Relationship of Sputum Color to Nature

and Outpatient Management of Acute

Exacerbations of COPD*
Robert A. Stockley, MD, DSc; Christine O’Brien, MRCP; Anita Pye, PhD; and
Susan L. Hill, PhD

Study objectives: To stratify COPD patients presenting with an acute exacerbation on the basis of
sputum color and to relate this to the isolation and viable numbers of bacteria recovered on
culture.
Design: Open, longitudinal study of sputum characteristics and acute-phase proteins.
Setting: Patients presenting to primary-care physicians in the United Kingdom. Patients were
followed up as outpatients in specialist clinic.
Patients: One hundred twenty-one patients with acute exacerbations of COPD were assessed
together with a single sputum sample on the day of presentation (89 of whom produced a
satisfactory sputum sample for analysis). One hundred nine patients were assessed 2 months later
when they had returned to their stable clinical state.
Interventions: The expectoration of green, purulent sputum was taken as the primary indication
for antibiotic therapy, whereas white or clear sputum was not considered representative of a
bacterial episode and the need for antibiotic therapy.
Results: A positive bacterial culture was obtained from 84% of patients sputum if it was purulent
on presentation compared with only 38% if it was mucoid (p < 0.0001). When restudied in the
stable clinical state, the incidence of a positive bacterial culture was similar for both groups (38%
and 41%, respectively). C-reactive protein concentrations were significantly raised (p < 0.0001)
if the sputum was purulent (median, 4.5 mg/L; interquartile range [IQR], 6.2 to 35.8). In the
stable clinical state, sputum color improved significantly in the group who presented with
purulent sputum from a median color number of 4.0 (IQR, 4.0 to 5.0) to 3.0 (IQR, 2.0 to 4.0;
p < 0.0001), and this was associated with a fall in median C-reactive protein level to 2.7 mg/L
(IQR, 1.0 to 6.6; p < 0.0001).
Conclusions: The presence of green (purulent) sputum was 94.4% sensitive and 77.0% specific for
the yield of a high bacterial load and indicates a clear subset of patient episodes identified at
presentation that is likely to benefit most from antibiotic therapy. All patients who produced
white (mucoid) sputum during the acute exacerbation improved without antibiotic therapy, and
sputum characteristics remained the same even when the patients had returned to their stable
clinical state. (CHEST 2000; 117:1638 –1645)

Key words: bacteria; COPD; exacerbations; myeloperoxidase; sputum

Abbreviations: CI ⫽ confidence interval; IQR ⫽ interquartile range

C and
OPD is a major cause of morbidity worldwide
affects ⬎ 14 million patients in the United
States alone.1 It is predicted to become the fifth
leading cause of death and disability worldwide by
the year 2020.2 Acute exacerbations of COPD rep-
*From the Department of Respiratory Medicine, Department of
Medicine, Queen Elizabeth Hospital, Edgbaston, Birmingham,
B15 2TH, UK.
Supported by an educational grant provided by Glaxo Wellcome plc.
Manuscript received May 18, 1999; revision accepted February
10, 2000.
Correspondence to: Robert A. Stockley, MD, DSc, Department of
Medicine, Queen Elizabeth Hospital, Birmingham, B15 2TH, UK
resent a major health-care burden for both the
primary and secondary health-care sectors.1,3 How-
ever, the management of such episodes is far from
clear, and this almost certainly reflects their ill-
defined nature. Acute exacerbations present as a
worsening of the previous stable state and include
some, or all, of such clinical features as increased
dyspnea, wheeze, cough, sputum volume, the pres-
ence or development of sputum purulence or chest
tightness, or of such systemic symptoms as lethargy
or pyrexia. Not surprisingly, the cause of such epi-
sodes is also variable, including increased airflow

