Acquired PAP (the most common type, see Pathogenesis and Classification subsequently here) usually presents as progressive dyspnea of gradual onset, at times associated with a minimally productive cough or fatigue (2, 7). Other variably associated features may include weight loss and low-grade fever, although marked fever usually indicates the presence of a complicating infection. Physical examination is often normal or reveals relatively minor and nonspecific pulmonary findings, and digital clubbing is uncommon (9, 35). Described biochemical abnormalities may include elevated serum levels of lactate dehydrogenase (LDH), other protein products of pulmonary epithelial cells, including carcinoembryonic antigen (36), cytokeratin 19 (37), and the mucin KL-6 (38, 39), and levels of the SP-A, SP-B, and SP-D (40, 41), although none of these findings are specific for PAP.

From an analysis of 410 published cases (Table 1)

Serum immunoglobulin levels have been reported to be reduced in 4% of patients tested (44, 48, 49); however, two of these were siblings with immunoglobulin A deficiency (49), and there was also a single instance of a low-level immunoglobulin M paraprotein without proven hematologic malignancy (15). Elevated cholesterol levels have been described in 19% of the patients tested. There were 52 cases with data provided for absolute neutrophil counts, and all but one were normal; one patient had mild neutropenia of 1.04 × 10⁹ per L (43). Bone marrow examinations were reported in 13 patients, and these showed no definite abnormalities other than erythroid hyperplasia attributable to chronic hypoxia (17, 43, 44, 50–58).

A number of these clinical features resemble those seen among patients with another autoantibody-mediated systemic disease with pulmonary manifestations, Goodpasture's syndrome (antiglomerular basement membrane disease). In this disorder, there is also a marked male predominance (approximately 6:1) (59) and a high proportion of smokers (up to 80% [60, 61]). However, in contrast to PAP, there is a bimodal age peak seen in the incidence of Goodpasture's syndrome (59). In Goodpasture's syndrome, the target antigen is a domain of the α3 chain of type IV collagen, preferentially contained within glomerular and pulmonary alveolar basement membrane. The predominance of smokers suggests that inhaled toxins may either expose or lead to conformational alterations of the type IV collagen, rendering these immunogenic, or may alternatively increase the permeability of lung capillaries, allowing access of preformed antibodies to the alveolar basement membrane (60). Similar hypothetical models can be proposed for the role of smoking in the pathogenesis of PAP, but the anatomic localization of GM-CSF is far less restricted. These similarities between PAP and Goodpasture's syndrome suggest other possible areas for future investigation, including the possible HLA associations with the development of PAP (62), the role of antibody titers in prognosis and therapeutic monitoring (63), and the possible utility of plasmapheresis and immunosuppressive therapies targeting antibody production (59).

Excluding patients younger than 10 years of age who were effectively nonsmokers by allocation, disease and demographic features have been reported for 121 smoking patients and 35 nonsmoking patients. Apart from the gender imbalance described previously here and a tendency for smokers to more often be reported from North America (66% of smokers versus 49% of nonsmokers, p = 0.08), there were no differences in age, duration of symptoms, hemoglobin, serum level of LDH, Pa_O2, or alveolar–arterial oxygen gradient ([A–a]Do₂) evident between these two groups (each p ⩾ 0.2).

Age at diagnosis is approximately normally distributed with a mean ± SD of 37.8 ± 13.3 years (Figure 1)

Figure 1. Histogram of the age at diagnosis for published cases of acquired PAP. Data are grouped into 5-year intervals and are shown with a superimposed comparative normal distribution curve of identical mean and SD (mean = 37.8, SD = 13.3, n = 408).

Download Figure | Download Powerpoint

[More] [Minimize]

The distribution of age at diagnosis among males (mean ± SD, 39.6 ± 12.3 years) closely resembled that of the population as a whole. However, there were notable differences in the pattern of age at diagnosis for females (Figure 2)

Figure 2. Histogram of the age at diagnosis for published cases of acquired PAP in females. Data are grouped into 5-year intervals and are shown with a superimposed comparative normal distribution curve of identical mean and SD (mean = 33.7, SD = 14.8, n = 109).

Download Figure | Download Powerpoint

[More] [Minimize]

The mean ± SD partial Pa_O2 at diagnosis was 58.6 ± 15.8 mm Hg (Figure 3)

Figure 3. Arterial Po₂ measurements at the time of diagnosis of acquired PAP.

