INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In 1958, Rosen and colleagues described a unique lung disorder characterized by excessive surfactant accumulation within the alveoli, which they called alveolar proteinosis (AP) (1). There are three apparently distinct forms of AP. The rare congenital form is transmitted in an autosomal recessive manner, most often through homozygosity for a frameshift mutation in the surfactant protein (SP)-B gene (2). AP also rarely complicates an underlying malignancy, usually of hematopoietic origin, which may reflect numerical deficiency and/or functional impairment of alveolar macrophages (AM) (3, 4). However, more than 90% of cases of AP represent a primary acquired disorder without familial predisposition, and remain of idiopathic etiology (1, 5, 6).

The only known effective therapy for AP is the periodic physical removal of accumulated surfactant through whole-lung lavage (5). Although never evaluated in a prospective study, lavage usually provides temporary symptomatic benefit, but requires prolonged general anesthesia, is complex to perform, is associated with potential morbidity, and fails to correct the primary defect in AP (5). More effective and less invasive therapies are required, but their development has been hampered by a poor understanding of the underlying pathophysiologic process in AP.

Gene-targeted mice lacking the hematopoietic cytokine granulocyte-macrophage colony-stimulating factor (GM-CSF) have impaired surfactant clearance, leading to AP (11, 12), which can be corrected by the delivery of GM-CSF through various means (13). Given the striking similarity between the pulmonary disease in mice lacking GM-CSF and human AP, we explored the hypothesis that the administration of GM-CSF may enhance AM function and thereby accelerate surfactant clearance in patients with idiopathic acquired AP (17).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Eligibility

The study was conducted from August 1995 to September 1998. Eligibility required a diagnosis of acquired AP confirmed by central pathologic review, symptoms attributable to AP, and the absence of active pulmonary infection. Patients were ineligible if they had secondary AP or had been exposed to cytotoxic agents, systemic or inhaled corticosteroids, or cytokines within 4 wk before screening. The study was conducted in accordance with the Declaration of Helsinki and approved by the institutional ethics committees of the participating institutions, and all patients provided written informed consent.

Prestudy Evaluation

In the 4 wk before commencement of the study, disease had to have shown no evidence of improvement, as assessed with a minimum of two arterial blood gas (ABG) samples and through reported pulmonary symptoms. Investigations required within 14 d of study entry included a full blood examination and assays of liver enzymes, serum lactate dehydrogenase (LDH), plasma levels of SP-A, and SP-B (samples stored and assayed centrally) (18), thoracic computed tomography (CT), and exercise testing either through bicycle ergometry or walk testing (19, 20). Within 24 h before treatment was begun, repeat ABG measurements and spirometry with measurement of the diffusing capacity of carbon monoxide (DL_CO) were performed according to institutional protocols (21, 22). GM-CSF-neutralizing autoantibodies were retrospectively assayed by Western blotting in pretreatment sera that had been stored at 70° C (23, 24).

GM-CSF Administration

Patients treated in Australasia and Europe (n = 11) received bacterially-synthesized recombinant GM-CSF (Leucomax; Schering-Plough, Baulkham Hills, Australia), with a specific activity of 0.66 to 1.66 × 10⁷ IU/mg protein and  25 Eu endotoxin/vial by the Limulus amoebocyte lysate assay, [data on file, Schering-Plough]), and patients treated in the United States (n = 4) received yeast-derived recombinant GM-CSF (Leukine; Immunex, Seattle, WA) with a specific activity 5.6 × 10⁶ IU/mg protein and  25 Eu endotoxin/vial [data on file, Immunex]). Administration of GM-CSF began at 3.0 µg/kg subcutaneously daily for 5 d, and increased to 5.0 µg/kg/d from Day 6 onwards. Patients self-administered all doses beyond Day 1. Dose reductions were specified if the total leukocyte count exceeded 30 × 10⁹/L (25). Peripheral blood counts were performed three times per week until stable, and then weekly. GM-CSF was to be ceased if any Grade 3 or 4 toxicity attributable to therapy developed. Patients without any improvement after 6 wk could be taken out of the study at the discretion of the treating physician. All other patients were to continue treatment for a total of 12 wk, at which time patients with significant improvement in the absence of toxicity could continue treatment until a maximum response was attained.

