Abstract

Background: Autoimmune pulmonary alveolar proteinosis (aPAP) results from impaired macrophage-mediated clearance of alveolar surfactant lipoproteins. Whole lung lavage has been the first-line treatment but recent reports suggest the efficacy of granulocyte–macrophage colony-stimulating factor (GM-CSF). We aimed to review the efficacy and safety of nebulised GM-CSF in aPAP.

Methods: We conducted a systematic review and meta-analysis searching Embase, CINAHL, MEDLINE and Cochrane Collaborative databases (1946–1 April 2022). Studies included patients aged >18 years with aPAP receiving nebulised GM-CSF treatment and a comparator cohort. Exclusion criteria included secondary or congenital pulmonary alveolar proteinosis, GM–CSF allergy, active infection or other serious medical conditions. The protocol was prospectively registered with PROSPERO (CRD42021231328). Outcomes assessed were St George's Respiratory Questionnaire (SGRQ), 6-min walk test (6MWT), gas exchange (diffusing capacity of the lung for carbon monoxide (DLCO) % predicted) and arterial–alveolar oxygen gradient.

Results: Six studies were identified for review and three for meta-analysis, revealing that SGRQ score (mean difference −8.09, 95% CI −11.88– −4.3, p<0.0001), functional capacity (6MWT) (mean difference 21.72 m, 95% CI −2.76–46.19 m, p=0.08), gas diffusion (DLCO % predicted) (mean difference 5.09%, 95% CI 2.05–8.13%, p=0.001) and arterial–alveolar oxygen gradient (mean difference −4.36 mmHg, 95% CI −7.19– −1.52 mmHg, p=0.003) all significantly improved in GM-CSF-treated patients with minor statistical heterogeneity (I2=0%). No serious trial-related adverse events were reported.

Conclusions: Patients with aPAP treated with inhaled GM-CSF demonstrated significant improvements in symptoms, dyspnoea scores, lung function, gas exchange and radiology indices after treatment with nebulised GM-CSF of varying duration. There is an important need to review comparative effectiveness and patient choice in key clinical outcomes between the current standard of care, whole lung lavage, with the noninvasive treatment of nebulised GM-CSF in aPAP.

Introduction

Pulmonary alveolar proteinosis (PAP) is a condition of impaired alveolar macrophage function characterised by excessive accumulation of surfactant within the alveoli [1, 2]. Disease progression leads to dyspnoea, respiratory failure, secondary infection, pulmonary fibrosis and death, although spontaneous resolution is described rarely in some subjects. The most common variant of PAP is autoimmune PAP (aPAP), associated with circulating polyclonal autoantibodies directed against granulocyte–macrophage colony-stimulating factor (GM-CSF). GM-CSF stimulates the differentiation, proliferation and survival of granulocytes and monocytes. Autoantibodies neutralise GM-CSF bioactivity on alveolar macrophages, preventing them from catabolising phagocytosed lipoproteins [3].

Clearance of lipo-proteinaceous material from the alveoli may be achieved by whole lung lavage (WLL) and this treatment has been suggested as first-line therapy [3–5]. However, WLL is resource-intensive and invasive, and requires a high level of operator expertise, intensive care unit admission, general anaesthesia and intubation. Furthermore, although reported as generally safe, significant potential complications include fever, pneumonia, hypoxaemia, fluid leak, pleural effusion, pneumothorax and death [4, 6].

A causal linkage with GM-CSF signalling in aPAP was suggested by studies of GM-CSF gene-deficient (knockout) mice, where such mice developed a lung phenotype mimicking PAP, in addition to impaired immunological responses to several microbial pathogens [7–9]. Further, the administration of human GM-CSF antibodies to nonhuman primates led to GM-CSF signal blockade and pathological lung infiltration consistent with human aPAP [10, 11]. Subsequent studies explored GM-CSF replacement as a primary treatment following the demonstration that inhaled GM-CSF led to clearance of surfactant proteins from the alveoli [12–15]. The pathological impacts of such autoantibodies remain to be elaborated.

Nebulised (inhaled) GM-CSF is a noninvasive, biologically cogent treatment option for aPAP with demonstrable impacts on alveolar clearance. Given the advantage of inhalation therapy over invasive procedures, the use of nebulised GM-CSF as a potential treatment of aPAP warrants further evaluation. We performed a systematic review and meta-analysis of published literature to explore the effects of nebulised GM-CSF treatment for aPAP. We sought to assess the effects of nebulised GM-CSF on symptoms and exercise function, gas exchange, spirometry, radiology and change in serum biomarkers associated with disease activity.