Download Figure | Download Powerpoint

[More] [Minimize]

Data on LDH values and applicable normal ranges were only reported in 36 cases (36, 43, 49, 56, 66–84) and are expressed as a percentage of the upper limit of the applicable normal reference range. PAP patients had a mean ± SD LDH level that was 168 ± 66% of the upper limit of the normal range. There was no difference in LDH level according to gender (p = 0.91), nor was there any correlation between LDH and age at diagnosis (p = 0.75). Serial measurements of serum LDH levels in individual cases have suggested that the level of this enzyme may be useful as an indicator of disease severity (70, 85). There were 27 patients reported who had values for both serum LDH and concurrent Pa_O2 (in 24 cases the [A–a]Do₂ was also known). LDH and Pa_O2 were moderately correlated (r² = 0.372, p = 0.001), but the correlation between LDH and [A–a]Do₂ was more significant (r² = 0.489, p = 0.0001) (Figure 4)

Figure 4. Correlation with superimposed regression line (r ² = 0.489, p = 0.0001) between the serum level of LDH and the alveolar–arterial oxygen gradient among 24 patients with acquired PAP where concurrent values were reported.

Download Figure | Download Powerpoint

[More] [Minimize]

On pulmonary function testing, the most common pattern seen is that of a restrictive defect, with a disproportionate reduction in diffusing capacity relative to modest impairment of vital capacity (2, 7, 9, 12). By plain chest radiograph, widespread bilateral patchy and asymmetrical airspace consolidation may be seen, without predilection for central or peripheral distributions, but many patterns are possible (9). The extent of radiologic abnormalities is frequently disproportionate to the relatively modest pulmonary symptoms and physical findings. High-resolution computed tomography scanning has a characteristic appearance of patchy or geographic air-space “ground-glass” opacities or consolidation with some thickening of the interlobular septa, resulting in a “crazy-paving” pattern (12, 13) (Figure 5)

Figure 5. Thin-section, high-resolution computed tomography scan of the thorax showing features typical of acquired PAP. Extensive opacity involves most of the lungs. The areas of ground-glass opacity demonstrate thickening of the interlobular septa leading to the characteristic “crazy paving” appearance typical of PAP.

Download Figure | Download Powerpoint

[More] [Minimize]

More than 90% of all cases of PAP occur as a primary acquired disorder of unknown etiology (2, 4, 15, 96), not associated with any familial predisposition. The first step toward understanding the possible pathogenesis came unexpectedly in 1994 from the field of experimental hematology through the development of gene-knockout mice lacking the hematopoietic growth factor GM-CSF (107, 108).

GM-CSF had been chemically purified in the late 1970s (109) and in 1984 was one of the first human cytokines to be cloned (110). GM-CSF became an intense focus of investigation through this period because of its potent capacity to stimulate the proliferation and differentiation of neutrophilic and monocyte/macrophage lineage hematopoietic cells in vitro, an action that had remained unexplained since its first recognition in 1964 (111). This capacity provides the basis for the clinical application of GM-CSF (112–114). GM-CSF shares some, but not all, of its actions with granulocyte colony-stimulating factor, the other major neutrophilic hematopoietic regulator currently in clinical usage (112, 113). The pharmacologic administration of recombinant GM-CSF consistently leads to a dose-dependent stimulation of myeloid hematopoiesis resulting in peripheral blood neutrophilia, monocytosis, and eosinophilia (112). Each of these actions requires engagement of GM-CSF with its high-affinity receptor complex, which comprises a GM-CSF–specific α chain and a common β chain (β_c). In both the human and the mouse, this β_c is also a component of the receptor complexes for interleukin-3 and interleukin-5 (115) and is expressed on both alveolar macrophages and alveolar type II epithelial cells (116). Pulmonary epithelial cells also produce GM-CSF (34).

