YKL-40 is an inflammatory biomarker associated with disease activity and mortality in patients with diseases characterized by inflammation and tissue remodelling. The aim of this study was to describe the prognostic value of YKL-40 in an unselected patient population.
In consecutive patients admitted to hospital during a 1-year period, blood was collected and information regarding final diagnosis and mortality was collected. Median follow-up time was 11.5 years.
District hospital, Copenhagen, Denmark.
A total of 1407 patients >40 years of age were admitted acutely.
Main outcome measure
Median YKL-40 was increased in patients (157 μg L−1, range 13–7704 μg L−1) compared to healthy controls (40 μg L−1, range 29–58 μg L−1; P < 0.001). Patients with YKL-40 in the highest quartile had a hazard ratio (HR) of 7.1 [95% confidence interval (CI) 4.2–12.0] for all-cause mortality in the first year and 3.4 (95% CI 2.8–4.2) in the total study period, compared to those in the lowest quartile (HR = 1). The HR for death for all patients with YKL-40 above the normal age-corrected 95th percentile was 2.1 (95% CI 1.6–2.7) after 1 year and 1.5 (95% CI 1.3–1.7) during the total study period, compared to patients with YKL-40 below the age-corrected 95th percentile. The results of multivariable analysis showed that YKL-40 was an independent biomarker of mortality; this was most significant in the first year. YKL-40 was a marker of prognosis in all disease categories. The HR for death was increased in patients with YKL-40 above the normal age-corrected 95th percentile in healthy subjects independent of type of disease (all P < 0.001).
The level of YKL-40 at admission is a strong predictor of overall mortality, independent of diagnosis and could be useful as a biomarker in the acute evaluation of all patients.
There is a great variation of patients admitted acutely to a general hospital for evaluation, diagnosis and treatment. In addition, it can be difficult to detect those patients with the highest mortality after discharge from hospital. Because inflammation is an important contributor to many disease conditions, biomarkers of inflammation could be important predictors of prognosis. YKL-40 (also known as chitinase-3-like-1 and human cartilage glycoprotein-39) is an inflammatory biomarker and a highly conserved member of the mammalian chitinase-like proteins [1, 2]. YKL-40 is expressed by several cell types of the immune system including neutrophils and macrophages [3-5], but also by vascular smooth muscle cells , cancer cells [2, 7] and arthritic chondrocytes  and is therefore not a disease-specific inflammatory biomarker. YKL-40 is an acute-phase protein, and its plasma concentration has been shown to increase reversibly in patients by more than 25% following an inflammatory stimulus .
Serum concentrations of YKL-40 are elevated (compared to healthy subjects) in patients with diseases characterized by inflammation, increased extracellular remodelling and ongoing fibrosis [2, 9] such as cancer , coronary artery disease [1, 10-13], heart failure [14-16], ischaemic cerebrovascular disease , diabetes [18-20], asthma [21, 22], chronic obstructive pulmonary disease (COPD) , pneumonia , sepsis , inflammatory bowel disease [26, 27], rheumatoid arthritis and liver fibrosis . Furthermore, high plasma levels of YKL-40 in the general population are associated with increased risk of early death from various disorders including cardiovascular disease and cancer . YKL-40 is also associated with risk of early death in patients with heart failure [14-16], cardiovascular disease , cancer , sepsis  and liver fibrosis .
The aim of this study was to determine whether serum concentrations of YKL-40 could be an independent risk factor for prognosis in a population of unselected all-comers above 40 years of age admitted acutely to hospital with various medical and surgical diagnoses.
Material and methods
Patients were recruited from the Copenhagen Hospital Heart Failure (CHHF) Study, which included all patients over 40 years of age who were admitted acutely to either the medical or surgical department at Amager Hospital in Copenhagen, Denmark, between 1 April 1998 and 31 March 1999 . Of 2506 patients admitted during this period, 2230 (89%) consented to participate in this study, the purpose of which was to follow all participants and record mortality.
Within the first 24 h after admission, a structured medical interview was conducted and self-reported medical history was recorded. Information from previous hospital admissions was also recorded. In addition, all patients underwent a clinical examination and echocardiographic evaluation of the heart.
After discharge, the final diagnosis was recorded from patients' files. Information regarding death was obtained using patients' unique national Danish Central Person Registry numbers. Patients were followed for a median duration of 4187 days (11.47 years), range 4015–4371 days (11.00–11.98 years) or until death. The cause of death or morbidity at the end of the study period was not registered; only whether the patient died and the date of death were recorded.
