| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the San Francisco General Hospital and University of California, San Francisco (Drs. Huang, Stansell, Osmond, Shafer, and Hopewell, and Ms. Turner), San Francisco, CA; Research Triangle Institute (Dr. Fulkerson), Durham, NC; Henry Ford Hospital (Dr. Kvale), Detroit, MI; University of California, Los Angeles (Dr. Wallace), Los Angeles, CA; Mount Sinai Medical Center (Dr. Rosen), New York, NY; Northwestern University (Dr. Glassroth), Chicago, IL; and University of Medicine and Dentistry of New Jersey (Dr. Reichman), Newark, NJ. Supported by contract nos. N01-HR7–6029, 6030, 6031, 6032, 6033, 6034, and 6035 from the National Heart, Lung, and Blood Institute and by the National Institute of Allergy and Infectious Diseases.
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
Design: Prospective, 64-month study.
Setting: Multicenter, ambulatory care.
Patients: 306 HIV-infected subjects enrolled in the Pulmonary Complications of HIV Infection Study who developed 467 episodes of new or worsening respiratory symptoms.
Measurements: Chest radiography followed by DLCO measurement, if the radiograph was normal or unchanged.
Results: An algorithm combining a chest radiograph followed by a DLCO measurement, if the radiograph was normal or unchanged, was effective and detected abnormalities that led to a diagnosis of PCP in 78 of 80 evaluable episodes (97.5%). The radiograph (specific parenchymal abnormality, number of lung zones involved) and the DLCO (degree of decrease, degree of decrease from baseline) also provided additional information on the probability of PCP.
Conclusions: In symptomatic HIV-infected patients suspected of having PCP, the diagnostic evaluation should begin with a chest radiograph, followed by a DLCO measurement, if the radiograph is normal or unchanged. If both of these tests are normal, it may be reasonable to conclude the evaluation rather than to proceed on to additional testing. This algorithm can serve as a benchmark for future comparisons.
Key Words: acquired immunodeficiency syndrome • human immunodeficiency virus • lung radiography • opportunistic infections • Pneumocystis carinii pneumonia • respiratory function tests
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
200
cells/µL, and the diagnostic approach should include an evaluation
for this disease.2 An evaluation for PCP is challenging for a number of reasons. First, the chest radiograph, the cornerstone test for detecting PCP, and a measurement of the diffusing capacity of the lung for carbon monoxide (DLCO), which is recommended when the radiograph is normal, have been reported to be normal in as many as 39% and 11% of cases of PCP, respectively.3 ,4 Furthermore, no radiographic abnormality or DLCO decrease is specific for PCP and, given a particular finding, little information exists defining the probability of PCP. Since neither of these tests is 100% sensitive or specific for PCP, it may be necessary to obtain additional tests, such as arterial blood gas, serum lactate dehydrogenase (LDH), exercise oximetry, and/or gallium scan as part of the evaluation.5
Algorithms incorporating the chest radiograph, the DLCO, and the gallium scan in a stepwise approach to detect PCP have been proposed.6 ,7 These reviews recommend a chest radiograph as the first test, followed by a DLCO measurement if the radiograph is normal, and a gallium scan if both the radiograph and the DLCO measurement are normal. They also propose that an evaluation for PCP conclude if the chest radiograph, DLCO measurement, and gallium scan are all normal. Other reviews have included tests such as arterial blood gases, serum LDH, and/or exercise oximetry in their approach.8 ,9 However, no study has compared one approach to another, and no study has validated a stepwise approach. It is unclear whether combining these tests increases the detection of PCP and whether the evaluation for PCP can conclude after multiple normal results. Because, for example, a patient with a normal chest radiograph might also be more likely to have a normal DLCO (and a normal gallium scan), using these tests in the recommended stepwise approach might delay diagnosis and therapy and increase costs. In an era of increasing medical cost consciousness, the question of whether such an approach is effective is significant.
To answer the question of whether a stepwise approach using the chest radiograph and the DLCO is effective and to establish this approach as a benchmark for future comparisons, we determined the proportion of PCP cases detected by an algorithm using these two tests among HIV-infected persons enrolled in the Pulmonary Complications of HIV Infection Study (PCHIS) who developed new or worsening respiratory symptoms. We also determined the proportion of PCP cases detected by each individual test and the probability of PCP given a specific finding. Finally, we examined the potential contribution of presenting symptoms, signs, and laboratory data as indicators of PCP.