To explore the innate physiologic role of GM-CSF, investigators used gene-targeting methods to generate mice lacking either GM-CSF (GM^−/−) (107, 108) or β_c (β_c^−/−) (117, 118). Surprisingly, these animals had no detectable abnormality of steady-state hematopoiesis but had impaired surfactant clearance by alveolar macrophages (119–121), leading to a condition that resembled human PAP, including a prominent lymphocytic infiltrate. In contrast to the diminished rate of surfactant clearance, the rates of synthesis of surfactant phospholipid and proteins were unperturbed (119–121). This abnormality of surfactant clearance in GM^−/− mice could be corrected by the local delivery of GM-CSF and did not require a systemic effect (122–125). Although type II alveolar epithelial cells are responsible for synthesis, secretion, and recycling of all surfactant components (126, 127), surfactant catabolism involves approximately equal contributions from both type II cells and alveolar macrophages (34, 128). Although current evidence suggests that the primary defect in surfactant clearance in the absence of GM-CSF activity is alveolar macrophage dysfunction (119, 121, 129), alternative non-GM-CSF–dependent surfactant clearance pathways, perhaps involving type II cells, may explain the less severe surfactant accumulation in β_c^−/− compared with GM^−/− mice (120). These conclusions are based on the demonstration of impaired surfactant catabolic capacity in isolated GM^−/− alveolar macrophages (121) together with complete correction of the PAP phenotype in GM^−/−, but not β_c^−/− mice, by bone marrow transplantation, which would not be expected to influence any type II cell defect (129, 130). Other incompletely investigated influences on surfactant homeostasis that may modulate the severity of PAP include the levels of alveolar SP-D (131–133) and activity of the transcription factor PU.1, through which the GM-CSF signal is transduced (134). Thorough overviews of the metabolic profiles of GM^−/− and β_c^−/− mice have been published recently (34, 120, 121, 135).

Although some naturally occurring immunodeficient mouse strains such as CB.17 scid/scid and the beige mouse have been reported to develop PAP (136, 137), in contrast to the GM^−/− and β_c^−/− animals, the penetrance of this phenotype is low, and the lung abnormalities occur late in the animals' life. To date, the cellular mechanisms for these observations have not been elucidated, and their relevance to the human condition of PAP remains unclear.

In addition to the PAP, GM^−/− mice manifest a number of more subtle, but important, extrapulmonary abnormalities. These include disturbed macrophage function (138–140), a propensity to develop systemic infections (141, 142), reduced fertility and impaired ovarian follicle maturation (141, 143–145), T-cell dysfunction (146, 147), a reduced number of cutaneous Langerhans cells (148), and ultimately reduced survival (141).

These observations provided an impetus to explore the possible existence of defects in either GM-CSF or its receptor and the potential therapeutic activity of exogenous GM-CSF in patients with PAP. A clinical study of GM-CSF therapy that began in August 1995 provided two important early observations (14). First, in contrast to expectations, GM-CSF administration did not result in any increase in peripheral blood neutrophil or monocyte counts, the only apparent hematopoietic response being a mild eosinophilia (149). Second, there was apparent improvement in the severity of PAP in the first treated patient (14). This attenuated hematopoietic response to GM-CSF has subsequently been observed in all patients with confirmed PAP treated by our own group (41), investigators at the Cleveland Clinic (150, 151), and other groups (152) (Ana Romero, personal communication, July 2001).

A series of experiments performed by Nakata and colleagues in Tokyo provided the likely explanation for this attenuated hematopoietic response to GM-CSF. They identified a GM-CSF–neutralizing autoantibody in the serum and BAL fluid of patients with acquired PAP that was not present in patients with congenital or secondary PAP (153–155). Other investigators have confirmed these findings (156). No clear defects in either the GM-CSF β_c receptor (41, 157) or GM-CSF gene sequence itself (157) have yet been identified in patients with acquired PAP. Although a study of two patients suggested normal basal GM-CSF secretion by alveolar macrophages (158), there have been some in vitro observations of impaired responsiveness to GM-CSF (41) or reduced GM-CSF secretion in response to various agonists (156, 159, 160), but these results are difficult to interpret in light of the likely presence of neutralizing GM-CSF antibody and the use of antigenic assays such as enzyme-linked immunosorbent assay rather than functional assays specific for biologically active GM-CSF. On present evidence, it is likely that the anti–GM-CSF antibody is pathogenic in the development of the disease through its ability to inhibit the activity of endogenous GM-CSF, leading to a state of functional GM-CSF deficiency, recapitulating the findings in the GM^−/− mouse.

