| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Thoracic Imaging |
1 From the Departments of Radiology at Columbia Presbyterian Center, New York-Presbyterian Hospital, 622 W 168th St, New York, NY 10032 (J.H.M.A., G.D.N.P., M.C.S., Y.M.B.); Weill Cornell Medical Center, New York-Presbyterian Hospital, NY (P.K.S.); University of Maryland Medical System, Baltimore (C.S.W., P.P.); and Albert Einstein College of Medicine, Montefiore Medical Center, Bronx, NY (L.B.H.). From the 2000 RSNA scientific assembly. Received November 26, 2001; revision requested February 8, 2002; revision received March 25; accepted April 29. Address correspondence to J.H.M.A. (e-mail: jha3@columbia.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: From 1993 to 2001, an observational cohort was identified that consisted of 40 instances of NSCLC evident retrospectively at chest radiography but undetected by a radiologist at a time when the cancer was potentially resectable for cure. Sizes and locations of the tumors were assessed. Pearson 
2 testing was performed to compare the sex distribution of lung cancer in the present series with population data for the sex distribution of lung cancer in the United States during the present study.
RESULTS: Twenty-five (62%) undetected NSCLCs were in men and 15 (38%) were in women, yielding a ratio not significantly different from that for the sex distribution of NSCLC according to national data (2 = 0.22, P = .64). Median patient age was 62 years (range, 37–87 years). Median diameter of the missed cancers was 1.9 cm. Missed cancers were most commonly located in the upper lobes (right, 45%; left, 28%; total, 72%), especially in the apical and posterior segments/subsegments (60% of all the missed cancers). A clavicle obscured 22% of the missed cancers. Eighty-five percent of the missed cancers were in peripheral locations.
CONCLUSION: Potentially resectable NSCLC lesions missed at chest radiography were characterized by predominantly peripheral (85%) and upper lobe (72%) locations and by apical and posterior segmental/subsegmental locations in an upper lobe (60%). Distribution by sex of the missed cancers was comparable to national data for NSCLC. The missed cancers had a median diameter of 1.9 cm.
© RSNA, 2003
Index terms: Diagnostic radiology, observer performance • Lung neoplasms, 60.3211, 60.3212, 60.3214, 60.3216 • Lung neoplasms, diagnosis, 60.30
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Despite advances in diagnosis and therapy, the 5-year survival rate for patients with non–small cell lung cancer (NSCLC) in the United States has increased from 14.5% to only 16.3% in the past 20 years (10). At the time of diagnosis of NSCLC in the United States today, distant spread is present in approximately 52%, regional spread is present in approximately 29%, and local disease is present in only approximately 19% of patients (10). When stage IA NSCLC is resected surgically, the 5-year survival rate is approximately 75% (range, 67%–84%) (11–16). When missed at chest radiography, NSCLC tends to progress from early-stage to late-stage disease, especially if several years pass between examinations (1,17).
Although much attention has recently been drawn to detection of early-stage lung cancer through low-dose computed tomographic (CT) screening of individuals at high risk(18–24), chest radiography is usually the first imaging modality used in the detection of lung cancer, whether early or late stage (1,2,24,25). Lung cancers missed at interpretation of chest radiographs were described by Austin et al (17), in a report of a series of patients observed during the 1980s, as usually being located in an upper lung zone (81%) (most commonly in the right upper lobe [56%]), as typically having a mean diameter of 1.6 cm ± 0.8 (SD), and as being more likely to occur in women than in men (17).
The present study was undertaken to assess whether there was a change in the 1990s in the failure of detection at chest radiography of potentially resectable NSCLCs compared with experience in the previous decade (17).
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Thirty-nine patients with 40 missed cancers met these criteria. Approval of all three institutional review boards was obtained; the need to obtain informed consent from the patients was waived by each review board.
The retrospectively positive radiographs of 20 patients with 20 missed cancers were obtained at Columbia Presbyterian Center, New York-Presbyterian Hospital, New York, NY, with 120–130 kVp, an 8:1 grid, a par-speed screen, Kodak T-L film (Eastman Kodak, Rochester, NY), and a 183-cm focus-to-film distance. The retrospectively positive radiographs of 16 patients with 17 missed cancers were obtained at the University of Maryland Medical System, Baltimore, with 120 kVp, MVP-60 (General Electric Medical Systems, Milwaukee, Wis) or Fuji AC-3 (Fuji Medical Systems, Tokyo, Japan) radiographic units, and interpretation of images in soft-copy format at a workstation. The retrospectively positive radiographs of three patients with three cancers were obtained at the Albert Einstein College of Medicine, Montefiore Medical Center, Bronx, NY; two of these patients were imaged with GE Amx 3 radiography units at 75–80 kVp, and one was imaged with a Fuji digital chest unit at 90–110 kVp.
