WO2015164772A1 - Circulating insulin-like growth factor (igf)-associated proteins for the detection of lung cancer - Google Patents

Abstract

A method and a kit for assessing risk of lung cancer versus benign disease in a subject are provided. The method includes obtaining a biological sample from the subject and determining a measurement for a panel of biomarkers in the biological sample. The panel includes at least one biomarker selected from the group consisting of IGFBP-1, IGFBP-2, IGFBP-3, IGFBP-4, IGFBP-5, IGFBP-6, IGFBP-7, IGF-1, and IGF-2 and at least one biomarker selected from the group consisting of IL-6, IL-1 ra, IL-10, SDF-Ια+β, TNF-a, MIP-1a, slL-2Ra, CA-125, Eotaxin, OPN, sEGFR, and sE-Selectin. The method further includes comparing the measurement to a reference profile for the panel of biomarkers, sorting the subject into a stratification group and determining whether the subject is at risk for lung cancer based on the stratification group.

Description

CIRCULATING INSULIN-LIKE GROWTH FACTOR (IGF)-ASSOCIATED PROTEINS FOR THE DETECTION OF LUNG CANCER

RELATED APPLICATIONS

[0001] This application claims the benefit of U.S. Provisional Application No. 61/984,507 filed April 25, 2014, which is incorporated by reference herein in its entirety.

TECHNICAL FIELD

[0002] The present invention relates to a method and a kit for assigning clinical significance to indeterminate lung nodules and for a primary screen in assessing high risk subjects, and in particular to a method and a kit for assigning clinical significance to indeterminate lung nodules or for a primary screen using a biomarker panel for detecting lung cancer.

BACKGROUND

[0003] Non-small cell lung cancer (NSCLC) has the highest prevalence of all malignancies worldwide and remains the most common cause of cancer-related mortality. In 2012, lung cancer represented 12.9% of newly diagnosed cancer with 1.8 million cases and 19.4% of cancer related deaths with 1.59 million cases. [1] In the United States alone during 2013, there were an estimated

228,190 cases of newly diagnosed lung cancer and 159,480 lung cancer related deaths. The overall 5-year survival rate for lung cancer averages 15% depending on the stage at diagnosis, and falls to 2% for advanced disease. Early diagnosis of NSCLC has been proposed to improve 5-year survival rates by 60-80%. [2] Recent single arm prospective uncontrolled studies on low-dose computed tomography (LDCT) of the chest for screening have yielded conflicting results. [3, 4] The National Lung Screening Trial (NLST) demonstrated a 20% relative reduction in lung cancer mortality. [5] Further advancements in early detection of NSCLC could confer additional benefits.

[0004] Advances in delineating the tumorigenesis cascade and growth factors implicated in lung cancer dissemination have produced novel therapeutic options. Research indicates that insulin-like growth factor I (IGF-I) and its receptor, IGF- 1 R, play a role in the development of various malignancies, including lung cancer.[6] Binding of IGF-I to its receptor can lead to insulin receptor

phosphorylation, which instigates protein kinase-mediated pathways that promotes cell growth, differentiation, and survival. Insulin-like growth factor binding proteins (IGFBPs), notably IGFBP-3, regulate this pathway by binding IGF in the serum and regulating the amount of free IGF available to bind with its receptor. [7] A disproportionate elevation of IGF-I relative to IGFBP-3 is theorized to promote the proliferative capacity and metastatic potential of lung cancer.[8, 9] Wang et al demonstrated increased expression of IGF-I in the setting of NSCLC with the degree of upregulation correlating directly with the extent of disease progression. [10] Cao et al showed that IGFBP-3 levels were inversely associated with lung cancer, noting a precipitous drop during tumoriogenesis.[1 1] The proposed mechanism for increased cancer cell dissemination involves an IGF- mediated upregulation of urokinase-type plasminogen activator (uPA), matrix metalloproteinase-2, and matrix metalloproteinase-9. IGFBP-5 has been linked with TGF-βΙ induced epithelial-to-mesenchymal invasion of breast cancer cultures. A decrease in the IGFBP-5 allows for unregulated TGF-βΙ action and increased cell migration. [12]

[0005] Biomarkers useful for distinguishing stage I NSCLC from benign disease are identified herein and applied to the refinement of a multianalyte classification algorithm. Previous work in this area demonstrated the prognostic ability of IGFBP-5 and IGFBP-7 to predict disease recurrence and outcomes in patients undergoing an anatomic resection for NSCLC. [13] IGF-related factors identified herein are shown to distinguish early-stage NSCLC from cases of benign disease in high-risk individuals with positive radiography. What is needed to improve the efficacy and cost-effectiveness of the LDCT screening paradigm is a biomarker panel for assigning clinical significance to indeterminate lung nodules that may be used as a companion test to help assign clinical significance to LDCT-detected indeterminate lung nodules or provide a risk of malignancy.

What is also needed is a biomarker panel assigning risk for the presence of lung cancer as a primary screen, possibly to indicate who should have further diagnostics (like LDCT imaging or biopsy) performed.

