Poor tumor oxygenation is one of the major factors for radioresistance and significantly contributes to the poor prognosis and unsatisfactory treatment outcomes of patients undergoing radiotherapy for lung cancer [1, 2, 3, 4]. Tumor hypoxia constitutes an interesting aspect for radiation oncologists due to its involvement in provoking an aggressive and rapidly progressing cancer phenotype which is resistant to treatment and features increased neo-angiogenesis [5]. Detection of intrinsic hypoxia markers (i.e. endogenous hypoxia-related proteins) have been suggested as a promising noninvasive and feasible approach of cost-effectively predicting clinically significant tumor hypoxia [6, 7, 8, 9].

Among the factors most consistently triggered by hypoxia is hypoxia-inducible factor (HIF)-1α and its downstream genes carbonic anhydrase IX (CA IX) and vascular endothelial growth factor (VEGF) [10, 11, 12]. Their overexpression has been linked to poor prognosis in human malignancies, including NSCLC [13, 14, 15]. Besides the aforementioned molecules, another new potential surrogate of (tumor) hypoxia is osteopontin (OPN) which has been shown to be associated with intratumoral pO2, which is prognostic in NSCLC [6, 16, 17]. Interestingly, OPN plasma levels were able to successfully predict tumor hypoxia and to identify head and neck cancer patients who benefitted from hypoxic radiosensitization during radiotherapy [18]. Although the prognostic value of OPN overexpression in NSCLC patients treated by surgery or chemotherapy is now indisputable [19, 20], its prognostic role in the radiotherapy of NSCLC is quite unclear. Similar considerations apply to VEGF and CA IX whose cooperative role in response to radiotherapy and whose interplay with OPN in the context of tumor hypoxia underlines the clinical potential of these biomarkers [21, 22, 23, 24]. Since there are few studies investigating the prognostic value of OPN, VEGF and CA IX (mostly single-marker-based, reported for head and neck cancer [25, 26, 27]) and because there is no equivalent data for radiochemotherapy of NSCLC, we aimed to assess the joint prognostic effects of the potential endogenous hypoxia-related proteins OPN, VEGF and CA IX and correlated their plasma levels with clinicopathological patient characteristics and outcome.

Materials and methods

Patients and treatment

Between 2008 and 2010, 55 patients (median age 63 years, range 47–86 years) with advanced NSCLC, treated with radiotherapy (n = 13) or radiochemotherapy (n = 42) at the Department of Radiation Oncology, Martin Luther University Halle-Wittenberg, Germany, were entered into the study. Inclusion criteria were (1) histologically confirmed NSCLC, clinical stage M0, (2) no prior surgery or radiotherapy, (3) indication for primary curative-intent radio(chemo)therapy, (4) age ≥ 18 years and (5) signed informed consent. A positive vote was given by the ethics committee of the Medical Faculty. Clinical stage was determined according to the UICC TNM classification, 7th edition. Chemoirradiation was usually given in two cycles of simultaneous cisplatin (20 mg/m2 on days 1–5) and vinorelbine (25 mg/m2/day 1) in treatment weeks 1 and 5. Radiotherapy consisted of a normofractionated (5 fractions/week) regimen with daily fractions of 2 Gy to a mean total dose of 63.9 Gy (range 51.7–72.1 Gy). Patients were followed-up regularly at the Department of Radiation Oncology, University Hospital Halle (initially 4–6 weeks after the end of radiotherapy and later at longer intervals) and survival status of patients was obtained and continuously monitored in cooperation with local citizen registration offices.

Plasma samples

Blood samples were obtained by venous puncture before the start of radiotherapy, EDTA-anticoagulated (Sarstedt monovette, Nümbrecht, Germany) and centrifuged at 4 °C for 10 min at 4000 rpm. Plasma was removed, aliquoted and stored at − 80 °C until assayed. Each sample was measured in duplicate, using a commercial ELISA system for OPN (Human Osteopontin Assay, IBL Ltd., Japan), VEGF (Quantikine Human VEGF, R&D Systems, USA) and CA IX (Human CA IX Quantikine ELISA Kit, R&D Systems, USA) according to the manufacturer’s instructions.

Statistics

All statistical analyses were performed with the SPSS PASW 18.0 software package for windows (SPSS Inc, USA.). The relationship of plasma marker levels and clinicopathological characteristics was evaluated using nonparametric Mann–Whitney’s U test or Kruskal–Wallis’ h test. Pearson’s test was applied to determine correlation between two markers and the survival curves were generated using Kaplan–Meier analysis with the log-rank test to test for differences. For the univariate and multivariate analysis, the Cox proportional hazard regression model was used to calculate the relative risk and hazard ratio and its 95 % confidence interval (CI) in the survival analysis. All p values were two-sided and p < 0.05 was considered statistically significant.

