| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Medical Imaging, North York General Hospital, 4001 Leslie St.,
Toronto, ON M2K 1E1, Canada.
2 Department of Medicine, North York General Hospital, Toronto, ON M2K 1E1,
Canada.
3 Toronto General Research Institute, University Health Network, 200 Elizabeth
St., Toronto, ON M5G 2C4, Canada.
Received January 8, 2004;
accepted after revision June 1, 2004.
Address correspondence to E. K. Y. Lai
(elai{at}nygh.on.ca).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Three radiologists who were blinded to the identity, diagnosis, treatment protocol, and outcome of each patient independently evaluated serial chest radiographs from 67 patients with confirmed SARS. In addition to the chest radiographic abnormalities and percentage of involvement, several quantitative improvement parameters, including the peak to 50% improvement time (PIT50), were collected. Correlation between PIT50 and clinical parameters (duration of fever, cough, dyspnea, oxygen supplementation, intubation, and death) were evaluated using Wilcoxon's rank sum testing and Spearman's correlation.
RESULTS. The most common initial findings were unifocal air-space disease in the periphery of the lower lungs occurring a mean of 3.6 ± 2.4 (SD) days from symptoms onset. Peak abnormalities were seen at 10.4 ± 2.9 days. PIT50 was dependent on disease severity, showing a strong linear correlation with the clinical parameter duration of oxygen supplementation (r = 0.44, p = 0.0015). Three patterns of disease were recognized: pattern A (severe, 29.9%) with PIT50 of more than 10 days, pattern B (typical, 44.8%) with PIT50 of 10 or fewer days, and pattern C (mild, 25.4%) with minimal findings throughout the course of the disease. This classification was supported by collaborative clinical parameters.
CONCLUSION. The quantitative radiographic parameter PIT50 has strong clinical correlation and can be used to differentiate severity of disease into severe, typical, and mild types.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
SARS-CoV-like viruses detected in Himalayan palm civets and a raccoon dog in a market in southern China suggest that the origin of SARS-CoV may have been from these or other wild animals [7]. Given the possibility that human or animal reservoirs of SARS-CoV still may exist and that SARS may have a seasonal predilection, there is concern that SARS may return in upcoming respiratory seasons. Current WHO guidelines emphasize the need for all countries to remain vigilant and to maintain their capacity to detect and respond to the reemergence of SARS if it occurs [8]. For this reason, it is important to understand the clinical and radiographic features of SARS.
SARS predominantly has been found to be a pneumonia with a prodrome of fever of 38°C or higher, headache, and myalgia followed by progressive nonproductive cough, dyspnea, and in many cases hypoxia [9, 10]. Chest radiography is one of the most important investigations in the diagnosis of SARS [11, 12]. Several studies have been published summarizing the radiographic features of SARS. These studies were mainly qualitative and found that the early radiologic manifestations of SARS are unilateral and bilateral areas of consolidation and groundglass opacities with findings progressing in most cases [13–15]. Wong et al. [13] proposed a radiographic classification of SARS consisting of four distinct patterns. Limitations of all these studies are that each is a single or consensus review of radiographs from patients with confirmed SARS without any control group. Ooi et al. [16, 17] found that the radiographic findings correlated with clinical parameters, including oxygen saturation and treatment response, and laboratory parameters in a patient population in Hong Kong, but these studies did not incorporate the different radiographic patterns of SARS in the analysis.
The aim of our study was to define quantitative radiographic parameters of SARS using chest radiographs from confirmed cases of SARS from a Toronto hospital using blinded, independent, multiobserver reviewing methods. Using these quantitative parameters, we developed a clinically useful classification system that correlates with clinical parameters and can be used to predict disease prognosis.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In all patients, the early chest radiographs were obtained in the anteroposterior projection with portable conventional radiography using the existing hospital equipment (Mobilett/Mobilett Plus HP, Siemens). Near the end of the study period, computed radiography was implemented, and several of the follow-up radiographs were obtained with computed radiography equipment (Centricity, GE Healthcare). Initially the radiographs were obtained daily from admission for the first 1–2 weeks and thereafter every 2–3 days until discharge. Follow-up images obtained during the first monthly clinic visits also were included when available.
Two chart abstracters reviewed charts to obtain the clinical parameters of all patients with confirmed SARS. The clinical database collected information about the presence and duration of fever, cough, exertional dyspnea, resting dyspnea, oxygen supplementation, intubation, and death. This study was approved by our institutional ethics review board; informed patient consent was not required.
