Introduction

Idiopathic pulmonary fibrosis (IPF) is a fibrosis interstitial pneumonia of unknown cause. IPF patients usually die within 3-4 years after the diagnosis [1,2,3]. The 5-year survival rate of IPF patients was 53.7%, with chronic respiratory failure being the leading cause of death in IPF patients and acute exacerbations(AEs) being the second leading cause of death in IPF patients (23.5%) [4]. It is characterized by high morbidity and high mortality [5], and the incidence tends to increase with age [GRADE evaluates the results

Grading of Recommendations, Assessment, Development and Evaluations (GRADE) [29] was used to evaluate the results of NMA. Refer to the previously published literature [30, 31] for specific methods: For direct comparisons, the estimated starting point of certainty was “high”, and for indirect comparisons, the starting certainty was reduced to “moderate”.

Data synthesis and analysis

In this study, all network meta-analyses were conducted using a random effects model. Odds ratio (OR) and 95% confidence interval (95% CI) were used for statistical analysis for dichotomous variables and mean difference (MD), and 95% CI were used for continuous variables. P<0.05 was statistically significant. When exact mean and SD values were not reported in the included articles, we used the following methods and referred to the previous literature [32]: for Median (IQR) and Median (range), we used online tool (https://www.math.hkbu.edu.hk/~tongt/papers/median2mean.html) for format conversion; for mean (SE) and mean (95% CI), we used the built-in data conversion tool in Revman 5.3 for format conversion. When the included data were sufficiently similar (heterogeneity test: P>0.1, I2<50%), the NMA can be performed. And consistency models were used simultaneously to evaluate the consistency and inconsistency between data. Due to the inclusion of dual arm studies that directly compare drugs with placebo, only the consistency could be tested rather than the inconsistency.

We ranked the treatment using the surface under the cumulative ranking curve (SUCRA), which is the cumulative relative probability of a treatment being the best option [33, 34]. The higher the rank of SUCRA shows, the higher the level of risk is, for example, a high all-cause mortality value of SUCRA indicates a high all-cause mortality. Influence analysis were performed when there was significant heterogeneity between studies; the funnel plot analysis was used to analyze the publication bias. All the statistical analysis above used Revman 5.3 and Stata 14.0 software.

Results

Literature retrieval results

PubMed, EMbase, CENTRA Land MEDLINE searched 756 literatures initially, 22 of them, containing 24 studies with a total of 6208 patients, were finally included after layer-by-layer screening, including 3387 in the experimental group and 2821 in the control group and including RCTs of 13 drugs (5 of Nintedanib [13, 14, 35, 36], 4 of Pirfenidone [12, 37, 38], 2 of Sildenafil [19, 20], 1 of Ambrisentan [26], 1 of Pamrevlumab [18], 2 of Bosentan [39, 40], 1 of Macitentan [41], 1 of Imatinib [21], 1 of GLPG1690 [23], 1 of Simtuzumab [42], 1 of Warfarin [25], 2 of PRM-151 [22, 43], 2 of N-acetylcysteine [24, 44]). Figure 1 is a literature screening flowchart developed according to the requirements of the PRISMA statement [27]. The basic characteristics of the included studies were shown in Tables 1 and 2.

Fig. 1
figure 1

PRISMA literature screening flow chart

Full size image
Table 1 Basic features of the included study (1)
Full size table
Table 2 Basic features of the included study (2)
Full size table

Quality assessment

The 24 included studies were RCTs, all of which mentioned the use of a randomization method and described the comparability of baseline data between the two groups, and there were no incomplete data being reported or data missing. All of them also described the treatment and outcome measures in the experimental and control groups, and 22 studies described specific allocation concealment methods and blinding. The modified Jadad scale [28] was used to evaluate the 24 included studies, of which 19 studies were with 7 points, 3 studies were with 6 points, 1 study was with 5 points, and 1 study was with 3 points. There were 23 high-quality studies and 1 low-quality study. The results of the quality evaluation were shown in Table S2 in supplemental content.

Cochrane risk of bias assessment results

The results showed that the low risk proportion of random sequence generation in the selection bias of the 24 studies was about 79.76%, the moderate risk was about 17.26%, and the high risk was about 2.98% (Fig. 2), so the included studies had less selection, implementation and measurement bias, and the bias statistics of each study were shown in Fig. 3.