1638 Clinical Investigations

Downloaded From: http://publications.chestnet.org/ on 10/09/2014

 obstruction, mucus plugging and retention, and fluid
retention, as well as bacterial and viral infection.
These varied causes reflect the difficulty in establish-
ing the optimal therapy for the acute episode, and,
for this reason, national guidelines have been devel-
oped.1,4,5
However, despite the publication of such guide-
lines, their validity depends on the quality of the
published literature that forms the evidence for their
basis, and often the indicators for antibiotic therapy
in exacerbations are vague. For instance, in the
British Thoracic Society guidelines,5 the manage-
ment of acute exacerbations of COPD in primary
care was summarized in box 10. This included
empirically adding or increasing bronchodilators
(which would be expected to reduce airflow obstruc-
tion) and giving antibiotics to treat bacterial infec-
tions and oral corticosteroids to reduce inflamma-
tion. However, if such treatments are to be used
wisely and appropriately, particularly given the sen-
sitivity regarding the use of antibiotic therapy, it is
imperative to classify and treat the exacerbations
according to their features.
Antibiotic therapy is widely used in the treatment
of acute exacerbations, but evidence of efficacy is
debatable, with some controlled studies showing a
clear benefit,6,7 whereas others do not.8,9 Such re-
sults are to be expected as the nature of the exacer-
bation is rarely defined. Although sputum culture
may be expected to clarify the role of antibiotics, the
results can also be confusing because, in the stable
clinical state, some patients have a sputum culture
that is positive for bacteria.10,11 Madison and Irwin12
highlighted the lack of a widely accepted definition
of an exacerbation and emphasized that this created
difficulties in the interpretation of studies and was
not helpful for the clinician. This view was also
promoted by Wilson and Wilson,13 who stated that
future studies required defined populations but that
placebo-controlled studies were probably no longer
ethically justified. However, in 1987, Anthonisen and
colleagues14 published what remains the most widely
referenced controlled trial, which indicated some
clinical features related to a significant benefit of
antibiotic therapy. These authors classified exacerba-
tions into three groups depending on the number of
clinical features present. Subjects with increased
breathlessness, sputum volume, and sputum puru-
lence showed a significant advantage of antibiotic
therapy, whereas those with only one or two of these
three features showed none.14 Nevertheless, the
results of this study have been incorporated specifi-
cally into the British Thoracic Society guidelines;
antibiotic therapy is recommended if two of the
three clinical criteria, outlined above, are present,
and, therefore, the production of purulent sputum is
Downloaded From: http://publications.chestnet.org/ on 10/09/2014
not mandatory.5 Other guidelines, however, are less
clear, suggesting the decision should be made clini-
cally1 or when sputum becomes purulent.4
In view of this relatively loose guidance, we de-
cided to embark on a prospective study of acute
exacerbations of COPD in an attempt to determine
whether a subgroup existed in which antibiotic ther-
apy was likely to play a more important role. In an
editorial review on the relationship of bacteria to
lung host defenses, it was suggested that it should be
possible to separate the presence of bacteria as
commensals in the airway from those causing an
infection.15 The latter would be expected to be
accompanied by activation of secondary host de-
fenses, which include increased neutrophil recruit-
ment to the airways. This neutrophil influx should be
associated with a change in secretions from mucoid
to purulent (because the myeloperoxidase from the
neutrophils is green), and the process would reverse
after antibiotic therapy that reduced or eliminated
the bacterial load, thereby leading to resolution of
the secondary host response. Such a concept is
consistent with the classic study of Anthonisen and
colleagues,14 because sputum purulence was one of
the three clinical features in the group that showed a
significant response to antibiotics. In addition, the
presence of sputum purulence would be consistent
with the reported increase in neutrophils seen mi-
croscopically by other workers in an assumed bacte-
rial exacerbation.16,17
We therefore decided to differentiate acute exac-
erbations at presentation into those with or without
purulent sputum assessed clinically. The microscopic
and bacteriologic characterization of these secretions
was also recorded, and data were compared with
samples obtained some 2 months later in the stable
clinical state. In addition, we measured serum C-re-
active protein as an independent marker of the
systemic effect related to the activation of host
defenses.
Materials and Methods
Patients presenting to their primary-care physicians with acute
exacerbations associated with sputum production and underlying
diagnoses of COPD were considered for the study. All had a
history of chronic bronchitis (daily sputum production for at least
3 months of 2 consecutive years) and the development of new
symptoms with sputum production that led to a consultation with