BRIEF SUMMARY

[0006] A method and a kit for assessing risk of lung cancer versus benign disease in a subject are provided. The method includes obtaining a biological sample from the subject and determining a measurement for a panel of biomarkers in the biological sample. The panel includes at least one biomarker selected from the group consisting of IGFBP-1 , IGFBP-2, IGFBP-3, IGFBP-4, IGFBP-5, IGFBP-6, IGFBP-7, IGF-1 , and IGF-2 and at least one biomarker selected from the group consisting of IL-6, IL-1 ra, IL-10, SDF-Ι α+β, TNF-a, M IP- l a, slL-2Ra, CA-125, Eotaxin, OPN, sEGFR, and sE-Selectin. The method further includes comparing the measurement to a reference profile for the panel of biomarkers, sorting the patient into a group and determining whether the subject is at risk for lung cancer based on the group.

BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 A-1 D illustrates representative 'Box and Whisker' plots indicating distribution of biomarker levels in subjects with benign nodules versus malignant nodules. Shown are IGFBP-3 (Panel A), IGFBP-5 (Panel B), IGF-1 (Panel C) and IGF-II (Panel D).

DETAILED DESCRIPTION

[0008] The present invention will utilize a panel of biomarkers measured in a biological sample obtained from a subject to identify subjects having lung cancer or to assess clinical significance to CT-detected indeterminate lung nodules. In some embodiments, the panel of biomarkers measured may be used to identify subjects having NSCLC or benign disease, possibly as a probability score or assignment of 'risk'.

[0009] The term "biomarker" as used herein, refers to any biological compound that can be measured as an indicator of the physiological status of a biological system. A biomarker may comprise an amino acid sequence, a nucleic acid sequence and fragments thereof. Exemplary biomarkers include, but are not limited to cytokines, chemokines, growth and angiogenic factors, metastasis related molecules, cancer antigens, apoptosis related proteins, proteases, adhesion molecules, cell signaling molecules and hormones.

[0010] "Measuring" or "measurement" means assessing the presence, absence, quantity or amount (which can be an effective amount) of a given substance within a sample, including the derivation of qualitative or quantitative concentration levels of such substances, or otherwise evaluating the values or categorization of a subject's clinical parameters. Alternatively, the term

"detecting" or "detection" may be used and is understood to cover all measuring or measurement as described herein.

[0011] The term "high-risk" is assessed according to NLST guidelines. (See Abele et al.) [28]

[0012] The terms "sample" or "biological sample" as used herein, refers to a sample of biological fluid, tissue, or cells, in a healthy and/or pathological state obtained from a subject. Such samples include, but are not limited to, blood, bronchial lavage fluid, sputum, saliva, urine, amniotic fluid, lymph fluid, tissue or fine needle biopsy samples, peritoneal fluid, cerebrospinal fluid, nipple aspirates, and includes supernatant from cell lysates, lysed cells, cellular extracts, and nuclear extracts. In some embodiments, the whole blood sample is further processed into serum or plasma samples. In some embodiments, the sample includes blood spotting tests.

[0013] The term "subject" as used herein, refers to a mammal, preferably a human.

[0014] Biomarker Panel

[0015] Biomarkers that may be used include but are not limited to cytokines, chemokines, growth and angiogenic factors, metastasis related molecules, cancer antigens, apoptosis related proteins, proteases, adhesion molecules, cell signaling molecules and hormones. In some embodiments, the biomarkers may be proteins that are circulating in the subject that may be detected from a fluid sample obtained from the subject. [0016] In some embodiments, the biomarker panel may include 2, 3, 4, 5, 6, 7, 8, 9, 10, 1 1 , 12, 13 14, 15, 16, 17 or 18 biomarkers. In some embodiments, the biomarker panel may include include ten or fewer biomarkers. In yet other embodiments, the biomarker panel may include 2, 3, 6 or 7 biomarkers. In some embodiments, the biomarker panel may be optimized from a candidate pool of biomarkers. By way of non-limiting example, the biomarker panel may be optimized for determining whether a subject has a specific disease. In some embodiments, the biomarker panel may be optimized to determine whether the subject has lung cancer and in some embodiments, the lung cancer may be NSCLC. The biomarker panel may be optimized for differentiating between NSCLC from benign disease using a candidate biomarker panel starting with eighteen candidate biomarkers selected from the group including Insulin-like growth factor binding proteins 1 , 2, 3, 4, 5, 6, 7 (IGFBP-1 , -2, -3, -4, -5, -6, -7), Insulin-like Growth Factor-1 and II (IGF-I, -II), Interleukin 6 (IL-6), Interleukin 1 receptor antagonist (IL-1 ra), Interleukin 10, (IL-10), Stromal cell-derived factor 1 a + b (SDF-Ι α+β), Tumor necrosis factor alpha (TNF-a), Macrophage inflammatory protein 1 alpha (MIP-1 a), Soluble interleukin 2 receptor antagonist (slL-2Ra), Cancer antigen 125 (CA-125), Eotaxin, Osteopontin (OPN), Soluble epidermal growth factor receptor (sEGFR), Soluble endothelial selectin (CD62E) (sE- Selectin). (See Table IV and Figure 1.) In some embodiments, the panel may include one or more biomarkers selected from the group including Insulin-like growth factor binding proteins 1 , 2, 3, 4, 5, 6, 7 (IGFBP-1 , -2, -3, -4, -5, -6, -7), Insulin-like Growth Factor-1 and II (IGF-I, -II) and one or more biomarkers selected from the group including Interleukin 6 (IL-6), Interleukin 1 receptor antagonist (IL-1 ra), Interleukin 10, (IL-10), Stromal cell-derived factor 1 a + b (SDF-Ι α+β), Tumor necrosis factor alpha (TNF-a), Macrophage inflammatory protein 1 alpha (MIP-1 a), Soluble interleukin 2 receptor antagonist (slL-2Ra), Cancer antigen 125 (CA-125), Eotaxin, Osteopontin (OPN), Soluble epidermal growth factor receptor (sEGFR), Soluble endothelial selectin (CD62E) (sE- Selectin).

