Hiroaki Satoh MD, Division of Respiratory Medicine, Institute of Clinical Medicine, University of Tsukuba, Tennodai 1-1-1, Tsukuba City, Ibaraki 305-8575, Japan. (fax: +81-29-853-3320; e-mail: firstname.lastname@example.org).
Objectives. KL-6 is a specific marker in patients with interstitial lung diseases (ILDs); however, the relationship between elevated levels of KL-6 and subsequent mortality is not well defined. To determine if elevated serum levels of KL-6 are associated with increased mortality, and to identify the most suitable cut-off level of KL-6 by which to distinguish between good prognosis and poor prognosis, we evaluated the prognostic significance of serum KL-6 levels in patients with stable-state ILDs.
Methods. Two hundred and nineteen patients diagnosed with ILDs (152 with idiopathic interstitial pneumonia and 67 with collagen disease-associated pulmonary fibrosis) at Tsukuba University Hospital from April 1999 to October 2005 were entered in this study. Serum KL-6 levels in patients with ILDs were measured with a commercially available enzyme immunoassay kit, and these patients were then followed up.
Results. During the follow-up period, 58 of the 219 patients died of respiratory failure. Patients who died during this period had higher levels of KL-6 than did those who did not (P = 0.0004). The receiver operating characteristic curve analysis showed 1000 U mL−1 as the most suitable cut-off level by which to distinguish between the two groups of patients. The 95% specificity serum KL-6 level with poor outcome was 2750 U mL−1. In univariate and multivariate analysis, elevated serum KL-6 (>1000 U mL−1) in the stable state indicated poor prognosis (P = 0.0005, log-rank test; P = 0.0001, Cox proportional hazard model).
Conclusions. Elevated KL-6 level may provide simple, yet valuable information by which to identify patients with ILDs who are at increased risk for subsequent mortality.
KL-6 is a mucin-like glycoprotein with a high-molecular weight and is strongly expressed on type II alveolar pneumocytes and bronchiolar epithelial cells . In interstitial lung diseases (ILDs), type II cells are regenerated over the alveolar basement membrane after the death of type I cells during the first stage of lung injury . Circulating serum KL-6 is believed to derive from type II cells, as the existence of KL-6 in regenerating type II cells was confirmed immunohistochemically [3, 4]. The elevation of serum KL-6 level is also associated with increased permeability of the alveolar capillary barrier . Several studies demonstrated that KL-6 is elevated in both the bronchoalveolar lavage fluid and serum of patients with various types of ILDs [3, 4, 6–17]. Therefore, measurement of serum KL-6 is now widely accepted as a diagnostic test to monitor the activity of ILDs [3, 4, 6–17].
Additionally, Kohno and co-workers  demonstrated that KL-6 might be a useful marker for evaluating therapeutic effect in rapidly deteriorating idiopathic pulmonary fibrosis (IPF). Thereafter, some authors reported that patients with various types of collagen disease-associated pulmonary fibrosis (CDPF) also had elevated serum levels of KL-6 [9, 11, 12, 16]. During an acute exacerbation of ILDs including idiopathic interstitial pneumonia (IIP) as well as CDPF, levels of KL-6 seem to increase in serum, making prediction of their therapeutic efficacy . However, it is not known whether KL-6 at the time of initial measurement in patients with stable-state ILDs could provide predictive information about subsequent mortality rather than short-term outcomes. In this study, we evaluated the prognostic significance of KL-6 level at the time of initial measurement and provided the cut-off KL-6 level for defining unfavourable prognosis.
Methods and materials
We conducted a prospective observational study of serum KL-6 levels in patients with ILDs. This study was approved by the institutional ethics committee of the University of Tsukuba. Consecutive patients diagnosed with IIP and CDPF at Tsukuba University Hospital from April 1999 to October 2005 were entered in this study. Patients with IIP were diagnosed on the basis of the American Thoracic Society (ATS)/European Respiratory Society (ERS) international multidisciplinary consensus classification of the IIPs . Amongst IIPs, diagnosis of IPF was made on the presence of all major ATS/ERS criteria as well as at least three of the four minor criteria . Diagnosis of nonspecific interstitial pneumonia (NSIP) was made on the presence of predominantly basal ground-glass opacity (GGO) and/or reticular pattern as well as the presence of major and minor criteria for diagnosis of IPF other than imaging features . Similarly, cryptogenic organizing pneumonia (COP), and desquamative interstitial pneumonia (DIP) were diagnosed on the presence of patchy peripheral or peribronchovascular consolidation (COP), as well as smoking-related lung diseases characterized by GGO and centrilobular nodules (DIP), and the presence of criteria for IPF other than imaging features . The patients with CDPF were clinically classified as having rheumatoid arthritis, polymyositis and dermatomyositis, systemic sclerosis, systemic lupus erythematosus, Sjogren's syndrome, or mixed connective tissue disease using the criterion for each disease [19–24]. In the present study, patients with acute exacerbation, which was defined as increased dyspnoea (an increase of more than 1 grade in the Hugh-Jones classification system or a decrease in PaO2 >10 torr during the prior 2 months) , were excluded from this study.