their general practitioner. These new symptoms included in-
creased dyspnea, cough, sputum volume, sputum purulence,
temperature, or malaise. The primary-care physician made the
initial diagnosis, but patients were only included in the study if
the diagnosis was confirmed by the research team. Patients were
excluded if they had recently received antibiotic therapy (previ-
ous 4 weeks), if the illness was ⬎ 5 days in length before the
consultation, or if the primary-care physician thought that oral
corticosteroid therapy or hospital admission was mandatory.
CHEST / 117 / 6 / JUNE, 2000 1639
 Patients were seen on the day of the consultation by a
respiratory research nurse. Demographic details were noted and
a postbronchodilator FEV1 was obtained when possible using a
bellows spirometer. A fresh sample of sputum was collected into
a sterile container during a 1-h period, as free from saliva as
possible. Samples containing more than minimal salivary contam-
ination were discarded, and the remainder were sent to the
Respiratory Research Laboratory. The sample was assessed by
one of three laboratory staff, and the macroscopic appearance of
the majority of the sample was allocated a sputum number by
reference to a standard color chart. This chart was based on the
principle that neutrophil myeloperoxidase concentrations in the
sputum reflect the number of neutrophils present and that this
would relate to the degree of yellow-green coloration of the
sample. Values of 1 and 2 reflected the nature of mucoid sputum
(opaque or milky), whereas values of 3 to 8 reflected increasing
yellow-green coloration up to the darkest color observed in
sputum from cystic fibrosis patients. Intrasubject assessment
demonstrated that individuals differed by no more than one
number category but that mucoid (grade 2) and mucopurulent
(grade 3) sputum were always identified correctly.
After macroscopic assessment, a sputum smear was prepared
for Gram’s stain. This was examined for the presence of a
predominant bacterial type under high-power microscopy. In
addition, the smear was examined under low-power microscopy
(⫻ 100) for the presence of neutrophils. These were counted in
a semiquantitative way, and the sample was classified as contain-
ing ⬍ 25, 25 to 50, 50 to 100, or ⬎ 100 neutrophils/low-power
field.
After classification of the nature of the sputum, those patients
with macroscopically mucoid samples did not receive antibiotic
therapy. However, for the purpose of the current study, the
ethics committee believed that antibiotic therapy was mandatory
for those with clearly purulent sputum on the basis of the
arguments presented in the introduction.
Quantitative sputum culture was then performed on an aliquot
of the sample as described previously.18 Finally, 10 mL of blood
was obtained from the patient, and the serum was collected and
stored at ⫺70°C for subsequent measurement of the C-reactive
protein concentrations by radial immunodiffusion using commer-
cially available prepoured plates and standards (Binding Site;
Birmingham, UK). In addition, repeat serum samples were
obtained ⱖ 14 days after the presentation with symptoms and
assessed for a rise in or positive complement fixation titer to viral
agents or atypical organisms.
When patients were reviewed in the stable clinical state,
sputum was again collected, when possible, and subjected to the
above procedure, and a final blood sample was drawn for further
measurement of C-reactive protein.
The study was approved by the University Hospital Birming-
ham NHS Trust Ethical Review Board. Statistical differences of
sputum sample characteristics between groups were determined
by Fisher’s Exact Test. Differences in C-reactive protein and its
change with resolution, together with sputum color change, were
assessed by Wilcoxon rank sum test for unpaired and paired data,
respectively. A p value of ⬍ 0.05 was taken as an indicator of a
difference in the data.
Results
Of the 148 patients referred during the 15 months
of the study, 1 patient was not entered in to the study
because clinical review indicated the presence of
pneumonia. Three patients were withdrawn within 7
days because of noncompliance, 3 refused entry, 6
1640
Downloaded From: http://publications.chestnet.org/ on 10/09/2014
were unable to provide a suitable sputum sample for
analysis, and 14 were not entered because it was
believed that they would be unable to comply with
the study or had received recent (in the previous 4
weeks) changes in therapy. The demographic fea-
tures of the remaining patients are summarized in
Table 1.
Eighty-two patients complained of increased
breathlessness at presentation, 83 had increased
sputum volume, and 59 noted a change in their
sputum color. All patients with purulent sputum and
86.5% of the mucoid group complained of at least
two of these symptoms.
There was no difference between the two groups
with respect to inhaled therapy (␤2-agonists or anti-
cholinergic agents). Only 6 patients were receiving
regular nonsteroidal anti-inflammatory drugs, and 21
were receiving prophylactic low-dose aspirin.
The sputum samples from the 121 patients en-
tered into the study were graded as mucoid (grade 1
or 2) for 34 patients, grade 3 for 12 patients, grade 4
for 42 patients, grade 5 for 29 patients, and grade 6