Clinical data (including patient age and sex, the time between the initial radiograph and the eventual detection of lung cancer, and histopathologic or cytologic diagnosis) were obtained by means of chart review. The cell type of each carcinoma was determined with percutaneous needle biopsy only (n = 16), thoracotomy and open biopsy only (n = 15), bronchoscopic examination only (n = 8), or bronchoscopic examination and thoracotomy (n = 1).
Lobar and segmental lesion locations were determined with chest radiography (n = 40), CT (n = 20), thoracotomy (n = 2), and/or anteroposterior cervical spine radiography (n = 1). Lateral radiographs were available for retrospective review for 28 of the cancers. The number of chest radiographic examinations at which each lesion was missed was also recorded. Each missed lesion was localized as hilar (ie, involving a bronchus, vessel, or node at the level of a main or lobar bronchus, including the origin of a segmental bronchus), juxtahilar (ie, adjacent to the hilum and/or at the level of a segmental bronchus, including the origin of a subsegmental bronchus), or peripheral (ie, any site not hilar or juxtahilar).
Review of Images
The initial and follow-up chest radiographs of each patient were subsequently reviewed together by subspecialists in thoracic radiology. For 20 of the cancers, the images were reviewed independently by two of three available subspecialists (G.D.N.P., M.C.S., Y.M.B.) at the institution; they then agreed on all findings in consensus. For 20 of the cancers, one available subspecialist (C.S.W. or L.B.H.) from each institution performed the review. For each patient, the subspecialist or subspecialists reviewed the imaging studies and agreed that the studies showed a lung cancer that had been initially missed.
The size of each lesion was calculated as the mean value of the transverse, vertical, and, if available, anteroposterior diameters, as measured on the initial radiographs to the nearest millimeter. The intensity of opacity (ie, "whiteness") of each missed lesion was assessed subjectively with a seven-point scale (minimal, minimal to slight, slight, slight to moderate, moderate, moderate to marked, and marked). Minimal intensity of opacity was defined as a minimal increase in opacity compared with the expected normal level of opacity for the site in question, and slight intensity was defined as intensity midway between expected normal intensity and moderate intensity. Moderate intensity of opacity was defined as soft-tissue intensity comparable to that of the heart, whereas marked intensity was comparable to that of calcified opacities of cortical bone.
If multiple views were available for review, it was first determined if the lesion was visible on each view. Whenever the assessed degree of opacity of the missed NSCLC differed between frontal and lateral projections, it was determined on which view the lesion was better seen—that is, on which view the greater degree of opacity was observed—and that greater degree of opacity was recorded. If any anatomic structures obscured a missed lesion, this was also recorded.
Statistical Analysis
Pearson 2 testing was performed to compare the sex distribution of patients in the present series with the sex distribution of lung cancers, according to population data, in the United States in 1996 (26) and 1997 (27)—the middle years in the present study—as well as to compare the sex distribution of lung cancers in the patients observed between 1981 and 1988 in the study of Austin et al (17) with the population data–based sex distribution of NSCLC during that time period (28). The unpaired t test and the Mann-Whitney (nonparametric) t test were used to compare the sizes of the missed cancers in the present study with the sizes of the missed cancers in the study of Austin et al (17). P 
.05 was considered to indicate a statistically significant difference.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
2 = 0.22, df = 1, P = .64) compared with the actual data in the present study, in which cancer was missed in 25 men (64%) and 14 women (36%). The cell types of the carcinomas were differentiated in 27 (68%) instances (18 of the lesions were adenocarcinomas, six were squamous cell carcinomas, two were bronchioloalveolar carcinomas, and one was an adenosquamous carcinoma) and were undifferentiated (ie, non–small cell) in 13 (32%). The mean interval between initial radiography and eventual detection of the cancer was 301 days (median, 152 days). Seventeen (42%) of the 40 cancers were missed only once at chest radiography, 16 (40%) were missed twice, and seven (18%) were missed more than twice.
