ANNALS
of Internal Medicine
SEPTEMBER 1980 • VOLUME 93 • NUMBER 3
Published Monthly by the American College of Physicians
Continuous or Nocturnal Oxygen Therapy in Hypoxemic Chronic Obstructive
Lung Disease
A Clinical Trial
NOCTURNAL OXYGEN THERAPY TRIAL GROUP*
At six centers, 2 0 3 patients with hypoxemic chronic
obstructive lung disease were randomly allocated to
either continuous oxygen ( 0 2 ) therapy or 12-hour
nocturnal 0 2 therapy and followed for at least 12 months
(mean, 19.3 months). The two groups were initially well
matched in terms of physiological and neuropsychological
function. Compliance with each oxygen regimen was good.
Overall mortality in the nocturnal 0 2 therapy group was
1.94 times that in the continuous 0 2 therapy group
(P = 0.01). This trend was striking in patients with carbon
dioxide retention and also present in patients with
relatively poor lung function, low mean nocturnal oxygen
saturation, more severe brain dysfunction, and prominent
mood disturbances. Continuous 0 2 therapy also appeared
to benefit patients with low mean pulmonary artery
pressure and pulmonary vascular resistance and those
with relatively well-preserved exercise capacity. We
conclude that in hypoxemic chronic obstructive lung
disease, continuous 0 2 therapy is associated with a lower
mortality than is nocturnal 0 2 therapy. The reason for this
difference is not clear.
PATIENTS WITH chronic obstructive lung disease and
hypoxemia have a poor prognosis in spite of treatment
regimens aimed at improving the mechanical function of
the lungs (1). Because of this, such patients are often
treated with supplementary oxygen on an outpatient ba-
sis. Early studies of this treatment, which compared pa-
tients before and after oxygen therapy, indicated that
chronic 0 2 therapy resulted in improved exercise toler-
ance, decreased pulmonary hypertension and erythrocy-
tosis, and improved neuropsychological function (2-6). A
trial in which oxygen was given to one group of patients
but withheld from a similar control group has not been
done in North America, but such a trial is underway in
the United Kingdom, and a preliminary report indicates
•For acknowledgment of contract support, a list of participants in the Nocturnal
Oxygen Therapy Trial, and lists of other supporting personnel, see Acknowledg-
ments at the end of this article.
Annals of Internal Medicine. 1 9 8 0 ; 9 3 : 3 9 1 - 3 9 8 .
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/27/2016
that oxygen administration was associated with a reduced
mortality (7).
Long-term oxygen administration is an expensive form
of treatment, particularly when ambulatory patients are
supplied with portable units. It is not clear whether con-
tinuous 0 2 therapy is necessary; patients suffer their most
severe hypoxemia while sleeping (8), and it is possible
that hypoxemic sequelae such as erythrocytosis and pul-
monary hypertension could be prevented by exclusively
nocturnal oxygen administration. Indeed, there is some
evidence that pulmonary hypertension can be reduced by
as little as 15 hours of oxygen administration per day (9,
10).
On the basis of this rationale, the Division of Lung
Diseases of the National Heart, Lung, and Blood Insti-
tute initiated a multicenter clinical trial comparing con-
tinuous 0 2 therapy with nocturnal 0 2 therapy in patients
with hypoxemic chronic obstructive lung disease (11).
Methods
The protocol for the trial has been published elsewhere in
detail (12) and therefore will be presented briefly here. Six treat-
ment centers recruited 203 patients over 27 months and fol-
lowed each surviving patient at least 1 year. The number of
patients at each center varied from 26 to 44. Entry and exclu-
sion criteria are shown in Table 1. The most important entry
criterion was, of course, hypoxemia. The protocol required that
this criterion be fulfilled on at least two occasions more than 1
week apart during a 3-week observation period while the sub-
ject was free of exacerbations and was managed without supple-
mental oxygen and with intensive bronchodilator therapy. In
practice, patients were initially identified as fitting the entry but
not the exclusion criteria, recruited in a preliminary way, and
observed for 3 weeks to ensure stability. At the end of this time,
if the patients still met these criteria, informed consent was
obtained and the patient was hospitalized for a week of baseline
studies. At the end of these studies, each patient was randomly
allocated by the Data Center to either continuous Oz or noctur-
nal 0 2 therapy. Randomization schedules were developed sepa-
391
 Table 1 . Entry and Exclusion Criteria
Entry criteria
Clinical diagnosis of chronic obstructive lung disease
Hypoxemia
Pa0., ^ 55 mm Hg
Pao2 ^ 59 plus one of the following:
Edema
Hematocrit ^ 55%
P pulmonale on ECG: 3 mm in leads II, III, aVf
Lung function*
FEV,/FVC < 707c after inhaled bronchodilator
TLC ^ 80% predicted
Age > 35
Exclusion criteria
Previous 0 2 therapy: 12 h/d for 30 days during previous 2 months
Other disease that might be expected to influence mortality, mor-
bidity, compliance with therapy, or ability to give informed con-
sent
* FEVi = forced expiratory volume in I second; FVC = forced vital capacity;
TLC = total lung capacity.
rately for each investigative center. Treatment assignments
were preset in blocks of four with an equal number of patients
receiving nocturnal Oz therapy and continuous 0 2 therapy in
each block. The order of treatment assignment was randomly
computer-generated within each block of four.
Baseline studies included a complete history and physical ex-
amination, blood count, urinalysis, measurements of blood urea
nitrogen and electrolytes, ECG, and chest radiograph. Lung
function was assessed by measuring forced expiratory volume in
1 second (FEV,) and forced vital capacity (FVC) spirometrical-
ly and functional residual capacity (FRC) with a plethysmo-
graph. Arterial blood was obtained while the patient, breathing
room air, rested semirecumbent and was analyzed for Po 2 , Pco 2 ,
and pH with appropriate electrodes. Exercise performance was
studied with bicycle ergometry, using a progressive multi-stage
technique. Right heart catheterization was done with measure-
ments of pulmonary artery and wedge pressures and of cardiac
output. Oxygen saturation during sleep was measured with an
ear oximeter while the patient breathed air and 0 2 on separate
nights. Neuropsychological function was assessed by study of
neuropsychological history and administration of a modified
Halstead-Reitan battery (13) of tests. These tests were evaluated
collectively, in terms of the Russell-Neuringer average impair-
ment index (14), and a clinical rating scale of 1 to 6 with 6
representing maximal impairment (15). This rating was made
by two neuropsychologists without knowledge of the treatment
assignment of the patient. The patient's quality of life was as-
sessed by the administration of the Minnesota Multiphasic Per-
sonality Inventory (16), the Sickness Impact Profile (17), and
the Profile of Mood States (18).
After 6 months of nocturnal or continuous 0 2 therapy, all
baseline tests except sleep studies were repeated. Thereafter, at
6-month intervals, all baseline studies were again repeated ex-
cept sleep studies, right heart catheterization, and neuropsycho-
logical testing.
After randomization, patients were instructed in the use of
their oxygen equipment and discharged for follow-up as outpa-
tients. Oxygen was administered by nasal prongs at a measured
flow rate of 1 to 4 L/min. Each patient received the lowest flow
in whole litres per minute that demonstrably increased resting,
semirecumbent arterial Po 2 at least 6 mm Hg and maintained a
resting arterial Po 2 of 60 to 80 mm Hg. This dose was increased
