CT can play an important role in the diagnosis of pulmonary alveolar proteinosis. Patients often present with vague, nonspecific signs and symptoms that can fluctuate in severity. Although chest radiographic findings are often inconclusive in patients with this condition, CT findings in these patients can show a pattern and distribution of disease that is sufficiently characteristic to suggest the diagnosis. Once the diagnosis is considered, the patient can undergo bronchoalveolar lavage, transbronchial biopsy, or open-lung biopsy. CT can be used to direct the thoracic surgeon or bronchoscopist to the areas of greatest concern [1]. The diagnosis is confirmed by finding lipoproteinaceous material in the air spaces that stains positively with periodic acid—Schiff staining.

Several reports have described the CT aspects of pulmonary alveolar proteinosis [2,3,4,5,6], but these reports have analyzed only a small number of CT scans. Although CT findings can suggest the diagnosis of pulmonary alveolar proteinosis, the CT features are not pathognomonic [1]. For radiologists, the term “pulmonary alveolar proteinosis” is misleading because the CT appearance is not purely alveolar. Although air-space, interstitial, or ground-glass opacities can dominate the CT appearance of pulmonary alveolar proteinosis, some combination of these findings is more common. In rare cases, fibrosis can be seen. Relatively few articles have described the effects of pulmonary lavage on the CT appearance [2, 3, 6], and just one of these articles examined more than one patient [3].

This article systematically reviews the intraslice patterns and zonal distribution of disease on CT from the largest number of CT scans of patients with pulmonary alveolar proteinosis that has yet, to our knowledge, been described. The relative frequency of different types of pulmonary opacities in pulmonary alveolar proteinosis and the influence of section thickness are examined. Also, the effect of whole-lung lavage on different types of CT opacities in a smaller group of patients is assessed.

Patients from our medical center with a confirmed diagnosis of pulmonary alveolar proteinosis from open-lung biopsy, transbronchial biopsy, or bronchoalveolar lavage fluid were identified by searching a pathology database. Patients with AIDS or a known infectious cause of pulmonary alveolar proteinosis were excluded. None of the patients had pulmonary alveolar proteinosis related to a known malignancy. The radiology records of the included patients were searched, and all the available electronically recorded or hard-copy chest CT scans were assembled.

The CT scans were not ordered at a consistent point in the disease process. Many scans were ordered for further characterization of the disease before the diagnosis was made or when the patient showed clinical deterioration. CT scans were not routinely obtained after pulmonary lavage, but postlavage CT scans were obtained in 12 instances for nine patients.

The collimation and slice intervals for the CT studies were not uniform. CT images obtained with a 5- to 10-mm collimation were considered thick-section images; CT images obtained with a 1- to 1.5-mm collimation were considered thin-section images. Images obtained with lung settings were chosen for assessment at three levels (the arch of the aorta, carina, and inferior pulmonary veins) and were marked for review. These levels were chosen as identifiable landmarks that were consistently present on all scans to compare thin- and thick-section images. These three levels were also representative of three different zones of the lung, allowing examination of zonal differences in distribution of disease. Scoring a limited number of slices reduced the number of observations to a manageable level.

Each lung was evaluated for the intraslice pattern of disease (anterior, central, diffuse, geographic, peripheral, posterior, and no disease) for each lung at each level. Each lung was also analyzed for the predominant zonal distribution of disease (uniform, upper, middle, lower, or no disease). Disagreements over intraslice pattern and zonal distribution of disease were resolved by consensus. In addition, each observer separately recorded any hilar or mediastinal lymph node enlargement, lung nodule or mass, pneumothorax, or pleural effusion.

Two observers estimated the degree of involvement of each lung at each level with six types of pulmonary opacities: ground-glass opacities, airspace opacities, fibrosis, interlobular opacities, intralobular opacities, and emphysema. Commonly accepted definitions of these opacities were used [7]. Honeycombing, distortion, and traction bronchiectasis were considered indicators of fibrosis. Each observer ranked the degree of involvement in both lungs at each level on a scale from 0% to 100% in 5% increments. Scores for each location were judged concordant if they differed by 20% or less. Observations that differed by 25% or more were resolved by consensus.