Imaging Evaluation and Analysis
Each of the radiographs was reviewed sequentially and independently by
three practicing general radiologists with a minimum of 17 years of
experience. The reviewers were blinded to the identity, diagnosis, treatment
protocol, and outcome of each patient. For each radiograph, the multiplicity
(unifocal, multifocal), appearance (air space, interstitial, mass), and
distribution (upper, middle, lower, central, peripheral) of the radiographic
abnormalities and any defined associated findings (lymphadenopathy, pleural
effusion, pneumothorax, and pneumomediastinum) were documented. In addition,
an approximate size estimate of the abnormalities based on the percentage of
lung involvement was recorded for each radiograph. Qualitative findings were
summarized by consensus.
For quantitative assessment of radiographic disease progression, the mean percentage of lung involvement reported by the three reviewers was calculated for each radiograph. All radiographs were referenced chronologically from the first day of symptoms onset, which was considered day 1. In addition, the intervals from symptoms onset to peak radiographic findings, 50% improvement, and 100% improvement and the time between peak and 50% improvement were obtained for each patient.
Statistical Analysis
Three categories of disease severity were described using the radiographic
parameters obtained. Correlation with the clinical parameters was made using
Wilcoxon's rank sum testing and Spearman's correlation. In addition, the
radiographic responses to different SARS therapies including ribavirin,
interferon alfacon-1, and corticosteroids were summarized using the
Kaplan-Meier method. Statistical analyses were performed using SAS software,
version 8.0 (SAS Institute).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The treatments received by the patients varied. Seventeen patients (25.4%) received subcutaneous interferon alfacon-1 and corticosteroids. Eight patients (11.9%) received ribavirin with or without corticosteroids. Twenty-nine patients (43.3%) received corticosteroids alone. The remaining 13 patients (19.4%) received supportive therapy only.
Eighteen of the patients were admitted to the ICU during the course of the disease, four of whom died. One additional death occurred 3 days after admission before anticipated ICU admission. All 67 patients had fever of more than 38°C, with a mean duration of 7.5 ± 6.2 days. Fifty-seven patients (85.1%) complained of coughing. A total of 46 patients (68.7%) had dyspnea: 24 patients showed both exertional and resting dyspnea; 21 patients, only exertional dyspnea; and one patient, only resting dyspnea. Unfortunately, the duration of cough and dyspnea could not be measured accurately because many of the ICU patients were intubated and ventilated. Thirty-six patients (53.7%) required oxygen supplementation during treatment for a mean duration of 17.9 ± 24.1 days. Fourteen patients (20.9%) required intubation during the course of treatment for a mean duration of 23.3 ± 22.8 days.
Radiographic Findings of SARS
The radiographic abnormalities observed early on chest radiographs were
subtle. Of the 67 study patients, 51 (76.1%) had initial radiographs obtained
less than 7 days from symptoms onset (Table
1). The earliest manifestations were unifocal air-space disease in
the periphery of the lower lungs. There was no predilection to the right or
left lung. From a subset of patients with initial radiographs obtained within
the first 2 days of symptoms onset (n = 21) and patients with normal
findings on the initial radiographs (n = 14), the first radiographic
abnormalities were found to appear a mean of 3.6 ± 2.4 days from
symptoms onset.
|
The typical progression of disease was toward bilateral multifocal air-space disease (Figs. 1A, 1B, 1C, 2A, 2B, 2C, 3A, 3B, 3C, 4A, and 4B). Pleural effusion was rare and seen in only three patients (4.5%) at the time of peak radiographic lung findings. Pneumothorax and pneumomediastinum were seen in six ICU patients (9.0%), likely representing complications secondary to mechanical ventilation. No lymphadenopathy was seen.
|
|
|
|
|
|
|
|
|
|
|
Quantitative analysis of disease progression was extracted from the average lung involvement estimates reported by the three reviewers. Although the individual estimates for each radiograph varied among the three reviewers, the overall disease trend for each patient was relatively constant among the reviewers. Consequently, we refrained from relying on the actual mean involvement percentage score for the assessment of disease progression. Instead, three more consistent quantitative parameters related to disease trends also were obtained for each patient. The first two parameters were the time from symptoms onset to peak radiographic abnormalities and complete radiographic resolution. Complete resolution could not be documented for all patients and, therefore, was not a practical end point for comparison. A significant number of patients (n = 27, 40.3%) had a variable amount of residual radiographic findings on the last radiographs obtained at discharge and follow-up clinic visits. Therefore, a third reference point also was chosen: 50% improvement from the peak radiographic changes. The time between peak and 50% improvement represented the rate of convalescence and was used to compare disease severity among patients.