Upper lobe predominance was noted in 29 (72%) of the 40 cancers: 18 cancers (45%) were located in the right upper lobe, and 11 cancers (28%) (percentages do not add up to 72% due to rounding) were located in the left upper lobe (Table 2, Fig 1). Within each upper lobe, apical or posterior segmental/subsegmental predominance was also noted; 24 (60%) of the 40 cancers were in these locations (Table 2). The apical segment of the right upper lobe contained 11 (27%) of the missed cancers. Four (10%) of the 40 cancers arose in the juxtahilar region, and two (5%) were hilar in origin (Fig 2). Peripheral predominance was also noted: 34 (85%) of the 40 cancers were located in peripheral areas of the lung (Fig 3).
|
|
|
|
|
|
|
The intensity of opacity was categorized as minimal, minimal to slight, or slight in 52% of the lesions; as slight to moderate in 22% of the lesions; and as moderate, moderate to marked, or marked in 25% of the lesions (Table 1). Only the lateral chest radiograph revealed the lesion in two patients (5%), and the lateral chest radiograph revealed the lesion distinctly better than the PA chest radiograph in one other patient (2%).
The mean diameter of the 40 missed NSCLCs at the time of initial radiography (at which they were not detected) was 2.1 cm ± 0.9 (SD) (median, 1.9 cm; range, 0.6–5.0 cm). The sizes of the missed lesions were slightly but significantly larger in the present series compared with those in the 1992 study of Austin et al (17) (mean diameter in the earlier study, 1.6 cm ± 0.8; unpaired t test, t = 2.10, df = 64, P = .04; Mann-Whitney test, P < .05).
The diameters of seven (18%) of the 40 cancers in the present series were greater than or equal to 3.0 cm (three of these cancers occurred in the left upper lobe, and two each occurred in the right upper and right lower lobes), whereas only four (10%) cancers were less than or equal to 1.0 cm in diameter (each of these cancers occurred in the right upper lobe; three occurred in the apical segment). Only two (5%) of the undetected cancers were less than 1.0 cm in diameter. Thirty-two (80%) of the 40 lesions were at least 1.5 cm in diameter. Fourteen (35%) of the 40 lesions were 1.9 cm or smaller and were located in the apical or posterior segment/subsegment of an upper lobe.
Chest radiography protocols in the present study incorporated the use of either screen-film (n = 28 [70%]) or digital (n = 12 [30%]) radiography. When we compared results obtained with the two techniques, we found that the diameters of the missed lesions were not significantly different (mean lesion diameter, 2.1 cm ± 0.9 on screen-film radiographs and 2.0 cm ± 0.8 cm on digital radiographs; P not significant, unpaired t test).
Prognosis was not necessarily adversely affected by the radiologic miss of an NSCLC. Nine (23%) of the patients in the present series remained alive and free of recurrent cancer at follow-up, although this was after variable intervals between the miss and subsequent detection (range, 0.25–5.0 years; median, 1.7 years).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Mean diameters of missed lung cancers at chest radiography have been reported to be in the range of 1.3–1.8 cm (1,4,17,24,29). We have no specific explanation for the fact that the mean lesion diameter in our study was larger than that in five previous studies, except to note that nine (22%) cancers in the present series that were 2.5 cm or larger (including one that was 5.0 cm in diameter) were located in either the apical or posterior segment of the right upper lobe or the apicoposterior segment of the left upper lobe—regions in which multiple overlying bones are often present and in which lesions may therefore be especially lacking in conspicuity. In comparison, Melamed et al (30) reported that of 169 stage I lung cancers detected in a chest radiography screening program, only 12 (7%) were more than 3 cm in diameter. Diameters of 8–10 mm have been described as the lower limits at which a radiologist can identify a nodule at chest radiography with a 90% probability of being correct (31).
Only two (5%) of the cancers in the present series were less than 1.0 cm in diameter, compared with 18% in the 1992 study by Austin et al (17). Quekel et al (1) reported an overall miss rate for lung cancer at chest radiography of 19% in general radiologic practice, but a miss rate of 71% for lesions 1.0 cm or smaller. To our knowledge, results of studies of three CT screening programs for detecting small peripheral lung cancers have been reported; chest radiography in those studies depicted only 23%–27% of the peripheral nodular lung cancers that were detected at CT (21–23). The mean diameter of lung cancers missed at CT has been described as ranging from 0.3 to 1.2 cm (32,33). Therefore, the size data of the present series support the concept that CT is distinctly more sensitive than chest radiography for the detection of small peripheral nodular lung cancers (21).