by 1 L/min for periods of exercise and sleep. Oxygen delivery
systems varied; oxygen concentrators, liquid oxygen systems,
and compressed gas were all used. Compliance with therapy
was checked in two ways. The patient and a family member
were required to keep written records of oxygen use, and oxy-
gen reservoirs or concentrators were fitted with timers that re-
corded the duration of gas flow. Because only stationary sys-
September 1980 • Annals of Internal Medicine • Volume 93 • Number 3
392
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/27/2016
tems had such timers, oxygen use was recorded accurately in
nocturnal 0 2 therapy patients but underestimated in some con-
tinuous 0 2 therapy patients who, in addition to a stationary
system, used a portable system such as liquid oxygen walkers or
small tanks of compressed gas.
Besides oxygen, all patients were treated with oral theophyl-
line and inhaled beta-2 agonists. Diuretics and antibiotics were
used as clinically indicated. Use of other drugs, such as steroids,
cardiac glycosides, sedatives, tranquilizers, antidepressants and
oral beta agonists, was discouraged. Therapeutic phlebotomies
were not done.
All patients were followed closely to assure compliance and
to assess changes in clinical state. For the first 6 months of the
study, they were visited weekly in the home by a nurse practi-
tioner, and were seen each month in outpatient clinic. After the
first 6 months, they were visited at home at least once a month
and were seen in the outpatient clinic at least every 3 months.
At 1-month intervals during the first 6 months and at 3-month
intervals thereafter, samples of arterial blood were obtained
while the patient breathed his prescribed dose of oxygen and the
dose adjusted if arterial P0 2 was not 60 to 80 mm Hg. Unsche-
duled clinic or emergency room visits were recorded. When a
patient became ill enough to be hospitalized, the assigned treat-
ment regimen was suspended but was resumed as soon as possi-
ble thereafter. Oxygen dose was reassessed after each hospitali-
zation. If a patient died, an attempt was made to obtain a post-
mortem examination.
All observations were made under specified conditions at pre-
determined time intervals and recorded on standardized data
forms to assure uniform collection of data by all participants.
Immediately after each recorded event, copies of the completed
data forms were mailed directly to a designated Data Center for
editing, analysis, and storage. All the data on the study forms
were subjected to an initial clerical review and were then key-
punched and verified. All keypunched information was subject-
ed to an extensive computer edit. Errors detected in this editing
process were sent to the clinics for correction. Measurements
based on analogue recordings, notably records of cardiac cathe-
terization and sleep studies, were verified by the submission of a
random sample amounting to 10% of the original records to
single readers. All neuropsychologic data were reviewed by ex-
perts who checked them for accuracy and consistency. An Ad-
visory Board including clinicians, epidemiologists, and other
experts periodically reviewed confidential interim data on the
progress of the trial. In particular, follow-up measurements and
other outcome criteria were examined so that the trial could be
terminated promptly if a clinically significant difference be-
tween treatments emerged. In fact, the scheduled end of the
trial coincided with the development of a difference in mortality
that would have necessitated termination of the trial.
Percentage of events in the nocturnal and the continuous 0 2
therapy groups as well as life tables calculated according to
Kaplan and Meier (19) are reported. Survivorship of all patients
as of 26 May 1980, regardless of the extent of their participa-
tion, was ascertained by individual follow-up observation.
For comparison of survival in the nocturnal O- therapy and
the continuous O therapy groups, the Cox actuarial procedure
(20) was applied. This statistical procedure, which is based on a
proportional hazard model, takes into account the ranking of
the times of follow-up and death in the two treatment groups.
Mortality data were monitored at regular intervals through-
out the trial to detect treatment differences as soon as possible.
Also, many other end points were evaluated in the final data
analysis. Because of this, if nocturnal 0 : and continuous 0 :
therapy regimens were alike in every respect, by chance alone
5% of the secondary outcome variables would show a difference
significant at the 5% level (21). It is clear that, because multiple
variables were evaluated and key variables analyzed during the
trial, one should be cautious in assessments of statistical signifi-
cance.
Results
T h e 203 patients w h o were recruited were selected
 from 1043 patients who were screened for the study. Of
these, 809 patients never entered the stabilization phase:
31% of them were found to have other major disease,
15% refused, 12% were discovered to have previously
received chronic 0 2 therapy, and in 21% the arterial Po2
rose so that the patient was no longer eligible. The re-
mainder of these 809 patients were ineligible for a variety
of reasons, such as living too far from a study center. In
all six centers, a total of only four patients fulfilled the
entry criteria but were judged by the investigators to be
"too sick" to undergo 3 weeks of observation off 0 2 .
Thirty-one patients entered stabilization but did not com-
plete it and were not randomized to the study. Failure of
completion was chiefly due to an absence of informed
consent or a rise in arterial Po2 during stabilization such
that the patient was ineligible.
Table 2 shows selected mean baseline characteristics of
both nocturnal 0 2 therapy and continuous 0 2 therapy
groups and indicates that they were comparable: The
randomization process was successful. The average pa-
tient age was more than 65 years; most patients were
male. They were hypoxemic with a slightly elevated he-
matocrit value and borderline C0 2 retention. Expiratory
flow was severely compromised, and they were hyperin-
flated. Maximum exercise performance, as tested by bicy-
cle ergometry, was sharply limited. The average patient
showed resting tachycardia, modest pulmonary hyperten-
sion, and increased pulmonary vascular resistance. Dur-
ing sleep, patients were hypoxemic while breathing air
but generally not while breathing their prescribed dose of
0 2 . Neuropsychological test results showed that the aver-
age patient had impaired brain function; patients on con-
tinuous 0 2 therapy appeared to be slightly less impaired
than those on nocturnal 0 2 , the difference approaching
significance when the overall clinical rating and the Rus-
sell-Neuringer average impairment index were consid-
ered. Quality of life measures indicated relatively low lev-
el of patient self-satisfaction, reduced physical and social
capabilities, and increased levels of depression, anxiety,
and hostility. More detailed analyses of these baseline
data will be presented in other reports.
Compliance was assessed both by timers and by exam-
ining patient logs. The former was objective but system-
atically underestimated 0 2 use by continuous 0 2 therapy
patients with portable systems. In nocturnal 0 2 therapy
patients, timer and log data were in excellent agreement,
both indicating that more than 82% of the nocturnal 0 2
therapy patients used 13 hours or less of 0 2 per day.
According to timers, nocturnal 0 2 therapy patients aver-
aged 12.0 h/d (SD = 2.5 h/d) whereas continuous 0 2
therapy patients averaged 17.7 h/d (SD = 4.8 h/d). Fig-
ure 1 shows data from patient logs and indicates that