for 4 patients by the laboratory research staff.
Of the 121 samples assessed at the start of the
study, 112 (92.6%) had ⬎ 25 neutrophils/low-power
field seen on the sputum smear. Seventy-nine sam-
ples (65.3%) contained a predominant bacterial type
seen on Gram’s stain, but a viable bacterial species
was cultured from 86 samples (71.1%). In 71 of the
samples, the number of the bacterial species cul-
tured was ⬎ 107 cfu/mL.
Sputum Characteristics
Mucoid Exacerbations (n ⫽ 34): At presentation,
the Gram’s stain appearance of samples showed ⬎ 25
neutrophils/low-power field in 26 samples and the
presence of a predominant bacterial type in 3 sam-
Table 1—Patient Demographics Obtained at Entry to
Study*
Variables Mucoid Purulent
Age (SD), yr 63.2 (8.2) 66.6 (8.9)
Sex (M:F) 16:18 53:34
Smoking, No.
CS 23 32
ES 11 49†
NS 0 6
FEV1 (SD), L 1.57 (0.62) 1.56 (0.77)
FEV1 (SD), % predicted 61.1 (22.4) 57.5 (24.5)
*Results are separated into the nature of the sputum at presentation.
The lung function data are from 29 subjects with mucoid samples
and 63 with purulent samples who were able to provide acceptable
traces. CS ⫽ current smoker, ES ⫽ ex-smoker, NS ⫽ never
smoked.
†p ⬍ 0.008, purulent vs mucoid.
Clinical Investigations
 ples (Fig 1). Thirteen samples (38.2%) subsequently
grew a putative pathogen on quantitative sputum
culture, which included 38.5% Haemophilus influen-
zae, 38.5% Haemophilus parainfluenzae, 15.4%
Moraxella catarrhalis, and 6.7% Neisseria meningi-
tidis. The median bacterial culture for these positive
samples was 7.5 ⫻ 106 cfu/mL (interquartile range
[IQR], 6⫻105 to 7.5 ⫻ 106), but only four samples
cultured ⬎ 107 cfu/mL (Fig 1).
Purulent Exacerbations (n ⫽ 87): In contrast to
the mucoid exacerbation samples, all samples except
one in this group contained ⬎ 25 neutrophils/low-
power field; in 34 samples, this was ⬎ 100/low-
power field. Indeed, there was a clear relationship
between the semiquantitative neutrophil count and
sputum color number (Fig 2). A predominant bac-
terial type was seen on Gram’s stain in 75 of the
samples, and a positive bacterial culture was ob-
tained from 73 (83.9%), which was significantly
greater than for the mucoid exacerbation samples
(odds ratio, 8.42; 95% confidence interval [CI], 3.43
to 20.67; p ⬍ 0.0001). The cultures with positive
results revealed a similar range of bacterial species to
those seen in the mucoid group: H influenzae
(56.2%), H parainfluenzae (13.7%), M catarrhalis
(15.1%), Streptococcus pneumoniae (9.6%), and oth-
ers (5.5%), which included Staphylococcus aureus
(one sample) and N meningitidis (four samples). The
median bacterial growth for these positive culture
samples (3.7 ⫻ 108 cfu/mL; IQR, 6.6 ⫻ 107 to
9.8 ⫻ 108) was significantly higher (p ⬍ 0.0001) than
for the mucoid samples. Furthermore, 67 of the 73
samples had a bacterial growth ⬎ 107 cfu/mL. Some
of these results are summarized in Figure 1.
Figure 1. Sputum characteristics are shown for samples classi-
fied as purulent and mucoid exacerbations at presentation.
Histograms are the proportion of samples showing ⬎ 25 neutro-
phils/low-power field (PMN), bacterial type seen on Gram’s stain,
positive bacterial culture, and samples with ⬎ 107 cfu/mL of a
putative pathogen.
Downloaded From: http://publications.chestnet.org/ on 10/09/2014
Figure 2. The relationship between average sputum color ⫾ SE,
shown on the vertical axis, compared with the semiquantitative
assessment of neutrophils in the sputum smear (see Methods).
Clinical Outcome
Mucoid Exacerbations: Thirty-two of the 34 pa-
tients showed resolution of their symptoms without
antibiotic therapy. However, two patients deterio-
rated within the first 8 days, and their sputum
changed from mucoid to purulent (which was asso-
ciated with a positive sputum culture of H influen-
zae); both subjects subsequently improved after a
14-day course of a broad-spectrum antibiotic. The
remaining 32 patients were studied again 2 months
after the start of the exacerbation when clinically
stable; 29 were able to provide a sputum sample. In
these patients, there was no significant difference in
the sputum characteristics in the stable clinical state
when compared with those at presentation. The
average sputum color number (median, 2.0; IQR, 2.0
to 3.0) was similar to that at presentation (median,
2.0; IQR, 2.0 to 2.0), and 22 of the sputum samples
(75.9%) contained ⬎ 25 neutrophils/low-power field
on Gram’s stain. The Gram’s stain revealed a pre-
dominant bacterial type in 8 samples (25.8%), and a
positive bacterial culture was obtained from 12
samples (41.4%). Of these positive cultures, eight
samples had a viable bacterial growth of ⬎ 107
cfu/mL (27.6% of all samples). These data are
summarized in Figure 3. The bacterial species ob-
served were similar to those at presentation, includ-
ing H influenzae (41.7%), H parainfluenzae (38.5%),
and M catarrhalis (16.7%), with other organisms (S
aureus and Enterobacterial species) accounting for
6.7% of the positive cultures. When organisms were
cultured, the numbers were similar to those at
presentation (median, 1.5 ⫻ 108; IQR, 1.0 ⫻ 106 to
1.3 ⫻ 109 cfu/mL).
Purulent Exacerbation: During the course of the
study, 77 patients had resolution of their symptoms
after antibiotic treatment and were seen when clin-
CHEST / 117 / 6 / JUNE, 2000 1641