The upper lobe predominance (72% of cases) of missed lung cancers in the present series is comparable to that (81% of cases) observed in the series of Austin et al (17). In particular, location of the missed cancer in the apical or posterior segment of the right upper lobe (38% of cases in the present study and 44% of cases in the study of Austin et al [17]) or in the apicoposterior segment of the left upper lobe (22% of cases in the present study and 15% of cases in the study of Austin et al) is striking. In sum, these three upper zone segments alone were the sites of missed cancers in 60% of cases in the present series and in 59% of cases in the series of Austin et al (17). The upper lobe to lower lobe ratio of lung cancer for the birth-year cohorts of the present series has been reported to be 2.85:1 (ie, 74% of lung cancer occurs in upper lobes in these cohorts) (34), which corresponds closely to the percentage of upper lobe cancers (72%) observed in the present series. To our knowledge, results of evaluation of segmental distribution of lung cancers in a large series have not been reported.
Given the distribution of missed lung cancers in each of these two studies (ie, the present study and the study of Austin et al [17]), it is tempting to speculate that the visual search patterns of radiologists do not result in adequate attention being paid to the upper lung zones. An alternate possibility is that lesions of low conspicuity abound in the upper lung zones and radiologists can be expected to fail to detect subtle lesions wherever they may occur (17,35,36). The finding that a clavicle obscured the cancer in the apical or apicoposterior segment of an upper lobe in 22% of the cancers in the present series (Table 1) supports this possibility.
The lack of sex bias observed in the present series is distinctly different from the results of Austin et al (17). In that series, the sex distribution was nine men to 18 women at a time when the ratio of men to women with NSCLC in the United States was approximately 62:38 (28). On the basis of that demographic ratio, if there had been no sex bias in the study of Austin et al (17), the expected ratio of men to women, transposed to a total number of 27 patients, would have been 17 (63%) men to 10 (37%) women. This same-size cohort analysis yields a significant difference (2 = 6.00, df = 1, P = .02) between the actual and the expected sex ratios. That is, the results of the study of Austin et al (17) did show a statistically significant association between female sex in the period between 1981 and 1988 and an increased likelihood (compared with male sex) that potentially resectable NSCLC would be missed at chest radiography by a radiologist.
In 1992, Austin et al (17) speculated that the radiologists in their series may have had a lower suspicion for lung cancer in women than in men. Now that lung cancer occurs in women and men in the United States in a 45:55 ratio (10) and is widely known among physicians to be the most common cause of death from cancer in women in the United States, it does not seem surprising that the radiologists in the present study did not show sex bias in failing to detect lung cancer at chest radiography. As the incidence of lung cancer in women continues to rise, epidemiologic reports suggest that the susceptibility of women to the carcinogens in contemporary cigarette smoke is probably greater than that of men (37,38). Although small cell carcinoma remains a major disease in women who smoke (37–39), the results of the present study show that radiologists at three institutions in the United States were not biased by the patient’s sex in the detection of NSCLC at chest radiography during the past decade.
The rate of missed NSCLC in our cohort cannot be assessed because of the observational design of the present study. However, it is tempting to speculate that because a total of approximately 400 new patients with NSCLC are seen per year at the three institutions of this study, the actual rate of missed NSCLC at chest radiographic interpretation may be much less than has been suggested in previous reports (1–9). It is worth noting that in contemporary thoracic surgical practice, the chest radiograph remains by far the leading imaging tool for the initial detection of resectable NSCLC despite the popularity of CT screening (25).
The role of the lateral radiograph in detection of lung cancer at chest radiography has been in dispute for over 30 years (1,9,17,40). Only the lateral radiograph revealed the cancer retrospectively in two patients (5%) in the present series, and the cancer was seen better on the lateral radiograph than on the frontal projection in one other patient (2%) in our series. These results are comparable to those of other series, which indicate a 2%–4% detection rate for lung cancer on the lateral compared with the frontal chest radiograph (1,5,17,41). The data of the present study confirm that the lateral chest radiograph remains of occasional but important use in detecting lung cancer.
Retrospective detection of a cancer does permit assessment of its change in size over time. The prognosis of NSCLC is multifactorial (42–44), but growth rate does tend to correlate with prognosis (45). Retrospective detection of a lung cancer may therefore offer prognostic value.
