Idiopathic pulmonary fibrosis (IPF) is a progressive fibrotic lung disease of unknown etiology with a median survival of 2-3 years from the time of diagnosis. At present, high-resolution computed tomography (HRCT) is an essential component of the diagnosis of IPF [
Krebs von den Lungen-6 (KL-6), surfactant protein-A (SP-A), and SP-D are type II pneumocyte-derived molecules which have been investigated by our group and other investigators for their usefulness as serum biomarkers of IPF [
In addition to the abovementioned type II pneumocyte-derived biomarkers, recent reports indicate that matrix metalloproteinase-7 (MMP-7) and CC-chemokine ligand 18 (CCL18) are potential diagnostic and prognostic markers of IPF. MMP-7 has been shown to be upregulated in the lungs in IPF, particularly in alveolar macrophages and hyperplastic epithelial cells [
The aim of this study was to perform direct comparisons of the abovementioned five serum biomarkers as disease markers for IPF. We evaluated the serum levels of the five biomarkers in patients with IPF and control subjects, which consisted of patients with bacterial pneumonia (BP) and healthy controls (HC), and determined the relative values of these biomarkers in discriminating IPF patients from control subjects. Moreover, we examined independent predictive values of serum markers for survival of patients with IPF and tested their additive predictive ability compared with clinical information using the C statistic.
Sixty-five patients with IPF, 31 patients with BP, and 101 HC were included in the present study. IPF was diagnosed by clinical features, laboratory findings, chest HRCT, and/or surgical lung biopsy, according to the ATS/ERS/JRS/ALAT statement [
Spirometric measurements, including vital capacity (VC) and forced expiratory volume in 1 second (FEV1), were performed according to the ATS/ERS recommendation [
Blood samples were taken at diagnosis and stored at –80°C until analysis. MMP-7, CCL18, SP-A, and SP-D were measured by commercially available enzyme-linked immunosorbent assay (ELISA) kits (Human Total MMP-7 Quantikine ELISA Kit, R&D Systems, MN; Human CCL18/PARC Quantikine ELISA Kit, R&D Systems, MN; SP-A Test Kokusai-F Kit, Sysmex, Japan; and SP-D EIA Kit Yamasa, Yamasa, Japan). Serum KL-6 levels were measured by sandwich-type electrochemiluminescence immunoassay (ECLIA) using a Picolumi 8220 Analyzer (Eidia, Tokyo, Japan), as previously described [
The results were expressed as the mean ± SD. Demographic characteristics and the levels of serum biomarkers were compared between the subject groups using Bonferroni’s test. The levels of serum biomarkers were further analyzed by receiver operating characteristic (ROC) curves to determine the cut-off levels that resulted in the optimal diagnostic accuracy for each marker between the 65 patients with IPF and the 132 control subjects, including BP and HC. The use of these cut-off levels allowed the calculation of sensitivity, specificity, diagnostic accuracy, and likelihood ratio of the five biomarkers for separating the IPF patients from the control subjects. A likelihood ratio above 10 indicates strong diagnostic evidence [
In the survival analysis of the IPF patients, another ROC curve analysis was conducted to find an optimal cut-off level for the prediction of 5-year survival. The 5-year mortality between two groups was compared using the Kaplan-Meier method and the log rank test. Univariate and multivariate Cox proportional hazards model was used to identify predictors of 5-year survival in IPF patients. Martingale residuals plots were employed to check for assumptions of the proportional hazards and the linearity of each biomarker. The plots were visually evaluated with the help of locally weighted regression scatterplot smoothing [
The mean age of the IPF patients was 69.3 years, and the patients with IPF were significantly older than the HC. The mean pack-years of smoking were 37.6 in IPF patients, which was significantly higher than those in the HC. There was no significant difference in age and smoking pack-years between the patients with IPF and those with BP. In the lung function analysis, the mean % VC in patients with IPF was significantly lower than that in the HC (Table
Subject characteristics.