Fig. 2
figure 2

Bias risk percentage

Full size image
Fig. 3
figure 3

Bias risk summary chart

Full size image

Security analysis

SAEs

Of the 24 studies included, a total of 19 studies reported SAEs in the treatment of IPF with 13 drugs, as shown in Fig. 4. Statistical analyses were performed with OR and 95%CI as the effect size, and the results of heterogeneity test showed I2=0%, P=0.623, thus it met the criteria of P>0.1, I2<50%, and the effect sizes could be combined for meta-analysis. The consistency model results showed that the effect sizes (Log OR) of all study were approximately between 0 and 2, indicating that the consistency of the results was credible, as shown in Figure S1 in supplemental content. The results of the network meta-analysis (NMA) showed (Table 3, Figure S2 in supplemental content) that there was no difference in the incidence of SAEs between the 13 drugs and placebo (P>0.05). Comparisons between drugs showed that Warfarin had a higher incidence of SAEs than Bosentan (OR = 2.49, 95% CI [1.06 to 5.89], low certainty of evidence) and GLPG1690 (OR = 16.75, 95% CI [1.06 to 263.92], low certainty of evidence); Ambrisentan had a higher incidence of SAEs than Bosentan (OR=1.88, 95% CI [1.04 to 3.38], low certainty of evidence). The SUCRA ranking showed: Warfarin (89.4) > Ambrisentan(81.6) > Pamrevlumab(80.1) > N-acetylcysteine(66.1) > Simtuzumab(54.1) > Pirfenidone(48.5) > Placebo(48) > Imatinib(44) > Nintedanib(42.7) > Sildenafil(38.7) > Macitentan(37.6) > PRM151(34) > Bosentan(29) > GLPG1690(6.1). Higher values of SUCRA indicate higher incidence of SAEs. As shown in Table 4 and Figure S3 in supplemental content.

Fig. 4
figure 4

Network evidence map of SAEs. A total of 19 studies reported SAEs in the treatment of IPF with 13 drugs: 1 of Ambrisentan, 2 of Bosentan, 1 of GLPG1690, 1 of Imatinib, 1 of Macitentan, 1 of N-acetylcysteine, 5 of Nintedanib, 1 of Pamrevlumab, 2 of Pirfenidone, 1 of PRM-151, 1 of Sildenafil, 1 of Simtuzumab, 1 of Warfarin

Full size image
Table 3 Results of network meta-analysis of the incidence of SAEs
Full size table
Table 4 SUCRA ranking of the incidence of SAEs
Full size table

All-cause mortality

Of the 24 studies included, 16 reported the all-cause mortality of IPF treated with 11 drugs, as shown in Fig. 5. The results of heterogeneity tests showed I2=28.7% and P=0.136, which could be combined for meta-analysis. The consistency model results showed that all study effect sizes (Log OR) were approximately between 0 and 2, indicating that the consistency of the results was credible, as shown in Figure S4 in supplemental content. The results of the NMA (Table 5, Figure S5 in supplemental content) showed that Warfarin had higher all-cause mortality than placebo (OR = 5.63, 95% CI [1.54 to 20.55], moderate certainty of evidence), and there was no difference in all-cause mortality compared with placebo for the remaining 10 drugs (P>0.05). Comparisons between drugs showed that Warfarin had higher all-cause mortality than Bosentan (OR = 4, 73, 95% CI [1.17 to 19.06], low certainty of evidence), Simtuzumab (OR = 5.84, 95% CI [1.44 to 23.61], low certainty of evidence), Imatinib (OR=7.18, 95% CI[1.39,37.05], low certainty of evidence), Pirfenidone (OR=8.17, 95%CI[2.10,31.82], low certainty of evidence), Nintedanib (OR=8.38, 95%CI[2.14,32.84], low certainty of evidence), Sildenafil (OR=11.26, 95% CI [1.31 to 97.22], low certainty of evidence) and Pamrevlumab (OR=11.26, 95% CI [1.62 to 78.32], low certainty of evidence); Ambrisentan had higher all-cause mortality than Pirfenidone (OR = 3.26, 95% CI [1.20 to 8.85], low certainty of evidence) and Nintedanib (OR = 3.34, 95% CI [1.22 to 9.15], low certainty of evidence). The SUCRA ranking showed: Warfarin (96.6) > Ambrisentan(82.9) > N-acetylcysteine(75) > Bosentan(60.9) > Macitentan(54.1) > Placebo(51.3) > Simtuzumab(48.2) > Imatinib(36.4) > Pirfenidone(25.6) > Nintedanib(24.2) > Sildenafil(23.7) > Pamrevlumab(21.2). Higher values of SUCRA indicate higher all-cause mortality. As shown in Table 6, Figure S6 in supplemental content.