Results

Biomarker plasma levels and their correlation in patient plasma

In patients with NSCLC, median plasma concentration of OPN before the start of radiotherapy was 817 ng/ml (range 299–2441 ng/ml). The median plasma concentration of VEGF was 92 pg/ml (range 0–1078 pg/ml) and for CA IX it was 105 pg/ml (range 22–420 pg/ml). Median haemoglobin concentration was 12.1 g/dl (range 8.3–15 g/dl). OPN significantly correlated with CA IX (r = 0.3; p = 0.03) and with VEGF (r = 0.3; p = 0.03). A highly significant inverse correlation was observed between OPN and haemoglobin concentration (r = − 0.5; p = 0.001). An inverse correlation was also found for VEGF and haemoglobin (r = − 0.3; p = 0.03). CA IX correlated positively with VEGF (r = 0.3; p = 0.02) but nonsignificantly with haemoglobin concentration (p = 0.08).

Bivariate analysis of biomarker plasma levels with clinical parameters

Patient demographics and the relationships between biological plasma marker levels and clinicopathological patient characteristics are given in Tab. 1.

Tab. 1 Relationship of plasma marker levels and clinicopathological parameters
Full size table

Bivariate analysis of plasma levels showed an association of high pretreatment OPN plasma levels with higher age (862 ng/ml vs. 713 ng/ml, p = 0.008), weight loss (1002 ng/ml vs. 721 ng/ml, p = 0.001) and poor lung function (871 ng/ml vs. 691 ng/ml, p = 0.002). Patients with weight loss (p = 0.001) and larger tumor volume (GTV above the median) had significantly elevated VEGF plasma levels (130 pg/ml vs. 52 pg/ml, p = 0.003) before radiotherapy. High plasma levels of CA IX were observed in patients with advanced T stage (T1/T2 78 pg/ml vs. T3/T4 149 pg/ml, p = 0.04).

Univariate survival analysis according to biomarker plasma levels

After a median follow-up of 37 months (range 27–48 months) in surviving patients, 78 % of patients had died. Median overall survival was 13 months; 3-year overall survival 19 %. In patients with high pretreatment single marker levels of OPN, VEGF or CA IX, respectively, Kaplan–Meier analyses showed a significantly inferior overall survival (Fig. 1 a, b, c). In a univariate Cox regression analysis, a significantly increased risk of death for patients with high plasma levels of either biomarker was observed (Tab. 2).

Fig. 1
figure 1

Univariate Cox regression analysis: association of single marker plasma levels (a, b, c) and combined triple marker levels (d; high above median, low below median) of OPN/VEGF/CA IX with overall survival for all 55 NSCLC patients. The p value was calculated with the use of an adjusted Cox proportional hazards model

Full size image
Tab. 2 Kaplan–Meier analyses and univariate Cox regression analyses: association of OPN, VEGF and CA IX plasma levels and median overall survival
Full size table

Interestingly, we found an additive prognostic effect when two biomarkers were combined, particularly an additively increased risk of death for patients with high plasma levels. Patients with high OPN/VEGF plasma levels had a significantly reduced overall survival (9.4 vs. 29.6 months, p = 0.03) and increased risk of death (rr = 3.8, 95 % CI 1.4–10.4, p = 0.004) compared to patients with low marker levels.

Overall survival of patients with high levels of OPN/CA IX was inferior to that of patients with low OPN/CA IX levels (6.3 vs. 26 months, p = 0.04) and the risk of death was also increased in the patient group with high plasma levels of both markers (rr = 3.9, 95 % CI 1.2–12.6, p = 0.02). The double marker combination VEGF/CA IX showed a reduced overall survival and an increased risk of death in patients with high levels of both markers; however, these findings were not significant (5.3 vs. 15.7 months, p = 0.22; rr = 2.8, 95 % CI 0.9–8.6, p = 0.07). Inverse marker combinations (i.e. plasma levels of one marker above vs. the other below the median) did not show significant prognostic effects (data not shown). The prognostic effects of elevated pretreatment biomarker plasma levels was most pronounced when the three biomarkers where combined (Tab. 3, Fig. 1 d). Patients with high OPN/VEGF/CA IX plasma levels before radiotherapy had a significantly shorter overall survival in comparison to patients with low levels of all three biomarkers (2.3 vs. 29.6 months, p = 0.007) and the risk of death was increased by a factor of 5 (95 %CI 1.5–17.2 p = 0.009) in patients with high triple marker plasma levels.

Tab. 3 Kaplan–Meier analyses and univariate Cox regression analyses for triple marker combinations of OPN, VEGF and CA IX plasma levels and their association with median overall survival
Full size table

Multivariate survival analysis of biomarker plasma levels

To test whether plasma marker levels can serve as independent prognostic factors for overall survival or merely are surrogates of clinical parameters, we carried out a multivariate Cox regression analysis, including the clinical characteristics N stage (p = 0.001), T stage (p = 0.04) and GTV (p = 0.009) which were significant predictors for overall survival in a univariate analysis in all 55 patients.