The reference values for plasma YKL-40 were determined in 3130 healthy subjects (1837 women and 1293 men) aged 20–80 years from the Copenhagen City Heart Study, a study of the Danish general population. Participants had no known disease at the time of blood sampling in 1991–1994 and remained healthy and alive during the 16-year follow-up period. The median plasma YKL-40 concentration in these healthy subjects was 40 μg L−1 (interquartile range 29–58 μg L−1) .
Blood was collected between 08.00 and 10.00 within 24 h of admission. Serum was isolated and stored at −80 °C until required for analysis.
Serum concentrations of YKL-40 were determined in duplicate using a commercially available enzyme-linked immunosorbent assay (ELISA; Quidel, Santa Clara, CA, USA) in May 2011. The detection limit was 20 μg L−1, and the intra-assay coefficients of variation were 5% (at 40 μg L−1), 4% (at 104 μg L−1) and 4% (at 155 μg L−1). The interassay coefficient of variation was <6%. YKL-40 is stable in blood stored at −80 °C .
N-terminal pro-brain natriuretic peptide (NT-proBNP) concentration was measured by ELISA (Roche Diagnostics, Mannheim, Germany). Levels of haemoglobin, C-reactive protein (CRP), sodium and potassium were measured by standard methods. The glomerular filtration rate was estimated (eGFR) according to age, creatinine and gender using the Modification of Diet in Renal Disease Study formula.
The study was performed in accordance with the Declaration of Helsinki and was approved by the regional scientific ethical committee of the city of Copenhagen (No. 01-320/97). All patients gave their written informed consent to participate before inclusion in the study.
Patients were divided into two groups, according to serum YKL-40: concentrations higher than the 95th percentile for age-matched healthy individuals and normal levels. This was performed using an equation from a previous report of plasma YKL-40 in 3130 healthy subjects in which the 95th percentile was used as a cut-off value to define an elevated YKL-40 level . The percentile of YKL-40 is expressed as a function of age (years) and YKL-40 concentration (μg L−1):
Continuous variables, in the two groups, are presented as means and 95% confidence interval (CI), whilst categorical data are presented as frequencies and percentages. Associations between the continuous variables were measured using one-way analysis of variance (anova). For categorical variables, associations were measured using the chi-squared test.
Patients were also classified according to quartiles of serum YKL-40. Continuous variables, in the four groups, are presented as means and 95% CI, whilst categorical data are presented as frequencies and percentages. Associations between continuous variables were measured using one-way anova, whereas a trend test was applied for categorical variables.
The hazard ratio (HR) for death according to elevated or normal serum YKL-40 level (dichotomized according to the 95th percentile of serum YKL-40 level in age-corrected healthy subjects ) was measured using the Cox proportional hazards model. HR was measured for all patients together and divided according to discharge diagnosis. Multivariable analysis was performed using the Cox proportional hazards model with backward elimination, with variables eliminated using P < 0.1 as the cut-off, after checking assumptions for proportionality. Thereby, adjusted HRs for serum YKL-40 above the normal age-corrected 95th percentile in healthy subjects were determined. Interactions between YKL-40 and other significant variables were tested in Cox proportional hazards models including interaction variables. The HR for death for all patients grouped according to quartiles of serum YKL-40 was measured using the same method.
Kaplan–Meier curves were used to estimate survival for different levels of YKL-40: (Fig. 1a) for all patients according to elevated versus normal serum YKL-40 (dichotomized relative to the age-corrected 95th percentile in healthy subjects ); (Fig. 1b) for elevated versus normal serum YKL-40 according to discharge diagnosis.
Table 1. Baseline variables for all patients according to normal versus elevated levels of serum YKL-40
Age-corrected YKL-40 ≤95 percentile for healthy subjects, n = 840
Age-corrected YKL-40 >95 percentile for healthy subjects, n = 567
To compare serum YKL-40 with other variables associated with mortality, C-statistics for prediction of 1 year as well as 10-year mortality were calculated. All statistical analyses were performed using pawstatistics, version 18.0 (SPSS Inc., Chicago, IL, USA). Differences were considered statistically significant at P < 0.05.
Blood samples were collected from 2294 patients. Serum samples for YKL-40 analysis were available from 1407 individuals (61%); these patients represent the current study population. The excluded patients, for whom serum was not available to measure YKL-40 levels, were randomly distributed across the entire study period. The study population was comparable to patients without available serum samples, but the two groups (included versus excluded) differed slightly with regard to medical history of pulmonary disease (49% vs. 51%, P < 0.001), mean NT-proBNP (411 vs. 281 pmol L−1P < 0.001), mean eGFR (79 vs. 86 mL min−1, P < 0.001), mean sodium concentration (136 vs. 137 mmol L−1, P = 0.04), mean systolic blood pressure (147 vs. 149 mmHg, P = 0.05) and mean diastolic blood pressure (83 vs. 84 mmHg, P = 0.02). Mean survival time was significantly different in the two groups, with an HR for death during the entire study period of 0.84 (95% CI 0.76–0.92) for the included patients. There were no other clinically relevant differences between the two groups (data not shown).