Fig. 5
figure 5

Network evidence map of all-cause mortality. A total of 16 studies reported the all-cause mortality of IPF treated with 11 drugs: 1 of Ambrisentan, 2 of Bosentan,1 of Imatinib, 1 of Macitentan, 1 of N-acetylcysteine, 3 of Nintedanib, 1 of Pamrevlumab, 3 of Pirfenidone, 1 of Sildenafil, 1 of Simtuzumab, 1 of Warfarin

Full size image
Table 5 Results of network meta-analysis of all-cause mortality
Full size table
Table 6 SUCRA ranking of all-cause mortality
Full size table

Combining the results from SAEs and all-cause mortality, the scatterplot showed (Fig. 6): the SUCRA values of Pirfenidone, Nintedanib, Sildenafil and Imatinib were lower than those of placebo, the SUCRA values of Simtuzumab, Macitentan and Bosentan were approximately equal to those of placebo, and the SUCRA values of Warfarin, Ambrisentan and N-acetylcysteine were higher than those of placebo. In addition, the SUCRA value of all-cause mortality of Pamrevlumab was lower than placebo, but the SUCRA value of the incidence of SAEs was higher than placebo.

Fig. 6
figure 6

The scatterplot combining the results of the incidence of SAEs and all-cause mortality (SUCRA values). The horizontal coordinate represents SUCRA values for all-cause mortality and the vertical coordinate represents SUCRA values for SAEs.If drugs are positioned further to the upper right of the graph, it means that the higher their SUCRA values, the higher their risk; if drugs are positioned further to the lower left of the graph, it means that the lower their SUCRA values are lower, the lower the risk

Full size image

Effectiveness analysis

FVC (L) absolute change from baseline

A total of 11 of the 24 included studies reported on 9 drugs to treat FVC (L) absolute change from baseline in patients with IPF, as shown in Fig. 7. Statistical analyses were performed by using MD and 95%CI as effect sizes. The results of the heterogeneity test showed I2=6.5% and P=0.382, which could be combined for meta-analysis. The consistency model results showed that all study effect sizes (MD) were approximately between -2 and 0, indicating that the consistency of the results is credible, as shown in Figure S7 in supplemental content. The results of the NMA (Table 7, Figure S8 in supplemental content) showed that the improvement of FVC (L) absolute change from baseline by Nintedanib (MD=-0.08, 95% CI [-0.12 to -0.04], high certainty of evidence) and PRM151 (MD=-0.13, 95% CI [-0.25 to -0.01], moderate certainty of evidence) was better than that by placebo and there was no difference in improvement with placebo for the remaining 7 drugs (P>0.05). The results of the comparison between drugs showed: Nintedanib (MD=-0.12, 95% CI [-0.22 to -0.02], low certainty of evidence) and PRM151 (MD=-0.17, 95% CI [-0.32 to -0.01], very low certainty of evidence) improved FVC (L) absolute change from baseline better than Warfarin. The SUCRA ranking showed: Warfarin (85.2) > Imatinib (73.2) > Placebo(72.6) > Macitentan(61.3) > Bosentan(54.8) > N-acetylcysteine(47.1) > Nintedanib(36.2) > GLPG1690(34) > PRM151(22.7) > Pamrevlumab(13). Higher values of SUCRA indicate higher values of decline in FVC (L) absolute change from baseline. As shown in Table 8 and Figure S9 in supplemental content.

