Inhaled granulocyte-macrophage colony stimulating factor for mild-to-moderate autoimmune pulmonary alveolar proteinosis - a six month phase II randomized study with 24 months of follow-up

Background Treatment of autoimmune pulmonary alveolar proteinosis (aPAP) by inhaled granulocyte-macrophage colony stimulating factor (GM-CSF) is considered safe and effective. Evidence of benefit from GM-CSG inhalation for mild to moderate aPAP patients is limited. Methods In this multicenter, randomized, open-labeled clinical trial, 36 aPAP patients with mild to moderate disease severity were randomized into either the GM-CSF treatment group or control group. Inhaled GM-CSF was prescribed for 6 months, and patients received follow-up for another 18 months without treatment. Physiological features of the patients were analyzed. Results There were 36 patients (19 in the treatment group, 17 in the control group) included. There were no significant differences in the primary endpoints as measured by the change of alveolar arterial oxygen gradient (A-aDO2) from the baseline values to the values obtained during treatment or during the following 18-month non-treatment observation period [control group vs. treatment group: 0.51 ± 12.09 mmHg vs. -0.35 ± 13.76 mmHg, p = 0.848 (3 month); 1.85 ± 11.21 mmHg vs. 7.31 ± 8.81 mmHg, p = 0.146 (6 months); 6.05 ± 11.14 mmHg vs. 6.61 ± 10.64 mmHg, p = 0.899 (24 months)]). Percentage of diffusion capacity predicted (DLCO%) and percentage of total lung capacity predicted (TLC%), however, were significantly improved in the treatment group by the end of the study (P = 0.010 and 0.027). St. George Respiratory questionnaire (SGRQ) scores were better after 6 months treatment with GM-CSF compared to the control group, and the benefits of treatment were maintained throughout the observation period. No severe side effects were observed during the study. Conclusion Six months of inhaled GM-CSF treatment had no effect on the alveolar–arterial oxygen gradient in patients with mild to moderate pulmonary alveolar proteinosis. There were changes in some clinical or laboratory measures, but no clinically important changes were noted at the end of study. (Clinical Trial Registry: NCT02243228, Registered on September 17, 2014, https://www.clinicaltrials.gov/ct2/show/NCT02243228?term=NCT02243228&draw=2&rank=1)


(Continued from previous page)
Results: There were 36 patients (19 in the treatment group, 17 in the control group) included. There were no significant differences in the primary endpoints as measured by the change of alveolar arterial oxygen gradient (A-aDO 2 ) from the baseline values to the values obtained during treatment or during the following 18-month nontreatment observation period [control group vs. treatment group: 0.51 ± 12.09 mmHg vs. -0.35 ± 13.76 mmHg, p = 0.848 (3 month); 1.85 ± 11.21 mmHg vs. 7.31 ± 8.81 mmHg, p = 0.146 (6 months); 6.05 ± 11.14 mmHg vs. 6.61 ± 10.64 mmHg, p = 0.899 (24 months)]). Percentage of diffusion capacity predicted (DLCO%) and percentage of total lung capacity predicted (TLC%), however, were significantly improved in the treatment group by the end of the study (P = 0.010 and 0.027). St. George Respiratory questionnaire (SGRQ) scores were better after 6 months treatment with GM-CSF compared to the control group, and the benefits of treatment were maintained throughout the observation period. No severe side effects were observed during the study.
Conclusion: Six months of inhaled GM-CSF treatment had no effect on the alveolar-arterial oxygen gradient in patients with mild to moderate pulmonary alveolar proteinosis. There were changes in some clinical or laboratory measures, but no clinically important changes were noted at the end of study. (Clinical Trial Registry: NCT02243228, Registered on September 17, 2014, https://www.clinicaltrials.gov/ct2/show/NCT02243228?term=NCT0224322 8&draw=2&rank=1) Keywords: Autoimmune pulmonary alveolar proteinosis, Granulocyte-macrophage colony stimulating factor, Inhalation Background Autoimmune pulmonary alveolar proteinosis (aPAP, previously known as idiopathic PAP) is a rare interstitial lung disease elicited by the formation of autoantibodies which neutralize the activity of granulocyte-macrophage colony stimulating factor (GM-CSF), consequently decreasing macrophage clearance of surfactant [1]. Currently, the standard treatment strategy for PAP is whole lung lavage (WLL). About 70% patients need another WLL within 3 years due to recurrence [2,3]. Patients who undergo WLL require general anesthesia and double-lumen endotracheal intubation, which means only hospitals with experienced physicians can perform the procedure. Considering the recurrence rate and the cumbersome procedure of WLL, whether or not patients with mild or moderate disease should obtain that treatment is a matter of controversy.
Inhaled GM-CSF therapy has become an attractive alternative option for aPAP patients not only due to its effectiveness and safety [4,5], but also because it is a convenient treatment method for patients who are reluctant to receive WLL. Previous studies included small sample sizes, and as a result, disease severity has not been stratified. Nevertheless, whether patients with mild or moderate disease will benefit from GM-CSF treatment over the long term is still unclear. We prospectively evaluated if inhaled GM-CSF would delay disease progression in patients with mild-to moderate aPAP over a two-year period.