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
500 cells/µL, 45% between 201
and 499 cells/µL, and 19% 200 cells/µL). The PCHIS followed subjects from November 1988 through February 1994 (64 months). Subjects had evaluations at enrollment and at regular intervals (subjects randomized to either 3- or 6-month intervals). Visits included history, physical examination, and CD4 lymphocyte count measurement. At predetermined intervals (subjects randomized to either 3- or 12-month intervals), routine screening chest radiographs and pulmonary function tests were performed.
Subjects who developed new or worsening respiratory symptoms—including a cough persisting for > 5 days, shortness of breath progressing for > 5 days, severe shortness of breath (inability to perform activities of daily living) for > 1 day, or unexplained fever (oral temperature > 38°C) for > 5 days—between these scheduled visits were instructed to present to their study site for evaluation for PCP. This analysis consists of 306 subjects who developed such symptoms (Table 1 ). The majority were male (87.9%), white (70.6%), and homosexual or bisexual (70.3%).
|
|
If the presenting symptoms consisted of fever without cough or shortness of breath, the patient was observed without further pulmonary evaluation or therapy. If the symptoms consisted of productive cough with purulent sputum (with or without shortness of breath or fever), the patient was treated for acute bronchitis with oral antibiotics (but not with trimethoprim-sulfamethoxazole because this therapy might also treat PCP). If the symptoms consisted of nonproductive cough or shortness of breath (with or without fever), the patient underwent pulmonary function testing. Pulmonary function test interpretation focused on the DLCO measurement and included a comparison with the baseline value obtained at PCHIS enrollment.
If the DLCO (corrected for hemoglobin) was 75% of the
predicted value or decreased 20% from baseline, the patient
underwent diagnostic evaluation consisting of sputum induction,
bronchoscopy, or both. If the DLCO was > 75% predicted,
the protocol again provided a number of options, including observation
without further pulmonary evaluation or therapy, treatment for airflow
obstruction if suggested by spirometry, or additional testing
(eg, gallium scan).
Overall, 467 evaluations were completed for an HIV-infected subject who reported new or worsening respiratory symptoms and met criteria for chest radiography. The number of evaluations ranged by site (range, 38 to 131). In 293 evaluations, the radiograph was normal or unchanged and, therefore, the patient met criteria for a DLCO. However, in 82 instances, a DLCO was not performed. The reasons for this varied from those permitted under the algorithm protocol—fever without cough or shortness of breath (n = 39, clinical observation) and productive cough with purulent sputum (n = 24, treated for acute bronchitis)—to those faced in "real world" conditions—patient refusal (n = 7, including one case of PCP) and patient inability to perform testing (n = 12, including four cases of PCP). Regardless of the clinical outcome (positive [PCP+] or negative [PCP-] for PCP), all 306 patients who presented for the 467 evaluations had a follow-up visit 30 days later. In addition, all subsequent patient hospitalizations and deaths were reviewed to determine whether documented PCP or an illness consistent with PCP had developed.
Chest Radiography and Pulmonary Function Testing
Radiographic examination of the chest included
posterior-anterior and lateral views. A study physician interpreted
each radiograph using a standardized approach that included an
assessment of parenchymal abnormalities (interstitial, alveolar, and
nodular infiltrate[s], nodule[s], and cavity[ies]) and their
location (right/left lung, upper/middle/lower lung zones). Each
radiograph was compared with prior films to assess whether the
abnormalities were new or unchanged.
Pulmonary function testing included measurement of lung volumes,
expiratory flows, and DLCO. The DLCO was
corrected for hemoglobin and expressed as a percentage of the predicted
value adjusted for sex, age, and height. Each DLCO was
compared with the baseline value obtained on entry into the PCHIS to
determine whether it had decreased by 20%.