Fig. 7
figure 7

Network evidence map of FVC (L) absolute change from baseline. A total of 11 studies reported the FVC (% predicted) absolute change from baseline in patients with IPF treated with 9 drugs:1 of Bosentan, 1 of GLPG1690, 1 of Imatinib, 1 of Macitentan, 1 of N-acetylcysteine, 3 of Nintedanib, 1 of Pamrevlumab, 1 of PRM-151, 1 of Warfarin

Full size image
Table 7 Results of network meta-analysis of FVC (L) absolute change from baseline
Full size table
Table 8 SUCRA ranking of FVC (L) absolute change from baseline
Full size table

FVC (% predicted) absolute change from baseline

Among the 24 included studies, a total of 9 studies reported the FVC (% predicted) absolute change from baseline in patients with IPF treated with 7 drugs, as shown in Fig. 8. The results of the heterogeneity test showed I2=42.5% and P=0.084, which could be combined for meta-analysis. The consistency model results showed that all study effect sizes (MD) were approximately between -5 and 0, indicating that the consistency of the results was credible (Figure S10 in supplemental content). The results of the NMA (Table 9, Figure S11 in supplemental content) showed no difference in FVC (% predicted)absolute change from baseline improvement with placebo (P>0.05) for 7 drugs and so are the comparisons between drugs (P>0.05). The SUCRA ranking showed: Ambrisentan (84.9) > Warfarin (79.3) > Placebo(70.3) > Sildenafil(62.5) > Nintedanib(36.3) > Pirfenidone(30.8) > PRM151(19.3) > Pamrevlumab(16.6). Higher values of SUCRA indicate higher values of decline in FVC (% predicted) absolute change from baseline. As shown in Table 10 and Figure S12 in supplemental content.

Fig. 8
figure 8

Network evidence map of FVC (% predicted)absolute change from baseline. A total of 9 studies reported the FVC (% predicted) absolute change from baseline in patients with IPF treated with 7 drugs: 1 of Ambrisentan, 1 of Nintedanib, 1 of Pamrevlumab, 2 of Pirfenidone, 1 of PRM-151, 2 of Sildenafil, 1 of Warfarin

Full size image
Table 9 Results of network meta-analysis of FVC (% predicted) absolute change from baseline
Full size table
Table 10 SUCRA ranking of FVC (% predicted) absolute change from baseline
Full size table

The proportion of patients with decline in FVC≥10% predicted

Among the 24 included studies, a total of 10 reported the effect of the proportion of patients with decline in FVC≥ 10% predicted of 5 drugs, as shown in Fig. 9. Statistical analyses were performed by using OR and 95% CI as effect sizes. Although the heterogeneity test results showed that I2=62.7%, P=0.004, which had some heterogeneity, the consistency model results showed that the log OR of all studies were roughly between 0 and 2, indicating that the consistency of the results was reliable, as shown in Figure S13 in supplemental content. The results of the NMA(Table 11, Figure S14 in supplemental content) showed that Nintedanib (OR=1.81, 95% CI [1.23 to 2.66], high certainty of evidence), Pirfenidone (OR=1.85, 95% CI [1.26 to 2.71], high certainty of evidence), and Pamrevlumab (OR=4.11, 95% CI [ 1.25, 13.58], high certainty of evidence) improved the proportion of patients with the decline in FVC ≥10% predicted better than placebo, and the improvement of the remaining 2 drugs were not different from placebo (P> 0.05). the results of comparison between drugs showed: Nintedanib (OR=2.76, 95% CI [1.21 to 6.30], low certainty of evidence), Pirfenidone (OR=2.81, 95%CI [1.23 to 6.42], low certainty of evidence), and Pamrevlumab (OR=6.26, 95% CI [1.54 to 25.40], low certainty of evidence) improved the proportion of patients with decline in FVC ≥10% predicted better than Ambrisentan. The SUCRA ranking showed: Ambrisentan (96) > Placebo(80.4) > Nintedanib(41.8) > Pirfenidone(40.4) > Warfarin(31.3) > Pamrevlumab(10.2). Higher values of SUCRA indicate that the proportion of patients with decline in FVC≥ 10% predicted is higher. As shown in Table 12 and Figure S15 in supplemental content.

Fig. 9
figure 9

Network evidence map of the proportion of patients with decline in FVC≥10% predicted. A total of 10 studies reported the effect of the proportion of patients with decline in FVC≥ 10% predicted of 5 drugs: 1 of Ambrisentan, 4 of Nintedanib, 1 of Pamrevlumab, 3 of Pirfenidone, 1 of Warfarin

Full size image
Table 11 Results of network meta-analysis of the proportion of patients with decline in FVC≥10% predicted
Full size table
Table 12 SUCRA ranking of the proportion of patients with decline in FVC≥10% predicted
Full size table

Combining the results of FVC (% predicted) absolute change from baseline and the proportion of patients with decline in FVC ≥10% predicted, the scatterplot showed (Fig. 10): The SUCRA values of Pamrevlumab, Pirfenidone and Ninedanib were lower than those of placebo, and the SUCRA values of Warfarin and Ambrisentan were higher than those of placebo.