[0017] Biomarker Panel Measurement [0018] Measurement of a biomarker panel generally relates to a quantitative measurement of an expression product, which is typically a protein or

polypeptide. In some embodiments, the measurement of a biomarker panel may relate to a quantitative or qualitative measurement of nucleic acids, such as DNA or RNA. The measurement of the biomarker panel of the subject detects differences in expression in subjects having lung cancer compared to subjects that are free from cancer. The expression levels of each individual biomarker may be higher or lower in the subjects having lung cancer compared to subjects that are free from cancer. A panel of a plurality of biomarkers provides an improved predictive value relative to a single biomarker.

[0019] Expression of the biomarkers may be measured using any method known to one skilled in the art. Methods for measuring protein expression include, but are not limited to Western blot, immunoprecipitation,

immunohistochemistry, Enzyme-linked immunosorbent assay (ELISA), Radio Immuno Assay (RIA), radioreceptor assay, proteomics methods (such as mass spectrometry), or quantitative immunostaining methods. Methods for measuring nucleic acid expression or levels may be any techniques known to one skilled in the art. Expression levels from the panel of biomarkers are measured in the subject and compared to the levels of the panel of biomarkers obtained from a cohort of subjects described below.

[0020] In some embodiments, Luminex-based xMAP^® multiplexed

immunoassays may be used to determine the expression levels of the panel of biomarker. (Luminex Corp.; Austin, TX.) The Luminex system uses

microspheres in a ninety-six well microplate. Each microsphere is dyed with red and infrared fluorophores at a range of independently varied concentrations of dye, creating unique absorbance signatures for each set of microspheres. Each of the microspheres is derivatized with antibodies having binding affinity for a particular type of molecular species. The subject sample is applied to a set of microspheres having different absorbance signatures, each carrying antibodies specific for a particular antigen. The antibodies on the beads then bind to the antigens present in the subject's sample. A secondary antibody may be applied in this system, followed by a streptavidin conjugated fluorophore reporter.

[0021] The beads, with their bound antigen and reporter are then sampled by an instrument. A detection chamber is used to detect the unique absorbance signatures and reporter fluorescence intensity, thereby identifying to which set of analytes each microsphere belongs, thus identifying each biomarker tested, and producing a quantitative fluorescent signal from the reporter. The fluorescence intensity of the observed signal is proportional to the quantity of antigen bound to the antibodies on the particular bead. Thus, it is possible to calculate the quantity of a particular biomarker in a sample.

[0022] In some embodiments, a kit may be provided with reagents to measure at least two of the panel of biomarkers. The panel of biomarkers to be measured with the kit may include two or more biomarkers from the group including Insulinlike growth factor binding proteins 1 , 2, 3, 4, 5, 6, 7 (IGFBP-1 , -2, -3, -4, -5, -6, - 7), Insulin-like Growth Factor-1 and II (IGF— I, -II) Interleukin 6 (IL-6), Interleukin 1 receptor antagonist (IL-1 ra), Interleukin 10, (IL-10), Stromal cell-derived factor 1 a + b (SDF-Ι α+β), Tumor necrosis factor alpha (TNF-a), Macrophage inflammatory protein 1 alpha (MIP-1 a), Soluble interleukin 2 receptor antagonist (slL-2Ra), Cancer antigen 125 (CA-125), Eotaxin, Osteopontin (OPN), Soluble epidermal growth factor receptor (sEGFR), Soluble endothelial selectin (CD62E) (sE-Selectin). The kit may include reagents to measure a panel of biomarkers that includes two, three, four, five, six, seven or more biomarkers combined together to measure a subject's biomarker panel. The kit may be provided with one or more assays provided together in a kit. By way of non-limiting example, the kit may include reagents to measure the biomarkers in one assay. In some embodiments, the kit may include reagents to measure the biomarkers in more than one assay. Some kits may include a 4-plex assay and a 2-plex assay while other kits may include different combinations of assays to cover all the biomarkers needed to be measured. The kit may also include reagents to measure a biomarker individually and other biomarkers in a 2-, 4-, or 8-plex assay. Any combination of reagents and assay may be combined in a kit to cover all the biomarkers needed.

[0023] Analysis of Biomarker Panel Measurements

[0024] In some embodiments, methods determining whether a subject is at risk for lung cancer is based upon the biomarker panel measurement compared to a reference profile that can be made in conjunction with a statistical algorithm used with a computer to implement the statistical algorithm to sort the subject into a group. In some embodiments, the statistical algorithm is a learning statistical classifier system. The learning statistical classifier system can be selected from the following list of non-limiting examples, including Random Forest (RF), Classification and Regression Tree (CART), boosted tree, neural network (NN), support vector machine (SVM), general chi-squared automatic interaction detector model, interactive tree, multiadaptive regression spline, machine learning classifier, and combinations thereof.