Antibodies capable of binding, and in some cases neutralizing, GM-CSF have been reported to occur in low titer in some immunocompetent patients treated with recombinant GM-CSF (161–163). The period of persistence of these therapy-induced antibodies and any possible in vivo activity remains unclear. In the absence of exposure to recombinant GM-CSF, spontaneous GM-CSF binding antibodies are far less common, being detected in only 4 of 1258 (0.3%) healthy volunteers (164, 165). There is a case report of the association of a GM-CSF–neutralizing immunoglobulin G antibody in a patient with acquired amegakaryocytic thrombocytopenia (166), but the causal relationship originally suggested is called into doubt by our current understanding of the redundancy of GM-CSF in steady-state hematopoiesis (107, 108). However, there does appear to be a higher frequency of GM-CSF–neutralizing antibodies among patients with autoimmune disorders (0.7%), specifically myasthenia gravis (1.9%), with at least one patient followed for more than 2 years without any pulmonary symptoms manifest (167).

This important discovery of GM-CSF–neutralizing antibodies in patients with PAP potentially also explains a number of earlier observations. It had been shown 20 years ago that the BAL fluid and serum from patients with PAP had “immunoinhibitory” activity in vitro, blocking the response of mononuclear cells to mitogens (168–170). Similarly, although high concentrations of surfactant are inhibitory to a number of macrophage functions (92), the observations of impaired phagocytic function (171, 172), chemotaxis (173), and microbial killing (174, 175) by alveolar macrophages derived from patients with PAP may be partly attributable to the actions of the anti–GM-CSF antibody. Furthermore, the systemic production of the antibody now provides an explanation for the recurrence of PAP following double-lung transplantation (176).

Although cases of PAP in infants had been sporadically reported in the 1950s and 1960s (15, 177–179), including cases in families with multiple affected siblings (177, 180), they were not considered initially to represent a process distinct from adult cases. Following the recognition of PAP as a rare cause of immediate-onset neonatal respiratory distress (181–183), the description in 1981 of a consanguineous family with four affected siblings (184) established “congenital” PAP as a distinct familial and likely genetic disorder.

It has been shown subsequently that most cases of congenital PAP are transmitted in an autosomal recessive manner (184, 185), most often caused by homozygosity for a frame shift mutation (121ins2) in the SP-B gene (186, 187), which leads to an unstable SP-B mRNA, reduced protein levels, and secondary disturbances of SP-C processing (188). The estimated gene frequency of the 121ins2 mutation is one per 1,000 to 3,000 persons in the United States (189). However, there is increasing recognition of molecular genetic heterogeneity among infants with congenital SP-B deficiency (187, 190–194), which may have phenotypic and prognostic correlations (195). Importantly, heterozygotes for the most commonly recognized SP-B mutation (121ins2) appear to have normal respiratory function into their 4th decade of life (196). If the mouse model of this condition is an accurate predictor of the human condition, such heterozygotes may be at risk for the development of reduced lung compliance and gas trapping with aging (197) and could be at increased risk of hyperoxic lung injury (198). It also has been recognized recently that mutations in SP-C can lead to similar forms of neonatal respiratory distress (199). In animal models, the deletion of SP-D also leads to accumulation of alveolar macrophages and increased surfactant pool size (131), but in both of these settings, the histopathology is distinct from that seen in congenital SP-B deficiency (200). Also, the SP-C processing defect of SP-B deficiency is not manifest in mice with SP-D deficiency (132).

There are a proportion of infants with the syndrome of neonatal onset PAP who do not have any recognized disturbance of SP-B expression, and abnormalities of the GM-CSF receptor β_c have been implicated in some of these cases (201). In a single report, four infants with congenital PAP were shown to have dramatically reduced β_c expression on peripheral blood mononuclear cells, greatly reduced binding of GM-CSF and impaired in vitro responsiveness to both GM-CSF and interleukin-3 (201). In one of these patients, a putative mutation in the GM-CSF β_c gene, which would lead to an amino acid substitution, was identified (201). To date, other investigators have been unable to identify any similar cases of β_c mutations (149, 202, 203), and five infants with congenital PAP in the absence of SP-B mutations have been treated with GM-CSF, all manifesting a normal hematopoietic response, which excludes the possibility of such a receptor defect (149). Conversely, preliminary findings from a British group have identified high levels of expression of a novel truncated form of β_c lacking the usual transmembrane domain, suggesting that this may function as a soluble inhibitory receptor (202). It is too early to be sure what proportion of otherwise unexplained cases of congenital PAP is due to defects in GM-CSF receptor expression, but these results clearly demonstrate that defects in this pathway can be associated with human disease states.