Participants
Patients with mild or moderate aPAP, aged between 18 and 80 years old, were enrolled at two hospitals, including Peking Union Medical College Hospital (PUMCH) and The Affiliated Drum Tower Hospital of Nanjing University Medical School in China.
The inclusion criteria included: (1) patients with a clinical diagnosis of PAP by high-resolution computed tomography (HRCT), further pathologically confirmed by testing for amorphous periodic Acid-Schiff (PAS)positive granules; (2) a positive serum GM-CSF antibody test which indicated an elevated serum GM-CSF antibody level; and (3) patients eligible for the trial should have progressive or unremitting PAP, defined as worsening or unchanging PaO 2 or A-aDO 2 over a 3-month period of observation. PAS-positive granules were found either in milky broncho-alveolar lavage fluid (BALF) or in alveolar structures of lung biopsy tissues which were obtained as follows: cytological findings of bronchial lavage fluid only (BAL) (n = 13), transbronchial lung biopsy only (TBLB) (n = 7), both TBLB and BAL (n = 12), percutaneous lung puncture biopsy (n = 3), and surgical lung biopsy only (n = 1). The GM-CSF antibody test was performed according to the method established by Uchida et al. [6,7]. Our hospital set the cutoff point at 2.39 g/mL, with measurements in excess of this value resulting a positive report [8].
Disease severity was assessed with a disease severity score (DSS), with patients with a DSS of 1 to 3 inclusive being included in our study. DSS scores were defined by Inoue et al. as follows [3]: Grade 1: No symptoms and an arterial oxygen partial pressure (PaO 2 ) ≥ 70 mmHg; Grade 2: PaO 2 ≥ 70 mmHg with symptoms; Grade 3: PaO 2 between 60 and 70 mmHg; Grade 4: PaO 2 between 50 and 60 mmHg; and Grade 5: PaO 2 below 50 mmHg.
Individuals were excluded if they met the following criteria: (1) patients had already received previous GM-CSF therapy or other forms of cytokine therapy, or had undergone lung lavage therapy within the 3 months prior to enrollment; (2) Individuals with PAP resulting from other conditions (e.g. occupational exposure to silica, underlying human immunodeficiency virus infection, respiratory infections, myeloproliferative disorder or leukemia); (3) Individuals with histories of severe allergic or anaphylactic reactions to humanized or murine monoclonal antibodies; (4) Individuals with chronic lung diseases or any other serious medical conditions, or (5) Women who were pregnant, lactating or planned to become pregnant during the study period.