Table 2 . Baseline Characteristics of Nocturnal Oa Therapy and Continuous Q2 Therapy Groups

Characteristics* Nocturnal 0 2 Continuous P Value

Therapy 0 2 Therapy

General and cardiopulmonary characteristics

Patients, no.] 102 101
Age, yrs 65.7 65.2 0.72
Male, %t 80.4 77.2 0.58
White, %t 78.4 77.2 0.84
Pao,, mm Hg 51.5 50.8 0.32
Pace, mm Hg 43.9 43.4 0.70
PH " 7.41 7.40 0.19
Hematocrit, % 47.3 47.7 0.60
FEVi, % predicted 29.9 29.5 0.82
FVC, % predicted 53.6 52.6 0.70
FRC, % predicted 177.6 175.7 0.78
Mean sleep Sao2, air, % 83.5 83.0 0.66
Mean sleep Sao«, 0 2 , % 94.0 94.1 0.83
Maximum workload, air, W 37.3 37.5 0.95
Heart rate, mirr1 92.6 93.1 0.83
Mean pulmonary artery pressure, mm Hg 29.0 30.0 0.58
Cardiac index, L/minm2 2.91 2.95 0.69
Pulmonary vascular resistance, dyne/scmb 330 333 0.91
Neuropsychiatry characteristics}:
Overall rating (3.5) 4.5 4.2 0.06
Halstead impairment index (0.63) 0.78 0.73 0.15
Russell-Neuringer average impairment index (1.8) 2.3 2.1 0.08
Brain age quotient (89) 75.4 80.4 0.11
Quality of lifej
MMPI, average scales 0.9 (54.5) 60.9 61.4 0.68
SIP
Physical scale (0.6) 20.5 19.8 0.78
Psychosocial scale (1.6) 23.9 20.5 0.22
POMS—mood disturbance (26.4) 48.4 49.7 0.76

* FEVi = forced expiratory volume in 1 second; FVC = forced vital capacity; FRC = functional residual capacity; MMPI = Minnesota Multiphasic Personality In-
ventory; SIP = Sickness Impact Profile; POMS •» Profile of Mood States,
t All values reported for the two groups are mean values except numbers of subjects, sex, and race.
X Normal values are shown in parentheses.

Trial Group • Oxygen Therapy for Obstructive Lung Disease 393

Downloaded From: http://annals.org/ by a University of California San Diego User on 12/27/2016