[0025] In some embodiments, the Random Forest algorithm may be used to identify the panel of biomarkers. The optimal multivariate panel of biomarkers was chosen based on variable selection algorithms performed within the random forests package in R. [29, 30] Liberal inclusion criteria were applied for the individual biomarkers (a Mann-Whitney p value smaller than 0.05 or an area under the ROC curve [AUC] higher than 0.60) to ensure that no biomarker with potential value in a multianalyte panel was prematurely excluded from this selection process based on a weak individual performance. Briefly, the Random

Forests package selects optimal combinations of biomarkers by growing numerous (1000 in the present study) cross-validated classification trees for each subpanel of biomarkers, with each tree used to predict group membership for each case. These are counted as the tree votes for that group. The forest chooses the group membership having the most votes over all the trees in the forest. Each such tree is grown by cross-validation; where a training set

(approximately two-thirds of the values) is randomly selected from the full data and each tree is grown on this training data to the largest extent possible (no pruning). The resultant tree is then used to predict the group membership for the remaining test cases, which is termed as an out-of-bag (OOB) prediction. This process is then repeated 1000 times; that is, another training set is randomly selected and a new tree is grown and used to perform another OOB prediction. The classification accuracy of the random forest is measured by the averaged error of the OOB predictions across the entire forest; this is termed the OOB error rate. The OOB error thus uses disjoint subsets of the data for model fitting and validation repeatedly. This cross-validation is also used to compute a variable importance for each biomarker included in the Random Forest analysis. The stepwise selection method sequentially searches for optimal subpanel of markers where the marker with the lowest variable importance score from the Random Forest are removed at each step. In some embodiments, the subject data set may include about 20, 30, 40, 50, 60, 70, 80, 90, 100, 1 10, 120, 130, 140, 150 or more subjects. In some embodiments, the subject data set may be about 100 subjects, about 50 having lung cancer and about 50 having benign disease. In some embodiments, the subject data set may be between about 130-140 subjects with about half the data set having lung cancer and about half the training set having benign disease. Other numbers of subjects in a data set may also be used in the training set. Two thirds of the data set randomly selected may be used for training and the remaining one third tests the algorithm and this process is repeated to select the optimum biomarkers generate a reference profile and to determine whether a subject has cancer.

[0026] In some embodiments, a Classification and Regression Tree (CART) algorithm may also be used. The optimal panel of biomarkers resulting from the

Random Forest variable selection process may then be used by a CART algorithm to model a classification tree for identifying a subject's true (pathologic) preoperative lymph node status. This analysis was performed using the RPART package of the R statistical software suite. [31] Briefly, classification trees determine a set of binary if-then logical (split) conditions that permit accurate classification of (in this case) the subject's nodal status. The CART algorithm discriminates between groups by splitting the range of values measured for each individual biomarker at all of its possible split points. The 'goodness of split criterion' is then used to determine the best split point for each biomarker for predicting nodal status. CART then ranks all of the best splits on each biomarker and selects the best biomarker and its split point for the split at the root node. CART then assigns classes to the two split nodes according to a rule that minimizes misclassification error. This process is continued at each nonterminal child node and at each of the successive stages until all observations are perfectly classified or the sample size within a given node is too small to divide (n ≤ a user-supplied number; such as 5). The final output of the resulting classification tree is a graphical display of decision criteria for each split, with the resulting predicted group memberships at the terminal nodes. The predicted probabilities of preoperative nodal status from the tree were used to obtain sensitivity and specificity across a range of cut-points for decision rules and the resulting ROC curve.

[0027] Treatment Stratification

[0028] In some embodiments, the analysis of the biomarker panel may be used to determine a treatment regime for the subject. In some embodiments, the measurement of one or more biomarkers in the panel may be used to determine whether to follow up at a later time point with the subject to repeat the

measurement of the one or more biomarkers in the panel before additional testing or treatment; to provide additional testing; to begin a treatment, to continue the same treatment or to modify the treatment regime for a subject. The treatment may be started or modified by administering a drug or changing the drug administered to the subject or to add an additional drug to an existing drug treatment regime, to change the dosage or other changes. In some

embodiments, other types of treatment regimes may be used such as radiation. The identification of patients at risk of lung cancer using the biomarker level may place the patient in a specific treatment, or an earlier treatment in the overall treatment strategy or identify subjects for further testing before beginning treatment. In some embodiments, the panel of biomarkers measured may be used to monitor subjects for post-surgical disease surveillance for early identification of disease recurrence. [0029] Subject Cohorts

[0030] Between 2004 and 201 1 , over 1200 patients that received an anatomic resection for suspected lung cancer were enrolled in our lung cancer

biorepository at Rush University Medical Center (Chicago, IL); 251 of these patients were enrolled in the present study. These cases were divided into the following cohorts: (a) high risk patients with benign disease on pathology (n=71 ) and (b) patients with stage I NSCLC (n=180). A collaboration with the Mayo Clinic

Rochester, MN provided patients with benign disease (n=52) and those with pathologically diagnosed stage I lung cancer (n=44) for the present study. All stage classifications are reported according to the American Joint Committee on

Cancer (AJCC) seventh edition criteria and confirmed by pathological evaluation

[14, 15]. Patients from Rush with benign pathology were accrued from an internal

LDCT screening program. Enrollment criteria for individuals in the LDCT program include age greater than 50 years or a smoker with greater than 20 pack years.

