Reference Values for Arterial Blood
Gases in the Elderly*
Jon A. Hardie, MD; William M. Vollmer, PhD; A. Sonia Buist, MD;
Ivar Ellingsen, MD, PhD; and Odd Mørkve, MD, PhD
Objectives: We present reference values for arterial blood gas measurements for persons > 70
years old. At the same time, we wish to examine how different criteria for exclusion from the
reference sample with regard to previous smoking and various comorbidities might influence
reference values.
Methods: After first screening a random sample of the general elderly population by postal
questionnaire, we selected 146 men and women without respiratory disease, significant dyspnea,
symptomatic heart disease or hypertension, or current smoker status. Arterial blood samples
were drawn from subjects while in the supine position.
Results: The mean (SD; lower limit of normal) PaO2 and arterial oxygen saturation (SaO2) for men
was 77.0 mm Hg (9.1; 62.0) and 95.3% (1.4; 93.0), respectively, and for women was 73.5 mm Hg
(8.4; 59.6) and 94.8% (1.7; 92.0). Mean (SD; upper limit of normal) PaCO2 was 39.4 mm Hg (3.3;
44.8) for both sexes. None of the blood gas variables were associated with age, smoking history,
or presence of various comorbidities.
Conclusions: The reference values for PaO2 and SaO2 in elderly persons are sex specific but age
independent. Ex-smokers and persons with nonpulmonary comorbidities who do not have
significant respiratory symptoms need not be excluded from the reference sample for arterial
blood gases.
(CHEST 2004; 125:2053–2060)
Key words: aged, 70 and over; blood gas analysis; frail elderly; reference; respiratory function tests
Abbreviations: ATS/DLD ⫽ American Thoracic Society/Division of Lung Disease; BMI ⫽ body mass index;
Fio2 ⫽ fraction of inspired oxygen; P(A-a)O2 ⫽ alveolar-arterial oxygen pressure gradient; Sao2 ⫽ arterial oxygen
saturation; So2 ⫽ oxygen saturation
of arterial blood gas tensions
T heandmeasurement
oxygen saturation (So ) is a cornerstone in
2
the evaluation of pulmonary disease severity. Reference values for arterial blood gases should be based
on subjects representative of the general healthy
population.1 With regards to the elderly population,
designating a reference group of subjects is difficult
due to extensive comorbidity and the large percent*From the Institute of Internal Medicine (Dr. Hardie), Deaconess Hospital, University of Bergen, Bergen, Norway; Kaiser
Permanente Center for Health Research (Dr. Vollmer), Portland,
OR; Division of Pulmonary and Critical Care Medicine (Dr.
Buist), Oregon Health Science University Portland, OR; and
Department of Thoracic Medicine (Drs. Ellingsen and Mørkve),
Haukeland Hospital, Bergen, Norway.
Funding was provided by the University of Bergen, Bergen,
Norway; Nasjonalforeningen for folkehelse, Oslo, Norway; GlaxoSmithKline, Oslo, Norway.
Manuscript received June 25, 2003; revision accepted January 6,
2004.
Reproduction of this article is prohibited without written permission from the American College of Chest Physicians (e-mail:
[email protected]).
Correspondence to: Jon A. Hardie, MD, Institute of Internal
Medicine, Deaconess Hospital, University of Bergen, Ulriksdal 8,
N5009 Bergen, Norway; e-mail: [email protected]
www.chestjournal.org
age who have previously been smokers.2 While it is
clear that current smokers should be excluded from
the reference sample, it has not been established
whether this applies to ex-smokers also.
Existing reference values for arterial blood gases in
the elderly are based on convenience samples or
extrapolation of values in younger age groups, and
therefore are not necessarily representative of the
general lung-healthy, elderly population. Two studies within the last decade3,4 have published reference
values for arterial blood gases in this age group. They
found no evidence of the age-related fall in Pao2 that
has been shown in numerous studies5–13 performed
in younger age groups. Although survival effect may
be the reason for this lack of age association,3
another explanation might be that previous studies
had more exclusions for comorbidities and previous
smoking.
Unfortunately, the accuracy of different blood gas
analyzers confounds the situation, so that reference
values established in one laboratory cannot be applied to other laboratories without adding in a
significant measure of interinstrument variation.14
CHEST / 125 / 6 / JUNE, 2004
Downloaded From: http://journal.publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22010/ on 06/14/2017
2053
Therefore, conclusions on how to establish reference
values need to be based on comparisons of subjects
tested at the same site.
The aim of this study was to present reference
values for arterial blood gas tensions (oxygen and
carbon dioxide) in supine persons ⱖ 70 years old
based on a random, general population sample. We
also wish to establish a standard on which selection
of an appropriate reference sample of the elderly
population can be based.