Although uncommon among adult PAP patients, there exist a number of recognized underlying causes for the secondary development of PAP. These conditions include lysinuric protein intolerance, acute silicosis and other inhalational syndromes, immunodeficiency disorders, and malignancies and hematopoietic disorders.

The rare genetic disorder “lysinuric protein intolerance” is attributable to a mutation in the “y+L amino acid transporter-1” gene (204–206), resulting in defective plasma membrane transport of dibasic amino acids and multisystem manifestations, including hematopoietic abnormalities leading eventually to PAP in a high proportion of the cases (207–210).

A rare acute-onset form of silicosis recognized in the 1930s (97, 211) was subsequently called “acute silico-proteinosis” to emphasize the histologic resemblance to PAP (98, 212, 213). This was associated with heavy short-term exposure to high concentrations of respirable free silica. With improved occupational health and safety standards, this condition has been reported rarely since the 1980s (42, 214, 215), except for cases of intentional inhalation of domestic scouring products (216). Very rarely, other inhaled environmental or industrial materials such as cement dust (217), cellulose fibers (218), aluminum dust (78), or titanium dioxide (219) have been associated with the development of PAP. Whether such associations are truly causal is not entirely clear in each of these circumstances.

Alveolar proteinosis also develops rarely as a complication of either underlying immunodeficiency disorders, such as thymic alymphoplasia (220), severe combined immunodeficiency disorder (221), or immunoglobulin A deficiency (49), or in the context of iatrogenic immunosuppression such as following solid-organ transplantation (222, 223). One patient has been reported to develop PAP in the setting of dermatomyositis, but this patient also had received prolonged steroid therapy (224). A single study has suggested that patients with the acquired immunodeficiency syndrome complicated by P. carinii pneumonia may have some features of PAP (225), although this observation has not been reproduced, and described cases of PAP associated with acquired immunodeficiency syndrome remain extremely rare (226, 227).

PAP also occurs in association with underlying malignancies, almost exclusively of hematopoietic origin (228–231). In these cases, the development of secondary PAP probably reflects numerical deficiency and/or functional impairment of alveolar macrophages, which are derived from blood monocytes (228, 232–234). In the myeloid leukemias and myelodysplastic syndromes, where secondary PAP is most commonly encountered, alveolar macrophages may be derived from the malignant clone itself and in some circumstances have been shown to carry specific defects that may explain the observed functional impairment of surfactant clearance (234). This suggestion is supported by resolution of the pulmonary process following restoration of normal hematopoietic function (230, 234). Similarly, there is some evidence implicating defects in GM-CSF signaling in patients with secondary PAP complicating acute myeloid leukemia. Three patients have been described where their leukemic cells lacked expression of β_c and were unresponsive to GM-CSF, with similar defects shown to be present in BAL-derived cells (234). These defects, together with the underlying PAP, were corrected in each of two cases where normal hematopoiesis was successfully re-established, suggesting that the leukemic clone with its impaired responsiveness to GM-CSF was responsible for replacing/displacing the functionally normal alveolar macrophages and potentially contributing to the manifestations of secondary PAP (234).

In the absence of a randomized trial or even a formal prospective study, the true impact of the advent of therapeutic lavage on the natural history of acquired PAP is difficult to ascertain. Furthermore, any comparison of survival rates for those patients who did, or did not, undergo lavage makes no allowance for the relative severity of the disease process itself, although it seems likely that those patients with more severe disease would have been more likely to undergo lavage. Nevertheless, from the analyzed literature, those patients who underwent lavage at any time during the course of their disease had a superior survival, with a 5-year actuarial survival rate ± SE from diagnosis of 94 ± 2% compared with 85 ± 5% for those not receiving such treatment (p = 0.04) (Figure 8)

Figure 8. Overall survival from the time of diagnosis of acquired PAP was significantly improved if patients had received therapeutic lavage at any time during their disease course (lavage, n = 146; no lavage, n = 85, p = 0.044).