All patients were followed with annual LDCT and remained cancer free for a minimum of two years. All patient data were obtained after the patient gave informed consent. The study was conducted in compliance with either the

Institutional Review Board at Rush University Medical Center or the Mayo Clinic.

[0031] Development of a multianalyte classification algorithm (plasma) was conducted with the identical cohort that was previously employed for this purpose. [16] Briefly, this consisted of 30 cases of benign disease and 64 cases of lung cancer from RUMC and a unique validation cohort from the Mayo Clinic, consisting of patients with benign disease (n=61 ) and those with pathologically diagnosed stage I lung cancer (n=20).

[0032] Measurement of Plasma Biomarker Concentrations

[0033] Pretreatment serum or plasma was prepared using standard

phlebotomy protocols and archived at -80 °C in 100 μί aliquots. All evaluable specimens were subjected to less than two freeze-thaw cycles. [17-20] Nine biomarkers associated with insulin uptake were evaluated in this study. [21 ] Nine biomarkers were part of either the MILLIPLEX® MAP Human IGF or Human IGF

Binding Protein Panels (EMD Millipore, Billerica, MA) and included the following assays: insulin-like growth factor-l (IGF-I), insulin-like growth factor-l l (IGF-I I), insulin-like growth factor-binding protein-1 (IGFBP-1 ), IGFBP-2, IGFBP-3, IGFBP-3, IGFBP-4, IGFBP-5, IGFBP-6, IGFBP-7. All assays were performed in a blinded fashion using a 384-well adaptation of the manufacturer's recommended protocols. All data was collected on a Luminex FlexMAP 3D system with concentrations calculated based on 7-point standard curves using a five- parametric fit algorithm in xPONENT v4.0.3 (Luminex Corp., Austin, TX).

[0034] Statistical Methods

[0035] A primary objective of this study was to evaluate the association of circulating biomarkers of IGF-signaling with the clinical significance of

indeterminate pulmonary nodules identified during LDCT-based screening studies. The Mann-Whitney Rank Sum U test was used to compare each of the 9 candidate biomarkers across the patient cohorts, as defined above. Receiver Operator Characteristics (ROC) curve was used for comparing group pairs and 'area under the curve' value greater than or equal to 0.6 was marked as relevant. The optimal multivariable panel of biomarkers for prognosticating clinical outcome was selected based on Random Forests based variable selection, as we previously described. [17] For the validation studies, a probability of malignancy cutoff of 0.50 was used to assign a case as a malignancy.

[0036] Results

[0037] A total of 347 patients were enrolled in the study from both Rush University Medical Center and the Mayo Clinic and placed into the following cohorts: (a) high risk with benign disease on pathology from either screening studies or surgical resections (n=123) or (b) patients that underwent anatomic resections with stage I disease with either non-squamous (n=180), with the majority of these being adenocarcinoma (72.4% overall; n=163), or squamous cell (n=45) histology. Those in the lung cancer cohort had Ti disease in 137 cases and T₂ in 87 cases. These cohorts are further delineated in Table I. [0038] Table I: Demographics for the discovery population.

Mayo Clinic

Rush Cohorts

Cohorts

Benign Benign Lung

Lung Be nig

Screenin Resect Cane

Cancer n

g ed er

(n=52 (n=44

Number (n=31 ) (n=40) (n=180)

) ) Gender

28

Male 14 (45) 17 (42) 83 (46)

(53)

24

Female 17 (55) 23 (58) 97 (54)

(46)

Age

Median 61 64 69 63 71

Range 51-83 20-81 43-91 30-83 50-87

Smoking History^b

Median 28 8 30 20

Range 7-1 10 5-60 0-180 0-90

Non-smoker 18

Nodule Size⁰

Median 25 10 19.5 12 18

Range 7-50 2-14 4-50 1-50 8-30

TNM

T^oMo 105 32

T₂N₀M₀ 75 12

Histologic Diagnosis

Non- 33

145 (80)

squamous* (75)

1 1

Squamous Cell 35 (19)

(25)

[0039] Notes: age in years; smoking history in pack-years; nodule size in millimeters (mm). Blood serum collected. ^* Both cohorts -90% adenocarcinoma.

[0040] IGFBP-3, IGFBP-6, and IGFBP-7 were found to be strongly significantly different across the genders (p<0.001 ) and IGF-II was marginally significantly different (p=0.037). Similarly, IGFBP-1 and IGFBP-2 were found to be

significantly different (p=0.020 and 0.044, respectively) across races with

Caucasians having higher levels in general when compared to non-Caucasians.