Monitoring during Treatment

Toxicity was assessed clinically, with serum creatinine and liver enzymes tested weekly, and was graded according to World Health Organization criteria (26). Pulmonary monitoring comprised continuous pulse oximetry for 4 h after the first dose of GM-CSF, with measurement of ABG values, plasma SP-A, and SP-B levels, serum LDH, spirometry, and measurement of DL_CO performed at 2-wk intervals, and with CT scanning and exercise testing at 6 and 12 wk.

All CT scans were performed with thin-section techniques, and were centrally reviewed by a single thoracic radiologist (J.M.V.) blinded to clinical information and using a semiquantitative grading scale. Anatomic levels representative of the upper, middle, and lower zones of the lung were prospectively defined. Both the extent and density of any abnormal opacity were determined at each level, with an overall summary conclusion generated.

Response Criteria

A complete response was defined as a normalization of the CT scan, spirometric parameters, DL_CO (using institutionally specified normal ranges), and arterial oxygenation (defined by the age-appropriate, calculated [A-a]DO₂ with breathing of room air ± 2 SD) (27). A partial response was defined as a  50% improvement in one or more of the following parameters during the study: (1) radiographically defined volume of pulmonary abnormality; (2) DL_CO; or (3) (A-a)DO₂. For example, a baseline (A-a)DO₂ of 20 mm Hg above the upper limit of normal was required to decrease to a level  10 mm Hg above the upper limit of normal to fulfil partial response criteria. The duration of response was taken from the date on which response criteria were first met to the date either of any subsequent therapeutic intervention or of deterioration of the (A-a)DO₂ gradient to pretreatment levels, whichever occurred first. Baseline oxygenation values were taken as the mean of all measured values at  4 wk before commencement, of therapy, and baseline spirometric and DL_CO values were those recorded immediately before commencement of therapy. An improvement in (A-a)DO₂ of  10 mm Hg was considered to represent a clinically useful therapeutic effect. Posttreatment results are those from assessments made within 14 d of the completion of GM-CSF treatment.

Dose Escalation

Early in the study it became evident that patients displayed an attenuated hematopoietic response to GM-CSF (25). Assuming that therapeutic activity required some degree of hematologic response to the administered GM-CSF, we modified the protocol to allow dose escalation in patients who did not improve with 5 µg/kg/d. For this purpose, a "hematologic response" was defined as a  1.5-fold increase in neutrophil count. To be eligible for dose escalation, patients were required to have no evidence of a hematologic response to previously administered doses of 5 µg/kg/d GM-CSF, no resolution of AP, no Grade 3 or 4 toxicity, and the approval of an independent thoracic physician. Single GM-CSF doses of 7.5, 10, 15, 20, and 30 µg/kg were sequentially administered with  3 d between doses. Blood counts were taken at baseline and at 4 h, 24 h, and 48 h after each dose. Dose escalation ceased once a hematologic response was attained, any toxicity of Grade 3 or more was seen, or the 30 µg/kg dose level was reached. The GM-CSF dose necessary to achieve a hematologic response was then continued daily for a further 6 wk.

Statistical Methods

If none of the first nine patients showed a response, the study was to have closed, having excluded a 30% overall response rate with 95% certainty. Otherwise, it was to continue for 3 yr or until 30 patients were enrolled. Changes with treatment were analyzed with paired t tests or Wilcoxon's two-sample tests. Comparisons between groups were made with the Fisher's exact test for categorical data and with the Mann- Whitney test for numerical data. Analyses were done with the Minitab 9.2 (Minitab Inc., State College, PA) or SYSTAT 8.0 (SPSS Inc., Chicago, IL, 1998) programs, and are presented as two-sided comparisons, with p < 0.05 regarded as significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Characteristics

Sixteen patients were screened and 15 were enrolled, one of whom was found to be ineligible, with endogenous lipoid pneumonia diagnosed on pathology review (Table 1). Most patients had undergone repeated therapeutic lavage with subsequent symptomatic disease recurrence. Only four patients were enrolled less than 6 mo after their most recent therapeutic lavage, all of whom had demonstrated deteriorating respiratory function during that period. All patients tested had significantly increased plasma levels of SP-A and SP-B.

