Lung cancer is the most frequently diagnosed tumour type and the leading cause of cancer mortality throughout the world. The prognosis of patients is strongly correlated with tumour stage,1 hence the importance of an accurate, easily usable diagnostic tool for early detection.

Low-dose spiral computed tomography (CT) has opened up interesting prospects for early diagnosis but is expensive, invasive and presents some limitations, especially in terms of specificity.2 Free circulating DNA levels have been shown to be higher in patients with lung cancer than in healthy individuals (including heavy smokers), suggesting that this marker could be an important tool to identify lung cancer in screening programmes.3 More recently, our group4 and other researchers5 have shown that plasma or serum free DNA is also significantly higher in patients with other cancer histotypes than in healthy donors. Cyclooxygenase-2 (COX-2) has been implicated in the early stages of lung oncogenesis,6 and increased levels are associated with a poor prognosis.7

In our case-control study on 128 cancer patients and 103 healthy donors, we aimed to define the potential of serum free DNA and COX-2 mRNA expression, determined singly or in combination, for non-small cell lung cancer detection. Serum free DNA levels and COX-2 mRNA expression in peripheral blood were significantly lower in healthy donors than in patients (8.8 vs 48.0 ng/ml, z = −11.17, p<0.001 for DNA; 1.5 vs 2.0, z = −6.02, p<0.001 for COX-2) and were not related to sex, age or smoking habits in either group. Moreover, no significant differences were observed in patients in relation to tumour histotype or stage (see table 1 in online supplement). The area under the ROC curve (AUC) for free DNA was 0.917 (95% CI 0.877 to 0.957) (see fig 1 in online supplement) and very high predictivity was observed for a large range of cut-off values (see table 2 in online supplement). In particular, using a cut-off value of 25 ng/ml, sensitivity and specificity were 83% and 92%, respectively, and values were not significantly different for the subgroups categorised by clinical characteristics or smoking habits (table 1). The AUC value for COX-2 mRNA was 0.758 (95% CI 0.697 to 0.819) (see fig 1 in online supplement), and at the cut-off of 2.5 we observed 38% sensitivity and 100% specificity (see table 2 in online supplement) which did not change as a function of clinical characteristics or smoking habits.

DNA and COX-2 were not significantly correlated in healthy donors (r_s = −0.04, p = 0.68) or patients (r_s = −0.05, p = 0.57) and were analysed in combination. At the cut-off value of 25 ng/ml for free circulating DNA and 2.5 for COX-2 mRNA, significantly higher predictivity was observed compared with single marker analyses, with 91% sensitivity and 92% specificity for the overall series as well as for subgroups categorised for clinical characteristics or smoking habits. Moreover, using the combined analysis, the greatest increase in sensitivity was seen for squamous cell carcinoma (SCC) and for stage I and II tumours (table 1), reaching 100% in the subgroup of 31 patients with stage I and II SCC.

When the markers were analysed as continuous variables in a logistic regression model, each marker provided independent and therefore additive diagnostic information: a unit increase in log DNA or log COX-2 was associated with a 9-fold and 13-fold increase in cancer detection, respectively (see table 3 in online supplement), and similar results were obtained in multivariate analysis including clinical characteristics and smoking habits. The ROC curve of the combined marker algorithm confirmed the significantly higher diagnostic accuracy (AUC 0.940, 95% CI 0.908 to 0.972) than that observed for DNA alone (AUC 0.917, 95% CI 0.877 to 0.957) (p = 0.035).

We conclude that the combined approach could represent an important test for the early diagnosis of lung tumours.

Acknowledgements

The authors thank Dr Daniele Calistri (Istituto Scientifico Romagnolo per lo Studio e la Cura dei Tumori) for his technical support, Gráinne Tierney for editing the manuscript and Dr Marta Mengozzi of the Department of Diseases of the Thorax (Morgagni-Pierantoni Hospital) for her assistance in sample collection.