lung function in the era of the global lung initiative · 2019-09-10 · z-scores to the rescue...
TRANSCRIPT
Lung Function in the Era of the
Global Lung Initiative: Changing
how we think about lung function
Larry C. Lands, MD, PhDProfessor of Pediatrics, McGill University
Director, Pediatric Respiratory Medicine, MCH-MUHC
“Making a clinical diagnosis is an art, where test
results help to confirm or reject the diagnosis.”
Quanjer et al, ERJ 2012
What is Normal?
Sodium 140 mmol/L
Hemoglobin 180 g/L
FEV1 77% predicted
FEV1 across Ages
Stanojevic et al, AJRCCM 2009
Males Females
Cross-Age Predictions
FVC
FEV1
FEF25-75
Stanojevic et al, ERJ 2010
The Bell Curve
Statistical Modeling for the GLI
Lower (Upper) Limit of Normal: LLN (ULN)
Comprises 5% of the population
Modeling based on height2.5
Takes into account variability between individuals
Uses z-scores (based on SD)
Variations Between Individuals
Coefficient of Variation = Standard Deviation/Mean
The SD normalized for its mean value
By tradition, we have assumed a 10% coefficient of
variation, so
2 X CV is ±20%, ie 80-120% considered as
NORMAL
WHAT IF THE CV WAS MORE THAN 10%?
Coefficient of Variation
FVC
FEV1
FEF25-75
Stanojevic et al, ERJ 2010
Z-scores To the Rescue
Z-score combines % predicted and the variation
between individuals “to give a single number that
accounts for age- and height-related lung function
variability expected within comparable healthy
individuals”
LLN of a z-score is -1.64 (lowest 5%)
ULN of a z-score is +1.64 (highest 5%)
Stanojevic et al, ERJ 2010
Going Global
Caucasians African-Americans
North East Asians South East Asians
Quanjer et al, ERJ 2012
Who’s Who?
White: Caucasian
Black: African-American
North-East Asian: North China, Korea
South-East Asia: South China, Thailand, Taiwan
Other: other groups or mixed ethnicity
Going Global-FEV1
Quanjer et al, ERJ 2012
Males Females
Going Global-Accuracy of Input Values
Quanjer et al, ERJ 2012
Entering 160 cm instead of 161 cm for height changes
the predicted value for FEV1 by 1.3%
Age needs to be to 1 decimal point: Using 14 years
instead 14.9 underestimates FEV1 by 4.7%
For a 14 year old male, combining these 2 errors would
result in a 6% difference in predicted value
Going Global-Coefficient of Variation
Quanjer et al, ERJ 2012
Males
Coefficient of Variation
FVC
FEV1
FEF25-75
Stanojevic et al, ERJ 2010
Going Global-5% LLN
Quanjer et al, ERJ 2012
Females
FEV1/FVC without GLI
Swanney et al, Thorax 2008; Hansen et al, Chest 2007
Problems with Fixed FEV1/FVC
Quanjer et al, ERJ 2012
FEV1/FVC with GLI
Quanjer et al, ERJ 2012
Males Females
Assessing Obstruction-Do we need FEF25-75?
Quanjer et al, ERJ 2014
% FVC, FEV1, FEV1/FVC
WNL and low FEF25-75
Assessing Obstruction-Do we need FEF25-75?
Lukic and Coates, Ped Pulmonol 2015
Going Global-Comparison with
Other Prediction Equations
Quanjer et al, ERJ 2012
Going Global-Test Interpretation
Quanjer et al, ERJ 2012
Using multiple tests that are physiologically linked and
applying the 5% LLN to each of them has a multiplicative
effect leading to more false positives
Combining FEV1, FVC, FEV1/FVC would give about
10% (10.4-10.6%) abnormal
Using the 2.5% (z-score: -1.96) reduces this to 5.6-
5.8% abnormal
Suggested:
For screening and case finding use -1.96 (2.5%)
For those with prior lung disease, borderline values
more likely to be associated with disease and so use -1.64
(5%)
Going Global-Test Interpretation
Culver et al, AJRCCM, 2017
Example 1
Mild Asthma with BD Response
Example 2
Moderate-severe Cystic Fibrosis
Example 3
Tracheal Rings
Lum S et al. Thorax 2012;67:A18
Copyright © BMJ Publishing Group Ltd & British Thoracic Society. All rights reserved.
UK Validation Study
Changes in Cormic Index (Sitting
Height/Standing Height)
Quanjer et al, Chest 201517 year old women in
Japan
FEV1
FVC
FEV1/FVC
Lum, PLOS ONE, 2016
South Asian Children
UK South Asian Children
FEV1
FVC
FEV1/FVC
Lum, PLOS ONE, 2016
South Asian Children (India)
South Asian Children
Centre B: Northern India
Arigliani, AJRCCM, 2017
African-American or
African
Oh, Canada!
Coates, Ann Am Thorac Soc, 2016
Hüls, PLOS ONE, 2016
Germany, Ja!
GLI May Not Always Apply
Abdullah et al, Respirology 2018
2003 2010 Δ/year
mean p mean p mean p
Fev1 (ml) Observed 1107.4 2435.9 180.4
Predicted 1123.2 2497.5 188.3
δ -15.8 NS -61.6 * -7.9 *
FVC (ml) Observed 1120.7 2824.1 233.4
Predicted 1216.0 2876.4 227.8
δ -95.2 ** -52.3 ** 5.6 **
FEV1/FVC (%) Observed 96.6 85.5 -1.57
Predicted 92.9 87.1 -0.8
δ 3.7 ** -2.4 ** -0.77 **
FEF25-75 (ml/s) Observed 1621.3 2766.9 152.8
Predicted 1602.6 2955.8 185.2
δ 18.7 NS -269.6 ** -32.3 **
Piccioni, BMC Pulm Med, 2015
Longitudinal Assessment
Piccioni, BMC Pulm Med, 2015
Longitudinal
Assessment
Armspan to Predict Height
Quanjer et al, ERJ 2014
Diffusing Capacity
Cooper et al, Breathe 2017
GLI DLCO for Caucasians
Compared to values frequently
used in devicesStanojevic et al, ERJ 2017
Ethnicity and Lung Volumes
Kirkby et al, Ped Pulmonol 2013
Conclusions
• Pulmonary function tests are clinical data with an
expected range of normal
• Generalized equations allows for measurement
across the age span and wider applicability
• More data on other population groups is required
• Using a 5% Lower Limit of Normal (LLN) allows for
more accurate classification of patients