Correlation Between Clinical and Radiographic Findings
The peak to 50% improvement time (PIT50) was found to be
directly proportional to the duration of oxygen supplementation
(Fig. 5) with relatively high
correlation (r = 0.44, p = 0.0015). Poor relationship was
found between PIT50 and the duration of fever. Given the relatively
small number of intubated patients, data to evaluate the relationship between
PIT50 and the duration of intubation were insufficient.
|
Disease Classification
Three patterns of radiographic disease progression were recognized
(Fig. 6 and
Table 2). Pattern A (severe)
had prolonged resolution with PIT50 of more than 10 days. Pattern B
(typical) had a bell-shaped progression with a defined radiographic
abnormality peak and relatively rapid resolution with a PIT50 of 10
or fewer days. The peak may be a single point in time or a short plateau with
persistent or minor fluctuating changes. Pattern C (mild) represented patients
with only minimal radiographic abnormalities throughout the course of the
disease. A subset of pattern C cases (C0) was defined by persistent
equivocal findings interpreted as normal by at least one of the reviewers
throughout the course of the disease.
|
|
Twenty patients (29.9%) had pattern A disease. All five patients who died were classified as pattern A cases. Three deaths occurred during the initial rising slope of the graph, and two deaths occurred during the protracted resolution phase. The remaining patients in this group had peak radiographic findings at 10.6 ± 2.9 days. Because patients with this pattern were considered to have more severe disease, only a small number of patients had achieved 100% or even 50% improvement over the duration of the study. Only six patients (30%) in this group showed complete radiographic resolution at 48.0 ± 11.0 days (range, 30–58 days). The mean age of patients in this group was 54.8 ± 14.8 years, with 13 females (65.0%). All patients but one in this group required oxygen during the course of treatment. All five patients who died and 15 of the 18 patients admitted to the ICU were classified as pattern A cases.
Thirty patients (44.8%) were found to have pattern B disease. The radiographic peak occurred 10.2 ± 3.0 days from symptoms onset. Complete resolution of the abnormalities was documented in 21 patients in this group. The mean age of patients with this pattern was 43.9 ± 15.3 years. Twenty patients (66.7%) were female. Sixteen patients (53.3%) with pattern B did not require oxygen during the study. Three of the 18 ICU admissions were classified as pattern B cases, but two had a short stay in the ICU and did not require intubation.
The remaining 17 patients (25.4%) had pattern C disease. Patients in this group had radiographic abnormalities that remained below 10% lung involvement throughout the course of the disease. Given this minimal degree of findings, accurate assessment of changes was difficult and the peak and improvement times obtained were limited by high variability. Fourteen patients in this group achieved 100% resolution by the end of the study. Three of the patients (4.5%) in this group had only questionable, equivocal findings throughout the course of the study and were subclassified as pattern C0 cases. The mean age in this group was 44.8 ± 14.0 years, with 14 females (82.4%). None of the patients with pattern C was admitted to the ICU. Only three patients in this group required oxygen supplementation with maximum duration of 4 days.
There was no statistical difference in age between the patients among the three disease patterns. The disease parameters for pattern C were highly variable, and therefore significance was limited. The duration of oxygen supplementation for the three patterns showed a trend corresponding well with radiographic disease classification, but no statistical significance was detected because of the high variability of the data (p > 0.1).
Assessment of Treatment and Outcome
Our classification of the patterns of radiographic progression did not take
into account the treatments received by the patients. Obviously, treatment
response is expected to alter the course of the disease. The distribution of
different treatments in patients with the three disease patterns in our
population is summarized in Table
3. Given the retrospective nature of this study, the different
treatments were not distributed equally. Generally, conservative treatment was
given to patients with less severe disease, such as pattern C. Therapeutic
treatments were given to patients with more severe disease, such as pattern A.
