Download Is a subgroup assessment sufficient?

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Is a subgroup assessment sufficient?
Department of Epidemiology
German Institute of Human
Nutrition Potsdam-Rehbrücke
Scenario
Criteria
Whole study
(n=30,000)
Random sample
(n=1,500)
Assessment
method
2 x 24 HDR + 1 x Food
Propensity Questionnaire
2 x 24 HDR + 1 x Food
Propensity Questionnaire
Staff
requirements
for 24 HDRs
25 interviewers
1 interviewer
€ 750.000
€ 30.000
Estimated total
costs of dietary
assessment
Objective
Comparison of food intake distributions derived
by different statistical methods
Study design
EPIC-Potsdam-Calibration Study II
• random sample of EPIC-Potsdam-Cohort:
– n= 393; 197 men and 196 women
• dietary assessment methods:
– 2 x 24HDRs by telephone, randomized over one year
– 1 x 102 food-item FFQ („food propensity questionnaire“) at the end
of the year
Nöthlings et al.(2007): Fitting portion sizes in self-administered Food frequency questionnaire. The Journal of Nutrition, 137: 2781-2786
Statistical analysis
food groups:
fresh fruits (often); fish (occasionally); breakfast cereals (rarely)
• 2-day-means
• Multiple-source method
– 2 x 24 HDRs and 1 x FFQ frequency as covariate
• Linear regression calibration methods
– Simple linear regression calibration
– Linear regression calibration with correction of intraindividual
variance
Hoffmann et al.(2002): Estimating the distribution of usual dietary intake by short-term measurements. European Journal of Clinical Nutrition, 56, Suppl 2, S53-S62
Results: Fresh fruits
Fresh fruits0
Method
P10
P25
0
7
103
234
395
513
67
102
145
239
350
Simple linear
Calibration 1
[g/day]
150
161
181
245
Linear
Calibration 2
[g/day]
74
100
150
232
2-day means
[g/day]
Multiple
Source
[g/day]
0
P5
P50
P75
P90
P95
Arithmetic mean
Standard deviation
678
265
203
469
544
264
148
324
400
432
265
93
346
470
542
265
154
N: 393
simple linear regression calibration
2 linear regression calibration with correction of intraindividual variance
1
Results: Fish
Fish0
Method
P5
P10
P25
P50
P75
P90
P95
Arithmetic mean
Standard deviation
2-day means
[g/day]
0
0
0
0
45
100
127
28
48
Multiple Source
[g/day]
4
7
12
19
44
61
71
28
23
Linear
Calibration 1
[g/day]
19
20
22
25
31
38
47
28
9
Linear
Calibration 3
[g/day]
2
3
10
19
37
59
87
28
27
0
N: 393
simple linear regression calibration
2 linear regression calibration with correction of intraindividual variance
1
Results: Breakfast cereals
Breakfast Cereals0
Method
P5
P10
P25
P50
P75
P90
P95
Arithmetic mean
Standard deviation
2-day means
[g/day]
0
0
0
0
0
6
25
3
11
Multiple Source
[g/day]
0
0
0
0
1
12
20
3
8
Linear
Calibration
[g/day]
0
0
0
0
3
9
18
3
6
0
0
0
0
3
11
22
3
8
1
Linear
Calibration 3
[g/day]
0
N: 393
simple linear regression calibration
2 linear regression calibration with correction of intraindividual variance
1
Correlations
Pearson Correlation Coefficients
Method
Multiple Source vs FFQ
Fresh fruits
Fish
Breakfast cereals
0.95
<0.0001
0.94
<0.0001
0.93
<0.0001
Summary
1.
There is a high correlation between FFQ and MSM
2.
Subgroup assessment applying the Multiple source
method and standardization of the FFQs of the whole
cohort according to the MSM distribution moments?
Thank you for your attention!
Related documents