Those in the screening cohort were younger on average than those with cancer

(61 vs. 70 years, p<0.001 ) and age was strongly associated (p<0.001 ) with

IGFBP-1 , IGFBP-2, IGFBP-3, IGFBP-6, IGFBP-7, IGF-I, and IGF-II. Smoking history was not significantly associated with any of the biomarkers tested. Of the 32 cases of type 2 diabetes mellitus on therapy, 14 received metformin, 8 were insulin dependent, and 10 received other glycomimetics; only IGFBP-1 was found to be associated (p=0.045) with metformin use in this cohort.

[0041] The nine IGF-related biomarkers were evaluated based on ultimate histology of the solitary nodule in the combined cohort. Levels of IGFBP-3, IGF-I and IGF-I I were significantly increased (p=0.001 , 0.002, 0.01 1 , respectively) in cases of malignancy versus benign nodules. Conversely, IGFBP-5 was notably decreased (p<0.001 ) in malignant cases. Molar ratios of IGF-I: IGFBP-3 and IGF- II: IGFBP-3 were both found to be significantly higher (p=0.001 and 0.015, respectively) in benign cases relative to the malignancy. Consideration of cases exclusively with nodules less than or equal to 2 cm (i.e. T1 a disease) again provided IGFBP-3, IGFBP-5, and IGF-I as significant (p=0.002, <0.001 , and 0.048, respectively) when comparing benign and malignancies, but failed to return significance for IGF-I I (p=0.164).

[0042] Malignant cases were further divided based on histology (squamous cell and non-squamous cell cancers) and contrasted to levels found in cases with benign lesions, as before. Overall, cases with non-squamous histology had median biomarker levels that were more dissimilar to the benign cases than those with squamous histology, such as IGFBP-3, and IGFBP-5 not only were IGFBP-3, IGFBP-5, IGF-I and IGF-I I also modulated in cases with squamous histology, but high levels of IGFBP-1 in cases of malignancy became strongly significant (p=0.006). 'Box and Whisker' plots of these findings are provided in Figure 1 , whereas the overall median values of biomarkers with significance are provided in Table II.

[0043] Table II. Patterns of serum levels of circulating IGF-related molecules.

IGFBP-2 Benign 1 18 5.91

Malignant 224 9.78 0.231

Non-squamous 777 8.10 0.509

Squamous 47 20.82 0.042

IGFBP-3 Benign 1 18 789.8

Malignant 224 723.0 0.001

Non-Squamous 777 776.0 <0.001

Squamous 47 735.4 0.109

IGFBP-4 Benign 1 18 39.33

Malignant 224 44.82 0.593

Non-squamous 777 43.76 0.338

Squamous 47 59.57 0.422

IGFBP-5 Benign 1 18 101.45

Malignant 224 28.16 <0.001

Non-Squamous 777 28.16 <0.001

Squamous 47 28.16 <0.001

IGFBP-6 Benign 1 18 132.55

Malignant 224 130.74 0.140

Non-squamous 777 130.34 0.068

Squamous 47 135.20 0.928

IGFBP-7 Benign 1 18 59.89

Malignant 224 64.35 0.378

Non-Squamous 777 63.60 0.675

Squamous 47 70.77 0.093

IGF-I Benign 123 68.16

Malignant 224 60.78 0.002

Non-squamous 777 59.98 0.007

Squamous 47 64.60 0.005

IGF-II Benign 123 149.85

Malignant 224 132.45 0.01 1

Non-Squamous 777 736.73 0.052

Squamous 47 116.45 0.003

IGF-I/ IGFBP-3³ Benign 123 0.122

Malignant 224 0.1 15 0.001

Non-squamous 777 0.777 0.001

Squamous 47 0.776 0.102

IGF-II/ IGFBP-3³ Benign 123 2.49

Malignant 224 2.35 0.015

Non-Squamous 777 2.36 0.037

Squamous 47 2.41 0.055

[0044] Notes: ^a molar ratio; Mann-Whitney (two-sided) rank sum test, all compared to benign cohort. [0045] Once trends were identified in the discovery cohort, biomarkers were reevaluated in a distinct set of plasma samples from Rush University Medical Center and the Mayo Clinic. These specimens were divided into benign

pathology (n=91 ) and stage I lung cancer (n=84). The validation cohort included a similar age distribution as our internal sample, with benign patients younger than those with carcinoma. Nodule sizes were similarly distributed as well, and histology again favored adenocarcinoma (Table III).

[0046] Table III: Demographics for the multianalyte algorithm refinement.

Mayo Clinic

Rush Cohorts

Cohorts (Discovery)

(Validation)

Lung Lung

Benign Benign

Cancer Cancer

Number (n=30) (n=64) (n=61 ) (n=20)

Gender

Male 12 (40) 26 (41 ) 30 (49) 12 (60)

Female 18 (60) 38 (59) 31 (51 ) 8 (40)

Age³

Median 61 .5 67 63 64

Range 51-82 48-88 30-83 49-82

Smoking History'

Median 36 35 25 35

Range 6-126 0-120 0-100 0-100

Non-smoker 0 13

Nodule Size⁰

Median 4 15 14 22

Range 2-17 7-121 3-50 8-80

TNM

Ti N₀M₀ 51 14 T₂N₀M₀ 13 5

Histologic Diagnosis

Adenocarcinoma 44 (72) 10 (50)

Squamous Cell 10 (14) 5 (25)