Mean serum YKL-40 concentration in the total population was 382 μg L−1 (95% CI 345–418 μg L−1), and the median concentration was 157 μg L−1 (range 13–7704 μg L−1, interquartile range 79–288 μg L−1). Patients had significantly increased median serum concentrations of YKL-40 compared to healthy subjects (157 vs. 40 μg L−1, P < 0.001) . In 41.5% (584/1407) of patients, the serum YKL-40 level was higher than the upper normal age-corrected level (cut-off of 95th percentile in the control group). The median level of YKL-40 in the study population was equivalent to the 91st percentile in a normal healthy population (data not shown).
Baseline variables (demographic characteristics, vital parameters, symptoms, medical history and laboratory examinations) in relation to elevated (95th percentile in the control group) or normal serum concentration of YKL-40 are presented in Table 1. Age, gender, systolic and diastolic blood pressure, heart rate, ejection fraction, history of heart failure or liver disease, eGFR and levels of sodium, haemoglobin, CRP and NT-proBNP differed significantly between patients with elevated and normal serum YKL-40 levels (Table 1). The same baseline variables, in relation to quartiles of serum YKL-40, are shown in Table 2. Age, diastolic blood pressure, ejection fraction, history of heart failure, eGFR and levels of sodium, haemoglobin, CRP and NT-proBNP differed significantly between quartiles of YKL-40.
Median follow-up time was 11.47 years (4187 days), range 11.00–11.98 years (4015–4371 days). Complete follow-up was possible for all except nine patients (0.6%) as a result of immigration: one patient after 1 year and eight during the complete observation period. All nine patients were censored at the time of immigration and were therefore not lost to follow-up. One-year mortality was 16.7% (235 patients), and the mortality rate during the entire study period 68.9% (969 patients).
Univariable HRs for death for all patients with YKL-40 above the 95th percentile for healthy age-matched subjects were 2.1 (95% CI 1.6–2.7) after 1 year and 1.5 (95% CI 1.3–1.7) for the entire observation period (Table 3). Multivariable Cox regression analyses were performed (Table 3) to determine the independent prognostic importance of serum YKL-40 and the baseline variables shown in Tables 1 and 2. Multivariable Cox regression analysis showed that elevated serum YKL-40 was an independent biomarker of short survival, with an adjusted HR for death of 1.5 (95% CI 1.1–2.0) after 1 year compared with 1.2 (95% CI 1.1–1.4) for the entire observation period. Thus, the HR was significantly higher in the first year compared to the long-term follow-up in the adjusted model (Table 3). Gender, age, heart rate, medical history of diabetes, previous myocardial infarction, known liver or pulmonary disease and blood levels of sodium, haemoglobin and NT-proBNP were also independent biomarkers of short-term overall survival. In similar analyses, with serum YKL-40 divided into quartiles (Table 3), YKL-40 quartiles II to IV were significantly different from quartile I (P < 0.001), with an increased HR for patients in the highest quartiles. The HR was highest in the first year, for both univariable and multivariable analyses (Table 3). Amongst all the variables included in the multivariable Cox analyses, gender, age, heart rate, medical history of diabetes, myocardial infarction, liver and pulmonary disease and levels of sodium, NT-proBNP and haemoglobin remained in the model.
Table 3. Unadjusted and adjusted hazard ratios for all patients according to (a) normal versus elevated levels and (b) quartiles of serum YKL-40
Patients are dichotomized according to the 95th percentile of serum YKL-40 level in age-matched healthy subjects .
Adjusted for those variables that remained significant in the final Cox proportional hazards analysis using backward elimination. The variables were gender, age, heart rate, medical history of diabetes, myocardial infarction, liver and pulmonary disease and levels of sodium, NT-proBNP and haemoglobin. Variables were considered significant at P < 0.05.
Normal level, n = 840 (ref.)
Elevated level, n = 567
Quartile 1 (≤79 μg L−1), n = 354 (ref.)