Prognosis is not necessarily adversely affected by the radiologic miss of an NSCLC. Nine (23%) of the patients in the present series and 33% of the patients in the series of Austin et al (17) were alive and free of recurrent cancer at follow-up, although this was after variable intervals between the miss and subsequent detection.
The results of the present study appear to be unrelated to the technical modality used to acquire the data (screen-film or digital radiography). Moreover, the results support the fact that the quality of digital radiography and the quality of screen-film radiography appear comparable with respect to the detection of subtle focal pulmonary parenchymal abnormalities (46–48). It has been suggested that digital radiography may be superior to screen-film radiography for depiction of malignant nodules in underexposed regions of the frontal radiograph (eg, the retrocardiac and sulcus regions) (46,48). However, a missed primary lung cancer in these sites at either screen-film or digital radiography, whether in the present study or in previous studies (6,17), was rare.
Perception analysis of radiologic detection of pulmonary nodules has indicated three major kinds of error: search (failing to look at the abnormality), recognition (looking at the abnormality but not identifying it), and decision making (identifying the abnormality and then deciding to ignore it) (35,36). Results of recent research into radiologists’ "stream of consciousness" while interpreting images that contained a simulated nodule have shown that the leading types of errors appear to be failures of recognition, followed by errors of search and errors of decision making (2,49). Radiographic detection of subtle focal pulmonary opacities can be a difficult endeavor (17). The results of the present study confirm that radiologists should interpret chest radiographs with special attention to the possible presence of subtle pulmonary nodules, especially in the upper lung zones.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, P.K.S., J.H.M.A.; study concepts and design, P.K.S., J.H.M.A.; literature research, J.H.M.A., P.K.S.; clinical studies, all authors; data acquisition and analysis/interpretation, all authors; statistical analysis, J.H.M.A.; manuscript preparation and definition of intellectual content, all authors; manuscript editing, J.H.M.A., P.K.S.; manuscript revision/review and final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  F. Li, R. Engelmann, K. Doi, and H. MacMahon Improved Detection of Small Lung Cancers with Dual-Energy Subtraction Chest Radiography Am. J. Roentgenol., April 1, 2008; 190(4): 886 - 891. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. Li, R. Engelmann, C. E. Metz, K. Doi, and H. MacMahon Lung Cancers Missed on Chest Radiographs: Results Obtained with a Commercial Computer-aided Detection Program Radiology, January 1, 2008; 246(1): 273 - 280. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Singh, S. Sethi, M. Raber, and L. A. Petersen Errors in Cancer Diagnosis: Current Understanding and Future Directions J. Clin. Oncol., November 1, 2007; 25(31): 5009 - 5018. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. K. Gould, J. Fletcher, M. D. Iannettoni, W. R. Lynch, D. E. Midthun, D. P. Naidich, and D. E. Ost Evaluation of Patients With Pulmonary Nodules: When Is It Lung Cancer?: ACCP Evidence-Based Clinical Practice Guidelines (2nd Edition) Chest, September 1, 2007; 132(3_suppl): 108S - 130S. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Das, G. Muhlenbruch, A. H. Mahnken, T. G. Flohr, L. Gundel, S. Stanzel, T. Kraus, R. W. Gunther, and J. E. Wildberger Small Pulmonary Nodules: Effect of Two Computer-aided Detection Systems on Radiologist Performance. Radiology, November 1, 2006; 241(2): 564 - 571. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Peldschus, P. Herzog, S. A. Wood, J. I. Cheema, P. Costello, and U. J. Schoepf Computer-Aided Diagnosis as a Second Reader: Spectrum of Findings in CT Studies of the Chest Interpreted as Normal Chest, September 1, 2005; 128(3): 1517 - 1523. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Monnier-Cholley, F. Carrat, B. P. Cholley, J.-M. Tubiana, and L. Arrive Detection of Lung Cancer on Radiographs: Receiver Operating Characteristic Analyses of Radiologists', Pulmonologists', and Anesthesiologists' Performance Radiology, December 1, 2004; 233(3): 799 - 805. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Tigges, D. L. Roberts, K. H. Vydareny, and D. A. Schulman Routine Chest Radiography in a Primary Care Setting Radiology, November 1, 2004; 233(2): 575 - 578. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
