Epidemiologic approach to identifying causes.

1
Childhood Asthma with an emphasis on disease
misclassification and synthetic bedding.
Anne-Louise Ponsonby 1,2 Terence Dwyer
2 Jennifer Cochrane 2 Andrew Kemp 3 David Couper
4 Allan Carmichael 5



February 2003
2
The dominant past clinical model for child asthma
is that of a largely allergic (atopic) aetiology
However, a systematic review of population based
studies found mean population attributable risk
for asthma due to atopy was only 38.
Pearce et al Thorax 2000
3
Disease misclassification has been a large
problem in asthma epidemiology
Airway Obstruction
viral-
Allergen-
induced
induced
HDM -
allergen
induced
irritant-induce
d
other
causes
small airway-
induced obstrution
4
Misclassification can occur between
- Asthma and other respiratory diseases -
Different types of asthma within the asthma
spectrum - Different types of allergy within
atopy (allergy)
5
Within the spectrum of child asthma, the subgroup
of children with severe disease are more likely
to have asthma that is attributable to atopy
With Respira tory Symptoms No.
OR (95 CI) for Respiratory Symptoms
Without Respiratory Symptoms, No.
Population Attributable Fraction
Respiratory Symptoms
With Atopy,
With Atopy,
13 episodes of wheeze in past 12 mo compared to
no wheeze episodes
49
3.27 (2.154.97)
38
526
66
119
412 episodes of wheeze in past 12 mo compared to
no wheeze episodes
3.44 (1.756.75)
38
68
40
49
526
12 episodes of wheeze in past 12 mo compared
to no wheeze episodes
8.70 (3.0724.55)
75
38
526
84
25
Ponsonby et al Chest 2002

6
In large studies focusing on allergic airway
disease, an improved signal-to-noise ratio will
be obtained by focusing on symptoms that indicate
severe or persistent disease
7
Atopic Disease
Subgroups also occur within the broad group of
atopic or allergic children because different
children are allergic to difference substances
8
Again, this issue is important to consider in
aetiological studies because the causal factors
involved in different allergen-specific atopy may
differ.
9
House dust mite
Ryegrass
69
29
34
34
44
15
9
34
6
Other

-
specific (n84)
HDM

Ryegrass
specific (n43)
-
Ponsonby et al PAI 2003 (in press)
10
69
29
44
15
9
34
Unadjusted OR for HDM-specific sensitisation

Adjusted OR for HDM-specific sensitisation (
trend,p0.80)



Unadjusted OR for ryegrass-specific sensitisation




Adjusted OR for ryegrass-specific sensitisation
(trend,p0.0001) Note, the difference in family
size effect is significant, p0.02





Ponsonby et al PAI 2003 (in press)
11
Bedding analysis Aim To examine the role of
infant upper bedding composition in the
development of house dust mite allergen induced
airway disease
12
Justification for main selected outcome measures

• Frequent wheeze (more than 12 wheeze episodes in
  the past 12 months
• this outcome is strongly linked to HDM atopy
  (RR19.6 (6.94, 55.56)).
• Night wheeze temporally related to exposure of
  interest (bedding).
• Asthma not used - only weakly associated with
  HDM atopy (RR 1.65 (1.30-2.09)).

Ponsonby et al J Clin Epi 2002
13
The 1995 Tasmanian Asthma Survey Full 1995
cross-sectional sample N 6,378 (92 of
eligible) with parental questionnaires
Methods
The 1988 to 1995 Tasmanian Infant Health Survey
(TIHS) 1988N 1,111 (81 of eligible) infants
participated in home interview and survived the
first year.
1995 Follow-up sample N 863(78), TIHS
children born in 1988 with 1988 home
interview Data, plus 1995 asthma data avail ble
and parental consent for record linkage.
Ponsonby et al Epidemiology 2003
14
Is the report of wheeze and wheeze frequency
valid?

• Previous validation work has shown the report of
  wheeze over the past 12 months has a sensitivity
  of 0.81 and a specificity of 0.78 for the
  physician diagnosis of asthma in childhood.
• Increasing wheeze frequency is associated with
  increasing deficits in child lung function.

Jenkins et al, Int J Epidemiol 1996 Ponsonby et
al, Chest 2002
15
Statistical Methods

• A generalized linear model with a log link
  function and binomial error structure was used.
• Age at onset discrete proportional hazard
  modelling.

