Clinical use of biomarkers of survival in pulmonary fibrosis

Interstitial lung diseases (ILD) [1] are a heterogeneous group of lung diseases that comprise more than 200 clinical pathological entities. Although clinically the different ILD have rather similar presentations with increasing shortness of breath, a restrictive lung function, impaired gas exchange and a widespread shadowing on the chest radiograph, they comprise a very wide spectrum of pathologies, clinical manifestations, and outcomes. Approximately two-thirds of ILD cases have no reported aetiology [2]. The remaining one-third is associated with or defined by various environmental or occupational factors including cigarette smoking, aspiration, certain drugs and radiation therapy [3, 4]. Despite their acknowledged complexity, there is little evidence about the best management of ILD. Morbidity of the ILD themselves and adverse events of the available treatments may be high, with potentially serious consequences therefore for mismanagement. Improved survival and cure from different forms of ILD are dependent on a better understanding of the pathophysiology of the disease, its diagnostic accuracy in clinical practice, and an analysis of possible biomarkers which can guide the clinician in their treatment [5].

The clinical course of individual patients with ILD is variable and can manifest long periods of stability, a steady gradual decline, and/or periods of acute deterioration [6]. Some forms of ILD respond well to therapy, others are insensitive to high doses of anti-inflammatory drugs (e.g. dexamethasone) or immunosuppressive agents. Lung transplantation is an option for some patients, but many patients are too old or die on the waiting list [7]. One potential reason for the high mortality on the lung transplant waiting list is the variable course of the disease, which makes it difficult to predict outcome. Consequently, the identification of predictors of survival is critical for physicians and patients [8].

In the search for an effective therapy, biologic predictors of survival in pulmonary fibrosis with a worse prognosis, more specifically in IPF, have been extensively studied in the last 10 years. These surrogate endpoints for survival are so-called biologic markers or biomarkers. They can be subdivided into those markers measured in serum, those measured by lung function testing and those by imaging techniques. Before discussing the different results of the studies performed on predictors of survival, an introduction about what a biomarker is, is given.

"A biomarker indicates a change in expression or state of a biologic measurement (e.g. concentration of a protein in serum, lung function measurement, amount of ground glass on HRCT, ...) at a given time point that correlates with the risk or progression of a disease, or with the susceptibility of the disease to a given treatment at a future time point [9]". Once a proposed biomarker has been validated, it is used to diagnose disease risk, presence of disease in an individual, or to tailor treatments for the disease in an individual [9]. In evaluating potential drug therapies, a biomarker may be used as a surrogate for a natural endpoint such as survival or irreversible morbidity [9]. If a treatment alters the biomarker, which has a direct connection to improved health, the biomarker serves as a surrogate endpoint for evaluating clinical benefit [9].

Use of such a biomarker as surrogate endpoint to assess a clinical endpoint (e.g. survival) has potential disadvantages. Biomarkers can be difficult to validate and require different levels of validation depending on their intended use [10]. If a biomarker is used to measure the success of a therapeutic intervention, the biomarker should reflect a direct effect of that intervention [10].

In summary, 4 pitfalls are present during the validation process of a biomarker as a diagnostic test [11]: (1) presence of a "spectrum bias", i.e. the study population on which the biomarker was validated has a different clinical spectrum (e.g. more advanced disease) than the population in whom the test has to be applied; (2) presence of a "selection bias", i.e. the test results of the validation study are related to test results (e.g. FVC < 90%); (3) presence of an "observer bias", i.e. observer (e.g. pathologist) is influenced by prior knowledge; and (4) presence of an "observer variability", i.e. the presence of variability in the interpretation of the results by the same observer (intra-observer variability) or by different observers (inter-observer variability).

At the end of the 1990s various serum markers were tested for their use in ILD, more specifically in IPF. Many markers have been studied and those with the most promise are surfactant proteins A and D (SP-A and SP-D), KL-6, and lactate dehydrogenase (LDH). In 2009 serum CC-chemokine ligand 18 (CCL-18) was presented as a potential biomarker in IPF [8]. Serum SP-A and SP-D are hydrophilic surfactant proteins produced and secreted by type II pneumocytes. Their concentration is elevated in certain inflammatory lung diseases, including IPF [12]. Why the concentrations of these proteins are elevated in these lung diseases is not precisely known. It is probably due to a combination of a loss of integrity of the epithelial barrier caused by lung injury and of an increased mass of type II cells due to hyperplasia [12]. Three studies tried to validate the use of these serum markers in an IPF population as a predictor for survival. The study of Takahashi et al. [13] concluded in a population of 52 IPF patients (mean follow up time 11.4 months for the subjects who died, more than 3 years for the survivors) that in the group of survivors (n = 10) the concentrations of SP-A and SP-D were within the normal range. In those subjects who died, around 25% also had protein concentrations within the normal range [13]. The second study by Greene et al. validated the use of SP-A and D in a population of 142 IPF patients as a predictor of survival. They used a Cox's proportional hazards model and found that an elevated concentration of SP-D (and not for SP-A) had a 56% elevated increase in death after adjusting for age, smoking status, TLC, FVC, FEV₁, D_Lco and resting P_A-a O₂ [12]. The third study was recently published by Kinder et al [14]. It was found in 82 IPF patients that each increase of 49 ng/mL in baseline SP-A level was associated with a 3.3 fold increased risk of mortality in the first year after presentation. They observed no statistically significant association with serum SP-D and mortality [14]. In conclusion SP-A and D are promising biomarkers, but can not yet be used as a biomarkers for survival. When reviewing the studies, large standard deviations of the measured surfactant concentrations were found. This questions the reproducibility of the measurements.