However, one of the pattern A patients received only conservative therapy, and
nine of the pattern C patients received ribavirin or ribavirin and
corticosteroids.
|
Because both the 50% improvement time and PIT50 depended on disease pattern and the proportion of patients with different patterns was not the same for each treatment group, comparison of treatment response between treatment groups was calculated for pattern B patients only (Table 4). The reasons for choosing pattern B were twofold. First, pattern B represented the largest patient group. Second, response to treatment would be detected most easily in pattern B patients given the relatively shorter disease course and better-defined radiographic end points. The PIT50, in increasing order, was found in pattern B patients receiving interferon and corticosteroids (3.8 days, n = 8), conservative supportive treatment (4.0 days, n = 3), corticosteroids only (5.1 day, n = 14), and ribavirin (6.0 days, n = 3). The difference in PIT50 between the interferon and corticosteroids group and the corticosteroids only group was statistically significant (p < 0.01).
|
Summary of Control Group
Our study included 39 control patients who were evaluated for SARS but had
alternate diagnoses and negative SARS-CoV serology or polymerase chain
reaction tests during the same study period. Sixteen patients had normal
findings on the initial and follow-up chest radiographs. Nine patients had
findings of pneumonia. Eight patients had equivocal findings that were
interpreted as either atelectasis or artifacts. Six patients had interstitial
pulmonary edema.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In our study, we found that chest radiographs might show normal findings during an initial period after symptoms onset, and the mean time of the initial appearance of the abnormalities was 3.8 days from symptoms onset. In their study of 138 SARS patients, Wong et al. [13] described a similar time of 3.1 days. In a recent study, Grinblat et al. [15] found that the initial radiographic abnormalities occurred 11.9 days after exposure. However, those researchers did not include symptoms onset as part of their study parameters. In our study, the initial findings of unilateral, peripheral, air-space disease in the lower lungs were the most common presentation. This is in agreement with previously published results [13]. We did not find predilection to either the right or left lung. In our study, multifocal disease with bilateral involvement with central, upper, and middle distribution was an uncommon initial presentation. Interstitial changes, masslike opacities, and pleural effusions also were unusual.
We describe a classification of radiographic findings of SARS into three categories: pattern A (severe), pattern B (typical), and pattern C (mild). There was correlation between these patterns and at least one of the clinical parameters—namely, oxygen supplementation. The peak abnormalities in patients with disease pattern A or B occurred at approximately the same time. This finding may suggest that the pathophysiology for the development of consolidation is similar, but the differences in the rate of resolution between the two patterns may depend on viral load and burden, preexisting comorbid diseases, or other factors. Age is also a likely contributing factor because the mean age in patients with disease pattern A was higher than that in patients with pattern B (Table 3), but the difference was not clinically significant.
In contrast to the four disease types described by Wong et al. [13], we recognized only three disease patterns. We could not show a distinct patient population with two radiographic peaks to correspond with type 2 disease in their study. Instead, we found that our pattern B (typical) patients may have a short plateau of persistent or minor fluctuated changes and, thus, may encompass both types 1 and 2 from the study by Wong et al. Their type 3 corresponds to our pattern C (minimal) with minimal radiographic changes documented throughout the course of disease. Finally, type 4 from their study had progressive radiographic deterioration, corresponding to the minority of pattern A (severe) patients. The remaining pattern A patients with protracted resolution were not described in their study. This difference is a significant one because two of the deaths in our patient population occurred during the pattern A resolution phase. The relative distribution of the patterns also differs between the two studies. Types 1 and 2 patients in their study represented most of their patients (87.3%), contrary to the 44.8% of pattern B patients in our study. These discrepancies could be secondary to differences in patient populations and duration of follow-up. The proposed radiographic progression of SARS is summarized in Figure 7.
|
To our knowledge, only a few studies describing quantification of chest radiographic abnormalities of SARS have been published [13, 16, 17]. These studies depended on the actual extent of lung involvement to determine the severity of disease. Our experience is that the degree of involvement is a subjective finding, particularly on frontal portable radiographs. This is secondary to the inherent subjective nature of radiographic size estimates and the inaccuracy of extracting 3D information from 2D images. Therefore, our study has introduced several quantitative parameters related to disease progression that are objective and reproducible between interpreting radiologists.
The use of 50% improvement time and PIT50 is a novel quantitative method of evaluating disease progression and comparing disease severity between different patient groups. These parameters minimize confounding changes from preexisting lung disease and require shorter follow-up than the time for complete resolution. We have shown good correlation between one of the clinical parameters evaluated and PIT50. By using the PIT50 and the overall extent of involvement, we could estimate the severity of disease with reasonable correlation to the clinical parameters examined. The two reports by Ooi et al. [16, 17] were the only current studies that included radiographic correlation with clinical parameters. Those researchers showed that the severity of lung abnormalities on chest radiographs correlates with clinical and laboratory parameters. However, those studies did not take into account the different patterns of SARS.