Materials and Methods
This blood gas study was the second part of a two-phase study
in which the first part was a postal questionnaire study on
respiratory symptoms in the elderly population. The study was
performed in the city of Bergen, Norway (population 209,000).
The sample for this study was drawn in two phases. In the first
phase, a sample was drawn from the general elderly population
for the postal questionnaire study (sample size, n ⫽ 2,871; response, n ⫽ 1,649). The questionnaire concerned respiratory
health and was based on the American Thoracic Society/Division
of Lung Disease (ATS/DLD) respiratory questionnaire.15 The
sample for the second phase (blood gas testing and spirometry)
was drawn from the healthy responders. All respondents reporting any chronic or current acute respiratory disease, ATS/DLD
dyspnea grade 4 (after 100 m on level ground), heart disease or
hypertension if complicated by ATS/DLD dyspnea grade 3 (ever
having to stop due to breathlessness on the level), or currentsmoker status were excluded. The phase-two sample size was 319
subjects. Of these 319 subjects, there were 153 participants who
could meet at the clinic and 146 participants from whom an
arterial blood sample was obtained. The seven persons who came
to the clinic but did not give blood samples are considered among
the nonresponders. The questionnaire also provides background
data on those included in the reference sample as well as the
nonresponders.
Arterial Sampling
Blood samples were drawn from the radial artery after the
subject had been in supine position (approximate 30° elevation of
the head and neck, horizontal thorax) for at least 10 min. Samples
of 1.5 mL were drawn with a self-filling, polypropylene syringe
containing 60 IU dry, electrolyte-balanced heparin (Radiometer
PICO 70; Radiometer; Copenhagen, Denmark). Air pockets were
expelled immediately after the sampling, and the sample was
thereafter mixed gently until testing on the analyzers (approximately 3 to 7 min). To ensure that each sample was of arterial
blood, we discarded samples in which the syringe did not fill
quickly, and drew a new sample. None of the samples were
obtained by aspiration. Local anesthesia was not used. None of
the subjects received supplemental oxygen prior to or during the
test period.
Blood Gas Analysis
The blood was analyzed sequentially by two analyzers (ABL
555 and ABL 625; Radiometer) in a prerandomized order. All
samples were analyzed on both machines within 10 min. The
average Pao2 and Paco2 of the two analyses on each sample are
reported. The arterial oxygen saturation (Sao2) was measured by
photometric method and only on the Radiometer ABL 625. The
alveolar-arterial oxygen pressure gradient (P[A-a]O2) was calculated using the following equation:
P(A ⫺ a)O2 ⫽ Fio2共bP ⫺ 47.18)
⫺ Paco2共Fio2 ⫹ 共1 ⫺ Fio2)/RQ兲 ⫺ Pao2
where Fio2 (fraction of inspired oxygen) ⫽ 0.2095, bP ⫽ measured barometric pressure, and RQ (respiratory quotient) ⫽ 0.83.
Quality Control
Calibration of the two blood gas analyzers was checked/
performed several times daily, automatically, and on a predetermined schedule. Quality control using tonometered bovine hemoglobin-based material (Equil Plus G/L; RNA Medical; Ft.
Devens, MA)16,17 at one or more of three different levels for Po2
and Pco2 was performed on each day of testing in connection
with this study. Quality control results are presented in Table 1.
Both of the analyzers systematically give Po2 values over target
for the tonometered blood with a mean over estimation of 2.3
mm Hg in the approximate pressure level applicable in this study.
Quality control of So2 measurements of Radiometer ABL 625
was performed daily using standardized aqueous solutions with
germicide added (Radiometer, Bergmann Diagnostic; Copenhagen, Denmark) at two of four different levels of So2.