Download Figure | Download Powerpoint

[More] [Minimize]

Patients who were treated with lavage were more likely to have been reported after 1969, reflecting increasing use of the newly described therapeutic procedure (Table 2)

In the literature cases, the interval between the diagnosis of PAP and the first application of therapeutic whole-lung lavage ranged from 0 (immediate lavage) to 210 months, with a median of 2 months (n = 92). The majority of patients who underwent lavage did so within 12 months of diagnosis (79%), but there was a continuing increase in the proportion of patients having received such therapy. In the era of availability of lavage after 1964, the likelihood of a patient with PAP remaining free from therapeutic lavage was only 37% at 5 years (Figure 9)

Figure 9. Proportion of patients surviving without receiving therapeutic lung lavage following the diagnosis of acquired PAP (37% [SE 4%] at 5 years).

Download Figure | Download Powerpoint

[More] [Minimize]

Among patients who had undergone therapeutic lavage, the median total number of procedures performed was two (range, 1 to 22), and the number of procedures performed was proportional to the duration of follow-up, such that 66% of patients followed for more than 1 year from diagnosis (median, 37 months) had required more than one lavage.

In 55 instances of reported response to lavage, there was information provided on the duration of benefit. The definition of disease recurrence varied between reports but included any of the following: the recurrence, or significant progression of respiratory symptoms attributable to PAP, or the application of further therapeutic interventions such as repeated lavage. Episodes of infection were not considered to represent disease recurrence. The median duration of clinical benefit from lavage was 15 months, with less than 20% of those patients followed beyond 3 years remaining free of recurrent PAP manifestations (Figure 10)

Figure 10. Duration of response to therapeutic lavage for patients with acquired PAP (median = 15 months; n = 55).

Download Figure | Download Powerpoint

[More] [Minimize]

Comparing the demographic and disease-related features of patients who did (n = 92) or did not (n = 18) respond to therapeutic lavage, there were no differences seen in gender, region of origin, duration of symptoms, smoking status, time from diagnosis to lavage, [A–a]Do₂, serum LDH, or year of publication (each p ⩾ 0.3). There was a tendency for nonresponding patients to be younger (median 35 versus 39 years, p = 0.1). When response rates to lavage were calculated within cohorts for age at diagnosis (20 years or less, 21–39 years, and 40 years or more), there was a significant difference observed: 58% (7 of 12), 84% (42 of 50), and 90% (43 of 48), respectively (p = 0.03).

To reduce “lead-time bias” associated with those patients who survived long enough to undergo therapeutic lavage many years after diagnosis, patients who underwent lavage more than 12 months after diagnosis were excluded from this analysis. Although there was a greater early mortality among the nonresponders (survival at 6 months was 87 ± 9% versus 97 ± 2%), overall survival did not differ significantly (actuarial 5-year survival rate 87 ± 9% versus 92 ± 4%, p = 0.3).

There were 47 patients reported with paired data prelavage and postlavage for Pa_O2, [A–a]Do₂, FEV_1.0, vital capacity, or diffusion capacity for carbon monoxide. A favorable therapeutic response was claimed for 85% of these patients. Overall, following therapeutic lavage there was a clear and significant improvement in all parameters analyzed (Figure 11

Figure 11. Paired prelavage and postlavage pulmonary function data from patients with acquired PAP. The p values shown are for the comparison of prelavage versus postlavage data for individual patients for each parameter using a two-sample t test. The heavy horizontal lines show the mean values. The number of evaluable patients for each parameter were: Pa_O2 (41), [A–a]Do₂ (21), FEV₁ (33), vital capacity (40), and diffusion capacity for carbon monoxide (25).

Download Figure | Download Powerpoint

[More] [Minimize]

The degree of change of each of the previously mentioned lung function parameters following lavage was not closely correlated. For example, there was no significant correlation between the increment in [A–a]Do₂ and either Δ diffusion capacity for carbon monoxide (r² = 0.011, p = 0.8) or Δ vital capacity (r² = 0.074, p = 0.3), nor was there a significant correlation between the Δvital capacity and Δ diffusion capacity for carbon monoxide (r² = 0.059, p = 0.2). One report also measured changes in pulmonary shunt fraction following lavage in 14 patients, and this improved from a mean ± SE of 20 ± 1% to 11 ± 1% (p < 0.001) (249).

Although the number of patients evaluable for some parameters was very small, those patients described in the original reports as “responders” had numerically greater improvements in each of the parameters analyzed than “nonresponders,” supporting the validity of the self-reported response categories (each p ⩽ 0.08). The median changes for “responding” patients were Pa_O2 +20.5 mm Hg (n = 34), [A–a]Do₂ −33.0 mm Hg (n = 17), FEV_1.0 +0.21 L (n = 27), vital capacity +0.52 L (n = 33), and diffusion capacity for carbon monoxide +4.5 ml/min·mm Hg (n = 20).