Figure 1 . Compliance with treatment regimens as assessed by pa-
tient log, expressed as a frequency distribution of daily 0 2 use.
Hatched bars represent patients on nocturnal O a therapy, open
bars, those on continuous 0 2 therapy. " M i s s i n g " indicates patients
in whom data are missing.
82% of nocturnal 0 2 therapy patients used 0 2 for 13
hours or less per day and that 56% of continuous 0 2
therapy patients used 0 2 for 19 or more hours per day.
The 203 patients were followed for an average of 19.3
months. Of the total group, 80 nocturnal 0 2 and 87 con-
tinuous 0 2 therapy patients were followed for 12 months
and 29 nocturnal 0 2 and 37 continuous 0 2 therapy pa-
tients were followed for 24 months. Two continuous 0 2
therapy patients were lost to follow-up 2 months before
the trial was ended, one after 12 months of treatment and
the other after 15 months. A total of 64 patients died, 41
in the nocturnal 0 2 therapy group and 23 in the continu-
ous 0 2 therapy group. Life table cumulative survival
rates, shown in Figure 2, indicate the pattern of mortality
throughout the study period. Based on the Cox model,
unadjusted for baseline characteristics, this difference in
survival was significant (P = 0.01). The 12-month mor-
tality was 20.6% (SE = 4.0%) in the nocturnal 0 2 thera-
py group and 11.9% (SE = 3.2%) in the continuous 0 2
therapy group, whereas the 24-month mortality was
40.8% (SE = 5.5%) and 22.4% (SE = 4.6%), respec-
tively. Overall mortality was 31.5% for all patients and
varied from 15.4% to 41.9% among centers. At all cen-
ters, mortality for the nocturnal 0 2 therapy group ex-
ceeded that for the continuous 0 2 therapy group. The
relative risk of death for the nocturnal 0 2 therapy group
compared with the continuous 0 2 therapy group was 1.94
with 95% confidence limits ranging from 1.17 to 3.24.
The above analyses were done according to treatment
assignment and did not consider compliance. However,
compliance was very good, and compliance failure could
not account for the difference between groups.
Mortality was examined in subgroups defined accord-
3 9 4 September 1980 • Annals of Internal Medicine • Volume 93 • Number 3
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/27/2016
ing to variables thought to be clinically important; the
median overall baseline values of these variables were
used to separate groups. Figure 3 shows mortality in pa-
tients with arterial Pco 2 equal to or greater than 43 mm
Hg, the median value for all patients at baseline; survival
was better in the continuous 0 2 therapy group, and this
difference was highly significant (P < 0.002). This and
other selected baseline values are related to mortality in
Table 3. In two other subgroups, those with low pH and
those who showed high levels of mood disturbance such
as depression and anxiety (Profile of Mood States test),
the continuous 0 2 therapy group demonstrated very sig-
nificantly greater survival than the nocturnal 0 2 therapy
group. Also, in a number of other subgroups mortality
differed (0.01 < P <0.05). Patients with low FVC, high
FRC, more severe nocturnal hypoxemia, low hematocrit
values, and more severe indexes of brain dysfunction all
tended to have lower mortality on continuous than on
nocturnal 0 2 therapy. The same was true of patients with
pulmonary artery pressure and pulmonary vascular re-
sistance values that were below the median and for those
whose work capacity on the cycle ergometer was above
the median.
Although patients on continuous 0 2 therapy tended to
be hospitalized less often and to have fewer long hospital-
izations than nocturnal 0 2 therapy patients, differences
were not statistically significant.
The effects of treatment on the physiological and psy-
chological variables listed in Table 2 were examined by
comparing baseline and follow-up data in individual pa-
tients. For some variables, such as those resulting from
cardiac catheterization, only 6-month follow-up data
were available. For other indexes, such as arterial Po 2
and Pco 2 , results at 6 months, 12 months, and 18 months
could be compared with baseline values. With the excep-
tion of hematocrit and pulmonary vascular resistance,
none of the variables listed in Table 2 showed statistically
significant changes that were dependent on treatment
regimen. Hematocrit values fell more in patients on con-
Figure 2 . Overall mortality. Ordinate is fraction of patients surviv-
ing; abscissa is time f r o m randomization or duration of treatment.
Open circles represent continuous 0 2 therapy group; squares rep-
resent nocturnal O a therapy group. Of the total group, 8 0 nocturnal
0 2 and 8 7 continuous 0 2 therapy patients were followed for 12
months, and 2 9 nocturnal 0 2 and 3 7 continuous 0 2 therapy pa-
tients were followed for 2 4 months.