Before beginning the data scoring, two board-certified chest radiologists were trained by grading 10 CT scans together and comparing results to establish standards for agreement. Then, the observers independently assessed all CT scans in random order without knowledge of clinical or laboratory information. After the CT scans were scored and concordant values were chosen for discordant data points, the scores were entered into a database and statistical analysis was performed.

A total of 139 CT scans (79 thick-section scans, 60 thin-section scans) from 27 patients were reviewed. Twenty-three patients had thick- and thin-section studies performed concurrently on at least one occasion.

Table 1 indicates the number and percentage of intraslice pulmonary patterns seen on thick- and thin-section scans. A geographic pattern was far more common, and a diffuse pattern was next most frequent. Pearson's chisquare test showed no significant difference in the intraslice pattern between left and right lungs; thick- or thin-sections; or upper, middle, and lower zones.

Table 2 shows that a substantial number of the CT scans revealed a uniform zonal distribution of disease. However, when a zonal predominance was present on thick- and thin-section images, lower zone predominance was significantly more frequent than middle zone predominance (p < 0.0001). Lower zone predominance was also significantly more frequent than upper zone predominance on the thick-section images (p < 0.001). On the thin-section images, lower zone predominance occurred more frequently than upper zone predominance, but this difference was not significantly different.

Table 3 illustrates the number and percentage of patients who had each type of opacity when all CT scans from each patient were considered cumulatively. Almost 30% of these patients had an element of fibrosis at some time. Of these patients, two had extensive regions of fibrosis of 70% or more, two had some regions of fibrosis up to 30%, and four others had minor regions of fibrosis ranging from 5% to 15%.

The respective proportions of different opacities observed on thick- and thin-section CT scans were as follows: ground-glass opacities, 100% and 100%; air-space opacities, 49.4% and 38.3%; fibrosis, 15.2% and 25.0%; interlobular opacities, 50.6% and 85.0%; intralobular opacities, 2.53% and 1.7%; and emphysema, 1.27% and 6.7%. Pearson's chisquare test indicated that the percentage of scans with interlobular opacities was significantly higher (p < 0.001) for thin-section images than for thick-section images, but no significant difference was detected for the other opacities. Pearson's chi-square test also showed no significant difference between the left and right lungs in the percentage of scans with each type of opacity for thick- and thin-section imaging at each location. The Jonckheere-Terpstar test of trend confirmed an increasing trend of interlobular disease from the upper to the lower lung images on the thick- and thin-section images (p < 0.002). No corresponding trend was found for the other opacities.

Follow-up studies were available for 20 patients. The average number of follow-up studies was 3.1 (range, 1-7 studies). The average follow-up time was 27 months (range, 2 weeks-97 months). Two patients with severe fibrosis had three and five follow-up scans. No consistent trend over time in either the type or the distribution of opacities was present.

No lung masses were observed in this study. One patient with pulmonary alveolar proteinosis proven at open-lung biopsy developed multiple cavitating lung nodules that were clearly an atypical finding for uncomplicated pulmonary alveolar proteinosis. The interlobular and ground-glass opacities present elsewhere in the lung were considered typical features of the condition. Mycobacterium kansasii was subsequently proven by culture of bronchoalveolar lavage fluid. One patient had a small left pleural effusion after bilateral whole-lung lavage. Another patient had a small pneumothorax related to recent transbronchial biopsy. Hilar lymph node enlargement was visible in the patient with M. kansasii and in another patient with pneumococcal pneumonia. Both of these patients also had mediastinal lymph node enlargement.

Mediastinal lymph node enlargement was visible in eight patients on 16 examination dates. Lymph node enlargement generally involved one or two lymph nodes and measured slightly more than 1 cm in the short-axis diameter. However, in the patients with pneumococcal pneumonia and M. kansasii, the lymph nodes were larger and more numerous. The lymph node enlargement related to pneumococcal pneumonia subsequently resolved; the lymph node enlargement related to M. kansasii persisted on later scans.