In 1989, Kohno et al. discovered a compound named KL-6, a mucin-like high-molecular weight glycoprotein, which is expressed on type II alveolar pneumocytes [15]. Concentrations of KL-6 in serum and broncho-alveolar lavage fluid are elevated in different forms of ILD [15]. However, an elevated concentration of KL-6 is not specific for alveolitis in ILD, as this is also seen in breast cancer [16], non small cell lung cancer [17], colorectal cancer [18] and pulmonary tuberculosis [19]. Yokoyama et al. published a study of 14 IPF patients who received a predefined therapy of a weekly pulse of high dose corticosteroids for at least 3 weeks. The concentration of KL-6 was measured one week before and, 1 and 3 weeks after start of treatment. They found that a decrease in concentration of KL-6 was a good predictor of survival (n = 8) [20]. Different studies have been published indicating that the presence of alveolitis is correlated with an elevated concentration of KL-6 [21–23]. However, most of these studies have been performed with small number of subjects and therefore use of KL-6 as a biomarker for prognosis or response to therapy in ILD still needs to be validated with an unbiased and representative study population.

Some authors suggest that LDH is a potential marker of disease activity and of response to therapy in different forms of ILD [24–27]. Cytoplasmatic enzymes, like LDH, when present in the extracellular space are indicators of cell damage or cell death [25]. A case report of an IPF patient suggests a good correlation between level of serum LDH and response to therapy [28]. In contrast the study of Thomeer et al. found no correlation between survival and concentration or change in concentration of serum LDH. Our results can, however, be biased by the fact that an increase of LDH is also seen with intake of azathioprine, a drug that was given to all patients in this study. Serum LDH is a sensitive marker for cell injury, but is a non specific test since the concentration is elevated in different circumstances of cell injury caused by ischaemia, as by excess heat or cold, starvation, dehydration, injury, exposure to bacterial toxins, after ingestion of certain drugs, and from chemical poisonings [25]. Consequently, serum LDH is difficult to use as a valid biomarker of disease activity in ILD since its concentration is defined by many factors.

CCL18 is a CC-chemokine produced by human myeloid cells [8]. CCL18 is known to trigger a biological response in vitro in T cells, B cells, dentritic cells, hematopoietic progenitor cells, fibroblasts, and potentially, in monocytes/macrophages but not in neutrophiles [29]. It is constitutively expressed at high levels in lung and at low levels in some lymphoid tissues such as lymph nodes, thymus, and appendix [29]. Macrophages that are activated in a specific way, produce CCL18 and play a role in tissue repair processes such as wound healing and fibrosis [8]. CCL18 regulates collagen production by lung fibroblasts and is abundantly produced by alveolar macrophages in patients with IPF [8, 30, 31]. Prasse et al. set up a prospective study of 72 IPF patients, followed during 24 months with a pulmonary function every 6 months, but without a standardised treatment. They found that baseline serum CCL18 concentrations predicted change in TLC and FVC at the 6-month follow-up. In the group with high serum CCL18 concentrations a higher incidence of disease progression was seen. A cut-off value of 150 ng/ml calculated by ROC analysis, predicts mortality (sensitivity 83%; specificity 77%, area under the curve 0.80), with a hazard proportional ratio adjusted for age, sex, and baseline pulmonary function data of 8.0 [8]. One may conclude from this single study that CCL18 is the first biomarker that predicts mortality in IPF in such a clear way. However the process of measuring CCL18 concentration by ELISA has some pitfalls and its reproducibility and internal validity are of concern. Prasse et al. suggest that a standardization of the entire procedure from drawing blood over freezing and thawing to ELISA measurement is mandatory to overcome this concern [8]. The findings of Prasse et al. induce important questions regarding the pathogenesis or the search for therapy of IPF. Is CCL18, that stimulates lung fibroblasts to produce collagen, the key to a potential new treatment in IPF [29]? Is the lack of an animal model for lung fibrosis explained by the fact that a rodent counterpart for human CCL18 does not exist [29]? Further studies are needed to elucidate these questions, and more precisely to discover a CCL18 receptor, which is still not found [29].