Study design
This was a multicenter, randomized, open-label clinical trial (clinical trial number: NCT02243228, Inhalation of granulocyte-macrophage colony stimulating factor for autoimmune pulmonary alveolar proteinosis) comprising three sequential periods: high-dose therapy for 3 months, low-dose therapy for 3 months and observation for 18 months. Study visits during treatment were designed at 0, 1, 3 and 6 months. Thereafter, patients were followed up by visits at 9, 12, 15, 18, 21 and 24 months (Fig. 1). Patients' safety questionnaires were reviewed by telephone at 1, 15 and 21 months. Before the therapeutic trial, all participants entered an initial 3-month observation period, during which disease severity and progression were evaluated. Participants that had their PaO 2 increase by 10 mmHg or more, or alveolar-arterial oxygen gradient (A-aDO 2 ) decrease by 10 mmHg or more were regarded as having undergone spontaneous improvement and were excluded from enrollment. It should be noted that if a participant was acquainted to the principal investigator as a patient with a welldocumented history showing an unremitting aPAP state, he/she could be enrolled into the study without this observation period. After 3-months of observation, all unremitting PAP patients underwent a stratified randomization based on their DSS at the time of enrollment to ensure equal representation of patients with various disease severities in both the treatment group and the placebo group using a random number table. The randomization was blinded to both the patients and the investigators.
Recombinant human GM-CSF (rhGM-CSF) was administered to patients in treatment group by inhalation as previously described [9]. 150 μg of rhGM-CSF was dissolved in 2 ml of sterile saline, and was inhaled as an aqueous aerosol using an LC-PLUS nebulizer with a manual interrupter valve connected to a PARI Turbo BOY compressor (PARI GmbH, Starnberg, Germany) [10]. Treatment was designed according to a previous study [11], including 3 months of high-dose GM-CSF administration (150 μg twice daily every other week) and another 3 months of low-dose administration (150 μg once daily every other week), serving as induction and maintenance therapy, respectively.
In the control group, patients were not prescribed any kind of treatment related to PAP (including GM-CSF, WLL or anti-CD20, etc.) but had the same follow up plan as the treatment group.
In previous published studies, patients inhaling GM-CSF had a mean change in A-aDO 2 of 11 mmHg [11]. Thus, the target sample size was 25, chosen to give a detection power of 90%, allowing for a 5% incidence of type-I error. After taking other outcome measurements and participant dropout into consideration, the study size was increased to 35-45 patients.
The study was approved by the institutional review board of PUMCH (Approval No. S-717) and reviewed by the institutional review boards at the Affiliated Drum Tower Hospital of Nanjing University Medical School. Fig. 1 The clinical trial comprised of three sequential periods: high-dose therapy for 3 months (150 μg twice daily every other week), low-dose therapy for 3 months (150 μg once daily every other week) and observation for 18 months. Study visits during treatment were designed at 0, 3 and 6 months. Thereafter, patients were followed up with using visits at 9, 12, 18 and 24 months. Patients' safety questionnaires were reviewed by telephone at 1. 15 and 21 months All participants provided informed written consent prior to enrollment.

Assessments
Clinical information including history, symptoms, serological (including lactate dehydrogenase, carcinoembryonic antigen levels) and radiological features, pulmonary function testing results and physical examination results were obtained at each visit during the study. Arterial blood gas analysis (ABG) tests were performed with patients that had been breathing room air for at least 15 min. A low dose quantitative computed tomography of the chest (in PUMCH) or high resolution CT (HRCT) of the chest (in The Affiliated Drum Tower Hospital of Nanjing University Medical School) was obtained before and after GM-CSF therapy and evaluated in a blinded fashion by a board-certified radiologist. The original CT measurements were collected, and the total lung volume and mean lung density were automatically calculated and post-processed with Pulmo 3D (syngo. Via, version VA 30, Siemens Healthcare, Germany) for the automatic segmentation of the pulmonary parenchyma by excluding the intrapulmonary vessels following the process published by one of our co-authors, Dr. Sui [12].
Intergroup differences in the change of A-aDO 2 from baseline to the end of treatment were defined as primary endpoints.
Other data, representing the efficacy of GM-CSF inhalation, were also evaluated as secondary endpoints, including pulmonary function test differences between the treatment group and the control group (forced vital capacity [FVC], total lung capacity [TLC], diffusing capacity for carbon monoxide [DLCO] or diffusing capacity for carbon monoxide corrected for alveolar volume [DLCO/ VA]), 6 min walking distance differences between the groups, and relapse time in the two groups. The definition of relapse was as follows: 1) a new requirement for whole lung lavage (WLL) or any other kind of treatment (including traditional medicine, subcutaneous injection or GM-CSF inhalation) due to disease progression; or 2) PAP death; or 3) reduction in PaO 2 of more than 10 mmHg, or increase in A-aDO 2 of more than 10 mmHg; or 4) a worsened chest HRCT independently confirmed by two physicians. Monitoring for adverse events was conducted during the study, looking for airway hypersensitivity, fever, mylagia, arrhythmia and potential effects on the circulatory system.
All blood tests were performed in laboratories affiliated with the two hospitals, both of which have China's quality management certification. Serum levels of GM-CSF antibody were tested at PUMCH.
All of the data was collected and stored in the database system founded by Beijing Yikang Healthcare Technology Co.

Statistical analysis
All statistical analyses were performed using SPSS 20.0 software. Numeric results were presented as either the mean ± SD or the median and inter-quartile range. Metric variables were shown as the mean and categorical variables were given in terms of frequencies and percentages. The X 2 test was used to analyze proportions of variables. For group comparisons, the unpaired t tests and Wilcoxon rank-sum test were used to evaluate the differences in normally distributed variables. Kaplan-Meier Curve analysis was used to analyze time for relapse in the two groups. All P values reported were two-sided.