Figure 3. Survival of patients with arterial Pco2 over 4 3 m m Hg.
Ordinate is fraction of patients surviving; abscissa is time f r o m
randomization or duration of treatment. Open circles represent
continuous O a therapy patients; squares represent nocturnal 0 2
therapy patients. Of these patients, 3 5 nocturnal O a therapy pa-
tients and 5 0 continuous 0 2 therapy patients were followed for 12
months, and 13 patients on nocturnal and 2 5 on continuous 0 2
therapy were followed for 2 4 months.
tinuous 0 2 therapy than in those on nocturnal 0 2 thera-
py: This was not significant at 6 months (P = 0.06) but
was significant at both 12 months (P = 0.005) and 18
months (P = 0.008). At 18 months hematocrit values
showed, on the average, a fall of 2.0% from baseline in 36
nocturnal 0 2 therapy patients and had decreased an aver-
age of 9.2% in 40 patients on continuous 0 2 therapy.
Pulmonary vascular resistance was measured in follow-
up only at 6 months; at that Hime 49 nocturnal 0 2 therapy
patients showed a mean increase of 6.5% while 52 con-
tinuous 0 2 therapy patients showed a mean decrease of
11.1% (P = 0.04).
When data for all patients were combined and baseline
and follow-up data compared, significant changes in he-
matocrit and pulmonary vascular resistance were ob-
served. Hematocrit values fell from a mean at baseline of
47.5% to a mean of 44.3% at 6 months. Pulmonary vas-
cular resistance fell from a mean of 322 dyne/s • cm5 at
baseline to a mean of 281 dyne/s • cm5 at 6 months. No
significant change occurred in arterial blood gas levels,
lung volumes or FEVj, maximum work attained, mean
pulmonary artery pressure, or cardiac index. With the
exception of the Minnesota Multiphasic Personality In-
ventory, all the measures of neuropsychological function
and quality of life listed in Table 2 improved significantly
when all patients were considered.
Discussion
The patients studied were probably representative of
patients in general with hypoxemic chronic obstructive
lung disease. The study centers attempted to recruit every
patient who fulfilled the entry criteria and presented no
reason for exclusion (Table 1). Over 1000 patients were
screened for the study; a large fraction of these were not
eligible for the study chiefly because of other disease and
failure to demonstrate persistent hypoxemia. Very few
patients did not enter the study because they were
thought "too sick" to endure 3 weeks' observation with-
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/27/2016
out oxygen. The only biases that operated in patient se-
lection were that the patient live near a study center, that
he or she have no serious extrapulmonary disease, and
that he or she not previously have been treated with long-
term oxygen therapy. Thus, except that they were largely
urban dwellers and did not have major disease of other
organ systems, these patients probably were representa-
tive of the type of person with obstructive disease cur-
rently thought to merit long-term low-flow oxygen thera-
py. As indicated in Table 2 the patients were randomized
successfully. At baseline, there was no significant differ-
ence between nocturnal 0 2 and continuous 0 2 therapy
groups in any of the variables listed, although in some
neuropsychological test results there were trends favoring
the continuous 0 2 therapy group. The effectiveness of the
prescribed dose of 0 2 in relieving hypoxemia was estab-
lished and repeatedly checked. Therapy other than that
involving 0 2 was the same in both groups and therefore
could not have biased the results. Finally, as indicated in
Figure 1, patients complied well with their prescribed
treatment regimens; few continuous 0 2 therapy patients
used 0 2 for 13 hours or less, and few nocturnal 0 2 thera-
py patients used it for 19 hours or more. This record of
compliance is unusually good (22) and probably occurred
both because the patients were followed very closely and
because they found it easy to believe that oxygen was
beneficial to them. In any event, this clinical trial ap-
peared to fulfill several important criteria: The patients
were representative of the general clinical population ad-
dressed by the trial, the treatment groups were similar at
the beginning of the trial, the treatment was shown to
achieve its short-term goal, and each group generally fol-
lowed its assigned treatment plan.
We believe that the trial gave a clear-cut answer in
terms of the variable of ultimate clinical importance:
mortality. Mortality in the nocturnal 0 2 therapy group
was nearly twice that in the continuous 0 2 therapy
group. The analysis of treatment-dependent differences in
overall mortality was the first statistic derived from the
trial data because it was considered the most important.
Therefore it cannot be argued that the difference in over-
all mortality was a chance result of many tests being ap-
plied to the data. The probability was truly 1% that the
difference in mortality between the nocturnal 0 2 and the
continuous 0 2 therapy groups was due to chance. Had
the trial been prolonged, an even more impressive statis-
tic might have emerged; this would have reflected an even
greater differential mortality that would have been ethi-
cally unacceptable. Had the trial been scheduled to last
longer, the Advisory Board would have terminated it at
the time that it did stop because of the mortality differ-
ence reported here.
In analyzing mortality, we analyzed all deaths. Non-
respiratory factors contributed to the deaths of some pa-
tients: Two committed suicide, one had leukemia, four
had carcinoma of the lung, and one had a cerebrovascu-
lar accident. These deaths were randomly distributed be-
tween the nocturnal 0 2 and the continuous 0 2 therapy
groups. When these patients were eliminated from con-
sideration, survival in the continuous 0 2 therapy group
Trial Group • Oxygen Therapy for Obstructive Lung Disease 395
 Table 3. Mortality According to Baseline Characteristics*

Characteristic! Number of Deaths P Value

Patients
Total Nocturnal 0 2 Continuous
Therapy 0 2 Therapy

no. < % >

Pa 0r mm Hg
<52 89 39.3 47.7 31.1 0.06
^52 113 25.7 34.5 16.4 0.06
Paco2, mm Hg
<43 96 31.2 34.6 27.3 0.78
$>43 106 32.1 46.0 19.6 0.002
PH
<7.40 95 28.4 42.2 16.0 0.004
^7.40 107 34.6 38.6 30.0 0.46
Hematocrit, %
<47.4 99 32.3 41.5 21.7 0.03
^47.4 102 31.4 38.8 24.5 0.20
FEVu L
<0.69 97 35.0 43.4 25.0 0.07
^0.69 101 29.7 39.1 21.8 0.08
FVC,£
<1.89 99 31.3 43.5 20.8 0.01
J; 1.89 99 33.3 39.6 26.1 0.20
FRC,L
<6.06 81 30.9 40.0 22.0 0.10
^6.06 82 32.9 42.9 22.5 0.05
Sleep, mean Sa0.„ air breathing, %
<85 89 37.1 50.0 24.4 0.02
2*85 92 23.9 31.1 17.0 0.14
Maximum work load, W
<35 85 42.4 48.8 35.7 0.13
5>35 113 23.9 33.9 14.0 0.02
Resting heart rate, beats/min
<92 101 28.7 38.8 19.2 0.08
£92 102 34.3 41.5 26.5 0.06
Mean pulmonary artery pressure, mm Hg
<27 86 27.9 37.0 17.5 0.03
£27 98 31.6 39.6 24.0 0.14
Pulmonary vascular resistance, dyne/scmb
<279 84 23.8 33.3 12.8 0.03
£279 84 36.9 45.2 38.6 0.11
Neuropsychological rating
<4.5 92 25.0 31.7 19.6 0.24
£4.5 94 39.4 48.0 29.5 0.04
Russell-Neuringer average impairment index
<2.17 85 25.9 31.7 20.5 0.30
£2.17 89 37.1 45.8 26.8 0.03
Halstead impairment index
<0.75 87 26.4 34.1 19.6 0.19
£0.75 87 36.8 43.8 28.8 0.05
Mood disturbance (POMS)
<43 89 29.2 30.2 28.3 0.80
£43 92 37.0 52.2 21.7 0.005

