Missed lung cancer is a source of great concernamong radiologists and an important medicolegalchallenge. Failing to diagnose lungcancer ranks second only to overlooking breast canceras a cause of litigation among radiologists in theU.S. Lung cancer may be missed on either chestradiography or CT.
Missed lung cancers carry medical, legal implications Both chest radiography and CT interpretations fall prey to mistakes in decision making, poor visualization, and atypical appearance
Missed lung cancer is a source of great concern among radiologists and an important medicolegal challenge. Failing to diagnose lung cancer ranks second only to overlooking breast cancer as a cause of litigation among radiologists in the U.S. Lung cancer may be missed on either chest radiography or CT. Whereas most malpractice cases involve lesions overlooked on the former, a small but presumably increasing proportion are related to chest CT.
The chest radiograph is a complicated projection of 3D structures on a 2D image. A lung lesion must be distinguished from bones, pulmonary vessels, mediastinal structures, and other complex anatomy projected over or adjacent to the nodule. This situation presents a formidable challenge to the radiologist, and it is not surprising that some lesions may be overlooked.
Investigators have attributed overlooked lung cancer to several factors. Observer error is probably the most important. In the 1970s, Kundel et al1 described scanning error, recognition error, and decision-making error. The first two refer to a failure to perceive an abnormality, and the last and probably most common describes mistaking a true lesion for a normal structure. More recently, satisfaction of search error has been recognized as well. This is characterized by an interesting but unrelated finding that distracts the interpreter from identifying a lung nodule.2
Lesion characteristics also play a critical role.3 These include the size of the nodule and its density and margination. Noncalcified lesions less than 4 mm in diameter cannot be visualized, and 50% or more of lesions 1 cm in diameter are not detected, depending on location. In several investigations of missed lung cancer on chest radiography, the average size of the missed lesion has exceeded 1.5 cm, however. Although density and margination are more difficult to quantify than diameter, these characteristics are important in determining the likelihood that a lesion will be detected.
The nature of the adjacent projecting structures also affects the visibility of the lesion. A nodule projecting behind a rib or the mediastinum will be easier to identify than a similar lesion that is encumbered by overlapping structures.
The term “lesion conspicuity” is used to describe the totality of lesion characteristics and related surrounding structures. Technical considerations, such as image quality and patient positioning and movement, also factor into the likelihood of missing lung cancer. The use of computed radiography permits a more uniform radiographic exposure, but it is not clear that this innovation has reduced the frequency of missed lung cancer.
There is a wide difference between a lung cancer that is merely overlooked and one for which medicolegal culpability exists. In the U.S., a standard of negligence applies. For negligence to be shown, three elements are necessary: A breach in the standard of care occurred as defined by the actions of a reasonable physician practicing in a prudent manner, an injury can be related back to that breach in care, and substantial harm was caused.
Most often, litigation centers on the first consideration. As breach in the standard of care is a subjective measure, litigation frequently becomes a contest between expert witnesses on opposing sides of the case. From previous investigations, it is clear that some lesions will be overlooked by a prudent physician, and thus not all missed lung cancer represents a breach in the standard of care. In general, an adverse medicolegal outcome is best correlated with high conspicuity of the overlooked lesion.4,5
Although it is difficult to eliminate missed lung cancer entirely, recent technologic advances may be valuable in decreasing its frequency. One approach is the use of computer-aided detection software to assist the radiologist in detecting subtle lesions.
Preliminary data suggest that this technique detects lung cancers that were missed on primary interpretation.6 A second promising approach is the use of dual-energy chest radiography. With this technique, a soft-tissue image is produced devoid of overlapping skeletal structures, rendering an underlying parenchymal lesion more visible. A third, less widely used method is temporal subtraction chest radiography.7
The development of multislice CT technology, accompanied by rapid increase in its use in the thorax, has led to a greater likelihood that nodules will be overlooked on chest CT. With 64-slice and greater technology, a far larger number of images is created. Moreover, recent advances allow radiologists to visualize pulmonary nodules using nontraditional techniques, such as off-axial reformatting, maximum intensity projection, and volumetric imaging.
Many of these techniques permit more confident separation of the lung nodules from adjacent confounding structures, such as pulmonary vessels. With the implementation of these strategies and concomitant increase in data, the issues of radiologist inattention and fatigue have become paramount. These factors increase the likelihood that overlooked lung cancer will occur on chest CT.