Two types of medical imaging are discussed. HRCT and nuclear imaging including imaging that measures the alveolar capillary membrane permeability of the lung by use of radioactive isotopes, pertechnegas and ^99mTc-diethylenetriamine pentacetate (DTPA), and positron emission tomography (PET).

It is nowadays unthinkable not to use HRCT in the diagnosis of ILD. The added value of HRCT depends upon its ability to increase confidence of a specific diagnosis, to alter patient management and, if possible, to influence outcome [4]. Research on the ability of HRCT scans to differentiate between active and inactive disease has been mainly confined to IPF and ILD associated with systemic sclerosis [4]. There is evidence that a predominant ground glass pattern is more likely to represent active inflammatory disease and to respond to appropriate therapy, particularly in fibrosing alveolitis, EAA, and desquamative interstitial pneumonia [63–65]. It is still unproven that a ground glass pattern precedes a reticular or honeycomb pattern, although this seems likely [4]. Not all ground glass change indicates cellular inflammation, however, as fine intralobular fibrosis may be indistinguishable from a cellular infiltrate on HRCT scans [64, 66]. The association of a ground glass pattern with traction bronchiectasis or bronchiolectasis is likely to indicate some associated fibrosis, whereas ground glass change without traction bronchiectasis usually indicates active inflammation [66]. Reticular and honeycomb patterns on HRCT scans correlate well with histological evidence of fibrosis [67, 68].

Can HRCT predict response to therapy in IPF? Gay et al. set up a study with 38 biopsy proven IPF patients. The study patients received 1 mg/kg prednisone daily during 3 months. The HRCT before treatment was scored (score from 0 to 5 for each lobe) for ground glass and fibrosis by 4 radiologists independently. They demonstrated that a fibrosis score of 2 or more has a 80% sensitivity and 85% specificity in predicting survival [69]. From this study it is not clear how many drop outs were present during the survival follow up and how long the time of follow up was, two factors that can induce a possible bias in their results. A spin off trial of the IFIGENIA study used the same scoring method proposed by Gay et al. The HRCT was scored before start of a predefined treatment of azathioprine and prednisone and at 6 and 12 months of treatment. The HRCT scores for fibrosis and ground glass were assessed by 3 radiologists independently. In this study 155 IPF patients had a median follow up of 2.5 yrs (SD 1.8). Only the fibrosis score at baseline was predictive for survival (HR 1.58 95% CI 1.15-2.17), whereas ground glass score or changes in ground glass score or fibrosis score over 6 and 12 months were not predictive for survival. A HRCT fibrosis score of more than 2 had a relative risk for death of 2.31 (95%CI 1.40-3.80). However the area under the curve for the fibrosis score was only 0.61 (95%CI 0.52-0.70), which means that the score had only a moderate to low sensitivity and specificity for survival.

The ILD are characterised by an acute or chronic inflammation of the interstitium, also called the alveolar capillary membrane [70]. A possible way to measure the alveolar capillary membrane permeability is by radionuclide aerosol lung imaging. The rate of the clearance of the aerosol is inversely related to the integrity of the alveolar capillary barrier [71–73]. Pertechnegas and DTPA have been studied as disease activity measure in different forms of interstial lung diseases, but, as far as we known, no studies has correlated rate of lung clearance with survival [73, 74].

Positron emission tomography (PET) imaging has recently emerged on the scene of biomarkers of ILD. A number of studies has suggested that ¹⁸F-FDG PET imaging may serve as a sensitive tool for the evaluation of disease activity in sarcoidosis, with higher sensitivity and interobserver agreement compared to the classical Gallium scintigraphy [75–77]. The potential value of ¹⁸F-FDG PET as a biomarker for disease activity in other ILD is less clear. In IPF the magnitude of ¹⁸F-FDG uptake in the lungs is usually low. As ¹⁸F-FDG is thought to assess the inflammatory burden and not the fibrosis, the finding of relatively low SUV in IPF can be regarded as confirmative for the concept that inflammation does not play a major role in the pathogenesis of this disease. No studies are present that correlates rate of ¹⁸F-FDG uptake with survival in specific forms of interstitial lung diseases.

Interstitial lung diseases are a diverse and complex collection of parenchymal lung diseases that vary widely in aetiology, histopathology, clinical radiological presentation, and clinical course [78]. There is an urgent need for biomarkers or markers of disease activity that would help the clinician in deciding whether or not to treat since treatment carries a potential risk for adverse events. These decisions are made easier if accurate and objective measurements of the patients' clinical status can predict the risk of progression to death. Serum biomarkers, and markers measured by medical imaging as HRCT, pertechnegas, DTPA en FDG-PET are not ready for clinical use to predict mortality in different forms of ILD. A baseline FVC, a change of FVC of more than 10%, and a decrement of more than 25 m in 6 MWD are predictors of survival. Measurements of FVC and 6 MWD are clinically easy to do, results are measured in no time and pose minimal effort for the patient.