* Groups defined according to the median value for each characteristic. Data percentages for the continuous and nocturnal therapy groups are for deaths in each group
divided according to the criteria for "Characteristic" and are not percentages of the totals in Column 2.
t FEVi = forced expiratory volume in 1 second; FVC = forced vital capacity; FRC = functional residual capacity; POMS = Profile of Mood States.

was still significantly (P = 0.03) greater than in the noc-
turnal 0 2 therapy group. We have included all patients in
our major analysis of mortality because we believe that
even in the cases cited above, it is impossible to argue
that the patient's severe respiratory disease did not con-
tribute to their death. An example of this complexity is
afforded by our experience with coronary artery disease.
Three patients (two on nocturnal 0 2 , one on continuous
0 2 therapy) died of arrhythmias after documented myo-
cardial infarctions. On the other hand, nine other pa-
3 9 6 September 1980 • Annals of Internal Medicine • Volume 93 • Number 3
tients had sudden deaths—they were found dead in bed—
some with a history of increasing respiratory difficulty
but most without. Presumably these patients had fatal
arrhythmias, but few would argue that their deaths were
not primarily respiratory.
In an effort to discern if certain types of patients bene-
fited from continuous 0 2 therapy, we split nocturnal 0 2
and continuous 0 2 therapy patients into subgroups ac-
cording to median values of baseline characteristics
thought to be clinically important. Many subgroups

Downloaded From: http://annals.org/ by a University of California San Diego User on 12/27/2016