IPF | BP | HC | |
---|---|---|---|
Subjects ( |
65 | 31 | 101 |
Age, yr | 69.3 ± 8.5 |
67.8 ± 15.0 |
55.9 ± 2.3 |
Sex, M/F | 50/15 | 21/10 | 76/25 |
Pack-years | 37.6 ± 35.4 |
21.5 ± 26.7 | 13.7 ± 21.0 |
Spirometry | |||
% VC, % | 74.5 ± 21.2 |
— | 109.5 ± 13.2 |
FEV1/FVC, % | 83.5 ± 17.0 | — | 80.6 ± 4.9 |
% D |
47.1 ± 15.8 | — | — |
IPF: idiopathic pulmonary fibrosis, BP: bacterial pneumonia, HC: healthy controls, VC: vital capacity, FEV1: forced expiratory volume in 1 second, FVC: forced vital capacity, and D
Data represent the mean
Significant differences versus the HC were evaluated using Mann-Whitney
Baseline serum levels of the five biomarkers in patients with IPF were significantly higher than those in the HC. Moreover, serum levels of MMP-7, KL-6, and SP-D in patients with IPF were significantly elevated compared with those in patients with BP. However, there was no significant difference in the serum levels of CCL18 and SP-A between patients with IPF and patients with BP. Moreover, serum levels of MMP-7, CCL18, SP-A, and SP-D were significantly elevated in patients with BP compared with the HC (Figure
Serum levels of (a) MMP-7, (b) CCL18, (c) KL-6, (d) SP-A, and (e) SP-D in patients with IPF, those with bacterial pneumonia (BP), and healthy controls (HC). The significant differences between the three groups were evaluated using Bonferroni’s test (
ROC curve analysis was used to evaluate the discriminating capability of the five serum biomarkers to differentiate IPF patients from control subjects (Figure
Cut-off values and the discriminatory ability of five biomarkers by ROC curve analysis, which distinguishes IPF patients (
MMP-7 | CCL18 | KL-6 | SP-A | SP-D | |
---|---|---|---|---|---|
AUC | 0.9638 | 0.7036 | 0.9957 | 0.7865 | 0.9242 |
95% CI | 0.9374–0.9901 | 0.6275–0.7815 | 0.9898–1.0020 | 0.7229–0.8501 | 0.8866–0.9619 |
|
|||||
Cut-off value | 5.56 ng/mL | 38.7 ng/mL | 476 U/mL | 44.0 ng/mL | 107.0 ng/mL |
Sensitivity | 87.7% | 66.2% | 96.9% | 66.2% | 84.6% |
Specificity | 93.2% | 67.4% | 98.5% | 76.5% | 88.6% |
Diagnostic accuracy | 91.4% | 67.0% | 98.0% | 73.1% | 87.3% |
Likelihood ratio | 12.9 | 2.0 | 64.0 | 2.8 | 7.5 |
ROC: receiver operating characteristic, IPF: idiopathic pulmonary fibrosis, MMP-7: matrix metalloproteinase-7, CCL18: CC-chemokine ligand 18, KL-6: Krebs von den Lungen-6, SP-A: surfactant protein-A, SP-D: surfactant protein-D, AUC: area under the curve, and 95% CI: 95% confidence interval.
ROC curve analysis in five biomarkers to distinguish IPF patients from control subjects which consisted of patients with bacterial pneumonia and healthy controls.
The median followup period in IPF was 31.0 (95% confidence interval: 26.6 to 35.4) months. To find an optimal cut-off level that could discriminate survivors from nonsurvivors, another ROC curve was drawn (figure not shown). Survival in IPF patients using biomarker levels above or below the cut-off level was estimated using the Kaplan-Meier method. Survival was significantly different between higher and lower levels of MMP-7, CCL18, and KL-6 (Figure S3).
In the univariate Cox analysis, decreased % VC, use of immunosuppressant drugs, and elevated serum levels of MMP-7 and KL-6 were associated with poor survival. In the multivariate analysis, only MMP-7 (hazard ratio (HR), 1.074;
Cox proportional hazards model to predict the 5-year mortality of patients with IPF.
Variables | HR | 95% CI |
|
---|---|---|---|
Univariate analysis | |||
MMP-7 (continuous) | 1.068 | 1.015–1.124 | 0.0109 |
CCL18 (continuous) | 1.007 | 0.999–1.014 | 0.0734 |
KL-6 (continuous) | 1.001 | 1.000–1.001 | 0.0005 |
SP-A (continuous) | 1.006 | 0.999–1.015 | 0.1143 |
SP-D (continuous) | 1.000 | 0.998–1.002 | 0.9180 |
Age | 1.032 | 0.982–1.085 | 0.2128 |
Sex, M | 2.163 | 0.734–6.370 | 0.1616 |
Smoking | 1.468 | 0.546–3.951 | 0.4471 |
% VC (continuous) | 0.965 | 0.942–0.989 | 0.0040 |
Medication |
2.730 | 1.177–6.333 | 0.0193 |
|
|||
Multivariate analysis |
|||
MMP-7 (continuous) | 1.074 | 1.060–1.147 | 0.0336 |
KL-6 (continuous) | 1.001 | 1.000–1.002 | 0.0042 |
% VC (continuous) | 0.981 | 0.954–1.009 | 0.1744 |
Medication |
2.066 | 0.667–6.399 | 0.2086 |
See legends of Tables
As shown in Figure
C statistic for Cox regression models predicting 5-year mortality of patients with idiopathic pulmonary fibrosis.