In addition to the previously mentioned efficacy data derived from published reports with individual patient data, there are a small number of single institutional reports that have evaluated the efficacy of lavage without providing individual patient data.

Through the period 1966 to 1976, Dr. Wasserman had performed therapeutic lavage on 19 patients, with efficacy data available from 11 of these (6). Reporting all results as median values for the evaluable patients, vital capacity improved from 78 to 89% of predicted normal values, diffusion capacity for carbon monoxide from 44% of predicted to 71%, and Pa_O2 from 66 to 85 mm Hg (each p ⩽ 0.005). In a subsequent publication from the same institution evaluating 21 patients (96), the median time to repeated lavage was 22 months, and after 5 years, 62% (13%) of patients had required a repeated procedure.

Through the period 1957 to 1983, a total of 29 patients with acquired PAP were seen, and 21 of these underwent some form of lavage procedure (8 transtracheal and 13 using more recent isolated whole-lung procedures) (4). Only three of these 21 patients (14%) were described as not obtaining any response to the lavage (type not specified). A major weakness in this report is that “postlavage” studies were performed many months after the lavage procedure, such that the mean interval between preprocedure and postprocedure tests was 30 ± 21 months (range, 2–135).

A review of 24 patients was recently published (2), 13 (54%) of whom underwent lavage on a median of two occasions (range, 1–8). There were limited objective response data presented for the first lavage, with the median vital capacity for three evaluable patients being unimproved, 71% and 66% of predicted, prelavage, and postlavage, respectively.

There was no difference in overall survival from the date of diagnosis according to gender (5-year survival rate 74% for males and 76% for females, p > 0.5). Also, there was no difference among women according to age, with a 5-year survival rate of 80 ± 6% for those aged above the median for women of 34 years and 71 ± 9% for those aged 34 years or less (p = 0.4).

However, within the entire cohort, using the median age of the entire group (39 years) as a cut point, there was evidence of inferior outcome for older patients (p = 0.006), attributable to both a greater direct mortality from PAP and a greater incidence of death from incidental causes. The risk of death from opportunistic infections did not differ appreciably between the two age groups. Excluding incidental causes of death, the magnitude of the survival difference (actuarial rates of 84 ± 3% versus 70 ± 5% at 5 years, p = 0.055) was reduced.

Given the higher than expected incidence of cases among young children less than 5 years of age, this subgroup was examined, revealing a poor outlook. The actuarial 5-year survival rate was 14 ± 13%, with just one of the seven patients surviving beyond 10 months.

There were just seven patients diagnosed at less than 5 years of age in the period before 1998. The adverse outlook in this age group appeared to be attributable in part to the reduced efficacy of therapeutic lavage. There were five such children diagnosed during the era of lavage availability (44, 177, 201, 292, 293), and three were treated (one repeatedly) without any evident benefit (44, 177, 293). Recent reports confirm this adverse experience (294, 295). This diminished effectiveness of therapeutic lavage mirrors that reported for patients with congenital disease where lavage also has limited efficacy (181, 185, 201, 296, 297) and suggests that some of these cases may have been misdiagnosed forms of congenital PAP. Some of these infants were also treated with corticosteroids, N-acetylcysteine, and ambroxol without any evidence of benefit.

Within the limitations of the small number of patients available in published case reports, there are similarities in the clinical presentation of apparent PAP in children below the age of 5 years with those of congenital disease. Four of the reported cases described prominent gastrointestinal symptoms at presentation (food intolerance, failure to thrive, progressive weight loss, and vomiting) (44, 177, 287, 293). Also, one infant presenting at the age of 4 years had a sibling who died at the age of 2 and a half months from a respiratory illness labeled as “viral pneumonia” without a lung biopsy or autopsy being performed, and another family had two affected children, suggesting a possible familial basis (293, 295). The majority of these cases were reported before the recognition of SP-B deficiency as a cause of congenital PAP, although in one instance, this diagnosis was excluded by the finding of SP-B in BAL fluid (201) and another by normal polymerase chain reaction for SP-B mRNA (295).