At study entry, the median arterial oxygen tension (Pa_O2) was 63.5 mm Hg and the median (A-a)DO₂ was 45.5 mm Hg. Repeated pretreatment ABG analyses were performed in the 28 d before the beginning of GM-CSF treatment in 12 patients; the median change () in (A-a)DO₂ was +1.6 mm Hg (p = 0.5), with results differing from the utilized baseline values by from 9.4 to +4.9 mm Hg.

IgG antibodies capable of neutralizing recombinant GM- CSF were detected in stored baseline serum samples from all patients tested (n = 12).

Delivery of Therapy

No patients experienced any significant desaturation or symptomatic exacerbation with the first dose of GM-CSF, and all were able to self-administer subsequent doses. There were no dose reductions required for excessive leukocytosis. Treatment was prematurely terminated in one patient after 13 d because of neutropenia (as discussed subsequently). Five patients ceased treatment after 6 wk without objective improvement, seven patients received from 10 to 12 wk of therapy, and one patient continued GM-CSF for 26 wk.

Therapeutic Effects

Paired baseline and posttreatment oxygenation data were available for 13 patients and spirometric data for all 14 treated patients (Figure 1). The single patient without posttreatment oxygenation data had symptomatic deterioration and was considered unresponsive to therapy. For the 13 patients with paired data, the mean improvement in (A-a)DO₂ was 8.1 mm Hg (p = 0.09) (Figure 1). According to the study criteria, five patients (36%) achieved partial responses with initial therapy, with a mean improvement in (A-a)DO₂ of 23.2 mm Hg (range: 13.1 to 46.2 mm Hg). Thus, all responses also fulfilled the post hoc criterion of clinical utility ( 10 mm Hg improvement in [A-a]DO₂). No patient was classified as a responder solely on the basis of radiographic or DL_CO criteria.

View larger version (29K): [in this window] [in a new window]   Figure 1.   Pre- and posttreatment values from responding (upper panels) and nonresponding (lower panels) patients with GM-CSF treatment at 5 µg/ kg/d for (A) arterial PaO2, (B) (A-a)DO2, (C ) DLCO, and (D) VC. Median values are shown by horizontal bars and p values are for comparison of pre- with posttreatment values for individual patients by paired t test within each category. Following therapy with 5 µg/kg/d GM-CSF, there were five responders and nine nonresponders. Because posttreatment oxygenation data were unavailable for one of the nonresponding patients only 8 data points are shown for this category in A and B. Each individual responding patient is represented by a separate color, which is used consistently throughout all figures.

Although there was no significant change in DL_CO for the entire cohort (median : + 1.8% predicted; p = 0.5), each of the five responding patients showed an increase in DL_CO (median : + 4.6% predicted; p = 0.06) (Figure 1). The functional significance of these responses was supported by concomitant improvements in maximal oxygen uptake (O_2max) during cycle ergometry in the four responders tested, for whom the median pretreatment O_2max was 19.9 ml/min/kg and the median posttreatment O_2max was 26.3 ml/min/kg.

Treatment did not affect spirometric values. The median  FEV₁ and VC values were 0.01 L and 0.02 L, respectively (both p > 0.8). GM-CSF therapy was not associated with changes in serum levels of LDH or plasma levels of SP-A or SP-B (each p  0.16) (data not shown).

Dose Escalation

Seven patients did not respond to initial treatment with GM- CSF at 5 µg/kg/d and were considered for dose escalation (see METHODS). Of these, one patient declined to proceed and two showed gradual resolution of disease manifestations within 3 mo of completing GM-CSF therapy at 5 µg/kg/d. In the remaining four patients, a hematopoietically active dose was reached at GM-CSF doses of 7.5, 7.5, 10, and 20 µg/kg (25), respectively. After 6 to 10 wk of daily treatment with these doses, three patients remained unresponsive and one attained a partial response. This last patient, a 14-yr-old girl, had a 16-mo history of severe AP, had undergone 10 whole-lung lavages with only minor improvements, and had not responded to 6 wk of GM-CSF at 5 µg/kg/d ([A-a]DO₂ = 0.7 mm Hg). However, with 20 µg/kg/d GM-CSF for 10 wk, her (A-a)DO₂ improved from 52.0 to 28.8 mm Hg, her DL_CO improved from 30% to 40% predicted, and her 6-min walk-test distance increased from 330 m to 429 m.