C index | 95% CI | |
---|---|---|
Covariates |
0.705 | 0.559–0.851 |
Covariates plus MMP-7 (continuous) | 0.741 | 0.605–0.876 |
Covariates plus KL-6 (continuous) | 0.769 | 0.643–0.895 |
Covariates plus MMP-7 + KL-6 (continuous) | 0.816 | 0.707–0.923 |
See legends of Tables
Kaplan-Meier analysis to evaluate the probability of 5-year survival among the three groups which were divided according to the serum levels of KL-6 and MMP-7. The cut-off levels of KL-6 and MMP-7 were 1040 U/mL and 9.67 ng/mL, respectively. The probability of 5-year survival was significantly different among them (
In the present study, we directly compared the diagnostic and prognostic value of five serum biomarkers—MMP-7, CCL18, KL-6, SP-A, and SP-D—in patients with IPF and control subjects. Multivariate Cox analysis showed that serum levels of MMP-7 and KL-6 were independent predictors of prognosis in IPF patients. In addition, IPF patients with elevated levels of both KL-6 and MMP-7 had worse survival rates, and the combination of the two markers with the baseline covariates provided the highest C index. These findings indicated that both MMP-7 and KL-6 were promising prognostic markers of IPF, and a combination of the two markers might improve the survival prediction in patients with IPF. Additionally, we showed that MMP-7 and KL-6 could clearly differentiate IPF patients from patients with bacterial pneumonia and healthy controls, suggesting their potential as diagnostic biomarkers.
In this study, MMP-7 and KL-6 were independent predictors of prognosis in patients with IPF, which was consistent with the results of previous reports [
The present results indicate that MMP-7 could be a promising diagnostic marker of IPF. The ROC curve analysis showed the excellent discriminative capability of MMP-7, as indicated by the high area under the curve (AUC) value (>0.90) and likelihood ratio (>10). It should be noted that the ability of a biomarker to discriminate IPF from BP does not necessarily indicate that the biomarker is sufficient, by itself, for diagnosing IPF, although low false positive rates in BP would be an important feature for diagnostic markers for ILDs as shown in our previous study [
Our study also demonstrated that serum levels of CCL18 had moderate discriminatory ability for differentiating IPF patients from the HC; however, CCL18 was found to be a poor indicator for distinguishing IPF from BP. An in vitro study reported that CCL18 had a chemotactic effect on lung fibroblasts and stimulated collagen production [
The present study had several limitations. First, this study is a retrospective review of patients with IPF prospectively recruited from one tertiary hospital, and only Japanese participants who agreed to join this study were included. Therefore, our results may not be generalized to all patients with IPF. Second, there are distinctions in age and smoking histories between patients with IPF and HC. Third, control groups in this study consisted of only patients with BP, an acute lung disease, and healthy subjects; patients with chronic lung diseases, especially other ILDs, were not included in the control group. Therefore, we did not fully evaluate the diagnostic utility of the serum biomarkers. It should be noted that no clinically useful biomarker for distinguishing IPF from other ILDs has been found. We have previously reported that serum levels of KL-6 were also elevated in patients with hypersensitivity pneumonia (HP) [
Our results showed that both MMP-7 and KL-6 might be a useful prognostic marker of IPF, and a combination of the two markers may improve survival prediction in patients with IPF. Additionally, we showed that MMP-7 and KL-6 could differentiate IPF patients from patients with bacterial pneumonia and healthy controls. These results indicate that the measurement of serum levels of KL-6 and/or MMP-7 could potentially support the diagnosis of IPF and would be useful for identifying vulnerable patients especially when the two markers are used in combination. Further large-scale investigation would be warranted to confirm this finding and to find the best method to use this combination of biomarkers of IPF.
Area under the curve
Bronchoalveolar lavage fluid
Bacterial pneumonia
CC-chemokine ligand 18
Diffusing capacity of the lung for carbon monoxide
Electrochemiluminescence immunoassay
Enzyme-linked immunosorbent assay
Forced expiratory volume in 1 second
Healthy controls
Hypersensitivity pneumonia
Hazard ratio
High-resolution computed tomography
Interstitial lung diseases
Idiopathic pulmonary fibrosis
Krebs von den Lungen-6
Matrix metalloproteinase-7
Nonspecific interstitial pneumonia
Receiver operating characteristic
Surfactant protein-A
Surfactant protein-D
Vital capacity.
Nobuoki Kohno holds a patent on KL-6. The remaining authors have no competing interests.
Kosuke Hamai had full access to all of the data in the study and ensured the data integrity and the accuracy of the data analysis. Kosuke Hamai contributed to data analysis and writing the paper; Hiroshi Iwamoto contributed to data analysis and revision and writing of the paper; Nobuhisa Ishikawa and Hironobu Hamada contributed to the study design and revision of the paper. Yasushi Horimasu, Takeshi Masuda, Shintaro Miyamoto, and Taku Nakashima contributed to the data acquisition, analysis, and interpretation; Shinichiro Ohshimo, Kazunori Fujitaka, Noboru Hattori, and Nobuoki Kohno contributed to the study concept and design.
This work was supported by grants-in-aid for scientific research from the Ministry of Education, Culture, Sports, Science, and Technology of Japan.