Peripheral venous blood samples collected from patients with IIP and CDPF were used for the KL-6 assay. The samples had been stored at −30 °C until use.
Measurement of KL-6 levels
Serum KL-6 levels were measured by a sandwich-type enzyme-linked immunosorbent assay technique using a KL-6 antibody kit (Eisai, Tokyo, Japan). The recommended cut-off value was determined at 500 U mL−1 from the levels of healthy individuals reported previously . The assay was performed by technicians who had no clinical information regarding the samples. Twenty-two control subjects (13 patients with chronic obstructive pulmonary disease; nine patients with bronchial asthma) with an average age of 69 years were also studied. The median serum levels of KL-6 in the control group were 306 U mL−1 (interquartile range: 214–356). There were no subjects showing elevated serum levels of KL-6 > 500 U mL−1.
The Mann–Whitney U-test was applied to elucidate the difference between the two independent groups, and proportion was compared by the chi-square test. To determine the most suitable cut-off level, we used the receiver operating characteristic (ROC) curve analysis . The Kaplan–Meier method was used to assess survival curves and the log-rank test to evaluate the statistical significance of differences found between the two groups. The multivariate Cox proportional hazard model was also used in this study. P < 0.05 was considered significant.
During the study period, 219 patients (152 patients with IIP and 67 patients with ILD-CD) were entered in this study (Table 1). On high-resolution computerized tomography (CT) scan, 183 patients showed an IPF/usual IP pattern, 30 an NSIP pattern, four a COP pattern and two a DIP pattern. Pathological confirmation on surgical lung biopsy specimens was performed in the patients with COP and DIP, respectively.
Table 1. Characteristics of patients with interstitial lung diseases
There was no statistical difference in the serum KL-6 levels of the 152 patients with IIP (median: 943 U mL−1; range: 182–9000) and those of the 67 patients with CDPF (median: 912 U mL−1; range: 105–6770; P = 0.4915).
The median follow-up period of the 219 patients was 20 months (range: 1–72). Within the study period, 58 patients died of exacerbation of ILD (nonsurvivors). Overall mortality was 26.5%. The median period from the measurement of KL-6 to death in the 58 patients was 9 months (range: 1–60). The ratio of carbon monoxide diffusion capacity to alveolar ventilation (DLco/VA) in nonsurvivors (mean: 45.7%) was not lower than that in patients who survived (survivors; mean: 53.4%; Mann–Whitney U-test, P = 0.2538). However, there was a statistical difference in vital capacity (VC) between them (mean: 59.1% vs. 78.2%; P = 0.0021). Serum KL-6 levels in the 58 nonsurvivors (median: 1330 U mL−1; range: 297–6770) were significantly higher than those of the 161 survivors (median: 823 U mL−1; range: 105–9000; P = 0.0004).
In the univariate and multivariate analysis, elevated levels higher than the recommended cut-off value of KL-6 (>500 U mL−1) indicated poor prognosis (P = 0.0174, log-rank test; P = 0.0213, Cox proportional hazards model), respectively.
To better define whether elevated levels of KL-6 at the time of initial measurement in patients with stable-state ILDs could predict the likelihood of death during the follow-up, we performed an ROC curve analysis. We examined the sensitivity and specificity of various cut-off values of serum KL-6 at the time of initial measurement for predicting survival. On the basis of ROC analysis, the optimal point on the ROC curves for discriminating between survivors and nonsurvivors corresponding to KL-6 was 1000 U mL−1. When we applied the optimal cut-off level of 1000 U mL−1, its sensitivity was 67.2% and specificity was 60.2%. The 95% and 80% specificity serum KL-6 level with poor outcome was 2750 and 1670 U mL−1, respectively. The respective sensitivity at 80% and 95% specificity was 44.8% and 20.7%.
Figure 1 depicts the Kaplan–Meier survival curves. The patients are divided into two groups according to the optimal cut-off level (1000 U mL−1) of KL-6. As illustrated in Fig. 1, the group with elevated serum KL-6 level (1000 U mL−1) at the time of initial measurement showed significantly higher mortality rates (P = 0.0005, log-rank test). Table 2 summarizes the relationship between clinical variables and survival. According to a multivariate Cox proportional hazards model, of the noninvasive variables, elevated serum KL-6 (>1000 U mL−1) at the time of initial measurement was an important predictor of outcome (odd ratio: 2.95, 95% confidence interval: 1.71–5.08, P = 0.0001).