MethodsData sources and search strategy

We performed a systematic review and meta-analysis according to Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines [16], using a protocol prospectively registered with PROSPERO (CRD42021231328).

We searched Embase, CINAHL, MEDLINE and Cochrane Collaborative databases for relevant studies, supplemented by manually searching reference lists, previous meta-analyses, grey literature and the ClinicalTrials.gov trials registry. The following search terms were used: “pulmonary alveolar proteinosis”, “lung alveolus proteinosis”, “granulocyte macrophage colony stimulating factor”, “filgrastim”, “lenograstim”, “sargramostim”, “molgramostim” and “GM-CSF”. All databases were searched from 1946 to 1 April 2022. The search strategies are included in supplementary tables S1–S5. We combined articles from all databases and excluded duplicates. Abstracts of studies identified were screened by two independent reviewers (M. Munsif, R.G. Stirling). Full-text articles were reviewed to determine inclusion suitability and lack of consensus resolved by a third reviewer (T.L. Leong).

Eligibility criteria

Inclusion and exclusion criteria were defined a priori. We included all studies of patients aged ≥18 years with aPAP (diagnosed by computed tomography (CT), biopsy or bronchoalveolar lavage and where positive GM-CSF serum antibodies were identified) that reported nebulised GM-CSF treatment. Studies for inclusion required an active nebulised recombinant human GM-CSF treatment and control comparator cohorts.

We excluded duplicate references and overlapping (previously reported) cohorts. From the overlapping cohorts, we included the larger cohort and/or the cohort with the most comprehensive data. Further exclusion criteria included participants with secondary or congenital PAP; treatment with GM-CSF, rituximab or cytokine therapy within 1 month of baseline; treatment with any investigational medicinal product within 4 weeks of screening; concomitant use of sputum-modifying drugs; history of allergic reactions to GM-CSF or aerosol-delivered agents; known active infection (viral, bacterial, fungal or mycobacterial); apparent pre-existing concurrent respiratory comorbidity; and any other serious medical condition. Only human studies and studies reported in English were included; studies based on questionnaires and single cases were excluded. Eligibility criteria are described further in supplementary table S6.

Outcomes

We used prespecified end-points based on expert consensus and published studies evaluating treatment with GM-CSF [1, 9, 17]. The end-points selected were respiratory symptoms and associated symptom scores (quality of life measures and dyspnoea scales) and exercise function (6-min walk test (6MWT)). Objective measures of respiratory physiology and lung function included change in forced expiratory volume in 1 s (FEV1) % predicted, forced vital capacity (FVC) % predicted, vital capacity (VC) % predicted, diffusing capacity of the lung for carbon monoxide (DLCO) % predicted, arterial blood gas (PaO2) and alveolar–arterial oxygen tension difference (PA−aO2).

Low-dose quantitative CT of the chest and/or high-resolution CT (HRCT) were used to evaluate total lung volume and mean lung density measurements before and after treatment with GM-CSF. The extent of ground-glass opacification was quantified visually and divided into three regions (upper, middle and lower) with a total HRCT score calculated as the sum of all values for the lung regions [9]. Changes in serum levels of biomarkers including Krebs von den Lungen 6 (KL-6), chemokine monocyte chemoattractant protein-1 (MCP-1), carcinoembryonic antigen (CEA), surfactant protein D, high sensitivity C-reactive protein (CRP) and lactate dehydrogenase were evaluated. KL-6 is a mucin-like glycoprotein that correlates with disease progression in PAP and various interstitial lung diseases [18] and MCP-1 is an important pro-inflammatory chemokine for macrophage recruitment, both of which are elevated in patients with PAP [19].

The safety end-points and adverse effects were separated into treatment effects, adverse events due to disease progression, treatment discontinuation and mortality.

Data extraction and assessment of risk of bias in included studiesData extraction

Two reviewers (D. Sweeney, T.L. Leong) independently extracted and tabulated the data using a predesigned standard form. The categories of data included 1) general information (study, first author's name, year of publication), 2) methodology of the study (study design/type, allocation concealment and blinding, number of sites), 3) baseline participant information, 4) interventions and comparisons (type of GM-CSF, dose and schedule) and 5) outcomes (for continuous variables we extracted mean±sd/se or median (IQR) and for categorical data we extracted number of participants for each outcome). Baseline and change in outcomes pre- and post-treatment were collated. Discrepancies and clarification of data were resolved by a third reviewer (M. Munsif).