After the blood was drawn, spirometry was performed using a
Vitalograph wedge bellows spirometer. All measurements were
performed and reported according to American Thoracic Society
recommendations.18 Normal values for spirometry are based on
our own never-smoking subjects.19
Statistical Analysis
Differences between the subject group and the nonresponders
were tested using Pearson ␹2 or adjusted ␹2 for categorical
Table 1—Quality Control Data for the Two Radiometer Blood Gas Analyzers Used in the Study
Po2, mm Hg
Pco2, mm Hg
Variables
Level I
Level II
Level III
Level I
Level II
Level III
Target
ABL 555 mean
Range
ABL 625 mean
Range
Mean deviation from target for both analyzers
Range
Deviation from target (mean deviation/target), %
100.7
103.2
100.8–110.1
104.3
100.8–106.2
3.0
0–9.3
2.98
69.9
71.8
70.7–73.5
72.5
70.9–74.8
2.3
0.83–5.0
3.29
40.8
41.9
41.2–43.5
42.3
41.1–43.5
1.3
0.30–2.7
3.19
70.8
69.1
66.8–70.4
69.4
67.8–70.9
⫺ 0.81
⫺ 3.3–0.83
1.1
41.0
40.3
39.5–41.1
41.1
40.5–43.5
⫺ 0.25
⫺ 1.5–2.6
0.61
19.6
19.4
19.0–19.8
19.4
18.9–19.9
⫺ 0.16
⫺ 0.60–0.38
0.82
2054
Downloaded From: http://journal.publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22010/ on 06/14/2017
Clinical Investigations
Table 2—Descriptive Data on the Reference Sample
Variables
Total
Male
(n ⫽ 146) (n ⫽ 79)
Female
(n ⫽ 67)
Neversmoker Ex-smoker
(n ⫽ 90) (n ⫽ 56)
Age, %
70–74 yr
27
25
28
27
27
75–79 yr
27
30
24
29
25
80–84 yr
21
24
18
23
18
85–89 yr
10
8
12
9
11
ⱖ 90 yr
15
13
18
12
20
BMI, mean 25.7 (3.8) 25.2 (3.2) 26.2 (4.3) 26.0 (3.7) 25.2 (3.8)
(SD)
variables, and independent-samples t test for continuous variables. We screened for predictors of blood gas values using
independent-samples t test for categorical predictors and the
Pearson correlation for continuous predictor variables. Multiple
linear regression was performed for the analysis of association
between dependent and predictor variables. All variables were
entered in a single step then removed in backward fashion based
on the probability of f statistic. Criterion for removing was
p ⬎ 0.05. Statistical analyses were performed with SPSS 9.0
(SPSS; Chicago, IL). Outlier analysis was performed graphically
using a scatter plot. Variable interaction terms were created as
the arithmetic product of two variables. These interaction variables were created for all interaction between the variables sex,
age, smoking history, Paco2, FEV1 as percentage of predicted,
FEV1/FVC ratio, and body mass index (BMI). Upper and lower
limits of normal (ie, 95th and fifth percentiles) were calculated as
mean ⫾ 1.65 ⫻ SD. The regional committee on medical ethics
approved the study.
Results
Of the 319 persons selected for participation in the
study, there were 146 persons in whom we were able
to obtain an arterial blood sample. The demographic
characteristics of the 146 subjects are presented in
Table 2. Our goal was to have equal number representation for all of the age groups and for each sex.
We fell short of this, ending up with a moderate
overrepresentation of the younger ages and male
subjects. Among the subjects, there were slightly
more ex-smokers than desired (one third).
Subjects and nonresponders are compared with
regards to demographic data and questionnaire response in Table 3. The 173 nonresponders were
significantly older, more frequently female, and were
approximately three times as likely to have a walking
disability and/or be living in a nursing home. Nonresponders were more likely to report having heart
disease and/or having had a stroke.
Blood gas and spirometry values by sex, smoking
history, and the presence of various comorbidities
are presented in Table 4. Mean Pao2 and Sao2 were
significantly higher—and mean P(A-a)O2 lower—in
male than in female subjects. Percentage of predicted FEV1 and FEV1/FVC ratio were significantly
higher in female subjects. The blood gas values were
virtually identical in the never-smoker and exsmoker groups. Ex-smokers tended to have lower
percentage of predicted FEV1 and FEV1/FVC ratio
than never-smokers, although this was not statistically significant. Blood gas values in each of the
comorbidity subgroups were close to the value in the
“none of the above” group, and none of the differences were statistically significant. With regards to
spirometry, the comorbidity subgroups tended to
have lower percentage of predicted FEV1, although
this was only significant for hypertension vs “none of
Table 3—Comparison of Sample Subjects to Nonresponders*
Variables
Sample Subjects (n ⫽ 146)
Nonresponders (n ⫽ 173)
p Value, ␹2
Age
70–74 yr
75–79 yr
80–84 yr
85–89 yr
ⱖ 90 yr
Male sex, %
Ex-smokers
Walking disability
Living in a nursing home
BMI, mean (SD)†
Heart disease
Hypertension
Using cardiovascular medications
Stroke
Muscle/joint disease
26.7
27.4
21.2
9.6
15.1
54.1
38.4
12.8
6.8
24.5 (3.35)
17.1
32.9
35.6
6.8
26.7
8.7
12.1
11.0
22.5
45.7
41.0
32.5
42.3
18.5
23.9 (3.7)
30.2
35.0
35.3
20.1
38.0
⬍ 0.001
0.024
0.29
⬍ 0.001
0.002
0.17‡
0.01
0.72
1.0
0.001
0.056
*Data are presented as % unless otherwise indicated.