Table 2. Multivariate analysis of prognostic factors of interstitial lung diseases
The primary cellular source of KL-6 is thought to be type II pneumocytes and respiratory bronchiolar epithelial cells, as KL-6 is expressed on these cells in normal lungs  and is strongly expressed on regenerating type II pneumocytes and alveolar macrophages in IIP lungs . The concentration of KL-6 was estimated to be extremely high in epithelial lining fluid [26, 27]. An increase in circulating KL-6 levels in patients with IIP is therefore thought to be due to an increase in KL-6 production by regenerating alveolar type II pneumocytes and/or to enhanced permeability following destruction of the air–blood barrier in the affected lungs . Therefore, KL-6 levels are thought to be significantly correlated with the extent of the opacities detected by chest CT scan in patients with ILDs irrespective of their aetiology.
Several recent reports have demonstrated that serum KL-6 levels in the acute phases of ILDs are a powerful predictor of outcome of treatment [10–12, 15, 16]. However, to our knowledge, there have been no reports showing a relation between serum KL-6 level at the time of initial measurement and subsequent mortality in patients with ILDs. The objective of this study was to determine whether elevated serum levels KL-6 at the time of initial measurement in patients with stable-state ILDs could predict subsequent mortality and provides the most suitable cut-off level for defining unfavourable prognosis. In this study, we found that patients who died within the 6-year observation period showed significantly higher levels of KL-6 than those who were still alive. In addition, ROC analysis documented that elevated levels of KL-6 (>1000 U mL−1) predicted subsequent death from respiratory failure with clinically acceptable sensitivity and specificity. These findings suggest that KL-6 assay is useful in identifying ILD patients with poor prognosis.
There have been no studies which clearly indicated the cut-off level of KL-6 for unfavourable survival in patients with stable-state ILDs. Kohno and co-workers  evaluated the serum KL-6 levels of 14 patients with rapidly progressive IIP. In their report, KL-6 levels in patients who responded to steroid therapy significantly decreased at weeks 1 and 3 after steroid treatment. On the other hand, KL-6 in patients who did not respond to the steroid therapy tended to increase at weeks 1 and 3 after the treatment. However, they could not show the difference in the initial KL-6 levels between the six patients who died and the eight who survived, and they did not describe the most suitable cut-off level . Fukaya et al.  reported that serum KL-6 levels in patients with CDPF who remained stable for more than 6 months varied widely between low (207 U mL−1) and high (3726 U mL−1). In patients with acute respiratory distress syndrome (ARDS), Ishizaka et al. indicated that the optimal cut-off KL-6 value for unfavourable outcome was 253 U mL−1 . In contrast, Sato et al.  showed successfully treated ARDS patients whose serum levels were nevertheless higher than the level recommended by Ishizuka et al. . In paediatric patients with fatal measles pneumonia, Narita et al. reported that all of the deceased patients had serum KL-6 levels of more than 2000 U mL−1, and that no patients with KL-6 levels of <1000 U mL−1 died . By contrast, it should be noted that the patients in the present study were patients with the most usual slowly progressive ILD, rather than the rapidly deteriorating patients who had been reported in those previous studies [10, 13, 28–30]. Therefore, the optimal cut-off level, which we recommend, is apparently different from that for such acute clinical conditions. Recently, Goto et al.  reported that a more than 1.5-fold increase in KL-6 level was associated with serious radiation pneumonia that was refractory to steroid therapy. Changes in serum KL-6 levels in serial measurement may have a great importance; however, we cannot say for certain whether the hypothesis will be proved, as we have no definite information on the serial measurements of KL-6 and changes in its level.
This study has several limitations that need to be addressed before serum KL-6 at the time of initial measurement can be used clinically to predict subsequent mortality. First, the number of patients in this study is small. Secondly, this study included not only patients with IIP, but also those with CDPF, although we showed no statistical difference in KL-6 levels in both ILDs. The differences in treatment of these patients may have affected prognosis. Therefore, these limitations may have biased the results of the present study.
In conclusion, we found that patients who died within the study period showed significantly higher levels of KL-6 than did those who were still alive. Additionally, ROC analysis documented that an elevated level of KL-6 (>1000 U mL−1) predicted subsequent death from respiratory failure. These findings suggest that KL-6 assay is useful for identifying ILDs patients with poor prognosis. Elevated serum KL-6 levels in stable-state ILDs may be a good predictor of subsequent mortality and a useful noninvasive marker for identifying the risk of death from respiratory failure because of ILDs. In other words, elevated KL-6 level may provide simple, yet valuable information for identifying patients with ILDs who are at an increased risk of subsequent mortality. In particular, patients with a KL-6 level of more than 2750 U mL−1 have a 5% chance of surviving 5 years or more. A careful and larger-sized clinical study will verify our results.
Conflict of interest statement
No conflict of interest was declared.
The authors special thanks to Ms F. Miyamasu for her valuable advices.