Pitfalls in CT diagnosis of lung cancer can be divided into three groups: misinterpretation of a normal structure or benign nodule as a lung nodule (false-positive diagnosis), recognition of an abnormality with inaccurate interpretation of a lung cancer as a benign process (false-negative diagnosis due to decision-making error), and failure to describe a lesion that proves to be lung cancer (false-negative result).8
On chest CT, a normal structure may be misinterpreted as representing a pulmonary nodule because of partial volume effect or lack of familiarity with normal or variant thoracic anatomy. A commonly cited example of this phenomenon is the misidentification of the first costochondral junction as a pulmonary nodule if a small part of the inferior aspect is included on an axial slice. Reformatted images clearly show the true nature of this structure. A second instance of this type of “overcall” is a pseudonodule that may be present beneath the right inferior pulmonary vein. This finding is related to loculated pericardial fluid. In addition to pseudonodules, benign lung nodules, such as granuloma, hamartoma, round pneumonia, hematoma, or arteriovenous malformation, can mimic a lung cancer nodule.8
A second pitfall is an atypical appearance of a lesion that is due to lung cancer. Similar to chest radiography, lung cancer on chest CT has multiple appearances. Airspace nodules, a larger ground-glass pattern, and an ill-defined area of consolidation are all well-described manifestations of lung cancer, particularly bronchioloalveolar cell or adenocarcinoma. Calcification associated with lung cancer is more frequent in lung masses. Other atypical manifestations of lung cancer are a cystic appearance that mimics a bronchogenic cyst and an air-crescent sign similar to a mycetoma.8
Missed lung cancer on CT is challenging and clinically relevant. CT-related missed lung cancer occurs in both the routine clinical setting and a chest CT screening protocol. In either instance, misses are attributed to lack of detection or to misidentification of the nodule as a normal or benign structure such as a vessel or parenchymal scarring. Detection of early lung cancer is particularly difficult in the setting of extensive underlying lung disease.
Missed lung cancer on chest CT reported in the routine clinical setting has been described as a mixture of central and peripheral lesions. Satisfaction of search error is also relevant for missed lung cancer on CT. Major distractors can include aortic aneurysm and old tuberculous scarring. The relatively large diameter of the average missed lesion on CT In studies of missed lung cancer associated with chest CT screening protocols, overlooked small nodules constituted the primary area of difficulty.8
There has been considerable interest in the detection rate of small lung nodules on CT. Naidich et al electronically inserted simulated nodules ranging from 1 to 7 mm into computerized CT data sets.9 Average detection rates were 91%, 82%, 48%, and 1% for nodules smaller than 7 mm, 4.5 mm, 3 mm, and 1.5 mm, respectively. Dense nodules that were located in the lung periphery were more likely to be detected. In a screening CT study by Swensen et al, only 74% of all nodules were identified prospectively in the prevalence scan; the remaining 26% were found on the follow-up incidence scan. The proportion of these nodules that proved to be malignant is not stated.10
Investigations of missed lung cancers on screening chest CT reported them as peripherally located and early stage. In one study, Li et al described 32 lung cancers that went undetected on 39 low-dose CT scans, concluding that 23 errors were due to detection failure and 16 were caused by misidentification of cancer as normal structures. Missed lesions ascribed to detection failure averaged nearly 10 mm in diameter, and 91% showed a substantial component of ground-glass opacity. The misidentified lesions were on average nearly 16 mm and often associated with severe underlying disease. The bulk of lesions (88%) were Stage IA.11
Malpractice cases involving missing lung cancer on chest CT are well documented. One notable case involved a lawsuit by a patient with advanced lung cancer whose 8-mm left lower lobe lung lesion had previously been overlooked on a lung cancer screening CT scan. A settlement of one million dollars was made, of which the radiologist assumed 90%.12
CAD is being investigated vigorously as an aid to the radiologist in the detection of early lung cancer using CT.13 Studies assessing lung nodules using CT CAD systems have reported sensitivities ranging from 38% to 95%. False-positive results are reported to be between one and 5.5 per CT quadrant or section, and from 2.8 to 11 falsepositive identifications per CT study. Investigators have demonstrated that using thinner axial data (1 to 2 mm) led to improved sensitivity of the CAD system and a decreased false-positive rate, and most current systems require the use of thin-section imaging (≤2 mm) and are designed to detect nodules ≥4 mm.