Figure 3. Sputum characteristics for both groups seen in the
stable clinical state.
ically stable 2 months later. Seventeen of the patients
were unable to produce sputum at this follow-up
visit. The remaining patients’ samples showed a
significant improvement in sputum color
(p ⬍ 0.0001), with a reduction in color number from
a median value of 4.0 (IQR, 4.0 to 5.0) to 3.0 (IQR,
2.0 to 4.0). In addition, there was a significant
reduction in the proportion of samples containing
⬎ 25 neutrophils/low-power field from 98.9 to 80%
(odds ratio, 21.5; 95% CI, 2.71 to 170.50;
p ⬍ 0.0001) and in the presence of a predominant
bacterial type on Gram’s stain from 86.2% to 26.7%
(odds ratio, 17.19; 95% CI, 7.45 to 39.66;
p ⬍ 0.0001). Fewer (38.3%) of the samples grew a
bacterial species on culture (odds ratio, 8.34; 95%
CI, 3.87 to 18.18; p ⬍ 0.0001), and fewer samples
(38.3%) yielded a bacterial growth of ⬎ 107 cfu/mL
(odds ratio, 7.82; 95% CI, 3.71 to 16.46; p ⬍ 0.0001).
These results were similar to those obtained for the
mucoid exacerbations in the stable clinical state (Fig 3).
The numbers of each bacterial species cultured
from the positive samples was similar to the results
obtained for the mucoid exacerbations in the clini-
cally stable state (median, 6.2 ⫻ 107; IQR, 7.6 ⫻ 106
to 8.6 ⫻ 108 cfu/mL). The range of bacterial species
identified was similar to that at presentation: 69.6%
H influenzae, 8.7% H parainfluenzae, 13.0% M
catarrhalis, and 8.7% either Pseudomonas aerugi-
nosa or S aureus.
Relationship of Sputum Number to Bacterial
Culture
At the start of the study when patients presented
with an acute exacerbation, the presence of purulent
sputum (ⱖ grade 3) was associated with 73 of 86
samples having a positive bacterial culture and 67 of
71 samples in which the bacterial growth was ⬎ 107
cfu/mL. These results gave an overall sensitivity of
1642
Downloaded From: http://publications.chestnet.org/ on 10/09/2014
84.9% and 94.4%, respectively (Table 2). However,
some of the purulent samples had cultures that were
negative for bacteria or had lower numbers of organ-
isms, giving a specificity of 83.9% and 77.0% for
positive bacterial culture and ⬎ 107cfu/mL bacterial
load, respectively. The comparable figures for spu-
tum neutrophilia and organism seen on Gram’s stain
are indicated in Table 2.
C-Reactive Protein
Serum C-reactive protein was measured on sam-
ples obtained from 108 of the patients who entered
the study. The median value for 33 of the patients
classified as having a mucoid exacerbation on the
macroscopic appearance of their sputum was 4.9
mg/L (IQR, 1.0 to 10.2). This value was significantly
lower (p ⬍ 0.005) than for 75 purulent exacerbations
in which the median value was 14.5 mg/L (IQR, 6.2
to 35.8). When seen in the stable clinical state, the
value for the mucoid exacerbation group (median,
2.7; IQR, 1.0 to 4.9) was no longer significantly
different from that for the patients who had a
purulent exacerbation (median, 2.7; IQR, 1.0 to 6.6).
Lung Function
Spirometry was measured (when possible) after
bronchodilator administration in 92 of the patients
on entry to the study. The FEV1 predicted for the
patient’s age and sex showed a wide range of abnor-
mality, with 33% having moderate to severe impair-
ment (⬍ 50% predicted). The average results for
both groups are summarized in Table 1. However, in
view of the unstable nature of lung function during
an exacerbation, the patients were retested in the
stable clinical state (when possible). The results for
104 patients were stratified into mild (FEV1 ⬎ 50%
predicted), moderate (35 to 49% predicted), and
severe (⬍ 35% predicted) groups. These stratifica-
Table 2—Sputum Characteristics at Presentation for
All Patients (n ⴝ 121)*
Positive
Culture Culture ⬎ 107 cfu/mL
Variables Number (n ⫽ 86) (n ⫽ 71)
PMN 112 85 70
GRM 79 72 67
Purulent 87 73 67
Mucoid 34 13 4
*The numbers of samples containing ⬎ 25 neutrophils/low-power
field (PMN), positive organism on Gram’s stain (GRM), and
purulent or mucoid sputum are shown. In addition, the numbers of
samples for each characterization that yielded a positive bacterial
culture or high bacterial load (⬎ 107 cfu/mL) are shown.
Clinical Investigations
 Table 3—Bacterial Isolates at Presentation for Patients Seen 2 Months After the Exacerbation When Clinically
Stable*