Fig. 10
figure 10

The scatterplot combining the results of FVC (% predicted)absolute change from baseline and the proportion of patients with decline in FVC≥10% predicted(SUCRA values). The horizontal coordinate represents SUCRA values for FVC (% predicted) absolute change from baseline and the vertical coordinate represents SUCRA values for the proportion of patients with decline in FVC ≥10% predicted. If drugs are positioned further to the upper right of the graph, it means that the higher their SUCRA values, the higher their risk; if drugs are positioned further to the lower left of the graph, it means that the lower their SUCRA values are lower, the lower the risk

Full size image

Publication bias

Figure 11 showed that the inverted funnel plots were symmetrical, suggesting that there’s no publication bias. The statistical results of Begg's test and Egger's test were used to detect bias for all outcomes, and the results showed that PBegg>0.05 and PEgger> 0.05, indicating that there was no obvious bias in this study. As shown in Figure S16 and Figure S17 in supplemental content.

Fig. 11
figure 11

The results of inverted funnel plots. a SAEs. b All-cause mortality. c FVC (L) absolute change from baseline. d FVC (% predicted)absolute change from baseline. e The proportion of patients with decline in FVC≥10% predicted. (PLA:Placebo; NIN:Nintedanib; PIR:Pirfenidone; SIL:Sildenafil; AMB:Ambrisentan; PAM:Pamrevlumab; BOS:Bosentan; MAC:Macitentan; IMA:Imatinib; GLPG:GLPG1690; SIM:Simtuzumab; WAR:Warfarin;PRM: PRM151;NAC:N-acetylcysteine.)

Full size image

Influence analysis

Influence analyses for the 5 outcomes (Table S3-S7 and Figure S18 in supplemental content) showed that none of the included studies had clear sensitivity, indicating that there was no significant difference in the results after excluding any of the studies (except FVC (% predicted) absolute change from baseline). It’s proved that the effect size sensitivity of these outcomes was low, and it had good stability, reliability, and stable and reliable analysis results.

Discussion

The total of 24 RCTs on the clinical efficacy of 13 drugs for IPF were included in this NMA with the aim of comprehensively assessing their safety and efficacy in the treatment of IPF and differences in this safety and efficacy. Our results found that Nintedanib and Pirfenidone improved lung function (FVC (L) absolute change from baseline or the proportion of patients with the decline in FVC ≥10% predicted) better than placebo, and they improved lung function( FVC (L) absolute change from baseline or the proportion of patients with the decline in FVC ≥10% predicted) better than Warfarin or Ambrisentan. It’s also found that Pirfenidone and Nintedanib had lower all-cause mortality than Warfarin and Ambrisentan. The SUCRA values for the efficacy and safety of Nintedanib and Pirfenidone were also lower than those of placebo and many other drugs. Therefore, Our result confirmed that Nintedanib and Pirfenidone can significantly slow the decline of lung function in IPF patients with better safety profile than placebo and many other drugs.

Cell signaling pathways activated by tyrosine kinases, such as VEGF, FGF, and PDGF, had been shown to be involved in the pathogenesis of IPF [45,46,47]. Nintedanib (formerly BIBF 1120) is an intracellular inhibitor that targets a variety of tyrosine kinases, including receptors such as VEGF, FGF, and PDGF [48]. Some studies have reported that it can reduce the decline rate of FVC in IPF patients [13, 14]. In a previous phase 2 RCT [TOMORROW], compared with placebo, Nintedanib 150mg bid can better improve the FVC change rate of IPF patients (P=0.01), and the incidence of SAEs was lower (27.1% vs.30.6%) [14]. In two repeated RCTs (INPUTLIS-1 and INPUTLIS-2), it also achieved good results, and the incidence of SAEs and all-cause mortality of patients in the Nintedanib group were also lower than those in the placebo group [13]. In addition, similar results have been reported in two recent clinical trials [35, 36]. Pifenidone is an orally bioavailable synthetic molecule. It regulates the activity of TGF-β and TNF-α [49,50,51,52,53], and can inhibit collagen synthesis and fibroblast proliferation [50, 53,54,55,56]. In one trial, there was a statistically significant difference in VC decline at 9 months between placebo (-0.13 L) and Pirfenidone (-0.03 L) (p = 0.0366). One of the five patients in the placebo group died after an exacerbation episode while there were no deaths in the Pirfenidone group during the 9-month study period, and no SAEs were reported in the Pirfenidone group [15]. One literature later reported on two simultaneous studies (CAPACITY 004 and 006): two studies combined showed the effect of Pirfenidone treatment on predicted percentage FVC at week 72 (p=0.005): -8.5% in the Pirfenidone 2403 mg/day group and -11.0% in the placebo group. In terms of security, the rates of SAEs and mortality were lower in the group with Pirfenidone than in the placebo group [37]. In addition, similar results have been reported in two clinical trials [12, 38]. All these findings showed that Nintedanib and Pirfenidone had good effect on slowing the decline of lung function in patients with IPF and had lower incidence of SAEs and all-cause mortality. Therefore, Nintedanib and Pirfenidone can continue to be vigorously promoted in clinical practice.