 showed significant survival benefit with continuous 0 2
therapy (Table 3), reflecting the reduced overall mortality
associated with this treatment. Generally patients with
hypoventilation and relatively poor pulmonary func-
tion—high arterial Pco 2 , low pH, low nocturnal oxygen
saturation, low FVC, and high FRC—showed increased
survival with continuous 0 2 therapy. The same was true
of patients with more severe brain impairment of neuro-
psychological testing and of patients who demonstrated
high levels of mood disturbance. These results could be
synthesized to reach the conclusion that the effect of con-
tinuous 0 2 therapy was most striking in the sickest pa-
tients. On the other hand, it was patients with the least
disturbed pulmonary hemodynamics—relatively low
mean pulmonary artery pressure and pulmonary vascular
resistance—and relatively well preserved exercise capaci-
ty who showed benefit from continuous 0 2 therapy. Thus
the composite patient showing the most benefit from con-
tinuous 0 2 therapy would have relatively severe derange-
ments of life quality and brain and lung function but
relatively mild disturbances of pulmonary hemodynamics
and exercise capacity.
Although there was overlap between subgroups show-
ing benefit from continuous 0 2 therapy, this overlap was
never total. For example, of the 95 patients with baseline
pH under 7.40, 24 also had arterial Pco 2 less than 43 mm
Hg, and of the 107 patients with baseline pH over 7.40,
35 had arterial Pco 2 greater than 43 mm Hg.
Most of the differences in mortality noted in Table 3
that we have cited as significant were significant at the
5% level. It might be argued that these differences are
not impressive considering that we made many statistical
comparisons and that some therefore might attain these
levels of significance in the absence of a difference be-
tween nocturnal 0 2 and continuous 0 2 therapy groups.
We do not believe that this explains the results shown.
Subgroups with high arterial Pco 2 , low pH, and more
serious mood disturbance showed differences in mortality
that were an order of magnitude more significant than
those likely to be achieved by chance alone. Groups of
independent variables, such as those concerning hypo-
ventilation and lung function, showed concordant differ-
ences in mortality that would have been unlikely had
they been due to chance. Finally, too many of the varia-
bles shown in Table 3 showed P < 0.05 to be due to
chance; the items shown were selected from analysis of
the characteristics shown in Table 2, and probabilities of
less than 5% occurred in many more than 5% of the
subgroups tested.
The reason for the decreased mortality associated with
continuous 0 2 therapy is unclear. Of the numerous physi-
ological and psychological variables measured during the
study, only two showed a significant treatment-related
change with time. Hematocrit value decreased in patients
on continuous 0 2 therapy but not in those on nocturnal
0 2 therapy. Although this result is in accord with the
results of animal studies (23), the decline in hematocrit
values in the continuous 0 2 therapy group was small,
averaging 9.2% at 18 months, and assigning any clinical
significance to a change of this magnitude is difficult.
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/27/2016
Overall mortality (Table 3) was very similar in patients
with high and low hematocrit values at baseline; having a
low hematocrit value at baseline was not associated with
increased survival. Further, though continuous 0 2 thera-
py lowered hematocrit values, it appeared to decrease
mortality in patients who entered the study with low he-
matocrit values (Table 3). It appears therefore that al-
though continuous 0 2 therapy decreased hematocrit val-
ues and increased survival, there is no evidence that these
results are related to one another.
Pulmonary vascular resistance also showed a differen-
tial effect of treatment, which is also consistent with ani-
mal studies showing that intermittent normoxia does not
reverse anatomical evidence of hypoxia-induced pulmo-
nary hypertension whereas continuous normoxia does
(23). The treatment-related change in pulmonary vascu-
lar resistance was more substantial than the hematocrit
change, and the 11% decline in resistance seen in the
continuous 0 2 therapy group may have been partially re-
sponsible for its longevity. This is not supported by exam-
ining mortality data in relation to pulmonary vascular
resistance (Table 3). Although patients who entered the
study with high pulmonary vascular resistances had rela-
tively high overall mortality, patients with a low baseline
pulmonary vascular resistance showed a mortality benefit
from continuous 0 2 therapy, whereas those with high re-
sistance did not. The effect of change in pulmonary vas-
cular resistance on mortality also suggested that changes
in resistance were not the cause of the lower mortality in
patients on continuous 0 2 therapy. The median change of
pulmonary vascular resistance for 101 patients was a de-
crease of 20 dyne/s • cm5 at 6 months. Of patients with a
greater decrease 14 subsequently died, nine in the noctur-
nal 0 2 therapy group and five in the continuous 0 2 thera-
py group. Of patients who demonstrated a smaller de-
crease in pulmonary vascular resistance, nine subsequent-
ly died, all from the nocturnal 0 2 therapy group. Thus,
when both groups were combined, patients with large de-
creases in pulmonary vascular resistance tended to have a
greater mortality than patients with small decreases. Fur-
ther, it appeared that continuous 0 2 therapy had more
striking beneficial effect in the patients with small de-
creases of pulmonary vascular resistance. These data
strongly suggest that although continuous 0 2 therapy re-
duced both mortality and pulmonary vascular resistance,
the two phenomena were not related.
The present study may be compared with others in
which the effects of oxygen therapy have been examined
through testing of patients before and after some months
of treatment (2-6, 9, 10). Qualitatively, our results are
similar. When nocturnal 0 2 and continuous 0 2 therapy
groups were combined, oxygen therapy apparently result-
ed in a fall in hematocrit value and pulmonary vascular
resistance and improvements in neuropsychologic func-
tion and the patients' perceptions of their quality of life.
However, changes were not large, averaging about 10%
in all instances, and whether even these changes were in
fact due to oxygen is uncertain. It is entirely possible that
these improvements, especially those of quality of life,
were due not to oxygen but to the more intensive medical
Trial Group • Oxygen Therapy for Obstructive Lung Disease 397
 and nursing care the patients received as study partici-
pants. Further, mean pulmonary artery pressure did not
change significantly, falling from 28.0 mm Hg to 26.4
mm Hg in 6 months, and exercise tolerance as tested by
maximum work achieved on a bicycle ergometer did not
improve, findings that differ from those of others. The
reasons are not clear for the relatively small impact of
oxygen therapy on the physiological and neuropsycholog-
ical functions that we measured. Our study had many
more patients than did previous ones and may have con-
tained subgroups that showed greater benefits, but the
nature of our mean results indicates that were there such
subgroups, there must have been others showing nearly
equal deterioration. As has been discussed above, there is