†Body mass index is based on self-reported height and weight on the questionnaire.
‡Significance of BMI differences with t test.
www.chestjournal.org
CHEST / 125 / 6 / JUNE, 2004
Downloaded From: http://journal.publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22010/ on 06/14/2017
2055
Table 4 —Blood Gas and Spirometry Results by Sex and Smoking History, and as Affected by the Presence of
Various Comorbidities Compared to No Comorbidities*
Variables
No.
Pao2,
mm Hg
Paco2,
mm Hg
P(A-a)O2,
mm Hg
Sao2, %
No.
FEV1 %
predicted, %
FEV1/
FVC, %
Men
Women
Ex-smokers
Never-smokers
No comorbidities
Heart disease
Hypertension
Using cardiovascular medications
Stroke
Muscle/joint
79
67
56
90
50
24
46
52
9
35
77.0 (9.1)‡
73.5 (8.4)
75.3 (8.4)
75.5 (9.3)
74.5 (8.8)
75.5 (9.1)
75.7 (9.1)
75.6 (7.7)
74.9 (6.2)
73.2 (8.4)
39.0 (3.0)
39.8 (3.6)
39.1 (3.3)
39.6 (3.3)
39.2 (2.8)
38.8 (3.6)
39.6 (3.4)
39.3 (3.6)
39.3 (1.9)
39.9 (3.9)
25.2 (9.2)‡
28.1 (8.0)
27.1 (7.9)
26.2 (9.2)
27.7 (8.8)
27.6 (8.2)
25.8 (9.5)
26.2 (7.7)
27.8 (4.9)
28.0 (7.8)
95.3 (1.4)‡)
94.8 (1.7)
95.0 (1.7)
95.2 (1.5)
95.0 (1.5)
95.2 (1.5)
95.1 (1.5)
95.1 (1.3)
95.2 (0.8)
94.7 (1.9)
66
49
47
68
44
16
37
36
6
27
94.3 (19.5)‡
103.4 (21.1)
94.8 (20.9)
100.5 (20.3)
100.4 (18.7)
94.9 (18.3)
91.7 (19.6)†
94.5 (19.2)
95.0 (39.4)
102.8 (19.2)
70.3 (6.4)‡
72.9 (6.8)
71.0 (7.0)
71.6 (6.5)
70.9 (6.1)
71.4 (6.8)
70.7 (7.1)
71.5 (6.1)
69.3 (6.4)
73.6 (6.9)
*Data are presented as mean (SD) unless otherwise indicated. Number of measurements is different for blood gases and spirometry reflecting
certain participants inability to perform spirometry.
†Significant at p ⬍ 0.05 by t test vs no comorbidities group.
‡Significant difference at p ⬍ 0.05 by t test vs women.
the above.” Values for FEV1/FVC ratio were similar
in all subgroups. There were no significant associations between age and any of the blood gas variables.
The scatter of Pao2 and Paco2 vs age is presented
in Figure 1. The lack of significant association to age
is consistent in all sex and smoking history subgroups. Pao2 and Sao2 showed a highly significant
inverse association to Paco2, rendering a Pearson
correlation (R) of – 0.23 (p ⫽ 0.005) and – 0.29
(p ⬍ 0.001), respectively. Pao2 showed a borderline
significant, inverse association to BMI (R ⫽ ⫺0.16,
p ⫽ 0.05). The association of BMI to the blood gas
variables varied greatly by sex with the inverse
association of Pao2 and Sao2 to BMI being significant only among male subjects.
There was a significant, inverse association between Pao2 and Sao2 by FEV1/FVC ratio
(R⫽ ⫺ 0.19, p ⫽ 0.04; and R ⫽ ⫺0.20, p ⫽ 0.03,
respectively) and positive association between
P(A-a)O2 and FEV1/FVC ratio (R ⫽ 0.20, p ⫽ 0.03).
Analysis by smoking subgroups showed nearly identical regression lines for never-smoker and exsmoker subgroups. However, the association of
FEV1/FVC ratio to the blood gas variables varied
greatly by sex, being significant only among female
subjects. Graphically evaluated, there were no obvious outliers driving the regressions. There were no
significant associations between percentage of predicted FEV1, percentage of predicted FVC, FEV1,
or FVC and any of the blood gas variables.
Sex, age, smoking history, Paco2, percentage of
predicted FEV1, FEV1/FVC ratio, and BMI along
with all interactions thereof were entered in multiple
regression models for prediction of Pao2 and Sao2.