Proportion
Disease Severity Bacteria Mucoid Purulent Purulent, %

Mild (FEV1 ⬎ 50%) H influenzae 4 22 63

H parainfluenzae 4 7
M catarrhalis 2 4
S pneumoniae — 2
Others 1 1
n ⫽ 26 n ⫽ 44
Moderate (FEV1 35 to 49%) H influenzae 1 8 79
H parainfluenzae — 2
M catarrhalis — 1
S pneumoniae — 2
Others — —
n⫽4 n ⫽ 15
Severe (FEV1 ⬍ 35%) H influenzae — 7 80
H parainfluenzae — 1
M catarrhalis — 2
S pneumoniae — 1
Others — 1
n⫽3 n ⫽ 12
*Patients are divided into the severity of their disease as indicated by the FEV1 as a percent predicted in this stable clinical state. The percentages
in each group presenting with a purulent exacerbation are shown.

tions are compared with the initial microbiology and
proportion of samples that were purulent at presen-
tation (Table 3).
Discussion
The verification of a bacterial cause of an acute
exacerbation of COPD is difficult. Undoubtedly,
bacteria play a role, although it requires large studies
or meta-analysis to demonstrate that antibiotics in-
fluence outcome.19 This is hardly surprising, because
many exacerbations will not have a bacterial origin,
and even when they do, spontaneous resolution can
occur.
Previous authors have suggested that the presence
of ⬎ 25 neutrophils/low-power field20 or a positive
Gram’s stain21 indicates a bacterial cause. However,
neutrophils are usually present in the secretions of
patients with COPD when clinically stable.22 Indeed,
in the current study, ⬎ 25 neutrophils/low-power
field were seen in the sputum from 68 of the 89
samples collected in the stable state. Thus, although
sputum neutrophilia was a highly sensitive and spe-
cific test to identify samples with a positive or high
bacterial culture (Table 1), it was present in 93% of
the samples at presentation. Similarly, Gram’s stain
usually predicted the positive cultures and large
bacterial load, although it has been shown to be less
reliable at detecting the presence of bacteria than
sputum culture itself.21 Nevertheless, both methods
require laboratory assessment, and this limits their
practical use.
Sputum purulence, on the other hand, is clinically
detectable and would be consistent with increased
neutrophil recruitment, indicative of a new or signif-
icant bacterial stimulus. In the presence of increased
breathlessness and sputum volume, this would fit the
clinical syndrome shown by Anthonisen et al14 to
benefit from antibiotics. However, purulence is a
subjective term and not further defined; hence,
treatment on the basis of this observation may also
be partly empirical.
In the current study, the nature of the sputum was
compared with a standard color chart of increasing
intensity. In most samples (117 of 121), this resulted
in a clear separation of mucoid samples from puru-
lent ones. In the remaining four samples, the sputum
was not homogenous in color, and three were labeled
as purulent (sputum grade 3) because more than half
the sample was colored. With this separation, the
purulent nature of the sputum remained highly
sensitive and specific for a positive bacterial culture
and high bacterial load (Table 1).
The improvement of the sputum characteristics
when the patients were clinically stable, with a
reduction in color and culture positivity, would
suggest that bacteria played an important role in the
increased symptoms of many of these patients. The
higher C-reactive protein concentration and its fall
confirm that the patients were systemically unwell,
consistent with the study reported by Dev and
colleagues23 in 50 patients who presented with pu-
rulent sputum but in whom only 29 patients (58%)
had a positive bacterial culture. The authors noted