Radiographic Assessment

Four responding patients showed radiographic improvement. A representative case is shown, together with the most complete case of radiographic clearance, in Figure 2. None of the nonresponding patients showed radiographic improvement.

View larger version (124K): [in this window] [in a new window]   View larger version (122K): [in this window] [in a new window]   View larger version (129K): [in this window] [in a new window]   View larger version (92K): [in this window] [in a new window]   Figure 2.   Illustrative radiographic changes in responding patients. Thin-section, high-resolution CT scans were performed through the entire thorax. For evaluation of response, three anatomic levels were defined to represent each lung zone: the upper zone at the level of the aortic arch, the middle zone at the subcarinal level, and the lower zone above the dome of the right hemidiaphragm. The extent and density of abnormal alveolar opacity were graded. (A) At baseline and (B) after therapy, through the middle zone of a patient who achieved a partial response. A demonstrates patchy, bilateral pulmonary infiltrates with a confluent area of opacity in the subpleural region of the left lower lobe posteriorly. The image made after therapy (B) shows partial clearing with residual patchy infiltrate. High-resolution CT images of the lower zone in a separate responding patient (C ) at baseline and (D) after therapy. Extensive opacity involving most of the lungs is seen in C. The areas of ground-glass opacity demonstrate thickening of the interlobular septa leading to the characteristic "crazy paving" appearance typical of AP. After therapy ( D ) there is complete clearing of the opacity.

Summary of Efficacy

Considering both the initial treatment and the subsequent dose-escalation schedule, six of 14 evaluable patients responded to GM-CSF (43% response rate; 95% confidence interval [CI] 18 to 71%), with responses lasting a median of 39 wk (range: 6 to 67 wk). Responses were observed among patients treated with either preparation of GM-CSF. Although two patients reported major symptomatic improvement within 10 d, improvements in their (A-a)DO₂ gradient were not evident until 4 to 6 wk (Figure 3), with maximal improvements not achieved until 6 to 10 wk of therapy.

View larger version (29K): [in this window] [in a new window]   Figure 3.   Kinetics of improvement in (A-a)DO2 for patients responding to their first effective course of GM-CSF therapy (5 µg/kg/d for five patients and 20 µg/kg/d for one patient). Color coding is as used in Figure 1, with the patient responding only at the higher dose of GM-CSF (see text for details) represented by a broken line with open circles. The patient represented by a line with solid circles had only one baseline ABG evaluation. For other patients, pretreatment values are shown as mean ± SEM for all ABG analyses performed within 28 d of commencing therapy (in two cases SEM are contained within symbols).

Predictors of Response

The analysis of variables associated with response to GM-CSF treatment was restricted to the initial treatment episode (n = 14) (Table 2). The pretreatment factors associated with response were a longer time from diagnosis, higher VC, normal serum level of LDH, and higher plasma level of SP-B (each p  0.04). All responding patients had undergone prior therapeutic lavage.

Among treatment-related variables, only the peak eosinophil count differed according to response (median: 0.63 × 10⁹/L versus 0.27 × 10⁹/L; p = 0.01), despite similar baseline values (p > 0.9). There were no differences between peak total leukocyte or neutrophil counts according to response, nor was there any difference in duration of GM-CSF treatment (each p  0.2). The association between treatment-related eosinophilia and therapeutic response was also observed both among patients who underwent dose escalation and responders who were subsequently retreated with GM-CSF (data not shown).