Baseline demographic information
Forty-two aPAP patients were screened and 36 patients were randomized (19 in the treatment group and 17 in the control group). After 24 months of follow up, 26 patients (72.2%, 15 from the treatment group and 11 from the control group) completed the study. The period of recruitment and follow up was from July 20, 2014 to July 6, 2018 after the last enrolled patient completed his 24 months follow up. In the treatment group, one patient deteriorated at 3 months and required rescue therapy (WLL). Another patient lost follow up at 1 month and two more patients withdrew at 6 months. In the control group, 4 patients deteriorated at 3 months and required rescue therapy (one received GM-CSF inhalation, two received WLL and one was prescribed traditional medicine). 2 patients withdrew at 21 months. (Fig. 2).
Demographic information of the 36 patients entered the study is shown in Table 1. There were no significant differences in demographic information between the treatment group and the control group including age and sex. No significant differences were found in patients' disease severity markers at baseline, including symptoms, ABG, pulmonary function tests, 6 min walking distance (6MWD) and anti-GM-CSF antibody levels between the two groups.

Primary endpoint: A-aDO 2
There were no significant differences between the treatment group and control group based on primary endpoints measured by the change of A-aDO 2 from baseline to 3 and 6 months treatment and during the following 18 months [control group vs. treatment group: 0.51 ± 12.09 mmHg vs. -0.35 ± 13.76 mmHg, p = 0.848 (at 3 months); 1.85 ± 11.21 mmHg vs. 7.31 ± 8.81 mmHg, p = 0.146 (at 6 months); 6.05 ± 11.14 mmHg vs. 6.61 ± 10.64 mmHg, p = 0.899 (at 24 months)] (Fig. 3a). The change of PaO 2 level from baseline to 3 and 6 months treatment, and during the following 18 months also showed no significant difference between the two groups (Fig. 3b). The actual level of A-aDO 2 and PaO 2 showed no differences during both the treatment period and follow up period as well ( Fig. 3c and d) (Tables 2 and 3).
The diffusion capacity and total lung capacity were improved by the end of study Significant differences in DLCO% and TLC% between the treatment group and the control group emerged by the end of the study (Fig. 4a and b (24 months)]. However, there were no significant differences between the treatment group and the control group in terms of other pulmonary function tests, including FVC, FEV 1 (data not shown) and DLCO/ VA, both during the 6-month treatment period and the 18-month follow-up period (Fig. 4c and d) ( Table 2 and supplemental table 1). The SGRQ scores increased after 3 months and 6 months of inhaled GM-CSF treatment and 18 months follow-up Meanwhile, we can see obvious differences in patients' quality of life between the treatment group and control group, as measured by SGRQ. Total SGRQ scores in patients in the treatment group was improved after 6 months of GM-CSF treatment compared to the no treatment group, and the benefits were nearly continuously maintained throughout the 18-month observation period. Similar trends can be observed in symptom score, activity score and effect score, although significant differences between the two groups were not observed at all times (Fig. 5).

Quantitative CT did not find difference after treatment
There were no significant differences in total lung volume and mean lung density between the treatment  Table 2 and supplemental  Table 1).

Time to rescue therapy during the 24-month study was not improved
There was no significant difference in time to rescue therapy between the treatment group and control group. Kaplan-Meier Curve analysis for the two groups was shown in Fig. 6 (P = 0.304).

Safety and tolerability of inhaled GM-CSF
None of our patients died during the trial. None of the patients in the treatment group complained of fever, wheezing or coughing over the duration of their inhaled GM-CSF treatment. An increase in transaminase levels during GM-CSF inhalation treatment was observed (P = 0.037). Fortunately, none of the patients required medical intervention. The highest observed level of transaminase in the GM-CSF treatment group was 123 U/L for alanine aminotransferase (ALT) and 63 U/L for aspartate aminotransferase (AST), while the highest observed level in the control group was 95 U/L for ALT and 77 U/L for AST respectively. All parameters remained stable or gradually declined after patients ceased alcohol consumption and stopped taking medications with possible interfering effects. Peripheral blood neutrophils levels were not obviously increased within the GM-CSF inhalation group when compared to the control group (P = 0.429) ( Table 4).
No other significant safety and tolerability differences were observed between the two groups during the study.
Other details of side effects that occurred during the study can be found in the supplementary data.