PCP and Non-PCP Diagnoses
Overall, there were 244 episodes of confirmed PCP (microscopic
visualization of P carinii cysts or trophozoites in induced
sputum or fiberoptic bronchoscopy specimens) diagnosed as part of the
PCHIS. None was diagnosed in an asymptomatic subject seen for a
scheduled visit. The majority of PCP episodes were diagnosed during
hospitalization; in hospitals, clinical information was not collected
in a standardized fashion and the algorithm was not consistently
followed. Thus, only the 85 PCP episodes (PCP+) that resulted from one
of the 467 evaluations were included in this analysis. Among the
remaining 382 evaluations (PCP-), 217 (56.8%) resulted in another
diagnosis; bacterial pneumonia (n = 78), acute bronchitis (n = 76),
and upper respiratory tract infections (n = 28) accounted for the
majority of non-PCP diagnoses. Importantly, none of these patients in
these 382 PCP- evaluations had a diagnosis of PCP established within
30 days of their evaluation.
Statistical Analysis
The overall performance of the algorithm and the performance of
each individual test was measured according to the presence or absence
of PCP (PCP+ or PCP-). Sensitivity was defined as the proportion of
PCP+ patients who had an abnormal test (eg, chest radiograph
with parenchymal abnormality). Specificity was defined as the
proportion of PCP- patients who had a normal test. For all analyses, a
p value of < 0.05 was considered significant.
Patients could present for multiple evaluations during the study (median = 1; range, 1 to 7 evaluations). Thus, evaluations were not independent events. To account for this nonindependence, the generalized estimating equations were used as implemented in Statistical Analysis System macro (version 2.02; U. Groemping; Dortmund, Germany). The generalized estimating equation was used to calculate the odds ratios and to perform logistic regression analyses. The p values reported are from the robust estimate of the standard error.
|
Results |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
75%
predicted in 18 of the 20 episodes (90.0%) for which the radiograph
was normal or unchanged and a DLCO was obtained (five PCP+
patients refused or were unable to perform the DLCO
measurement). The combination of these two tests failed to identify two
patients with PCP. Both patients had a normal radiograph. One had a
DLCO of 78% predicted; the other had a DLCO of
> 80% predicted, but decreased 18% from baseline. Neither patient
underwent additional testing (eg, gallium scan). Instead,
both underwent sputum induction because of a strong clinical suspicion
for PCP.
Chest Radiograph and DLCO
Both the chest radiograph and the DLCO provided
additional information on the probability of PCP, beyond whether the
test was abnormal or not. Patients who had a radiograph that revealed a
new or worsening parenchymal abnormality were more than four times as
likely to be PCP+ than PCP- (p < 0.0001; Table 2
). In addition, the specific abnormality and the number of lung zones
involved were both associated with the probability of PCP. Patients who
had parenchymal abnormalities with interstitial infiltrates were more
than four times as likely to be PCP+ than PCP- (p = 0.0001), and
those who had more than two lung zones involved with any parenchymal
abnormality were four times as likely to be PCP+ (p = 0.0001).
Furthermore, the combination of the specific abnormality and the number
of lung zones involved was important. Among patients who had
abnormalities with interstitial infiltrates, involvement of five or six
lung zones was strongly associated with PCP (p = 0.002). Similar
results were seen among patients who had abnormalities with alveolar
infiltrates and involvement of five or six lung zones (p = 0.03).
However, among patients who had abnormalities with an alveolar
infiltrate, involvement of only one or two lung zones was associated
with not having PCP (p = 0.0008). Given a parenchymal abnormality on
chest radiograph, the finding with the highest positive predictive
value for PCP was that of interstitial infiltrates involving five or
six lung zones. Twenty of the 28 subjects with this finding had PCP
(71.4%).
|
75% predicted were more
than 11 times as likely to be PCP+ than PCP- (p = 0.005); those with
a DLCO 65% predicted were four times as likely to be
PCP+ (p = 0.02). Similar results were obtained using the baseline
DLCO for comparison and defining an abnormal
DLCO as either a DLCO 20% decreased from
baseline or 30% decreased from baseline (p < 0.0001 and
p = 0.006, respectively). Comparison of the sensitivity and
specificity of these two methods for defining an abnormal
DLCO demonstrated that using a DLCO of 75%
of the predicted value was a more sensitive but less specific cut-off
for PCP than a DLCO with a 20% decrease from
baseline. Ninety percent of the PCP+ patients had a DLCO of
75% predicted, whereas 75% had a DLCO with
a 20% decline. In contrast, of the PCP- patients, 52.9% had a
DLCO of > 75% predicted, compared with 79.6% who had a
DLCO < 20% decreased from baseline.