Neuroendocrine 10 (14) 5 (25)

[0047] Notes: age in years; smoking history in pack-years; nodule size in millimeters (mm) Blood plasma collected. [0048] The identical cohort of plasma specimens from our previous study on the topic of classification algorithm development was also evaluated with the intent of using IGF-related factors to help refine our 7-analyte classification algorithm. [16] Overall, the IGF-related factors performed favorably relative to the biomarkers we originally evaluated. Most notably, IGFBP-5 possessing a Mann- Whitney p-value of 0.001 and an area under (AUC) the receiver operator characteristics curve of 0.708 and IGFBP-4 with Mann-Whitney p-value 0.004 and AUC of 0.682. Also of note was IGFBP-7 with Mann-Whitney p-value 0.019 and AUC of 0.651 and IGF-II, with an AUC of 0.619. No other IGF-related factor was found to have significance in this distinct cohort of plasma specimens from Rush University.

[0049] The Random Forest based variable selection method was applied to the Rush University discovery cohort for classification, as we previously described. [16] Using this method a new 7-analyte classification panel was formulated that is composed of IL-6, IL-10, IL-1 ra, SDF-Ια+β, IGFBP-4, IGFBP-5, and IGF-II (see Table IV below). This panel differentiated patients with NSCLC from patients with benign disease with a cross-validated accuracy of 90.4%, which is an improvement from the 76.5% of a panel including IL-6, IL-10, I L- 1 ra , slL-2Rc SDF-Ια+β, TNF-a, and MIP-1 [16] Our new panel provided 24 cases of true negatives, 61 cases of true positives, 6 cases of false positives, and 3 cases of false negatives for a calculated sensitivity of 95.3%, specificity of 91.0%, and a negative predictive value of 89%. These parameters contrast to the calculated sensitivity of 100%, specificity of 52.2% from the panel including IL-6, IL-10, IL- 1 ra, slL-2R_a, SDF-la+β, TNF-a, and MIP-1a.

[0050] Our new Random Forest generated 7-analyte panel (defined above) was then used to predict the disease status in the validation cohort from the Mayo Clinic in a blinded fashion. The panel differentiated patients with lung cancer (n=20) from patients with benign disease (n=61 ) and provided 20 cases of true positives, 10 cases of true negatives, 51 cases of false positives, and no instances of false negatives. The receiver operator characteristics curve calculated sensitivity of 100% and a negative predictive value of 100%. The area under the curve was calculated to be 0.614 (curve not shown).

[0051] Emerging LDCT-based screening programs are designed to efficiently identify individuals with early lung cancer with the intent of decreasing the number of late stage lung cancer cases and, thereby, provide potentially curative treatment options. Based on current NLST inclusion criteria for LDCT scanning approximately 7 million individuals would qualify for a LDCT screening program in the US. [5] Of those 7 million, it is estimated 1.6-3.5 million indeterminate nodules would be identified and carry a high false positive rate of 94.5-96.4%. [16, 22] With this increased diagnostic burden of false positive cases, the International Association for the Study of Lung Cancer (IASLC) and the Strategic CT

Screening Advisory Committee (SSAC) published a position statement on LDCT screening to help address the increased diagnostic burden presented with LDCT screening. [23] The IASLC-SSAC recognized the need for further work on blood- based biomarkers as an adjunct to LDCT screening to decrease the high false positive rate. [23]

[0052] Although our program for discovery of blood-based biomarkers for lung cancer detection started in parallel to LDCT scanning, we have recently reinvigorated this effort and now hope to develop a useful adjunct in the diagnostic workup of high-risk patients with indeterminate nodules. Our objective in this study is to provide a simple and cost-effective companion diagnostic that can alleviate the impending diagnostic bottleneck when LDCT-based screening is fully deployed, improve cost-effectiveness of the overall paradigm, and mitigate the risks and anxieties associated with the serial scan approach.

[0053] As described herein, IGF-I/II and IGFBP1 -7 were selected as biomarkers to investigate the theorized sustained proliferative signaling and active metabolism. Cellular metabolism is a complex physiological process that is normally regulated by insulin and its associated proteins and factors (IGF-I/II and

IGFBP-1/7). IGF-1 has been studied and its function implicated in cell growth and apoptosis. [26] The dysregulation of these factors has been implicated in tumorigenesis. [7, 8, 10] IGF-I can increase expression and activity of normally inactive enzymes including urokinase-type plasminogen activator, matrix metalloproteinase-2 (MMP) and MMP-9. [27] Increased enzymatic activity has been linked to local and metastatic spread. Additionally, IGFBP's have been implicated in the epithelial-to-mesenchymal transition (EMT) of tumorigenesis. Specifically, normal levels of IGFBP-5 are linked with decreased migration of tumor cells mediated via TGF-31.[12]

[0054] The current findings have led to an improvement in our ability to assign clinical significance to indeterminate nodules identified via LDCT from the last reported iteration, which carried a sensitivity of 100% and a specificity of 52%. The cross-validated sensitivity of 95% and specificity of 80.0% in the discovery cohort and in the validation cohort a sensitivity of 100%, but a low specificity of 16.4%. However, the negative predictive value (NPV) was 100%, which would be of particular benefit to those individuals with indeterminate nodules found on LDCT screening. The application of this simple, inexpensive serum test may lead to decreased diagnostic burden, morbidity and mortality for the 1 .6-3.5 million indeterminate cases.