It’s also found that Pamrevlumab improved the proportion of patients with decline in FVC≥10% predicted better than placebo and Ambrisentan, and it had lower all-cause mortality than Warfarin. In addition, the SUCRA values for the lung function and all-cause mortality of Pamrevlumab were lower than placebo and all other drugs, but the SUCRA value of the incidence of SAEs was higher than placebo. Pamrevlumab (FG-3019) is a fully human recombinant monoclonal antibody against CTGF [57]. In the mice with pulmonary fibrosis, Pamrevlumab improved lung function and lung airway remodeling, and inhibited collagen production [58, 59]. In an open-label study, 89 patients with IPF were given two doses of Pamrevlumab every 3 weeks for 48 weeks, and good results were shown in lung function and quantitative HRCT changes [60]. In another phase 2 RCT, Pamrevlumab can delay the decline rate of the predicted FVC percentage of patients with IPF at 48th week (Pamrevlumab:–2.9% vs. Placebo:–7.2% ). Of the 3 (6%) deaths in the Pamrevlumab group and 6 (11%) deaths in the placebo group, none were considered as treatment-related. However, a higher proportion of treatment-induced urgent SAEs occurred in the Pamrevlumab group than in the placebo group (Pamrevlumab: 24% vs. Placebo: 15%) [18]. Our study confirmed that Pamrevlumab had a very good slowing effect in the declining of lung function and low all-cause mortality in patients with IPF compared to other drugs, and it’s currently in Phase 3 development and may become an important drug for the treatment of IPF in the future. However, it’s also confirmed in our study that the incidence of SAEs is still relatively high, therefore it should be used clinically according to the actual situation of patients.

The results showed no difference in lung function improvement with placebo for Imatinib (FVC (L) absolute change from baseline) and Sildenafil (FVC (% predicted) absolute change from baseline) , and the SUCRA values were also roughly equal to placebo. It’s also found that Imatinib and Sildenafil had lower all-cause mortality than Warfarin, and the SUCRA values for the safety of Imatinib and Sildenafil were also lower than those of placebo and some other drugs. Sildenafil is a phosphodiesterase-5 inhibitor that has been approved by the U.S. Food and Drug Administration for the treatment of idiopathic pulmonary hypertension (IPAH) [61]. It stabilizes cyclic guanosine monophosphate, the second messenger of nitric oxide, which leads pulmonary vasodilation in patients with IPF and thereby improve gas exchange [62]. In addition, sildenafil also reduced the production of superoxide in the mouse model of pulmonary fibrosis [63]. A previous study reported that Sildenafil (20–50 mg orally 3 times daily for 3 months) resulted in an improvement in 6-MWT distances of patients with IPF [64]. However, in one RCT, there was no difference in FVC (% predicted) changes in 89 patients with advanced IPF in the Sildenafil group compared with 91 patients in the placebo group (mean change=0.32, 95% CI [−1.12 to 1.76], P=0.66), and there was also no significant difference in SAEs and all-cause mortality [19]. In another RCT, there was no significant difference in FVC (% predicted) in IPF patients (P=0.79) between Sildenafil and placebo, and there were few SAEs [20]. Imatinib was approved by the U.S. Food and Drug Administration in 2001 for the treatment of chronic myeloid leukemia and has proven to be very effective. It is a TKI with activity against the PDGFR, DDR, c-kit and c-Abl [65, 66]. Current studies have demonstrated that imatinib inhibits bleomycin-induced IPF [67, 68], and the mechanism of action is closely related to inhibition of lung FMT and inhibition of the ECM produced by PDGF and TGF-β signaling [69]. One phase 2 RCT reported no significant difference between Imatinib and placebo at 96 week’s follow-up in terms of time to disease progression (predicted 10% reduction in FVC percentage from baseline) or time to death, and there was no significant difference in SAEs and mortality [21]. Thus, Sildenafil and Imatinib appear to have lower rates of SAEs and all-cause mortality, but do not have much effect in improving FVC in patients. Our study confirmed that Sildenafil and Imatinib had a better safety profile than other drugs, but their efficacy was less obvious than placebo, so further clinical studies are needed to confirm their effectiveness.