|
200 cells/µL were all associated with
PCP, as was a chest radiograph that revealed a new or worsening
parenchymal abnormality, a radiograph with interstitial infiltrates,
and a DLCO 75% predicted (Table 4
). A serum LDH of > 220 U/L exhibited a trend toward being associated
with PCP (p = 0.08). However, because LDH measurement was not a
formal part of the algorithm protocol, a number of evaluations failed
to include a measurement.
|
200 cells/µL (odds ratio [OR], 10.9;
p < 0.0001), a history of prior PCP (OR, 4.8; p < 0.0001), a
chest radiograph with new or worsening parenchymal abnormalities (OR,
4.7; p < 0.0001), and the symptoms shortness of breath (OR, 3.7;
p = 0.008) and fatigue (OR, 3.2; p = 0.006). Among persons with an
abnormal radiograph, the independent predictors were the specific
finding of interstitial infiltrate(s) on the radiograph (OR, 4.5;
p = 0.0001), a CD4 count of 200 cells/µL (OR, 6.8;
p < 0.0007), and a history of either shortness of breath (OR, 3.2;
p = 0.008) or prior PCP (OR, 2.7; p = 0.006). Finally, among
persons with a normal or unchanged radiograph whose DLCO
was measured, a CD4 count of 200 cells/µL (OR, 11.6; p = 0.02)
and DLCO of 75% predicted (OR, 8.0; p = 0.008) were
independent predictors of PCP.
|
|
Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
75% predicted identified 97.5% of the evaluable
PCP episodes. Thus, very few patients with PCP (2.5%) would
potentially have been identified by additional testing (eg,
gallium scanning).6
,7
,8
,9 This study shows that the chest radiograph is an effective first test, but it has a lower sensitivity for PCP than has been reported in most previously published studies. In this study, the sensitivity of an abnormal radiograph for PCP was 71%, which is within the range of published sensitivities (61 to 100%) but is lower than most.11 ,12 ,13 In the largest report, all 104 PCP cases presented with an abnormal radiograph (sensitivity, 100%).13 One possibility for the observed difference is that these previous studies were conducted before the routine use of PCP prophylaxis. Patients who develop PCP while on prophylaxis may present with milder disease and, therefore, would be more likely to present with a normal radiograph. In our study, 67% of the patients who developed PCP reported current prophylaxis use. Another explanation may lie in the patients studied. Patients who are ambulatory may have milder disease and, therefore, would be more likely to present with a normal radiograph. In our study, patients were ambulatory at the time of evaluation. Furthermore, because these were subjects enrolled in a prospective cohort study of lung disease, they were encouraged to present for evaluation if respiratory symptoms developed or worsened. In the only report that clearly identified the study population (hospitalized patients), 55 of the 59 PCP patients had an abnormal radiograph (sensitivity, 93%).12 Our result is consistent with the study by Opravil et al,3 in which 39% of PCP patients had a normal chest radiograph, and a normal radiograph was significantly associated with a lower serum LDH, higher PaO2, lower alveolar-arterial O2 difference, and lower mortality, all factors implying milder disease.
This study is the first to provide information regarding the probability of PCP given specific radiographic abnormalities and the degree of these abnormalities. For example, the presence of interstitial infiltrates, involvement of more than two lung zones, and involvement of five or six lung zones with either interstitial or alveolar infiltrates were all associated with an increased probability of PCP. In contrast, the presence of alveolar infiltrates involving one or two lung zones suggested another diagnosis (eg, bacterial pneumonia). The ability of specific radiographic findings to indicate the probability of PCP or other pneumonias supports the use of chest radiography as the initial test.