CHEST / 117 / 6 / JUNE, 2000 1643

Downloaded From: http://publications.chestnet.org/ on 10/09/2014

 that a positive bacterial culture is not the most
dependable factor of an acute exacerbation and
proposed that a positive C-reactive protein may help.
However, the C-reactive protein measurement is not
readily available, particularly in primary care, and
our study would suggest that careful assessment of
sputum color is more effective in identifying samples
that are likely to have a positive bacterial culture
(84% of the samples).
The increased color observed in sputum samples
represents the presence of myeloperoxidase (the
green-colored enzyme from the neutrophil azurophil
granules), and our original hypothesis was based on
the presumption that bacteria in the bronchial tree
would not be the cause of acute symptoms unless
there was activation of secondary host defenses
leading to significant neutrophil recruitment. This
concept is consistent with the study described by
Monso and colleagues,10 who obtained protected
brush specimens of the lower respiratory tract during
exacerbations of COPD. These authors confirmed in
their invasive study that a positive bacterial culture
was more likely during an acute exacerbation (asso-
ciated with purulent sputum) but was still present in
some 25% of stable patients. In addition, their study
indicated that the bacterial load was increased dur-
ing the exacerbation (as seen in our results). These
observations were endorsed in the review by Cho-
dosh20 and supported by Medici and Chodosh,17 who
also noted the increase in neutrophils seen micro-
scopically during a bacterial exacerbation, although
such episodes were not defined clinically.
Our second group of patients, those with mucoid
exacerbations, did not have purulent sputum as
assessed by the color chart. Thirty-two patients
(86.5%) had at least two symptoms consistent with
the criteria of Anthonisen et al14 for an acute exac-
erbation and would thus require antibiotic according
to the British Thoracic Society guidelines.5 However,
despite a positive bacterial culture in 38% of the
samples, the viable bacterial numbers were low, and
only two patients deteriorated, requiring antibiotics
(both associated with the subsequent development
of purulent sputum, grades 4 and 5). These mucoid
exacerbations were different from those classified as
purulent from the sputum characteristics (Gram’s
stain, positive culture, and high microbial load). In
addition, these patients demonstrated little evidence
of a systemic effect of the illness (the C-reactive
protein concentrations were low), and the sputum
characteristics were unchanged when the patients
were restudied in the stable clinical state.
The nature of these mucoid exacerbations remains
to be determined. It is possible that these represent
nonbacterial inflammation or increased airflow ob-
struction. Certainly, steroids can influence the out-
1644
Downloaded From: http://publications.chestnet.org/ on 10/09/2014
come in acute exacerbations,24 although, again, the
episodes are also poorly defined. It is possible that
the major benefit of steroids is in the mucoid
exacerbations described here. However, further
studies will need to be performed to determine this
possibility.
We considered that our initial classification could
have identified two different subsets of COPD pa-
tients. For this reason, we have analyzed and re-
ported the sputum data when the patients were
reviewed in the stable clinical state. The results
emphasize the similarity between both groups when
clinically stable as well as the clear difference in the
purulent exacerbations at presentation and the lack
of change in the group with mucoid exacerbations. In
addition, the C-reactive protein concentrations were
higher and clearly fell in the purulent group to the
point at which they were no longer different from
those of the mucoid group, providing further support
for the similarity of both groups in the stable clinical
state. Indeed, the only difference we have noted
between the groups was that more of those present-
ing with purulent sputum were ex-smokers
(p ⬍ 0.008).
The organisms isolated from these outpatients at
presentation and when clinically stable are similar to
those described by others,25,26 but dissimilar to those
isolated from patients requiring admission.27 This
may partly reflect our exclusion of patients who had
recently had antibiotics or required steroid therapy,
thus reflecting the milder nature of the episode.
Nevertheless, the patients did have a wide spectrum
of airflow obstruction, suggesting that FEV1 is not
the determinant of the bacterial cause of exacerba-
tions in nonhospitalized patients. It is of interest to
note, however, that a greater proportion of the
patients with moderate and severe airflow obstruc-
tion presented with purulent sputum (Table 3).
Conclusions
In summary, we believe that acute exacerbations
of COPD are heterogeneous as described in the
extensive study by Macfarlane and colleagues28 and
the review by Madison and Irwin.12 Subdivision of
the exacerbations by sputum color identifies a group
in whom recovery occurs without antibiotic therapy.
The presence of mucoid sputum should be con-
firmed, however, as 15 of the patients (40%) subjec-
tively reported that the color had changed. Compar-
ison with a color chart indicated that this was to
milky-white and not yellow. The nature of these
mucoid exacerbations, as well as their treatment,