It’s found that PRM151 improved FVC (L) absolute change from baseline in patients with IPF better than placebo and Warfarin. In addition, the SUCRA values for the lung function (FVC (L) absolute change from baseline or FVC (% predicted) absolute change from baseline) and SAEs of PRM151 were lower than placebo and most other drugs. Current clinical study had found very low plasma concentrations of Pentraxin 2 in patients with IPF [70] and had demonstrated in experimental studies that PRM-151 (Recombinant human pentraxin-2) inhibited TGF-β1 and bleomycin-induced pulmonary fibrosis [70,71,72]. Its mechanism of improving pulmonary fibrosis was closely related to inhibition of TGF-β1 production and inhibition of monocytes differentiation into pro-inflammatory macrophages and prefibrous fibroblasts [73,74,75]. In a phase 1 RCT of increasing doses of PRM-151, the results showed a trend towards improvement in FVC and 6-MWT in the combined dose group of PRM-151 and no SAEs [43]. In another phase 2 RCT, PRM-151 was also found to improve FVC from baseline to week 28 as a percentage of predicted value (difference, +2.3 [90% CI, 1.1 to 3.5], P = 0.001) in patients with IPF, and a lower incidence of SAEs [22]. Therefore, PRM151 appears to improve lung function well in patients with IPF, as well as a low incidence of SAEs, and our findings confirmed these views, and it may be a good targeted new drug for the treatment of IPF. However, due to the lack of data on all-cause mortality, further clinical studies are needed to confirm its safety and efficacy.

The results showed no difference in FVC (L)absolute change from baseline improvement with placebo for GLPG1690, but it had lower SUCRA values than placebo and many other drugs. It’s also found that GLPG1690 had lower incidence of SAEs than Warfarin, and GLPG1690 had lower SUCRA values than placebo and all other drugs. GLPG1690 is a potent and selective autoclassifier protein inhibitor and is well tolerated orally in humans [76, 77]. The results showed that GLPG1690 could improve the Ashcroft fibrosis score of mice with pulmonary fibrosis well and inhibit the profibrotic mediator in IPF fibroblasts [76, 78, 79]. In a phase 2a RCT, patients with IPF received either placebo (n=6) or oral GLPG1690 600 mg (n=7) once daily for 12 weeks, and the results showed that at week 12, patients in the GLPG1690 group had an average change in FVC of 25 mL versus -70 mL of placebo. SAEs occurred in 2 patients in the placebo group and 1 in the GLPG1690 group, and no patients died [23]. Combined with our findings, these data showed that although GLPG1690 had a low incidence of SAEs and had a certain effect on improving lung function, the patient sample size was too small, and there’s a lack of data on all-cause mortality, therefore further clinical studies are needed to confirm its safety and efficacy.

The results showed no difference in FVC (L)absolute change from baseline improvement with placebo for N-acetylcysteine (NAC), but it had lower SUCRA values than placebo and some other drugs. It’s also found that there’s no difference in the incidence of SAEs and all-cause mortality of N-acetylcysteine treated with IPF compared to placebo. In addition, the SUCRA values for the incidence of SAEs and all-cause mortality of NAC were also higher than those of placebo and many other drugs. Current studies had shown that NAC was able to directly scavenge oxygen radicals [80], and can inhibit TGF-β signaling in IPF [81]. No difference in NAC 600 mg tid versus placebo was found in one RCT in improving FVC and mortality in people with IPF [24]. Another multicentre RCT also found no difference in efficacy or safety between inhaled NAC and placebo [44]. In addition, a RCT(IFIGENIA), compared NAC + therapy including Prednisone and Azathioprine with Prednisone + Azathioprine + Placebo, found that NAC group can better improve VC and DLCO in IPF patients [82]. However, another study (PANTHER-IPF ) reported that the simultaneous use of these three drugs increased mortality and the incidence of SAEs in patients with IPF [83]. Our study also showed that NAC alone wasn’t evident in terms of effectiveness and there may be a number of adverse events, thus NAC alone is not recommended clinically for IPF.