This study also shows that the DLCO is an effective second
test, as it detects a significant proportion of PCP cases for which the
radiograph is normal or unchanged. In this study, a DLCO of
75% predicted identified 90% of the PCP episodes for which the
radiograph was normal or unchanged from the previous one and a
DLCO measurement was performed. This proportion is similar
to previous studies reporting a DLCO sensitivity for PCP
ranging from 89 to 100%.4
,14
,15
However, ours is the
first study to determine prospectively the sensitivity in patients with
a normal or unchanged radiograph. Our finding that the DLCO
is a sensitive test for PCP even in patients with a normal radiograph
supports its use as an effective second test. Importantly, a baseline
DLCO is unnecessary. Comparison of the DLCO
with a baseline value and using a decrease of 20% to define an
abnormal result had a lower sensitivity (although a higher specificity)
than the DLCO of 75% predicted. However, not all
patients were able to perform the breath hold maneuver necessary for a
DLCO measurement, and pulmonary function testing may not be
available in all clinical settings. Both of these are potential
limitations to its use.
The sensitivity and specificity of the chest radiograph and the DLCO measurement are likely to be influenced by the severity of symptoms and the population studied. Our study focused on symptomatic ambulatory patients. In this analysis, the sensitivity and specificity of the radiograph for PCP were 71% and 70%, respectively, while the sensitivity and specificity of the DLCO measurement for PCP were 90% and 53%, respectively. These numbers are higher than those from earlier PCHIS reports that demonstrated little utility of either the chest radiograph or the DLCO as screening tests in asymptomatic ambulatory patients.16 ,17 The observed difference between these earlier studies and our current one may be explained by the difference in the populations studied (asymptomatic vs symptomatic). Furthermore, the performance of these tests may also be different in admitted patients (ambulatory vs hospitalized), and our results cannot necessarily be extrapolated to this population.
This analysis focused on two tests used to evaluate patients for suspected PCP. While use of the chest radiograph as the initial test is widely accepted, the choice of the DLCO as a second test is debated. Our reason for selecting the DLCO as the second test was that this was the test used at many of the PCHIS centers at the time of study design. The main purpose of our study was to validate the effectiveness of a stepwise approach using a combination of tests and to establish a benchmark that could be used for future comparisons. This study specifically avoided attempting to define the optimum approach, and the questions of which combination of tests has the highest diagnostic sensitivity and specificity and which combination is most cost-effective remain unanswered.
In conclusion, an algorithm that uses the chest radiograph followed by a DLCO measurement if the radiograph is normal or unchanged is effective in detecting PCP in HIV-infected patients with new or worsening respiratory symptoms. In patients who have normal findings on both of these tests, it may be reasonable to conclude the evaluation rather than to continue on to additional testing such as gallium scanning.
|
Appendix 1 |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
|
Acknowledgements |
|---|
|
Footnotes |
|---|
A list of the institutions and investigators participating in
the Pulmonary Complications of HIV Infection Study is located in the
Appendix.
Correspondence to: Laurence Huang, MD, FCCP, AIDS Program Ward 84, San Francisco General Hospital, 995 Potrero Ave, San Francisco, CA 94110; e-mail: lhuang@sfaids.ucsf.edu
Abbreviations: DLCO = diffusing capacity of the lung for carbon monoxide; LDH = lactate dehydrogenase; OR = odds ratio; PCHIS = Pulmonary Complications of HIV Infection Study; PCP = Pneumocystis carinii pneumonia; PCP+ = positive for PCP; PCP- = negative for PCP
Received for publication June 17, 1998. Accepted for publication October 29, 1998.
|
References |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
This article has been cited by other articles:
|
|
 |
|
  L. Huang, A. Morris, A. H. Limper, J. M. Beck, and on behalf of the ATS Pneumocystis Workshop Partici An Official ATS Workshop Summary: Recent Advances and Future Directions in Pneumocystis Pneumonia (PCP) Proceedings of the ATS, November 1, 2006; 3(8): 655 - 664. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W. T. Miller Jr. and R. M. Shah Isolated Diffuse Ground-Glass Opacity in Thoracic CT: Causes and Clinical Presentations Am. J. Roentgenol., February 1, 2005; 184(2): 613 - 622. [Full Text] [PDF]
|
|
|
|
 |
|
 K. M. O'Neil The Changing Landscape of HIV-Related Lung Disease in the Era of Highly Active Antiretroviral Therapy Chest, September 1, 2002; 122(3): 768 - 771. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