Armitage P, Berry G. Statistical Methods in
Medical Research, 1994
16
Statistical Methods cont.
Etiologic fraction of wheeze attributable to
synthetic bedding P (aRR-1)/aRR 16 of
moderate wheeze attributable to synthetic
bedding 79 for frequent wheeze attributable
to synthetic bedding Statistically
attributable, causality not yet fully proven.
Ponsonby et al Epidemiology 2003
17
Infant synthetic pillow use and subsequent child
asthma symptoms at age seven, 1995 follow-up
sample
Infant synthetic pillow use and respiratory
symptom
Infant synthetic pillow use
N
 N

ARR 95CI
RR 95CI
No wheeze in past year
Moderate wheeze (1-12 episodes in past year)
Frequent wheeze (More than 12 episodes in past
year)
28
2.8 (1.3-6.1)
9
32
2.5 (1.2-5.5)
No night wheeze
Night wheeze
Ponsonby et al Epidemiology 2003
18
(No Transcript)
19
Child wheeze frequency over the past 12 months by
current use of synthetic bedding , full 1995
cross-sectional sample
Distribution of children by number of wheeze
episodes ()
N
Bedding
Frequent wheeze (more than 12 Episodes vs. none)
1-12
More than 12
 0
ARR
(95CI)
Neither pillow nor quilt synthetic
1.00
0.6
18
82
450
Only pillow synthetic
(1.5, 25)
6.03
2.0
19
80
3091
Only quilt synthetic
4.4
2.7
21
76
148
(1.1, 18)
Both pillow and quilt synthetic
6.4
2.7
24
73
2373
(1.2, 35)
Ponsonby et al Epidemiology 2003
20
An investigation of non-causal explanations for
the synthetic-bedding wheeze association

• Not due to parents introduction of synthetic
  bedding as part of an allergen reduction strategy
  (synthetic bedding not associated with markers of
  active allergen avoidance, e.g.
  allergen-occlusive mattress covers)
• No evidence parents of children with an-at-birth
  family history of asthma were selecting synthetic
  bedding. (Even among children with no family
  history of asthma, the consistent use of a
  synthetic pillow in early life was associated
  with night wheeze (aRR3.2 (1.0-10.1)
• No evidence that synthetic bedding was linked to
  under-management of wheeze. (Among children with
  asthma, synthetic bedding asthma medication
  use).

Ponsonby et al Epidemiology 2003
21
Features indicative of a causal relationship

• High strength of association
• Dose-response patterns
• Temporality
• prospective association evident
• earlier use of bedding earlier disease onset
• Ecological coherence
• Between 1978 and 1991, declining use of feather
  pillows could account for 20 of increase in
  current wheeze overtime
• 
  Butland Thorax 1997
• Consistency
• Many cross-sectional studies. These results are
  consistent with one other birth co
• hort (feather bedding protective)
  Nafstad CEA 2002

22
Causal Features

• Biological plausibility
• The adverse effect of synthetic bedding is more
  evident among atopic than non-atopic children,
  thus atopic mechanisms may be involved. The
  adverse effect of non-feather bedding is
  particularly evident among HDM allergic children
  (see next slide).
• HDM and other allergens are much higher in
  synthetic than feather bedding
• Direct allergen loading near the face
• Volatile organic compounds
• Lack of protective endotoxin products

23
A cross-sectional assessment of the combined
effect of HDM sensitization and feather quilt use
on severe asthma symptoms and lung function
Adjusteda rate ratio (95 CI) for severe symptoms
over the past year
HDM Sensitization
P Value
P value
Feather quilt use
Lung function FEV1/FVC ratio change () (95 CI)b
0.00 (reference)
1.00 (reference)
C
-
-
Ö
0.54
1.45 (0.45-4.65)
C
C
-0.37 (-1.80 to 1.06)
0.61
-0.31 (-1.33 to 0.71)
0.06
1.79 (0.97-3.27)
Ö
Ö
0.54
-0.68 (-1.24 to 0.13)
6.38 (2.51-16.23)
0.02
C
Ö
0.0001
Ponsonby et al J Clin Epi 2002
24
Causal Features cont.

• Experimental data from randomised controlled
  trials are not yet available.
• An RCT on this issue is underway in
  Australia.

25
Conclusions

• Careful consideration of disease
  misclassification within the spectrum of asthma
  is required for all studies investigating the
  aetiology of asthma.
• Synthetic bedding is prospectively associated
  with the subgroup of asthma that represents house
  dust mite related airway disease.

26
Conclusions cont

• Observational studies have demonstrated
  several causal features with regard to the
  synthetic bedding-frequent child wheeze
  association.
• Randomised trials are required, however, to
  fully exclude selection bias with regard to
  family choice of child bedding.