The same variables, minus Paco2—a factor in the
calculation of P(A-a)O2—were entered in to a model
for prediction of P(A-a)O2. The results of each
regression after solving for sex are presented in
Table 5. Paco2 remained in the model as a significant predictor of Pao2 and Sao2 with no significant
differences between the sexes. Among female subjects, there was a highly significant inverse association between FEV1/FVC ratio and Pao2 and Sao2,
and a significant positive association to P(A-a)O2.
Among male subjects, there was a highly significant
inverse association between BMI and Pao2 and Sao2,
and a significant positive association to P(A-a)O2.
Age and smoking history remained nonsignificant
also in the multivariate analysis.
Practically applicable reference values for Pao2,
P(A-a)O2, Sao2, and Paco2 are presented in Table 6.
These reference values are presented only specific to
sex, as it would not be practical in daily practice to
adjust blood gas values for FEV1/FVC ratio, BMI,
and Paco2. In this age group, ⱖ 70 years, Pao2
values as low as 62 mm Hg in men and approximately
60 mm Hg in women can be considered as normal.
Sao2 values as low as 93% in men and 92% in women
may be normal. Correspondingly, upper limits of
normal for Paco2 would be approximately 45 mm
Hg for both sexes, and for P(A-a)O2 approximately
40 mm Hg for men and 41 mm Hg for women.
Discussion
We present reference values for Pao2, Paco2,
P(A-a)O2, and Sao2. All except Paco2 are presented
as sex specific. Pao2 and Sao2 were significantly
lower, and P(A-a)O2 significantly higher, among
women. The blood gas values were virtually unaffected by smoking history (never-smoker vs ex-
2056
Downloaded From: http://journal.publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22010/ on 06/14/2017
Clinical Investigations
Figure 1. Arterial oxygen and carbon dioxide pressures by age and sex.
smoker) or by having any of the designated comorbidities. The blood gas values showed no association
to age. There was a significant inverse association
between Pao2 and FEV1/FVC ratio that was driven
by the female portion of the sample (nonsignificant
in the male portion alone). In the male portion of the
sample, there was a significant inverse association
between Pao2 and BMI.
The primary goal of this study was to establish
reference values for arterial blood gases based on a
general population sample that was appropriately
representative of the healthy elderly. The mean and
lower limits of normal for Pao2 in our study were
lower than previous studies. Cerveri et al3 gave 83.4
mm Hg as mean Pao2 and 68.4 mm Hg as lower limit
of normal for the same age group. Guenard and
Table 5—Multiple Regression Analysis: Predictors of PaO2, P(A-a)O2, and SaO2*
Variables
Pao2, mm Hg
Male
Female
P(A-a)O2, mm Hg
Male
Female
Sao2, %
Male
Female
BMI
Adjusted R2
0.032 (0.16)
⫺ 0.45 (0.18)†
⫺ 1.08 (0.31)†
0.24 (0.29)†
0.15
0.09 (0.16)
0.46 (0.17)†
1.21 (0.31)†
⫺ 0.21 (0.28)
0.16
0.004 (0.03)
⫺ 0.088 (0.03)†
⫺ 0.14 (0.05)†
0.051 (0.05)
0.15
Constant
Paco2, mm Hg
FEV1/FVC, %
123.5
122.2
⫺ 0.56 (0.26)†
⫺ 0.56 (0.26)†
0.89
⫺ 1.14
102.9
101.5
⫺ 0.11 (0.04)†
⫺ 0.11 (0.04)†
*Data are presented as coefficient (SE) unless otherwise indicated. The coefficients are based on a single model for each dependent variable
subsequently solving the equation for each sex. There were significant interactions between sex and FEV1/FVC as well as sex and BMI, although
these terms are accounted for when solving for sex. Other variables that were entered in the model but did not reach statistical significance were
age, percentage of predicted FEV1, smoking history, and all interactions thereof. None of the entered variables were significant predictors of
Paco2. Stated coefficients are from regression with only significant variables in the model.
†p ⬍ 0.05.
www.chestjournal.org
CHEST / 125 / 6 / JUNE, 2004
Downloaded From: http://journal.publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22010/ on 06/14/2017
2057
Table 6 —Reference Values for Arterial Blood Gases*
Men
Variables
Mean
(SD)
Lower Limit
of Normal
Pao2, mm Hg
P(A-a)O2, mm Hg
Sao2, %
Paco2, mm Hg
77.0 (9.1)
25.2 (9.2)
95.3 (1.4)
39.0 (3.0)
62.0
Women
Upper Limit
of Normal
40.4
93.0
44.0
Mean
(SD)
Lower Limit
of Normal
73.5 (8.4)
28.1 (8.0)
94.8 (1.7)
39.8 (3.6)
59.6
Upper Limit
of Normal
41.3
92.0
45.7
*Lower limit of normal (estimated fifth percentile) ⫽ mean ⫺ 1.65 ⫻ SD; upper limit of normal (estimated 95th percentile) ⫽
mean ⫹ 1.65 ⫻ SD.