Quartile 2 (≥80 and ≤157 μg L−1), n = 352
Quartile 3 (≥158 and ≤288 μg L−1), n = 351
Quartile 4 (≥289 μg L−1), n = 350
Analyses of patients with the different disease categories and serum YKL-40 above the 95th percentile for healthy age-matched subjects are shown in Table 4. The HR for death was highest in the first year after admission compared to the entire observation period for almost all disease categories and worst for patients with gastrointestinal (HR = 3.9, 95% CI 1.4–10.6), pulmonary (HR = 2.6, 95% CI 1.0–7.3) and heart diseases (HR = 2.5, 95% CI 1.3–4.9; Table 4).
Table 4. Univariable hazard ratio for death (short- and long-term) for all patients, according to discharge diagnosis
Number of patients
Above the 95th percentile, n (%)
HR (95% CI) for patients with elevated versus normal YKL-40
Patients are dichotomized according to the 95th percentile of serum YKL-40 level in age-matched healthy subjects .
HR, hazard ratio; CI, confidence interval.
Figure 1a shows Kaplan–Meier curves for overall survival in the 1407 patients with elevated or normal serum concentrations of YKL-40 (dichotomized according to the age-corrected YKL-40 95th percentile in healthy subjects). Overall mortality was significantly higher in patients with elevated YKL-40 during the entire period, and mortality was particularly high within the first year. Figure 1b shows Kaplan–Meier survival curves according to quartiles of serum YKL-40. The same tendency was found, with significantly better overall survival in patients with serum YKL-40 in the lowest quartiles, and survival declined significantly with increasing serum YKL-40 level (P < 0.001; Fig. 1b).
Figure 2 shows Kaplan–Meier curves for overall survival according to discharge diagnoses and serum YKL-40 level (dichotomized according to the age-corrected YKL-40 95th percentile in healthy subjects) at the time of acute hospitalization. Elevated serum YKL-40 concentration was related to reduced overall survival in all disease groups.
Receiver operating characteristic analyses showed that the area under the curve (AUC) for elevated age-adjusted serum YKL-40 concentration to predict 1-year and 10-year mortality was 0.60 and 0.56, respectively. Including in the analyses all variables from the final Cox model, we found an AUC for 1-year mortality of 0.79. This increased to 0.81 if elevated age-adjusted serum YKL-40 was included in the model. The same analyses performed for 10-year mortality showed an AUC of 0.86, which increased to 0.87 if serum YKL-40 was included; however, these differences were not significant (P > 0.05 for both).
The results of the present study demonstrate that an elevated serum concentration of YKL-40 at the time of hospital admission, compared to healthy age-matched subjects, identifies patients with a high risk of short- and long-term mortality independent of diagnosis at discharge. Serum YKL-40 remained an independent biomarker of mortality, even after adjusting for known risk factors such as age and serum levels of CRP and NT-proBNP. As expected, the HR for mortality was highest in the first year of follow-up in all disease categories, except for orthopaedic and endocrine disorders, compared to the total study period. The explanation for this observation may be that, if a patient survives the first year after acute admission, the risk of death overall declines. Our results indicate that a high serum concentration of YKL-40 at admission may reflect the severity of several disease conditions that could be fatal within the first year after hospital discharge and that serum YKL-40 can predict both short- and long-term outcome. If confirmed by others, these data suggest that serum YKL-40 measured at the time of hospitalization could be a clinically useful biomarker with the ability to identify patients at high risk of early death. A high serum concentration of YKL-40 at admission could indicate that close follow-up and intensive treatment are needed, independent of diagnosis.
The analyses for patients in the different disease categories showed the same tendency with a higher HR for death in the first year after admission compared to the entire observation period, except for patients with orthopaedic and endocrine disorders. The HR for death in the first year was worst for patients with gastrointestinal, pulmonary and heart diseases. This difference between disease categories could be explained by differences in the importance of the inflammatory component and by the natural course of each type of disease. Patients with endocrine disorders, for example, diabetes, may have a longer disease history and may eventually die from a different cause. Of note, the cause of death was not registered in this study. By contrast, patients admitted with pulmonary disease may have a stronger inflammatory component, hence a higher serum YKL-40 level and stronger correlation with death in the first year.
We found that serum YKL-40 concentration increased with age, which is consistent with findings from earlier studies [1, 2, 10, 12, 32]. In a large study of 3130 healthy individuals from the Danish general population, it was shown that plasma YKL-40 increases exponentially with age . It is therefore important to adjust for age when evaluating serum concentrations of YKL-40 in patients. However, when classifying patients according to normal or elevated serum YKL-40, based on the age-corrected 95th percentile cut-off value in healthy control subjects, serum YKL-40 remained a predictive biomarker of mortality in all disease groups. Furthermore, mortality was still highest during the first year compared to long-term follow-up. Recently, it was also demonstrated, in a large cohort of subjects from the general population followed for 16 years, that high plasma YKL-40 was independently associated with risk of early death .