CT scans were obtained from 2 to 57 days before and after bilateral lung lavage in 12 instances in nine patients. Only thick-section images were consistently available for evaluation of the effects of lavage. Evaluation of ground-glass and interlobular opacities by paired Student's t test indicated a significant reduction (p < 0.05) in each of these opacities after lavage (Table 4). The other opacities were not significantly changed. In three instances, the proportion of the lung involved with ground-glass opacities showed more than a minimal increase after lavage. In two patients there was substantial conversion of air-space opacities to ground-glass opacities, but the overall extent of disease clearly improved. A third patient underwent left-lung lavage followed by right-lung lavage 7 days later. Between the sequential lavages the patient had a clinical presentation consistent with pneumonia, although all cultures showed negative findings. An increase in ground-glass opacities on the CT scan obtained 2 days after the right-lung lavage could be related to the clinically suspected pneumonia or to retained lavage fluid.

Six observations about the intraslice pattern of disease were recorded for each scan (834 total). In 134 (16.1%) of 834 instances the observers disagreed. The zonal distribution analysis involved two observations per scan (278 total). Disagreement about zonal distribution occurred in 34 (12.2%) of 278 observations. The third type of observation was the proportion of lung with opacities. By definition, the observers were in agreement if the recorded percentage of involvement of lung opacities was 20% or less. The observers evaluated 139 scans and made 12 observations at three levels for each scan for a total of 5004 observations per observer; observers differed substantially on 190 values (3.8%).

Pulmonary alveolar proteinosis can be understood as a syndrome of altered surfactant homeostasis, leading to a pathologic accumulation of surfactant. Because surfactant homeostasis is complex, there are many potential points of disruption. Pulmonary alveolar proteinosis can be a consequence of overproduction of phospholipids by type II pneumocytes, impairment of clearance of phospholipids by macrophages, or both [8, 9]. In some cases, the altered homeostasis is caused by another condition such as malignancy (leukemia or lymphoma), chemotherapy, infection, or AIDS rather than by a primary process [10]. Secondary pulmonary alveolar proteinosis usually responds to treatment of the underlying disease [10].

The clinical presentation of pulmonary alveolar proteinosis is nonspecific. A representative patient would be a man in his early 40s, but one of four patients is a woman and the condition affects patients of all ages [11]. Patients with pulmonary alveolar proteinosis have an insidious onset of symptoms that typically includes dyspnea, fatigue, and nonproductive cough, but patients with the condition can be asymptomatic [1, 10, 11]. The most frequent findings on physical examination are diffuse rales, cyanosis, and clubbing [1, 10, 11]. Most patients have restrictive lung disease with decreases in total lung capacity, forced vital capacity, forced expiratory volume in 1 sec, and diffusion capacity for carbon monoxide [1, 11].

Chest radiographs of patients with pulmonary alveolar proteinosis can vary in appearance. Chest radiographs can show diffuse or patchy bilateral air-space disease with central or peripheral distribution [9,10,11]. Occasionally, involvement is unilateral [9]. Interstitial disease can also be seen, especially in longstanding cases [1, 9]. Pulmonary involvement can worsen, remain stable, or spontaneously improve [1].

Previous studies have described numerous different patterns of pulmonary alveolar proteinosis on CT [2,3,4,5,6]. A patchy or geographic distribution has often been observed, but involvement was sometimes predominantly central or peripheral [2, 5]. In our experience the most common intraslice patterns were geographic or diffuse, and peripheral sparing was rare. However, the pattern of pulmonary involvement can be inconsistent in a given patient (Fig. 1).

Serial scans frequently show variations in severity and pattern of disease [2]; there can be either generalized worsening or improvement or some areas of involvement can improve while others worsen. The condition can resolve spontaneously or can progress relentlessly, sometimes despite bronchoalveolar lavage. In this group of patients, changes in pattern of distribution and severity of disease over time were more common, but some patients had stable involvement over prolonged periods.

Other CT studies of pulmonary alveolar proteinosis have not identified a specific zonal distribution [10]. Our most common observation on thin-section images was a uniform zonal distribution (63%) (Fig. 2A,2B,2C). However, a lower zone predominance was present in 22% (Fig. 3A,3B,3C). The lower zone predominance is related to the predilection of interlobular opacities to involve the lung bases. Enlarged lymphatic ducts at the lung bases, where the lymphatics are more developed, could account for this gradation.