Retreatment Following Recurrence

Five of six responding patients subsequently developed deteriorating pulmonary function, and four were retreated with GM- CSF (4 to 20 µg/kg/d for 12 to 38 wk, with one patient receiving a "maintenance" schedule of 20 µg/kg thrice weekly). Three of these patients again responded (one patient is not yet evaluable). In these four patients, (A-a)DO₂ showed improvements within 21 d, and ultimately declined from a median of 46.2 mm Hg before treatment to 9.4 mm Hg after treatment (p = 0.06). This was accompanied by varying degrees of functional improvement; in one patient the 6-min walk-test distance improved from 333 m to 479 m, and in two patients VO_2max improved from 23.8 ml/min/kg to 25.8 ml/min/kg, and 17.1 ml/min/kg to 27.5 ml/min/kg, respectively. These responses to retreatment were ongoing at 15+, 72+, and 112+ wk.

Toxicity

Overall, therapy was well-tolerated (Table 3). There was no acute deterioration in oxygenation with the first dose of GM- CSF at 5 µg/kg. One patient developed asymptomatic Grade 3 neutropenia during treatment, which resolved rapidly after ceasing GM-CSF (25). Overall, there were four patients who developed a symptom complex resembling the "first dose effect" (28) but which occurred after varying periods of treatment (range: 1 to 35 d), with fevers, chills, and nausea/vomiting within 4 h of dosing. Two of the four responding patients who were retreated with GM-CSF underwent dose escalation, and both manifested the symptom complex just described, with additional transient hypoxemia in one (O₂ saturation: 89%). These toxicities developed at doses of 10 µg/kg and 20 µg/kg, with the patients later tolerating treatment with 8 µg/kg/d and 12.5 µg/kg/d without toxicity.

Long-Term Follow-up

At a median follow-up of 16 mo (range: 10 to 45 mo), all patients were alive, and there were no hematopoietic disorders or any clinically or radiographically evident instances of pulmonary fibrosis.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The phenotypic characterization of gene-targeted mice lacking the hematopoietic regulator GM-CSF (11) or the signal-transducing -common chain of its receptor (_c) (29) provided the rationale for exploration of GM-CSF as a potential therapy for AP. Despite the established capacity of GM-CSF to stimulate the proliferation and differentiation of myeloid hematopoietic progenitor cells (30), these animals had no detectable disturbance of their steady-state hematopoiesis, but had impaired surfactant clearance (11, 12), disturbed macrophage function (31, 32), a propensity to develop pulmonary and other infections (29, 33), and reduced fertility (33).

In the present study, we noted attenuated hematopoietic responsiveness to the administered GM-CSF (25). Subsequently, bronchoalveolar lavage fluid and serum from patients with AP have been shown to contain GM-CSF neutralizing activity due to the presence of an IgG autoantibody (23, 24). Indeed, all patients examined in the present study had such antibodies detectable in their pretreatment serum. Although relatively few cases have yet been studied (24), it appears that most patients with idiopathic acquired AP may have autoantibodies against GM-CSF, suggesting that the underlying pathogenetic process in AP may be inhibition of endogenous GM-CSF activity, leading to functional GM-CSF deficiency. Alternative processes that would similarly lead to a state of functional GM-CSF deficiency include defective GM-CSF _c expression, which has been described in cases of congenital (34) and secondary AP (4), impaired signalling, or defective GM-CSF production (35).

The assay for GM-CSF-neutralizing autoantibodies used in the present study appears to have both a high sensitivity and specificity for idiopathic AP (24), but until its widespread availability, assessment of the hematopoietic response to GM- CSF may provide a surrogate functional measure. However, GM-CSF-binding antibodies are occasionally seen in healthy persons (36), and may be induced after treatment with recombinant GM-CSF (37). Until the utility of such assays is determined, the diagnosis of AP should continue to be based on conventional criteria.

One of the difficulties in evaluating therapeutic interventions in patients with AP is the variable natural history of the disorder, and its possible "spontaneous resolution," which has been suggested to occur in up to one-third of patients (1, 6). However, a review of the literature to 1997 identified only 24 instances of spontaneous resolution among 303 published cases with follow-up information (8%), and in only seven cases (2%) was this supported by normalization of arterial oxygenation (Seymour, JF, unpublished data). Thus, although spontaneous resolution of AP undoubtedly occurs, it would appear to be infrequent. In the present study, no patients showed any improvement in the 4 wk prior to therapy, making coincidental spontaneous resolution an improbable explanation for the improvements observed during treatment.