Marthan4 gave 84.3 mm Hg as mean, and based on
published SD would have a lower limit of normal of
71.7 mm Hg (mean ⫺ 1.65 ⫻ SD). There may be
several reasons for these different means and lower
limits of normal. Likely most significant is the use of
different blood gas analyzers.14,20 Cerveri et al3
reported having used a Radiometer ABL30, and
Guenard and Marthan4 used an Instruments Laboratory IL613 (Instruments Laboratory; Milan, Italy).
It is unclear what form of quality control was used,
although neither reported using tonometered blood,
which provides the best basis for intercenter comparison.16,17,21 A study by Hansen and Casaburi14
showed that different blood gas analyzers may differ
by up to 6.8 mm Hg in the measurement of 78.6 mm
Hg of oxygen standardized gas solution. In that
study,14 there were both systematic differences between different manufactures and between different
models from the same manufacturer. All of the
centers in the study by Hansen and Casaburi14 were
members of a national blood gas testing proficiency
program and were using tonometered blood as quality control routinely.
The position of the subject during arterial blood
sampling, whether sitting or lying down, is also an
important difference between our study and the two
previous studies. Both Cerveri et al3 and Guenard
and Marthan4 performed the blood sampling with
the subject sitting, while in our study the subjects
had been supine for at least 10 min. A previous study
by the authors22 has shown in a sample of 46 healthy
elderly a mean difference of sitting minus supine
Pao2 of approximately 6.0 mm Hg (p ⬍ 0.05). Knowing this, we chose the supine position for the present
study because most of our work with elderly patients
is when they are lying supine in a hospital bed.
Finally, there were differences in selection procedures between our study and the previous studies
that may also have influenced the results. Cerveri et
al3 recruited subjects from persons visiting the clinic
for check-ups or for minor surgery. Potential subjects were screened and excluded based on a set of
extensive criteria: symptoms or diagnosis of pulmo-
nary, cardiovascular, metabolic, renal, hepatic, or
CNS disease; smoking history; blood screens; ECG;
chest radiography; and occupational exposure.3 Of
the 1,283 subjects screened, 194 subject were included in the reference sample, 74 of which were
ⱖ 70 years old. Guenard and Marthan4 recruited
elderly from a retirement home after having them
screened for history of pulmonary or cardiovascular
disease and ECG or radiographic abnormalities.
Each of these sets of selection criteria could possibly
cause a selection bias toward “super-healthy” elderly.
The exclusionary criteria in our study were specifically set narrow to exclude only those with symptoms
or conditions known to affect Pao2. To the degree it
was possible based on questionnaire data, we have
attempted to analyze the effect of having nonpulmonary comorbidities on the blood gases. We found no
sign of any such effect in either direction. In light of
this, it would seem that differences in inclusionexclusion criteria have little to do with the different
levels of Pao2 found between the present study and
previous studies.
In previous studies5– 8 on Pao2 in younger populations, there has been a significant inverse association to age. The majority of these studies have had
only a few subjects ⬎ 70 years old. Cerveri et al3
and Guenard and Marthan4 each studied ample
numbers of elderly subjects, and both have found
no association to age ⬎ 74 years and ⬎ 70 years,
respectively.3,4 Having also studied subjects aged
from 40 to 74 years, Cerveri et al3 was able to
demonstrate that Pao2 starts to plateau between the
ages of 70 years and 74 years. These previous studies
are in agreement with our own, in which the regression of Pao2 by age is virtually flat. The reason for
this lack of age association may be an unavoidable
selection bias. The two likely components to this bias
are survival bias and a feeble nonresponder bias.
Neither of these forms of bias is easy to resolve.
Arterial blood samples have a short duration of
freshness that makes testing any distance from the
laboratory nearly impossible, and bringing the weak
and feeble in for testing will likely be equally diffi-
2058
Downloaded From: http://journal.publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22010/ on 06/14/2017
Clinical Investigations
cult. As for the other component, survivor bias, it is
well known that patients with COPD and resting
Pao2 levels ⬍ 55 to 60 mm Hg have a higher 5-year
mortality than those receiving long-term oxygen
therapy.23,24 Whether such low levels of Pao2 also
imply a higher 5-year mortality for otherwise healthy
elderly is unknown, though it seems likely. Obviously, little can be done in the general population to
avoid this bias. Both of these factors make the
cross-sectional study design poorly suited to answering the question of age vs Pao2 in the individual.