There were 103 patients with available survival data who were smokers at the time of onset of symptoms (including three who then ceased prior to the confirmation of the diagnosis) and 24 who were nonsmokers at the first onset of symptoms (including six who had ceased between 1 and 10 years earlier). The actuarial 5-year survival rate from the time of diagnosis was identical at 82% for these two groups. There are insufficient data available to comment on any possible influence of continued smoking on infection risk or survival.

In the 159 patients where arterial oxygenation data were reported, Pa_O2 at presentation, using either the median of the cohort (58 mm Hg) or quartiles, was not associated with any differences in disease-specific survival (p ⩾ 0.2). However, there was a trend for an improved outcome among patients with an [A–a]Do₂ below the median of 50 mm Hg, where 5-year actuarial survival rates were 98 ± 2% compared with 90 ± 4% for those 50 mm Hg or more (p = 0.09).

Serum LDH was not predictive of survival, whether expressed as a percentage of the upper limit of normal or simply as elevated compared with normal.

Grouping patients in 10-year cohorts by year of initial publication, there has been a dramatic and consistent improvement in survival with time (trend p < 0.0001). The 5-year actuarial survival rates for patients reported in the years 1958–1967, 1968–1977, 1978–1987, and 1988–1997 are 52%, 72%, 93%, and 100%, respectively. This improvement, at least in part, is due to a major reduction in the number of patients diagnosed at autopsy. However, the same trend is evident using disease-specific survival and restricting the analysis to patients diagnosed antemortem, with 5-year actuarial survival rates of 77 ± 6%, 91 ± 4%, 96 ± 3%, and 100%, respectively (trend p = 0.0002).

There was no statistically significant difference observed between cases reported from North America (n = 140), Europe (n = 41), or Asia (n = 15); 5-year actuarial survival rates were 89 ± 3%, 87 ± 5%, and 100%, respectively (p = 0.4).

Analysis of 241 published articles describing over 400 separate individual cases of PAP has revealed a number of important features of this rare lung condition. The great majority of PAP cases are of the acquired variety occurring in adults. From these reports, a “typical patient” was a male smoker aged 30 to 50 years, and females aged 25–40 are under-represented. PAP is also a congenital disorder in a minority of cases, whereas other cases are secondary to other conditions, notably hematologic malignancy.

A number of important milestones mark the key advances of medical insight into PAP. Early among these was the 1965 description of whole-lung lavage, which remains today's standard therapy. Within 5 years from diagnosis, almost two-thirds of PAP patients had received such lavage, with more than 80% attaining significant, but transient, benefit.

More recently, major advances since 1994 have resulted in the discovery of animal models of PAP based on deletion of genes for GM-CSF itself or the GM-CSF receptor. Thus deficiency of GM-CSF has become a strongly suspected pathogenetic mechanism for adult acquired PAP, and GM-CSF administration with therapeutic intent has been evaluated in several small studies with benefit in approximately 50% of all patients so far treated. In the last 2 years, the demonstration of neutralizing anti–GM-CSF antibodies in all cases of acquired PAP has suggested that the disease may have an autoimmune pathogenesis, with the pulmonary disease resulting from a blockade of endogenous GM-CSF action. That GM-CSF therapy may reverse both pulmonary and extrapulmonary abnormalities found to be present in PAP suggests that these blocking antibodies, even if pathogenic, may in some cases be overcome or circumvented by pharmacologic administration of GM-CSF.

Spontaneous clinical resolution clearly occurs in a small minority of adult PAP patients, although the available evidence makes the existence of spontaneous cure less certain. There are no simple biochemical or clinical parameters so far described that are of use as prognostic variables applicable to most patients, although serum LDH levels correlate with the degree of impairment in oxygenation. Fortunately, only 10–15% of PAP patients may die directly from PAP-induced pulmonary failure. This mortality risk is perhaps less among those diagnosed in the last decade, probably because the prognosis of PAP patients with respiratory impairment is improved by whole-lung lavage. Those rare children aged less than 5 years diagnosed with apparently acquired PAP have a very adverse prognosis, and these cases may represent atypical manifestations of congenital PAP.

Much remains to be learned about this fascinating disease, and it is likely that we are on the threshold of a new era in the understanding of PAP. With the recent recognition of its possible autoimmune nature, it is to be hoped that improved pathophysiologic insights will translate rapidly into targeted therapeutic strategies, which may render redundant the physical removal by lung lavage of accumulated surfactant.