Procoagulases may directly stimulate fibrosis through cell surface receptor-mediated responses [84]. A previous unblinded study reported a 1-year survival benefit with anticoagulation (Heparin + Warfarin) in patients with IPF [85]. However, in a later RCT, warfarin was associated with an increase in all-cause mortality (14 warfarin versus 3 Placebo deaths; P = 0.005), and the study had to be terminated early due to excessive mortality [25]. Data from preclinical models suggested that the expression of endothelin receptors in IPF lung tissue increased while antagonizing endothelin receptors may reduce the severity of pulmonary fibrosis [86, 87]. Ambrisentan is a selective ETA receptor antagonist that had been approved for treatment of pulmonary arterial hypertension(PAH) [88]. One RCT reported that Ambrisentan treatment for IPF increased mortality (Ambrisentan: 7.9% vs. Placebo: 3.7%) and an increase in the proportion of patients with IPF with decreased lung function (Ambrisentan: 16.7% vs. Placebo: 11.7%). As a result, the study was terminated early [26]. Our findings also confirmed that the safety and efficacy of Warfarin and Ambrisentan were very poor compared to other drugs and placebo. Combing with previous findings that Warfarin and Ambrisentan should not be used clinically for IPF. In addition, several other drugs (Bosentan, Macitentan and Simtuzumab) are less outstanding than placebo in terms of safety and efficacy, and further clinical studies are needed to confirm their efficacy.

Limitations of Inclusion: The 24 studies we included described the outcomes of the experimental group and the control group in detail, but there are some problems remained:1) All the 24 studies reported on randomization, but 5 studies only mentioned randomization and did not give a clear randomization method; 2 studies did not report allocation concealment and 1 did not use blinding. These factors will influence the overall quality of the study to some extent; 2) The evaluation outcomes of the 24 studies were not exactly the same, and the criteria for evaluating the results were not exactly the same, so the results of this study may have some heterogeneity and sensitivity; 3) There were some small studies in the included studies, and there was some clinical heterogeneity (different doses administered, different administration methods, different courses of treatment, different course of disease, different interventions and different disease severity, etc.), which all affected the reliability of the results; 4) Some studies did not perform statistical analysis of important observation outcomes such as FEV1%, FVC%, TLC%, 6MWD, SGRQ, HRCT, inflammatory factors, pulmonary fibrosis factors, etc., which will affect the overall quality of the literature included in the statistics; 5) All the studies were in English instead of other languages.

Conclusion

Nintedanib and Pirfenidone can significantly slow the decline of lung function in IPF patients and have a better safety profile, they can continue to be vigorously promoted in clinical practice. Pamlumumab has a good slowing effect on lung function decline and low all-cause mortality in IPF patients and is currently in phase 3 development, it may become an important drug for the treatment of IPF in the future. Sildenafil and Imatinib have a good safety profile, but their effectiveness is not obvious, and further clinical studies are needed to confirm their effectiveness. Both PRM151 and GLPG1690 seem to have the effect of improving lung function of IPF patients, and the incidence of SAEs is low. However, due to the lack of data on all-cause mortality, further clinical studies are needed for comprehensive evaluation. N-acetylcysteine alone is not evident in terms of efficacy, and there may be a number of adverse events, so NAC alone is not recommended clinically for IPF. Warfarin and Ambrisentan have poor safety and efficacy, therefore they are not recommended for clinical use in the treatment of IPF.

In addition to the above drugs, some other drugs such as Nalbuphine [89], pembrolizumab [90], and Treprostinil [91] have shown good efficacy in clinical and animal experimental studies in IPF or interstitial lung disease. In the future, if there are high-quality RCTs of these drugs for IPF, they can be included and evaluated comprehensively. In addition, in future network meta-analysis, various molecular biomarkers in the field of precision medicine can be considered as efficacy evaluation indicators [92] to screen out more effective drugs for the treatment of IPF.