The treatments received by the patients in our population were dictated predominately by clinical severity at admission and during the course of disease. However, nine of the patients with mild disease received therapeutic treatments, and one of the patents with severe disease received only conservative treatment (Table 3). Although it is conceivable that the nine patients with pattern C disease developed mild disease from good response to the therapeutic treatments, the single patient with pattern A disease most likely would have benefited from more aggressive therapy. This supports the need for reliable quantitative parameters to evaluate disease severity and select appropriate treatment options.
We also used the PIT50 as a method to evaluate treatment effectiveness in this and other studies [19]. In this study, the calculation was limited to only patients with disease pattern B. Ideally, the PIT50 should be calculated for each pattern of disease in each treatment group, but our sample size was too small to generate any statistical significance in the other disease patterns. The data from our study indicate that interferon and corticosteroids yielded the shortest PIT50, followed by conservative treatment, corticosteroids alone, and finally ribavirin. The numbers of pattern B patients with an available PIT50, in the conservative and ribavirin treatment groups were too small to generate statistically significant data. However, there was statistical difference (p < 0.01) between the interferon and corticosteroids group and the corticosteroids alone group, implying that the combined treatment of interferon and corticosteroids resulted in earlier improvement in pattern B patients.
The control group in our study was selected randomly among patients being assessed for SARS. Although many of the control subjects had normal radiographs (38.5%), some of the observed findings could mimic the abnormalities of SARS. For example, those subjects in the control group with pneumonia (23.1%) had a radiographic appearance similar to that of patients with pattern A or B in our study. This emphasizes the accepted notion that radiographic findings of SARS are nonspecific and overlap greatly with other causes of atypical pneumonia [13]. However, the radiographic changes are important for the evaluation of disease progression.
Several limitations to our study should be noted. First, the estimation of lung involvement on 2D frontal radiographs is subjective and relatively inaccurate. This limitation is particularly relevant in cases with minimal findings, such as patients with pattern C (mild) disease. We attempted to minimize this limitation by using three independent reviewers in our study. In addition, our analysis was based mostly on the more consistent disease progression parameters rather then the actual involvement score. Second, some of the associated findings, such as pleural effusion and pneumothorax, are relatively difficult to evaluate on portable radiographs and may have been under- or overreported in our study. Third, although our study represented the second largest number of patients among the major radiographic studies published to date, our patient population received nonuniform treatment. As a result, data for several of the patient groups were insufficient to generate statistically significant results. Lastly, the 50% improvement time was chosen arbitrarily but with due consideration. In our patient population, 50% improvement occurred between 2 and 4 weeks after symptoms onset when most patients were hospitalized still and improvement could be estimated more easily. Other potential quantitative end points are possibly useful, but we think the results thus obtained would likely not be significantly different.
In conclusion, our study describes a comprehensive classification for the interpretation of the appearance of SARS on chest radiographs in a North American perspective. The timing and descriptions of early changes and progression of disease correlated well with previously published results. From our study, we propose three distinct radiographic patterns of disease and novel quantitative parameters of evaluating disease progression and treatment comparison—namely, the 50% improvement time and the PIT50. We found that the PIT50 shows good correlation with clinical parameters and can be used to predict prognosis, triage treatment options, anticipate length of convalescence, and evaluate treatment efficacy. Similar principles could be applied to classify other respiratory infectious diseases to evaluate disease progression.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  V. C. C. Cheng, S. K. P. Lau, P. C. Y. Woo, and K. Y. Yuen Severe Acute Respiratory Syndrome Coronavirus as an Agent of Emerging and Reemerging Infection Clin. Microbiol. Rev., October 1, 2007; 20(4): 660 - 694. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. J. Cowling, M. P. Muller, I. O. L. Wong, L.-M. Ho, S.-V. Lo, T. Tsang, T. H. Lam, M. Louie, and G. M. Leung Clinical prognostic rules for severe acute respiratory syndrome in low- and high-resource settings. Arch Intern Med, July 24, 2006; 166(14): 1505 - 1511. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. C. Arroliga, E. Diaz-Guzman, and H. P. Wiedemann Severe Acute Respiratory Syndrome, Pulmonary Function Tests, and Quality of Life: Lessons Learned Chest, September 1, 2005; 128(3): 1088 - 1089. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