We found subcutaneous GM-CSF treatment to be easily delivered. Only one patient ceased treatment because of an adverse effect (asymptomatic neutropenia). Toxicities at 5 µg/kg were infrequent and mild, the most common being local erythema and induration in 36% of patients. Indeed, patients with AP also appear to have a lower incidence of adverse effects of GM-CSF than anticipated (38). The presence of GM-CSF-neutralizing antibodies may result in a reduced effective biologic dose.

We did not observe any late toxicities attributable to GM- CSF, with follow-up approaching 4 yr. Patients with AP infrequently develop late pulmonary fibrosis (6), and the development of pulmonary fibrosis in an adenovirus-mediated GM- CSF transgenic mouse model (39) raises the concern that GM- CSF therapy may enhance this risk. However, the fibrosis is probably attributable to the vector used, since other GM-CSF-transgenic animals have no such propensity (15, 16, 40, 41).

With 5 µg/kg/d doses of GM-CSF, five of 14 (36%) patients showed pulmonary function improvement, confirming our initial observation (17). Objective improvements in oxygenation were evident within 4 to 6 wk, and a therapeutic trial of this duration would be reasonable before concluding that a patient is unlikely to respond. The responses obtained were of sufficient magnitude to provide meaningful clinical benefit to patients, with symptomatic improvements, and serial assessments of DL_CO and exercise capacity supported functional gains. Although walk-testing and cycle ergometry were used at different institutions, their results are highly correlated (42).

Given the detection of antibodies to GM-CSF in all patients in our study, it is not apparent why therapeutic responses were variable. Although no data is yet available, the effective antibody titer, reflecting total GM-CSF-neutralizing capacity, may be lower in responding patients. This could explain the relationship between peak eosinophil count during therapy and functional improvement (Table 2). Given the numerous cytokines capable of inducing some degree of neutrophilia (43), minor variations in neutrophil counts during treatment may not necessarily reflect the activity of administered GM-CSF.

If the GM-CSF-neutralizing antibody present in patients with AP can abrogate the activity of endogenous GM-CSF, these patients may display systemic features of GM-CSF deficiency, which could also be reversible with GM-CSF treatment. Consistent with this proposal is that patients with AP have an increased risk of specific opportunistic infections, both pulmonary and extrapulmonary, most notably with Nocardia asteroides (44). The effect of AP on the fertility of affected women is unknown, but the impaired reproductive capacity of GM-CSF-deficient mice suggests that this should be further investigated (33).

Improvements in oxygenation are achieved more rapidly after whole-lung lavage than with GM-CSF; however, issues such as resource availability, comorbid conditions, and symptom severity will also influence treatment considerations. It appears that the magnitude of the therapeutic effect achievable with either approach may be similar. Four available reports provide specific data on pre- and postlavage arterial Pa_O2 levels in patients with AP (5, 8) and these describe a mean improvement in arterial PaO₂ of 12 to 19 mm Hg with lavage (overall mean: 14.5 mm Hg), whereas the mean improvement in arterial Pa_O2 for all GM-CSF treatment courses in our study was 9.7 mm Hg, and among responding patients was 23 mm Hg.

Although we have shown the reproducible therapeutic activity of GM-CSF in patients with idiopathic acquired AP, many aspects of this novel therapy require further investigation. Even within this limited series, some patients required significant dose escalation to attain therapeutic effects. The sequential application of increasing doses of GM-CSF in a quest for evidence of biologic activity, such as eosinophilia, may identify appropriate dose ranges for further evaluation. Although the subcutaneous route of administration was used in this study, on the basis of its proven safety and efficacy in other settings (30, 38), recent animal work suggests that inhaled GM-CSF may have greater pulmonary effects (13), but conversely may not ameliorate any extrapulmonary aspects of AP. Although aerosolized GM-CSF is well tolerated in humans (45), the issues of pharmacokinetics and drug delivery remain unresolved.

The present study suggests that GM-CSF therapy may provide an alternative approach to treating patients with idiopathic AP. Regardless of its ultimate place in the therapy of idiopathic AP, which can only be established through well- designed clinical trials, investigation of the pharmacologic effects of GM-CSF in these patients has clarified the pathogenesis of this enigmatic disease.