Parameters normally associated with heart failure, such as high New York Heart Association class, high serum concentration of NT-proBNP and low ejection fraction, were more often present in patients with high serum concentrations of YKL-40, compared to those with normal levels. There were also more patients with diagnosed heart failure amongst those with elevated serum YKL-40. There may be associations between the presence of a degree of heart failure, high serum YKL-40 and mortality, even in patients with undiagnosed heart failure who were admitted to and discharged from hospital with another diagnosis such as pulmonary disease. Studies have shown that heart failure may be underestimated in patients admitted with COPD and vice versa, due to an overlap of signs and symptoms [33, 34]. Indeed, a prevalence of 21% of previously unknown heart failure was reported in patients with COPD . Serum YKL-40 is elevated in patients with heart failure [14-16], and high serum YKL-40 predicted adverse outcomes including cardiac death and rehospitalization . Furthermore, serum YKL-40 is associated with all-cause mortality in patients with heart failure  and may be an indirect marker of poor left ventricular recovery after myocardial infarction . The results of the present study indicate that serum YKL-40 has a prognostic impact on mortality in patients with diseases of the heart, but the type of heart disease was not specified. Elevated serum YKL-40 was also associated with reduced eGFR. This may indicate that adjustment for eGFR could be of value for clinical application of serum YKL-40 concentration in patient evaluation. However, eGFR did not remain a significant parameter in the adjusted model in this study.
Interleukin-6 stimulates production of YKL-40,  and furthermore, YKL-40 is regarded as an acute-phase protein due to an increase in concentration of more than 25% following an inflammatory stimulus . An element of inflammation may be present in the acute course of various diseases with which patients in this study were diagnosed. Serum CRP and YKL-40 are both biomarkers of inflammation, but with different origins; CRP is secreted by hepatocytes in the liver, whereas YKL-40 is mainly produced by inflammatory and cancer cells [2, 7]. Thus, serum YKL-40 may be a more direct marker than CRP of inflammatory disease activity. We have recently reported that serum YKL-40 is more independent of changes in cholesterol levels than CRP and is better able to monitor inflammation in patients with coronary artery disease . The results of the present study showed that patients with high serum YKL-40 concentrations have a poor prognosis and are characterized by more general features of disease such as high levels of serum CRP and NT-proBNP, a low haemoglobin level and reduced eGFR.
In contrast to serum YKL-40, CRP was not significantly associated with mortality in the adjusted model in this study. This is interesting, as a close correlation between these two markers of inflammation has previously been found, and CRP has been shown to be an independent prognostic biomarker in other populations .
The exact mechanism underlying the increased risk of death observed in patients with a high serum YKL-40 concentration is not known. YKL-40 regulates vascular endothelial growth factor , has a role in inflammation [2, 4, 7], angiogenesis [38-43], cell proliferation and differentiation [44, 45] and remodelling of the extracellular matrix  and protects against apoptosis .
A high serum concentration of YKL-40 is an independent biomarker of mortality in acutely hospitalized patients with different diseases. This suggests that an elevated level of serum YKL-40 may identify patients at high risk of early death independent of diagnosis, and thus, serum YKL-40 could be an important biomarker in the acute evaluation of all patients.
We are grateful to the organizers of the Copenhagen Hospital Heart Failure Study (CHHF) for allowing the use of their clinical data and serum biobank. We thank the team of Biomedical Laboratory Scientists, Tonni Løve Hansen, Dorthe Mogensen and Ulla Kjaerulff-Hansen, at Herlev Hospital for excellent technical assistance with serum YKL-40 analysis. We also thank the staff at the Department of Clinical Biochemistry at Amager Hospital for measurements of sodium, potassium, haemoglobin and serum CRP and NT-proBNP.
The study was funded by grants from the Arvid Nilssons Foundation and from the Joint Proof-of-Concept Fund, the Ministry of Science, Technology and Innovation, Denmark.
Conflicts of interests statement
A European patent (No. PA 2008 00089 ‘Classification of individuals suffering from cardiovascular diseases according to survival prognoses as found by measuring the levels of biomarker YKL-40') was issued on 23 January 2008 and is exclusively licensed to Jens Kastrup. Quidel provided some of the YKL-40 ELISA kits. The study sponsors had no role in: the design and conduct of the study; the collection, management, analysis and interpretation of the data; or the preparation, review and approval of the manuscript. The authors had full access to all study data and were responsible for the decision to submit the manuscript for publication.