no reason to believe that our patient sample was biased so
as to minimize the effects of oxygen therapy. A more
likely cause for the relatively small oxygen effect ob-
served is that our protocol required a 3-week observation
period before baseline studies and entrance into the
study. During this period, arterial blood gas levels in
many patients improved to the point that the patient was
no longer eligible for the study. Most previous studies
have been less meticulous in assuring patient stability,
raising the question that some of their subjects were un-
stable on entry and therefore likely to improve whether
given oxygen or not.
In this trial, mortality was lower in continuous 0 2
therapy patients than in nocturnal 0 2 therapy patients,
particularly in those with more severe derangements of
lung and brain function and high levels of mood distur-
bance. We believe such patients should be treated with
continuous 0 2 . This does not necessarily mean that there
is no place for nocturnal 0 2 therapy in the treatment of
hypoxemic chronic obstructive lung disease. Indeed, the
current British trial that compares supplemental 0 2 with
no supplemental 0 2 involves 15 hours of 0 2 administra-
tion a day. This regimen is closer to our nocturnal 0 2
regimen than to continuous 0 2 therapy and is reported to
be associated with a lower mortality than that experi-
enced by patients receiving no outpatient oxygen (7).
Thus it appears that some oxygen is better than none and
that continuous oxygen is better than nocturnal, at least
in severely ill patients.
ACKNOWLEDGMENTS: Supported by contracts N01-HR-6-2942, 2943,
2944, 2945, 2946, and 2947 from the Division of Lung Diseases, National
Heart, Lung, and Blood Institute, National Institutes of Health, U.S. De-
partment of Health and Human Services.
Participants in the Nocturnal Oxygen Therapy Trial were the following.
Clinical Centers: Henry Ford Hospital, Detroit, Michigan: Paul A. Kvale,
M.D.*; William A. Conway, M.D.; E.O. Coates, Jr., M.D.; George C. Bow-
er, M.D.; Fareed U. Khaja, M.D.; Kenneth M. Adams, Ph.D.; Julia A. Lee,
M.S.; Marcia Kopacz, R.N., M.S.N.; and Jennifer Reeves, R.N.
Northwestern University, Chicago, Illinois: David W. Cugell, M.D.*;
Norman Solliday, M.D.; John Campbell, M.D.; William Kuczerpa, M.D.;
Diane Judge, R.N.; and Carmen Zych, R.N.
University of Manitoba, Winnipeg, Manitoba: Nicholas R. Anthonisen,
M.D.*; Morley M. Lertzman, M.D.; John A. Fleetham, M.D.; Janet Clarke,
R.N.; and Ann McBurney, R.N.
University of California and Scripps Clinic Research Foundation, San
Diego, California: Richard M. Timms, M.D.*; Patty Nordgren, R.N.; Rich-
ard Bordow, M.D.; Dale Kocienski, M.D.; and Connie Neveu, M.T.
University of Colorado, Denver, Colorado; Thomas L. Petty, M.D.*;
Thomas A. NefT, M.D.; Enrique Fernandez, M.D.; Louise M. Nett, R.N.,
RRT; Robert Maulitz, M.D.; Ruth Harada, M.D.; and Michael D . Baird,
B.S.
University of Southern California, Los Angeles, California: C. Thomas
3 9 8 September 1980 • Annals of Internal Medicine • Volume 93 • Number 3
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/27/2016
Boylen, M.D.*; John Mohler, M.D.; Hugo Chiodi, M.D.; Daniel Kanada,
M.D.; Elaine Layne, R.N.; Sylviane Herzog, M.S.; and Peggy Pegg, R.N.
Data Center: University of Michigan, Ann Arbor, Michigan: George W.
Williams, Ph.D.; and Sandy M. Snedecor, B.S.
Neuropsychology Centers: Veterans Administration Medical Center and
University of California, San Diego: Igor Grant, M.D.; Robert Reed, M.S.
University of Colorado, Denver, Colorado: Robert H. Heaton, Ph.D.; and
Susan Heaton, B.A.
Quality of Life Assessment: University of West Virginia, Morgantown,
West Virginia: A. John McSweeny, Ph.D.
Pathology Center: University of Manitoba, Winnipeg, Manitoba: William
M. Thurlbeck, M.D.
Advisory Board: Marvin A. Sackner, M.D. (Chairman); Stephen M.
Ayres, M.D.; B. William Brown, Ph.D.; Millicent Higgins, M.D., Dr. P.H.;
Philip Kimbel, M.D.; Jeanne K. Malchon; Harold Menkes, M.D.; William
F. Miller, M.D.; and Louis Vachon, M.D.
National Heart, Lung, and Blood Institute: Lynn H. Blake, Ph.D.; David
DeMets, Ph.D.; Claude Lenfant, M.D.; Hannah Peavy, M.D.; and Richard
Sohn, Ph.D.
•Principal Investigators
References
1. BURROWS B, EARLE RH. Course and prognosis of chronic obstructive
lung disease, a prospective study of 200 patients. N Engl J Med.
1969;280:397-404.
2. LEVINE BE, BIGELOW DB, HAMSTRA RD, et al. The role of long-term
continuous oxygen administration in patients with chronic airway ob-
struction with hypoxemia. Ann Intern Med. 1967;66:639-50.
3. PETTY TL, FINIGAN MM. The clinical evaluation of prolonged ambula-
tory oxygen therapy in patients with chronic airway obstruction. Am J
Med. 1968;45:242-52.
4. ABRAHAM AS, COLE RR, BISHOP JM. Reversal of pulmonary hyper-
tension by prolonged oxygen administration to patients with chronic
bronchitis. Ore Res. 1968;23:147-57.
5. KROP HD, BLOCK AJ, COHEN E. Neuropsychologic effects of continu-
ous oxygen therapy in chronic obstructive pulmonary disease. Chest
1973;64:317-22.
6. STEWARD BN, HOOD CI, BLOCK AJ. Long term results of continuous
oxygen therapy at sea level. Chest. 1975;68:486-92.
7. FLENLEY DC, DOUGLAS NJ, L A M B D. Nocturnal hypoxemia and long
term domiciliary oxygen in blue and bloated bronchitics. Chest.
1980;77:305-7.
8. K o o KW, SAX DS, SNIDER GL. Arterial blood gases and pH during
sleep in chronic obstructive lung disease. Am J Med. 1975;58:663-70.
9. STARK RD, FINNEGAN P, BISHOP JM. Daily requirement of oxygen to
reverse pulmonary hypertension in patients with chronic bronchitis. Br
Med J. 1972;3:724-8.
10. STARK RD, FINNEGAN P, BISHOP JM. Long-term domiciliary oxygen
in chronic bronchitis with pulmonary hypertension. Br Med J.
1973;3:467-70.
11. LENFANT C. Twelve- or 24-hour oxygen therapy: why a clinical trial?
JAMA. 1980;243:551-2. Editorial.
12. NATIONAL H E A R T , LUNG, A N D BLOOD INSTITUTE, DIVISION OF L U N G
DISEASE. Protocol for Nocturnal Oxygen Therapy Collaborative Pro-
gram. Bethesda: National Institutes of Health; 1978. (Available from the
National Heart, Lung, and Blood Institute.)
13. REITAN RM, DAVIDSON LA. Clinical Neuropsychology: Current
Status and Applications, New York: John Wiley and Sons; 1974.
14. RUSSELL EW, NEURINGER C, GOLDSTEIN G. Assessment of Brain
Damage: a Neuropsychological Key Approach. New York: John Wiley
and Sons; 1970.
15. G R A N T I, H E A T O N RK, M C S W E E N E Y AJ, A D A M S KM, TIMMS RM.
Brain dysfunction in COPD. Chest. 1980;77 (suppl.):308-9.
16. DAHLSTROM WG, WELSH GS, DAHLSTROM LE. An MMPJ Hand-
book. Vol. I. Minneapolis: University of Minnesota; 1972.
17. BERGNER M, BOBBIT RA, POLLARD WE, M A R T I N DP, GILSON BS.
The sickness impact profile: validation of health and status measure.
Med Care. 1976;14:57-67.
18. WASKOW IE, PARLOFF MB. Psychotherapy Change Measures. Wash-
ington, D.C.: U.S. Government Printing Office; 1975.
19. KAPLAN EL, MEIER P. Non-parametric estimation from incomplete
observation. J Am Statist Assoc. 1958;53:457-81.
20. C o x DR. Regression models and life tables. / R Stat Soc. [Series BJ.
1972;35:187-220.
21. ARMITAGE P. Sequential Medical Trials. Springfield, Illinois: Charles C
Thomas; 1970.
22. Report of the Committee for the Assessment of Biometric Aspects of
Controlled Trials of Hypoglycemic Agents. JAMA. 1975;231:583-608.
23. K A Y JM. Effect of continuous and intermittent normoxia on chronic
hypoxic pulmonary hypertension, right ventricular hypertrophy, and
polycythemia in rats. Am Rev Respir Dis. 1980; in press.