Risk of bias assessment

The assessment of the methodological quality of the included studies was evaluated using the Risk of Bias 2 (RoB2) tool, a revised Cochrane risk of bias tool for randomised trials [20]. This tool assesses risk of bias in five domains including bias in randomisation process, bias due to deviation from intended interventions, bias due to missing outcome data, bias in measurement of outcomes and bias in selection of the reported result. A summary score of overall risk of bias is generated and reported as low, of some concern or high risk of bias.

Data synthesis and statistical analysis

All analyses were conducted using Review Manager (RevMan version 5.3, The Cochrane Collaboration, Copenhagen). Effect measures in meta-analysis were reported for St George's Respiratory Questionnaire (SGRQ), 6MWT, DLCO (% predicted) and PA−aO2 (mmHg) using mean differences (post-intervention mean minus baseline mean) with sd. Where mean difference±sd were unreported, mean difference was calculated by subtracting post-intervention mean from baseline mean and sd was imputed using the mean sd for the same outcome measure from other included studies. Subset analysis was performed to include evaluation of continuous and interrupted, week on week off, intermittent treatment. Statistical heterogeneity was assessed using the Chi-squared and I2 tests [21]. An I2 value <50% was regarded as low heterogeneity, 50–75% was regarded as substantial heterogeneity and an I2 value of ≥75% as considerable heterogeneity.

ResultsSearch results and study selection

Our search identified 1070 unique records. A total of 80 records were deemed suitable for full-text evaluation after screening titles and abstracts. A total of 74 articles were excluded for reasons shown in the PRISMA flow diagram (figure 1), including eight articles which had duplicate/overlapping population reports over several studies. A total of six studies were included in this review with three contributing to the meta-analysis [9, 22–26].

FIGURE 1

Preferred Reporting Items for Systematic Reviews and Meta-analysis flow diagram for study inclusion.

Study characteristics and risk of bias assessment

The characteristics of the six included studies are detailed in table 1. Three of the six studies were randomised (two were double-blind placebo-controlled randomised controlled trials (RCTs)), two studies were small retrospective cohorts and one study was a prospective cohort. The six studies included a total of 288 participants who received a form of nebulised recombinant human (rh) GM-CSF. A total of 238 participants were evaluated in RCTs, randomised to either sargramostim (Leukine) (n=33), molgramostim (continuous n=46, intermittent n=45) or rh-GM-CSF (n=19) with varying dosing intervals of up to 24 weeks. Two of the RCTs used GM-CSF in the induction phase at 125 μg twice daily for 8 days with no therapy on days 9–14 over six cycles of 2 weeks’ duration. All the studies included outcome data on gas exchange with varying data on lung function, 6MWT, dyspnoea scores and imaging changes (table 2).

TABLE 1

Characteristics of studies

TABLE 2

Parameters assessed

Risk of bias assessment for the RCTs identified low risk of bias in two RCTs [24, 26] and some concerns in the third study in which control participants received no treatment (figure 2) [25]. An additional concern related to per protocol data analysis with a disproportionate number of participants defaulting from the trial prior to 6 months with the potential to bias data outcomes and this study was therefore classified as having some risk of bias [25].

FIGURE 2

Risk of bias assessment. D1: bias arising from the randomisation process; D2: bias due to deviations from intended interventions; D3: bias due to missing outcome data; D4: bias in measurement of the outcome; D5: bias in selection of the reported result.

Clinical outcomesRespiratory symptoms and dyspnoea scores

The presence of respiratory symptoms such as dyspnoea was assessed in five studies, with two studies reporting the SGRQ score [25, 26]. The SGRQ is a 76-item questionnaire divided into three sections of respiratory symptoms, social and psychological impact, and activities, measured on a scale of 0–100 with a higher score indicating a poorer quality of life [27]. Tian et al. [25] reported a significant reduction in the SGRQ score in the treatment group; however, the absolute reduction was not quantified. Trapnell et al. [26] reported a meaningful reduction in SGRQ scores in both the continuous (n=45) and the intermittent GM-CSF treatment regimens (n=44). The continuous treatment group had a mean±sd point reduction of 12.3±14.3 and the intermittent treatment group of 12.0±15.1. Tazawa et al. [9] reporting on 39 participants described a categorical resolution of dyspnoea in 54.3% (p<0.001). Papiris et al. [22] used the BORG score to assess symptoms and noted an overall reduction in the scores; however, no further information was provided. Tazawa et al. [24] used the modified Medical Research Council (mMRC) dyspnoea score and noted the score reduced by 0.07 in the treatment group (95% CI 0.01–0.13).

