours, social support
and strain, and allostatic load?
Demographic and
life history factors
Parenthood
history
Wealth, health-related
behaviours, social support
and strain
Allostatic
load
Health
The model to be tested
Is the association between parenthood history and health
mediated by wealth, health-related behaviours, social support
and strain, and allostatic load?
Demographic and
life history factors
Parenthood
history
Wealth, health-related
behaviours, social support
and strain
Allostatic
load
Health
Wave 1
Wave 2
Wave 3
Wealth
-0.10 (0.028)
-0.16 (0.024)
-0.55 (0.054)
-0.40 (0.040)
Allostatic
load
Physical
activity
Children
4 vs. 2
-0.19 (0.053)
-0.61 (0.047)
0.37 (0.109)
0.46 (0.090)
0.09 (0.021)
Limiting
long-term
illness
Smoking
0.11 (0.021)
0.30 (0.084)
Social
strain
Figure 1. Path model for all women in ELSA (n=6123). Model adjusted for age, education, being married,
marital disruption and childhood health. Significant paths are shown (unstandardized estimate and
standard error).
Wave 1
Wave 2
Wave 3
Wealth
-0.59 (0.064)
-0.13 (0.030)
-0.13 (0.027)
Physical
activity
-0.35 (0.047)
-0.14 (0.043)
0.26 (0.121)
Children
4 vs. 2
0.54 (0.122)
Smoking
Allostatic
load
0.12 (0.023)
0.62 (0.099)
0.23 (0.085)
-0.62 (0.054)
0.05 (0.023)
0.39 (0.097)
Social
strain
Figure 2. Path model for all men in ELSA (n=5110). Model adjusted for age, education, being married,
marital disruption and childhood health. Significant paths are shown (unstandardized estimate and
standard error).
Limiting
long-term
illness
Wave 1
Wave 2
Wave 3
Wealth
-0.10 (0.031)
-0.48 (0.061)
-0.14 (0.026)
Allostatic
load
0.29 (0.114)
Early
childbirth
0.09 (0.023)
0.46 (0.100)
0.33 (0.110)
Smoking
-0.23 (0.040)
0.13 (0.103) NS
Limiting
long-term
illness
-0.40 (0.043)
-0.64 (0.051)
Physical
activity
-0.24 (0.121)
Late
childbirth
Figure 3. Path model for parous women in ELSA (n=5219). Model adjusted for age, education, being married,
marital disruption, childhood health, and coresidence with child. Significant paths are shown (unstandardized
estimate and standard error).
Wave 1
Wave 2
Wave 3
Wealth
-0.24 (0.063)
-0.12 (0.033)
-0.13 (0.029)
0.29 (0.121)
Early
childbirth
Allostatic
load
0.12 (0.025)
0.33 (0.114)
Smoking
Limiting
long-term
illness
0.30 (0.149)
0.62 (0.109)
-0.35 (0.052)
-0.13 (0.040)
-0.29 (0.071)
-0.65 (0.060)
Physical
activity
0.35 (0.108)
-0.11 (0.047)
Late
childbirth
0.06 (0.026)
Social
strain
Figure 4. Path model for parous men in ELSA (n=4256). Model adjusted for age, education, being married,
marital disruption, childhood health, and coresidence with child. Significant paths are shown (unstandardized
estimate and standard error).
Conclusions & Discussion
• Socio-economic position, health-related behaviors and social
strain mediate the association between high parity and later
life health. They also partially mediate the association
between early childbirth and later life health. Of these socioeconomic position was the strongest mediator.
• So, as hypothesised, biosocial pathways from parenthood
history to health involve economic position, social strain and
health related behaviours
→ need now to examine in more detail pathways to particular fertility
trajectories- especially childhood SES and broader environmental
influences (e.g. support from the state) and other potential mechanisms
(e.g. moderation).