Disease involvement can range from air-space consolidation to ground-glass opacities to interstitial disease [2,3,4,5,6]. Although a single type of opacity can dominate the CT appearance of pulmonary alveolar proteinosis, more commonly there is a combination of types (Figs. 2A,2B,2C,3A,3B,3C,4). Intralobular opacities are an unusual manifestation of interstitial involvement that can be observed occasionally in pulmonary alveolar proteinosis (Fig. 5A,5B). Regions of emphysema were not commonly observed in patients with this condition.

The type of opacity that is observed in pulmonary alveolar proteinosis is partly a function of slice thickness. Often, the use of thin-section imaging can reveal interlobular opacities that are exhibited as ground-glass opacities on thicker sections. Also, air-space opacities on thick-section images can appear as ground-glass opacities on thin-section images.

A small number of cases of clinically important fibrosis associated with pulmonary alveolar proteinosis have been reported [1, 12]. The infrequency of these cases implies either that fibrosis is a coincidental finding or that only a small number of cases of pulmonary alveolar proteinosis progress to fibrosis (Fig. 6A,6B). Although a total of eight patients (30%) in our study group had some element of fibrosis, only two (7%) had substantial regions of fibrosis. Both had fibrosis in 70% or more of a lung at some level. Two patients had small areas of involvement, and four had minimal or questionable involvement. Fibrosis was not an important part of the disease in six of these patients.

In assessing the intraslice pattern and zonal distribution of disease and the proportions of involvement of the lung by various opacities, the two radiologist observers occasionally disagreed. Overall, there was good general agreement in all categories. Some images were more difficult to assess than other images. Some CT images were technically limited. In some patients, the extensive nature of disease made classifying the opacities and determining the percentage of involvement of each type of opacity more difficult. On retrospective review, the observers had no difficulty in reaching consensus on any point.

Many patients with pulmonary alveolar proteinosis have a crazy paving pattern of ground-glass opacities or air-space opacities on a background of smooth interlobular thickening in a geographic distribution. This crazy paving pattern has been described as part of a characteristic appearance of pulmonary alveolar proteinosis [4] (Fig. 5A,5B). Although this pattern is suggestive of the diagnosis of pulmonary alveolar proteinosis, this pattern can be seen in several air-space and interstitial diseases [13]. However, the clinical features of these other diseases are often more defined and thus unlikely to be confused with pulmonary alveolar proteinosis [13]. Several other diseases can simulate pulmonary alveolar proteinosis, including pulmonary edema, alveolar hemorrhage, hypersensitivity pneumonitis, and bronchioalveolar cell carcinoma [1, 2]. If CT findings in pulmonary alveolar proteinosis are interpreted in the proper clinical context, a radiologist can often suggest the diagnosis.

Pulmonary alveolar proteinosis is often a diagnostic dilemma. When patients with this condition present, the clinical differential diagnosis can be broad. One of the advantages CT has over chest radiography is its ability to narrow the differential diagnosis or even suggest a diagnosis other than pulmonary alveolar proteinosis. Ideally, a baseline CT scan is helpful for classifying disease distribution and severity. Once pulmonary alveolar proteinosis is considered or if bronchoscopic or open-lung biopsy is contemplated, CT can be used to guide the bronchoscopist or thoracic surgeon to a region of greater involvement.

In patients with pulmonary alveolar proteinosis, radiographic detection of superimposed conditions such as pneumonia, edema, or nodules can be difficult. When a patient with known pulmonary alveolar proteinosis becomes more dyspneic, develops more rales, or shows laboratory evidence of worsening gas exchange, CT can help differentiate progressive pulmonary alveolar proteinosis from superimposed disease. Chest radiographic findings that are suggestive of possible complicating disease could also warrant a CT scan.

Chest radiographs are routinely obtained for follow-up after lavage. However, chest radiographs are insensitive for detecting small amounts of disease, and a chest CT scan can establish the severity of disease remaining after lavage. Also, chest radiographs of the occasional patient who has a poor response to bronchoalveolar lavage can be difficult to interpret. CT scans can help differentiate pleural effusion, atelectasis, edema, or pneumonia in these complicated postlavage cases.