Two RCTs including 106 GM-CSF-treated patients were available for comparison of the effect of inhaled GM-CSF on SGRQ score after 6 months’ treatment [25, 26]. The mean difference in SGRQ score was −8.09 (95% CI −11.88– −4.3), which is a greater than the minimal clinically important difference (MCID) (4 units decrease) from the control-treated groups (p<0.0001) [28]. There was minor statistical heterogeneity between trials (Chi-squared p=0.69, I2=0%) (figure 3a).

FIGURE 3

Forest plot analyses of a) St George's Respiratory Questionnaire at 6 months, b) 6-min walk test (metres) at 6 months, c) diffusing capacity of the lung for carbon monoxide (DLCO) (% predicted) at 6 months and d) alveolar–arterial oxygen tension difference (PA−aO2) (mmHg) at 6 months. GM-CSF: granulocyte–macrophage colony-stimulating factor; IV: inverse variance.

6MWT

A 6MWT was performed in four studies [9, 24–26]. Tazawa et al. [9] reported a significant improvement of mean±se walk distance from 393±27 m to 444±24 m (p<0.005) before and after treatment, respectively.

Three RCTs with 142 GM-CSF-treated patients were available for comparison of effect of inhaled GM-CSF on mean difference in 6MWT pre- and post-treatment [24–26]. The mean difference in 6MWT between inhaled GM-CSF-treated patients and placebo-treated patients was 21.72 m (95% CI −2.76–46.19 m, p=0.08). There was minor statistical heterogeneity between trials (Chi-squared p=1.48, I2=0%) (figure 3b).

Pulmonary function tests

Five studies reported at least one pulmonary function test parameter. Three of the studies reported FVC or VC findings, five reported DLCO and one reported total lung capacity. There was a nonsignificant improvement in FVC % predicted after GM-CSF therapy with an average of 3.7–14% difference between the groups (p=0.29). The estimated improvement in DLCO % predicted after therapy was significant in all five studies after treatment (p<0.05). In Tazawa et al. [24], the estimated improvement in DLCO % predicted between GM-CSF therapy versus placebo was 6.87% (95% CI 0.62–13.05%). The mean change from baseline in DLCO % predicted was greater in the continuous treatment group than in the placebo group at 7.8% (95% CI 2.3–13.3%).

Three RCTs with 136 GM-CSF-treated patients were available for comparison of effect of inhaled GM-CSF on mean difference in DLCO % predicted pre- and post-treatment [24–26]. The mean difference in DLCO % predicted was 5.09% (95% CI 2.05–8.13%) greater than in placebo-treated patients (p=0.001). There was minor statistical heterogeneity between trials (Chi-squared p=1.02, I2=0%) (figure 3c).

Gas exchange

All studies evaluated gas exchange using measures of arterial oxygenation (PaO2) and the difference in the alveolar–arterial gradient (PA−aO2). PA−aO2 measures the difference between the concentration of oxygen in the alveoli and the arterial system and widening of the PA−aO2 is used as a measure of pulmonary dysfunction. There was an average change (narrowing) in PA−aO2 of −5.4 mmHg in the intervention group compared to the control group. The mean±sd change in the continuous treatment group was −12.2±14.7 mmHg compared to −10.6±15.7 mmHg in the intermittent treatment group. The average PaO2 improvement was 6.2 mmHg after treatment compared to 2.9 mmHg in the placebo group.

Three RCTs with 140 GM-CSF-treated patients were available for comparison of effect of inhaled GM-CSF on mean difference in PA−aO2 pre- and post-treatment [24–26]. The mean difference in PA−aO2 was −4.36 mmHg (95% CI −7.19– −1.52) greater than in the placebo-treated patients (p=0.003). There was minor statistical heterogeneity between trials (Chi-squared p=0.21, I2=0%) (figure 3d).