When single marker levels of OPN, VEGF or CA IX were entered into the aforementioned model, only N stage remained prognostically significant (p = 0.01, data not shown). With the double marker combination included into the same baseline model, consisting of N stage, T stage and GTV, the combination OPN/VEGF (p = 0.07) and OPN/CA IX (p = 0.09) but not VEGF/CA IX (p = 0.2) showed a prognostic trend for overall survival along with N stage which significantly predicted overall survival throughout models with either biomarker combination. GTV and T stage remained not significant (data not shown). Interestingly, triple biomarker plasma levels of OPN, VEGF and CA IX were independent predictors for overall survival (p = 0.03) with a 2.9-fold increased risk of death in a multivariate model adjusted for GTV (n.s.), T stage (n.s.) and N stage (p = 0.03, Tab. 4).

Tab. 4 Multivariate Cox regression model for overall survival in curative-intent (M0) NSCLC patients
Full size table

Discussion

The relevance of tumor oxygenation for the radiotherapy of cancer was first implicated by Thomlinson and Gray in 1955 [28]. More recently, the HIF-1α system and its target genes VEGF and CA IX as well as other non-HIF-1α-regulated factors such as OPN have been discussed as endogenous hypoxia markers for their potential relation to tumor oxygenation [7, 9, 17, 18, 29, 30]. In our study, OPN, CA IX and VEGF were linearly correlated and an inverse correlation of VEGF and haemoglobin concentration was found, which is in concordance with other studies [31, 32, 33]. However, an association of high OPN levels with low haemoglobin concentration has not been reported so far. These findings could indicate a poor oxygenation status of the patient with a rapidly progressing tumor which features extensive hypoxia and enhanced neo-angiogenesis, reflected by overall elevated biomarker plasma levels [34].

Elevated circulating biomarker levels were associated with different clinical parameters of advanced disease stage including weight loss and poor lung function (OPN), tumor stage (CA IX) and gross tumor volume (VEGF), which is in line with the current literature [35, 36, 37]. An association of VEGF plasma levels with gross tumor volume (GTV) has, to the authors’ knowledge, not been published so far and is an interesting finding in the context of conventionally fractionated radiotherapy of NSCLC [38, 39].

Here, single protein levels of OPN, VEGF or CA IX significantly predicted overall survival and elevated protein levels of either plasma marker were accompanied by an increased risk of death. Interestingly, biomarker combinations resulted in an additive prognostic effect which was most pronounced for the triple-marker combination where high plasma levels of OPN–VEGF–CA IX were associated with a 5-fold increased risk of death. The latter combination remained an independent prognostic factor in an exploratory multivariate Cox proportional hazard model including N stage, T stage and GTV which indicates that OPN, VEGF and CA IX plasma levels were not merely surrogates of known prognostic factors such as metastatic disease, considering that only M0 stage NSCLC patients were studied. This finding is in agreement with the current literature where elevated single plasma biomarker levels were associated with unfavorable prognosis, however, in patients treated by surgery or chemotherapy for lung cancer [19, 20, 40, 41, 42].

Conclusion

This pilot study demonstrated that high pretreatment plasma levels of OPN, CA IX and VEGF are additively correlated with prognosis in M0-stage NSCLC patients receiving radical radiotherapy. Despite the homogeneity of the studied patient cohort, the relatively small patient number clearly constitutes a limitation of this study and underlines its exploratory character. Nevertheless, our results generated preliminary evidence that a combination of the studied plasma biomarkers is more robust in predicting overall survival in the curative-intent radiotherapy of NSCLC than a single marker, which supports the notion that protein concentration in plasma should be considered together for prognostic evaluation. Yet, further investigations are needed to confirm the hypotheses generated by this study and evaluation of a larger patient collective, utilizing pre-hoc biometric study panning, is needed. Consequently, validation studies, preferably in the setting of a multicenter study, are required and the results should be confirmed in an independent data set. To our knowledge, this is the first prospective clinical study in the curative-intent radiotherapy of (M0 stage) NSCLC patients to assess the prognostic value of a combination of the three plasma proteins OPN, VEGF and CA IX which have been associated with a highly malignant phenotype [6, 13, 14]. Although the aforementioned proteins cannot be considered direct surrogates of tumor hypoxia or specific hypoxia markers [43], their potential value for the radiotherapy of (lung) cancer may be seen in their possible indirect relation to tumor oxygenation and malignancy [7, 44, 45, 46]. In this context, our study in principle justifies the hypothesis that OPN, VEGF and CA IX might be promising candidates for further evaluation and correlation with other methods of detection of clinically significant tumor hypoxia such as exogenous hypoxia markers or dynamic hypoxia PET-imaging during radiotherapy [47, 48, 49]. Respectively, an individual hypoxic patient profile based on hypoxia-related plasma proteins (i.e. plasma hypoxia score [50, 51]) and hypoxia-specific (PET) imaging could integrate information on both tumor malignancy and hypoxic tumor burden. This might help identifying patients with largely hypoxic tumors, featuring an unfavorable and highly malignant phenotype. These “high-risk” patients could thus be selected for (hypoxia-) specific targeted therapies [52, 53] in order to improve outcome in the curative-intent radiotherapy of advanced NSCLC.