In our subjects, BMI is inversely associated with
age (R ⫽ ⫺ 0.21, p ⫽ 0.01) and at the same time
inversely associated with Pao2 (R ⫽ ⫺ 0.16,
p ⫽ 0.05). Thus, one might expect the age-related
fall in BMI to explain the lack of association between
age and Pao2. However, in the multiple, linear
regression analysis, addition of BMI to the regression
of Pao2 vs age had no discernable effect on the
coefficient for age, thus failing to support this possible explanation in our data.
Another possible explanation for the lack of age
dependence would be a physiologic compensation
that stops the falling Pao2 before reaching detrimentally low levels. An evaluation of this possible mechanism would require a longitudinal study design,
though no such study has been found in the literature. However, even though on an individual basis
Pao2 may physiologically continue to fall with increasing age, there might exist a lower limit for Pao2
below which hypoxia becomes too unhealthy and
morbidity increases sharply. Values below these
lower limits of Pao2, however physiologic they may
be, probably should not be considered as reference
values. Thus, for the purpose of establishing reference values, the cross-sectional design may still be a
valid approach.
As we did not screen for occult respiratory illness
by chest radiograph or CT scan, we have no guarantee that some of the subjects did not have occult,
asymptomatic illness. However, it seems to us unlikely that there is any systematic bias toward illness.
On the contrary, as approximately 50% of the sample
selected for blood gas testing failed to meet at the
clinic, there may be a healthy-responder bias in the
sample. There seems little reason to believe that
the nonresponders should have higher Pao2 levels as
they reported significantly more comorbidities than
the participants (Table 3).
The fact that a large majority of elderly men (along
with a significant minority of women) have smoked
at some time during their lives makes choosing a
reference population difficult.25 Traditionally, smokers have been avoided when establishing reference
values for various pulmonary function tests because
of the well-known effects of smoking on pulmonary
www.chestjournal.org
function.1 Still, it has not been previously established
whether it is necessary to exclude healthy ex-smokers
from reference samples with regards to arterial blood
gases. Thus, we chose to include these persons in the
sample. Our results show that the ex-smokers are
indistinguishable from the never-smokers with regards to blood gases. Smoking history failed to have
significance both in univariate and multivariate analyses. This finding needs to be seen in the light of
having already excluded persons with acute or
chronic pulmonary disease, severe dyspnea, or cardiovascular disease with moderate dyspnea. By this
exclusion, most of those ex-smokers having had
detrimental smoking effects were excluded. From
this we conclude that it really is not necessary to
exclude ex-smokers from the reference sample as
long as persons with pulmonary or symptomatic
cardiovascular disease are excluded first.
The finding that FEV1/FVC ratio was inversely
associated to Pao2 and Sao2 and positively associated
to P(A-a)O2 seems to be counterintuitive. This association was found both in never-smokers and exsmokers, although by sex it holds true only for
women. The mechanism behind this association is
not clear, though the consistency in the smoking
subgroups and for each of the oxygen variables
suggests that the finding is not spurious. This is
reinforced by the findings in a previous study by the
authors,22 in which Pao2 in the supine position (but
not in the sitting position) was inversely associated to
FEV1/FVC ratio. The fact that the association is
present only among women is also interesting in that
it may point to some fundamental differences in the
way men and women age. It is possible that there is
some degree of occult, asymptomatic restrictive pulmonary function deficit in these subjects resulting in
a tendency toward hypoxia, although in the absence
of a measure for total lung capacity this will remain
speculation. Otherwise, the authors have no immediate explanation for this finding and believe it may
warrant further investigation.
Conclusions
We have presented reference values for arterial
blood gases in supine elderly persons. The reference
values for Pao2 and Sao2 in elderly persons are sex
specific but age independent. Due to systematic
differences in Po2 values rendered by different
blood gas analyzers, the clinical application of these
reference values may need to be validated for each
laboratory. Ideally, reference values should be established at each laboratory. In selection of the reference sample for the elderly, it is not necessary to
CHEST / 125 / 6 / JUNE, 2004
Downloaded From: http://journal.publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22010/ on 06/14/2017
2059
exclude ex-smokers, nor persons with heart disease,
hypertension, or stroke as long as they do not report
respiratory symptoms.
ACKNOWLEDGMENT: We thank Eli Nordeide and Randi
Espelid for technical assistance in this study.