After treatment of pulmonary alveolar proteinosis with whole-lung lavage, marked improvement in disease is usually apparent (Fig. 7A,7B). Once a substantial amount of abnormal surfactant has been removed, the alveoli reexpand. Improvement in interlobular opacities presumably reflects clearance of edema and lymphocytes from the interstitium. Pleural effusion is not ordinarily seen in untreated pulmonary alveolar proteinosis but can be seen after whole-lung lavage, usually with resolution in 24-48 hr [1]. In an uncomplicated lavage, pulmonary opacities usually completely or almost completely resolve. Patients might not undergo scanning after lavage because they show substantial clinical improvement.

Lymph node enlargement is uncommon in patients with pulmonary alveolar proteinosis. Identification of lymph node enlargement or lung nodules should prompt consideration of underlying malignancy or infection. Finding one or two mildly enlarged lymph nodes on a CT scan is a common occurrence in patients scanned for reasons unrelated to primary chest disease. It is not surprising that some patients had mildly enlarged lymph nodes as an incidental finding. In two of our patients, more extensive lymph node involvement was related to infection.

This study was subject to some limitations related to its retrospective design. Patients with more severe involvement with this condition were more likely to be included because patients were usually referred for clarification of a clinical problem. The change in opacities after lavage could be skewed toward patients who had a less than optimal response to bronchoalveolar lavage because CT examinations were more likely to be performed in patients who failed to show complete clearing after lavage. Despite these inherent biases, the sample size was substantially larger than those described in previous studies and the patients exhibited a wide variety of CT presentations of pulmonary alveolar proteinosis.

An evaluation of a possible trend of interstitial disease progressing to fibrosis was limited by the retrospective experimental design. Patients were typically scanned when their clinical symptoms had progressed, often before lavage. Some, but not all, patients underwent CT before and after lavage. Most patients who underwent whole-lung lavage underwent chest radiography instead of CT after lavage. These radiographs typically showed complete resolution of disease and no evidence of fibrosis, suggesting that no trend toward fibrosis was present. However, a prospective study of scans obtained at regular prescribed intervals would be better suited for evaluating a possible tendency of interstitial disease to progress to fibrosis in patients with pulmonary alveolar proteinosis.

The diagnosis of pulmonary alveolar proteinosis is made by pathologic examination of material obtained from open-lung or transbronchial biopsy. Recent advances now allow diagnosis of pulmonary alveolar proteinosis by examination of lavage fluid [14]. The pathologic appearance of pulmonary alveolar proteinosis was first described by Rosen et al. [15] in 1958 as “a granular and floccular acidophilic material, which filled large groups of alveoli, with minimal or no changes in the interalveolar septa.” Although septal involvement was not identified in this initial description, subsequent investigations have detected edema, cellular infiltration, or fibrosis in the interlobular septa [12, 16,17,18].

Primary pulmonary alveolar proteinosis has a variable natural history. In one group of patients, 24% had spontaneous remission of disease [19]. Of the patients who completed long-term follow-up, 79% responded favorably to lung lavage [19]. If sufficient material accumulates to cause dyspnea or hypoxemia, the treatment of choice for pulmonary alveolar proteinosis is whole-lung lavage [11]. In some patients a single treatment is sufficient; other patients require repeated whole-lung lavage. Whole-lung lavage therapy has substantially improved patient morbidity and mortality. Formerly, Nocardia pneumonia or other pneumonias often occurred in patients with untreated pulmonary alveolar proteinosis [2], but now these complications are relatively infrequent [9, 19].

In summary, although the term “pulmonary alveolar proteinosis” suggests that the condition would present with air-space disease, the CT pattern in pulmonary alveolar proteinosis can range from air-space disease to ground-glass opacity to interstitial disease. Typically, patients with pulmonary alveolar proteinosis present with a combination of ground-glass and interlobular opacities, often in a crazy paving pattern with geographic involvement. Although a crazy paving pattern can be associated with other diseases, other conditions often have characteristic presentations that are unlikely to be confused with pulmonary alveolar proteinosis.

In a small number of patients, pulmonary fibrosis is the dominant feature. Large focal areas of air-space disease are unusual in uncomplicated pulmonary alveolar proteinosis and should raise the question of superimposed infection. Bulky lymph node enlargement is an uncommon finding in pulmonary alveolar proteinosis and should prompt further investigation for possible infection or malignancy. Chest CT scans obtained after lavage typically show substantial improvement in the extent of disease, particularly with a decrease in ground-glass and interlobular opacities.

Address correspondence to J. M. Holbert.