Radiology

Four studies reported methods of quantifying radiological involvement of PAP; two studies used visual quantification scoring of ground-glass opacification on HRCT [9, 26]. A higher score (ranging from 0 to 15 points) indicated a higher proportion of the area of the lung parenchyma affected by ground-glass opacification. These studies showed that continuous treatment with GM-CSF led to a greater improvement in the ground-glass opacification score (2.5 points) compared to placebo [26]. The baseline CT in one study revealed more ground-glass opacity in the upper zones. The improvement post-treatment was more marked in the middle (p<0.01) and lower zones (p<0.01) compared to the upper zones [9]. Two studies used mean lung density or CT density values from density signals (Hounsfield units) and pixel numbers for each image slice [24, 25]. There was no significant difference in total lung volume (p=0.81) or mean lung density (0.80) between the treatment and control group [25]. The density of lung field (Hounsfield units) was better in the GM-CSF group with a difference of −36.08 units (95% CI −61.58– −6.99) compared to the control group [24]. The differences in radiological quantitation of aPAP involvement between studies makes between-study comparison difficult.

Use of rescue WLL

Rescue WLL was described in two studies, with Trapnell et al. [26] reporting a reduction in WLL rescue from 0.82 procedures per patient year to 0.42 procedures per patient year in the continuous molgramostim group during the blinded period. Tian et al. [25] described no difference in time to rescue WLL between study groups.

Autoantibodies against GM-CSF

The development of an immune response against rh-GM-CSF may lead to delayed impacts on treatment efficacy. Tazawa et al. [24] reported a greater increase in autoantibodies against GM-CSF in sargramostim-treated patients than placebo-treated patients (8.6±24.9 µg·mL−1versus −4.9±10.4 µg·mL−1) with no change in serum antibody neutralising capacity. Autoantibody levels were similar between treatment groups at baseline and neutralising capacity was unchanged following molgramostim treatment [26].

Biomarkers

Tazawa et al. [24] observed minor reductions in KL-6 and MCP-1 levels following GM-CSF treatment but no change in surfactant proteins (SP-A and SP-D), CEA or CRP.

Safety of nebulised GM-CSF

No deaths were reported in these trials. Tazawa et al. [24] reported no significant difference between adverse events in GM-CSF-treated and placebo-treated patients, including six serious adverse events likely unrelated to disease treatment with a single case of PAP progression. Trapnell et al. [26] observed a higher proportion of chest pain reported in patients treated continuously, but not intermittently, with GM-CSF versus placebo-treated patients (22% versus 4% versus 2%), but any adverse event and severe adverse event rates were similar between groups. Tian et al. [25] observed a slight increase in serum alanine aminotransferase and aspartate aminotransferase to <2–3× normal which declined or stabilised after dietary and medication modifications. No severe adverse events were reported in observational trials of 63 patients [9, 22, 29, 30].

Discussion

This systematic review and meta-analysis in patients with aPAP treated with inhaled GM-CSF demonstrates significant improvements in clinical symptoms, dyspnoea scores, lung function, gas exchange and radiology indices after treatment with nebulised GM-CSF of varying duration with limited evidence of significant adverse events.

The clinical significance of quantitative change in outcomes warrants consideration. The MCID in SGRQ is suggested to be 4.0 units with a moderate difference estimated as 5.7 units [28, 31], and in this study the mean difference of 8.09 units appears clinically significant. In COPD, the MCID in the 6MWT is ∼25 m (95% CI 20–61 m), slightly longer than the 21.72 m (95% CI −2.76–46.19 m) observed in this review [32].

Anchor methods suggest a DLCO MCID of 11% differentiates levels of severity in COPD patients [33]. While this level is certainly greater than the 5.09% (95% CI 2.05–8.13%) change observed in this study, a decline in DLCO appears to be the strongest predictor of decline in exercise capacity in COPD [34] and suggests that DLCO measurement may help prognosticate exercise tolerance.

PA−aO2 measures the difference between oxygen concentration in the alveoli and arterial circulation and reflects alveolar–capillary dysfunction but may be affected by physiological ventilation–perfusion mismatch, diet, weight and posture [35]. While the MCID for PA−aO2 has not been established, the improvement is promising and may be useful for comparison with other treatments including WLL.