References
1 American Thoracic Society. Lung function testing: selection
of reference values and interpretative strategies. Am Rev
Respir Dis 1991; 144:1202–1218
2 Cavalieri TA, Chopra A, Bryman PN. When outside the norm
is normal: interpreting lab data in the aged. Geriatrics 1992;
47:66 –70
3 Cerveri I, Zoia MC, Fanfulla F, et al. Reference values of
arterial oxygen tension in the middle-aged and elderly. Am J
Respir Crit Care Med 1995; 152:934 –941
4 Guenard H, Marthan R. Pulmonary gas exchange in elderly
subjects. Eur Respir J 1996; 9:2573–2577
5 Crapo RO, Jensen RL, Hegewald M, et al. Arterial blood gas
reference values for sea level and an altitude of 1,400 meters.
Am J Respir Crit Care Med 1999; 160(5 pt 1):1525–1531
6 Sorbini CA, Grassi V, Solinas E, et al. Arterial oxygen tension
in relation to age in healthy subjects. Respiration 1968;
25:3–13
7 Mellemgaard K. The alveolar-arterial oxygen difference: its
size and components in normal man. Acta Physiol Scand
1966; 67:10 –20
8 Gunnarsson L, Tokics L, Brismar B, et al. Influence of age on
circulation and arterial blood gases in man. Acta Anaesthesiol
Scand 1996; 40:237–243
9 Yamasawa F, Kawashiro T, Yokoyama T, et al. Standard
values and normal limits for arterial blood gases in healthy
elderly Japanese subjects [in Japanese]. Nippon Kyobu Shikkan Gakkai Zasshi 1992; 30:430 – 434
10 Blom H, Mulder M, Verweij W. Arterial oxygen tension and
saturation in hospital patients: effect of age and activity. BMJ
1988; 297:720 –721
11 Cardus J, Burgos F, Diaz OR, et al. Increase in pulmonary
ventilation-perfusion inequality with age in healthy individuals. Am J Respir Crit Care Med 1997; 156(2 pt 1):648 – 653
12 Raine JM, Bishop JM. A-a difference in O2 tension and
physiological deadspace in normal man. J Appl Physiol 1963;
18:284 –288
13 Malmberg P, Hedenstrom H, Fridriksson HV. Reference
values for gas exchange during exercise in healthy nonsmoking and smoking men. Bull Eur Physiopathol Respir 1987;
23:131–138
14 Hansen JE, Casaburi R. Patterns of dissimilarities among
instrument models in measuring Po2, Pco2, and pH in blood
gas laboratories. Chest 1998; 113:780 –787
15 Ferris BG. Epidemiology Standardization Project (American
Thoracic Society). Am Rev Respir Dis 1978; 118(6 pt 2):1–
120
16 Larsson L, Sandhagen B, Kallner A. Quality assurance of
blood gas analysis: a medical risk zone [in Swedish]. Lakartidningen 1999; 96:2368 –2370,2373
17 Mahoney JJ, Wong RJ, Van Kessel AL. Reduced bovine
hemoglobin solution evaluated for use as a blood gas qualitycontrol material. Clin Chem 1993; 39:874 – 879
18 American Thoracic Society. Standardization of spirometry,
1994 update. Am. J Respir Crit Care Med 1995; 152:1107–
1136
19 Hardie JA, Buist AS, Vollmer WM, et al. Risk of overdiagnosis of COPD in asymptomatic, elderly never-smokers.
Eur Respir J 2002; 20:1117–1122
20 Burki NK. Arterial blood gas measurement [editorial]. Chest
1985; 88:3– 4
21 Burnett RW, Covington AK, Maas AH, et al. IFCC document
stage 3, draft 1, dated 1989 02 01. An approved IFCC
recommendation. IFCC method (1988) for tonometry of
blood: reference materials for pCO2 and pO2. International
Federation of Clinical Chemistry Scientific Division, Committee on pH, Blood Gases and Electrolytes. Clin Chim Acta
1989; 185:S17–S24
22 Hardie JA, Morkve O, Ellingsen I. Effect of body position on
arterial oxygen tension in the elderly. Respiration 2002;
69:123–128
23 Medical Research Council of Great Britain. Long term
domiciliary oxygen therapy in chronic hypoxic cor pulmonale
complicating chronic bronchitis and emphysema: Report of
the Medical Research Council Working Party. Lancet 1981;
1:681– 686
24 Continuous or nocturnal oxygen therapy in hypoxemic
chronic obstructive lung disease: a clinical trial; Nocturnal
Oxygen Therapy Trial Group. Ann Intern Med 1980; 93:391–
398
25 Ronneberg A, Lund KE, Hafstad A. Lifetime smoking habits
among Norwegian men and women born between 1890 and
1974. Int J Epidemiol 1994; 23:267–276
2060
Downloaded From: http://journal.publications.chestnet.org/pdfaccess.ashx?url=/data/journals/chest/22010/ on 06/14/2017
Clinical Investigations