Discussion
In this study, we prospectively evaluated the effects of inhaled GM-CSF on mild-to-moderate autoimmune pulmonary alveolar proteinosis (aPAP) patients. In contrast to previous reports, no obvious effects were found in our study. During the 6 months of treatment and 18 month of subsequent observation, the primary endpoint, A-aDO 2 remained unchanged. Health-related quality of life, measured using SGRQ improved after 3 months of treatment, with these improvements and maintained to 24 months. Marginal improvement was also noted in terms of TLC and DLCO by the end of the study. This research provides valuable clinical data and experience for inhaled GM-CSF treatment in aPAP patients who do not meet the criteria for WLL.
The current therapy for PAP patients involves the physical removal of surfactant using a procedure in which the lungs are repeatedly filled with saline and emptied -WLLwhich is invasive, inefficient, and is not widely available. Some authors reported that fever, hypoxemia, fluid leakage and other complications occurred in patients treated with WLL [13]. Additionally, the median time to next WLL is around 15 months [14], and about 30-57.6% of patients require further therapy after their first WLL [15,16]. Though there is currently no consensus, most physicians believe that patients with PaO 2 less than 70 mmHg when breathing room air or an alveolar-arterial [A-a] oxygen gradient of more than 40 mmHg, or patients with disease progression should receive WLL as treatment [13]. In a cohort study from our center, 33% of patients are stable or experience spontaneous remission [16], while the spontaneous remission rate varies from 8 to 18% in different reports [11,14,16,17]. Considering the occurrence of spontaneous remission in some patients, it becomes a critical question whether GM-CSF inhalation could become a primary treatment for mild to moderate aPAP patients.
After GM-CSF was confirmed to play an important role in the disease mechanism of aPAP, the efficacy of exogenous GM-CSF replacement was assessed in a previous paper. The response rate to this treatment varied, with the efficacy rate being between 62 and 100% when using inhaled GM-CSF [5,11,18] while the efficacy rate was 43-75% when using subcutaneously administered GM-CSF [19,20]. Due to its better responsiveness and tolerance, the use of inhaled GM-CSF is generally recommended [4].
In previous studies, inhaled GM-CSF treatment was prescribed to patients with moderate to severe disease [9,[21][22][23][24], with the mean PaO 2 level in a large prospective study of inhaled GM-CSF treatment on aPAP patients being 61.7 ± 1.4 mmHg [11]. During the preparation of our manuscript, a randomized placebocontrolled study of inhaled GM-CSF was published, with A-aDO 2 and CT density quantitative measurement being Abbreviations: see Table 1 significantly improved, though they concluded that clinical benefits were not significant [25]. There are two major differences in design between our study and Tazawa et al's study as follows: (1)  We believe GM-CSF could be beneficial for those with a PaO 2 of less than 70 mmHg. Additionally and critically, a significant difference in DLCO, which is a relatively reliable parameter to track improvement in aPAP patients, was observed after 24 months. This suggests that inhaled GM-CSF should be provided to patients in a personallytailored manner, providing the drug for as long as is necessary for patients to respond to the drug [22,26]. Longer treatment periods may be necessary for future clinical trials to confirm the efficacy of the inhaled GM-CSF in aPAP over the long term. We found that inhaled GM-CSF therapy is a welltolerated choice for aPAP patients as previous studies showed [9,11,21,22]. Although more than half of our Fig. 4 There were significant differences in the DLCO% and TLC% between the treatment group compared to the control group at the end of the study (P < 0.05, respectively, a and b). However, there were no significant differences between the two groups in terms of other pulmonary function tests, including FVC and DLCO/VA, both during the 6-month treatment period and the 18-month follow-up period (c and d patients in the GM-CSF group were found to have slight increases in amino-transferase levels, and a number of abnormal liver function results were observed in the GM-CSF treatment group, the increases in transaminase levels were all slight and no medical intervention was needed for any of the patients. All patients remained in stable condition or gradually improved after the cessation of alcohol and stopping intake of possibly related combination medicines. Therefore, inhaled GM-CSF therapy is a safe and convenient choice for patients. Our research has some limitations. Firstly, the sample size of the study was small, and there were not enough DSS3 patients for analysis. Our estimated target sample size was based on prior results, done using a patient population with greater disease severity [11], which may have caused an underestimation of the sample size actually needed. Secondly, more patients from the control group, compared to patients from the GM-CSF group, dropped out of the study during the observation period, which may have affected the evaluation of effectiveness when comparing these two groups. Thirdly, the patients in our study did not receive a tailored dosage of GM-CSF treatment, nor did they receive prolonged therapy after the initial 6 months of treatment, which may have caused some latent responders, requiring higher dosages or longer treatment time for a positive response, to remain hidden. The SGRQ score during the 24-month study period. a: total SGRQ score; b: SGRQ symptom score; c: SGRQ activity score; d: SGRQ effect score. *:P < 0.05, **: P < 0.01 Fig. 6 There was no significant difference between the treatment and control groups in terms of no disease progression rate (measured by time to rescue therapy) over the 24 month duration of the study. P > 0.05