There was minor study design heterogeneity in the studies included in the meta-analysis. One study excluded patients if they had undergone previous GM-CSF treatment and included a symptom severity stratification ensuring randomised matching of disease severity between treatment arms [25], and a second stratified on receipt of WLL or GM-CSF within 2 months of enrolment [26]. The trial agents were rh-GM-CSF biosimilars and included sargramostim (Sanofi Genzyme, Cambridge, MA, USA) [24], molgramostim (Savara Pharmaceuticals, Austn, TX, USA) [26] and Huabei Jimuxin (North China Pharmaceutical Corporation, Shijiazhuang, China) [25]. Two studies were placebo controlled [24, 26] and one study an open label trial in which control patients received no treatment [25]. Drug delivery engaged the eFlow [26] and LC-Plus nebulisers [24] (both PARI Gmbh, Germany), which are known to have different dose delivery [36] and peripheral lung drug deposition [37], likely to result in difference in efficacy. While the trial duration of 24 weeks was similar, the GM-CSF dose was 300 µg once daily in one study [26], 125 µg twice daily on alternate weeks in another [24], and 150 µg twice daily on alternate weeks reducing to 150 µg once daily after 3 months in the third study [25]. These differences in GM-CSF exposure may lead to differences in macrophage activation and disease response to treatments.

Limitations should be considered when interpreting the results of this systematic review. There were only six studies that met the inclusion criteria, with significant heterogeneity in the data. A few of the studies had incomplete or limited information on end-points and statistical data on outcome measures. The trial designs were heterogenous, used different formulations of rh-GM-CSF and had varying duration of follow-up. This review was restricted to therapy with GM-CSF after a specific time-frame post WLL; however, in practice, there may be a short interval between treatment options, and this was not considered in those studies.

WLL has been suggested to be the treatment of choice in aPAP and yet no head-to-head evidence of comparative effectiveness is available to confirm this recommendation. Patient-centred care encourages patient choice and patient choice is likely to be influenced by treatment invasiveness, duration, side-effects and effectiveness [38]. It is possible that different treatment types may have different impacts at different disease time points, such as induction and maintenance of remission in the management of aPAP. It is important to consider and explore the relative impacts of WLL, inhaled GM-CSF, plasmapheresis, rituximab and cholesterol-lowering agents in both the induction and maintenance of disease remission [39–41].

This is one of the first systematic reviews to evaluate the impact of nebulised GM-CSF therapy on the treatment of aPAP. Survival for PAP (of any cause) from date of diagnosis is 75% and 68% at 5 years and 10 years, respectively [1]. The cause of death is usually progressive respiratory failure or uncontrolled infection [30]. In aPAP, the clinical course is often unpredictable with progressive deterioration, spontaneous resolution or stable but unremitting disease. Therapeutic management of PAP is challenging in the current era of the SARS-CoV-2 pandemic. WLL is a resource-intensive and invasive procedure. Safe, alternative and efficacious treatment options such as nebulised GM-CSF therapy need to be considered.

Currently WLL is considered to be the standard treatment for PAP however, there are no head-to head trials comparing the efficacy and safety of nebulised GM-CSF therapy to WLL. Future studies need to define the indication, patient choice and timing of therapy with a focus on respiratory and quality of life outcomes with standardisation of treatment protocols to enable comparison between groups.

In conclusion, this systematic review suggests GM-CSF is a clinically important treatment consideration in aPAP. It is a noninvasive treatment that may have clinical benefits in improving SGRQ scores, lung function parameters, gas exchange and radiology in affected patients. Inhaled GM-CSF has an acceptable adverse risk profile and may be preferred in certain patients with limited cardio-respiratory reserve as an alternative to WLL.

Points for clinical practice

The most common variant of pulmonary alveolar proteinosis (PAP) is autoimmune PAP (aPAP), associated with circulating polyclonal autoantibodies directed against granulocyte–macrophage colony-stimulating factor (GM-CSF).

Currently, whole lung lavage (WLL) is considered to be the standard first-line treatment for PAP. There is clear need to assess the impacts of noninvasive nebulised GM-CSF in aPAP to enable exploration of the comparative effectiveness of invasive and noninvasive treatments in this disease.

This is the first systematic review and meta-analysis focused on evaluating the impact of noninvasive nebulised GM-CSF therapy treatment of aPAP.

Our data from 289 patients in a systematic review and 239 patients in a meta-analysis demonstrates significant improvements in clinical symptoms, dyspnoea scores, lung function, gas exchange and radiology indices after treatment with nebulised GM-CSF of varying duration. No serious adverse events were reported in the studies.

Questions for future research

WLL has been suggested to be the treatment of choice in aPAP and yet no head-to-head evidence of comparative effectiveness is available to confirm this recommendation. Patient-centred care encourages patient choice, which is likely to be impacted by treatment invasiveness, duration, side--effects and effectiveness. Future studies need to define the indication, patient choice and timing of therapy with a focus on respiratory and quality of life outcomes with standardisation of treatment protocols to enable comparison between groups.