needs to be clarified. Similar to Macfarlane and
colleagues,28 we found few of either group (⬃ 10%)
Clinical Investigations
 had a rise in or a positive complement fixation titer to
viral agents and atypical organisms (data not shown),
indicating that invasive viral infections or atypical
organisms were not the cause. Finally, green sputum
was nearly always associated with the presence of a
significant bacterial load, which reduced in the clin-
ically stable state, suggesting that bacteria play an
important role in such episodes. The choice of
antibiotics for such episodes will depend specifically
on local susceptibility patterns.
Classification of exacerbations by sputum color
may enable antibiotic therapy to be withheld at
present in some patients. In addition, the current
study provides data to enable the design of future
antibiotic studies in COPD to determine their role
and comparative efficacy more clearly.
References
1 Standards for the diagnosis and care of patients with chronic
obstructive pulmonary disease: American Thoracic Society.
Am J Respir Crit Care Med 1995; 152(suppl):S77–S121
2 Murray CJL, Lopez AD. Alternative projections of mortality
and disability by cause 1990 –2020: global burden of disease
study. Lancet 1997; 349:1498 –1504
3 McCormick A, Fleming D, Charlton J. In: Morbidity statistics
from general practice: fourth national study 1991–1992.
London: Office of Population Censuses and Surveys, 1995
4 Siafakas NM, Vermiere P, Pride NB, et al. Optimal assess-
ment and management of chronic obstructive pulmonary
disease (COPD). Eur Respir J 1995; 8:1398 –1420
5 BTS guidelines for the management of chronic obstructive
pulmonary disease. Thorax 1995; 52(suppl 5):S1–S28
6 Berry DG, Fry J, Hindley CP, et al. Exacerbations of chronic
bronchitis treatment with oxytetracycline. Lancet 1960;
1:137–139
7 Pines A, Rafat H, Greenfield JS, et al. Antibiotic regimens in
moderately ill patients with purulent exacerbations of chronic
bronchitis. Br J Dis Chest 1972; 66:107–115
8 Nicotra MB, Rivera M, Aura RJ. Antibiotic therapy of acute
exacerbations of chronic bronchitis. Ann Intern Med 1982;
97:18 –21
9 Elmes PC, King TKC, Langlands JHM, et al. Value of
ampicillin in the hospital treatment of exacerbations of
chronic bronchitis. BMJ 1965; 2:904 –908
10 Monso E, Ruiz J, Rosell A, et al. Bacterial infection in chronic
obstructive pulmonary disease: a study of stable and exacer-
bated outpatients using the protected specimen brush. Am J
Respir Crit Care Med 1995; 152:1316 –1320
11 Riise GC, Larsson S, Larsson P, et al. The intrabronchial
Downloaded From: http://publications.chestnet.org/ on 10/09/2014
microbial flora in chronic bronchitis patients: a target for
N-acetyl cysteine therapy. Eur Respir J 1994; 7:94 –101
12 Madison JM, Irwin RS. Chronic obstructive pulmonary dis-
ease. Lancet 1998; 352:467– 473
13 Wilson R, Wilson CB. Defining subsets of patients with
chronic bronchitis. Chest 1997; 112(suppl):303S–309S
14 Anthonisen NR, Manfreda J, Warren CP, et al. Antibiotic
therapy in exacerbations of chronic obstructive pulmonary
disease. Ann Intern Med 1987; 106:196 –204
15 Stockley RA. Role of bacteria in the pathogenesis and pro-
gression of acute and chronic lung infection. Thorax 1998;
53:58 – 62
16 Fagon J-Y, Chastre J, Trouillet J-L, et al. Characterization of
distal bronchial microflora during acute exacerbation of
chronic bronchitis: use of the protected specimen brush
technique in 54 mechanically ventilated patients. Am Rev
Respir Dis 1990; 142:1004 –1008
17 Medici TC, Chodosh S. The reticuloendothelial system in
chronic bronchitis. Am Rev Respir Dis 1972; 105:792– 804
18 Pye A, Stockley RA, Hill SL. Simple method for quantifying
viable bacterial numbers in sputum. J Clin Pathol 1995;
48:719 –724
19 Saint S, Bent S, Vittinghoff E, et al. Antibiotics in chronic
obstructive pulmonary disease exacerbations: a meta-analysis.
JAMA 1995; 273:957–960
20 Chodosh S. Sputum cytology in bronchial disease. Adv
Asthma Allergy 1977; 4:8 –27
21 Valenstein PN. Semi-quantitation. of bacteria in sputum
Gram stain. J Clin Microbiol 1988; 26:1791–1794
22 Keatings VM, Collins PD, Scott D, et al. Differences in
interleukin-8 and tumor necrosis factor-␣ in induced sputum
for patients with chronic obstructive pulmonary disease or
asthma. Am J Respir Crit Care Med 1996; 153:530 –534
23 Dev D, Wallace E, Sankaran R, et al. Value of C-reactive
protein measurements in exacerbations of chronic obstructive
pulmonary disease. Respir Med 1998; 92:664 – 667
24 Thompson WH, Nielson CP, Carvalho P, et al. Controlled
trial of oral prednisolone in out patients with acute COPD
exacerbation. Am J Respir Crit Care Med 1996; 154:407– 412
25 De Abate CA, Henry D, Bersch G, et al. Sparfloxacin vs
ofloxacin in the treatment of acute bacterial exacerbations of
chronic bronchitis. Chest 1998; 114:120 –130
26 Langan CE, Zuck P, Vogel F, et al. Randomized, double-
blind study of short-course (5 day) grepafloxacin versus 10
days clarithromycin in patients with acute bacterial exacerba-
tions of chronic bronchitis. J Antimicrob Chemother 1999;
44:515–523
27 Eller J, Ede A, Schaberg T, et al. Infective exacerbations of
chronic bronchitis: relation between bacteriologic etiology
and lung function. Chest 1998; 113:1542–1548
28 Macfarlane JT, Colville A, Guion A, et al. Prospective study of
aetiology and outcome of adult lower-respiratory-tract infec-
tions in the community. Lancet 1993; 341:511–514
CHEST / 117 / 6 / JUNE, 2000 1645