[0055] The practice of the present invention will employ, unless otherwise indicated, conventional methods for measuring the level of the biomarker within the skill of the art. The techniques may include, but are not limited to, molecular biology and immunology. Such techniques are explained fully in the literature.

See, e.g., Sambrook, et al. Molecular Cloning: A Laboratory Manual (Current

Edition, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, N.Y.); Current

Protocols in Molecular Biology (Eds. A Ausubel et al., NY: John Wiley & Sons,

Current Edition); DNA Cloning: A Practical Approach, vol. I & II (D. Glover, ed.);

Oligonucleotide Synthesis (N. Gait, ed., Current Edition); Nucleic Acid

Hybridization (B. Hames & S. Higgins, eds., Current Edition); Transcription and

Translation (B. Hames & S. Higgins, eds., Current Edition).

[0056] The above Figures and disclosure are intended to be illustrative and not exhaustive. This description will suggest many variations and alternatives to one of ordinary skill in the art. All such variations and alternatives are intended to be encompassed within the scope of the attached claims. Those familiar with the art may recognize other equivalents to the specific embodiments described herein which equivalents are also intended to be encompassed by the attached claims. REFERENCES

1. Ferlay J, S.H., Bray F, Forman D, Mathers C, Parkin DM. , Cancer

Incidence and Mortality Worldwide: IARC Cancer Base No 10. 2013, International Agency for Research on Cancer: Lyon, France.

2. Port, J.L., et al., Tumor size predicts survival within stage I A non-small cell lung cancer. Chest, 2003. 124(5): p. 1828-33.

3. International Early Lung Cancer Action Program, I., et al., Survival of

patients with stage I lung cancer detected on CT screening. N Engl J Med, 2006. 355(17): p. 1763-71 .

4. Bach, P.B., et al., Computed tomography screening and lung cancer

outcomes. JAMA, 2007. 297(9): p. 953-61 .

5. National Lung Screening Trial Research, T., et al., Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med, 201 1 . 365(5): p. 395-409.

6. Warshamana-Greene, G.S., et al., The insulin-like growth factor-l receptor kinase inhibitor, NVP-ADW742, sensitizes small cell lung cancer cell lines to the effects of chemotherapy. Clin Cancer Res, 2005. 11 (4): p. 1563-71 .

7. Lee, H.Y., et al., Insulin-like growth factor binding protein-3 inhibits the growth of non-small cell lung cancer. Cancer Res, 2002. 62(12): p. 3530-7.

8. Guerra, F.K., et al., Varying patterns of expression of insulin-like growth factors I and II and their receptors in murine mammary adenocarcinomas of different metastasizing ability. Int J Cancer, 1996. 65(6): p. 812-20.

9. Douglas, J.B., et al., Serum IGF-I, IGF-II, IGFBP-3, and IGF-l/IGFBP-3 molar ratio and risk of pancreatic cancer in the prostate, lung, colorectal, and ovarian cancer screening trial. Cancer Epidemiol Biomarkers Prev, 2010. 19(9): p. 2298-306.

10. Wang, Z., et al., Expression and clinical significance of IGF-1, IGFBP-3, and IGFBP-7 in serum and lung cancer tissues from patients with non- small cell lung cancer. Onco Targets Ther, 2013. 6: p. 1437-44. Cao, H., et al., Association between circulating levels of IGF-1 and IGFBP- 3 and lung cancer risk: a meta-analysis. PLoS One, 2012. 7(1 1 ): p.

e49884.

Vijayan, A., et al., IGFBP-5 enhances epithelial cell adhesion and protects epithelial cells from TGFbetal-induced mesenchymal invasion. Int J Biochem Cell Biol, 2013. 45(12): p. 2774-85.

Shersher, D.D., et al., Biomarkers of the insulin-like growth factor pathway predict progression and outcome in lung cancer. Ann Thorac Surg, 201 1 . 92(5): p. 1805-1 1 ; discussion 181 1.

Goldstraw, P., et al., The IASLC Lung Cancer Staging Project: proposals for the revision of the TNM stage groupings in the forthcoming (seventh) edition of the TNM Classification of malignant tumours. J Thorac Oncol, 2007. 2(8): p. 706-14.

Groome, P.A., et al., The IASLC Lung Cancer Staging Project: validation of the proposals for revision of the T, N, and M descriptors and consequent stage groupings in the forthcoming (seventh) edition of the TNM

classification of malignant tumours. J Thorac Oncol, 2007. 2(8): p. 694- 705.

Daly, S., et al., Development and validation of a plasma biomarker panel for discerning clinical significance of indeterminate pulmonary nodules. J Thorac Oncol, 2013. 8(1 ): p. 31 -6.

Borgia, J.A., et al., Establishment of a multi-analyte serum biomarker panel to identify lymph node metastases in non-small cell lung cancer. J Thorac Oncol, 2009. 4(3): p. 338-47.

Farlow, E.C., et al., Development of a multiplexed tumor-associated autoantibody-based blood test for the detection of non-small cell lung cancer. Clin Cancer Res, 2010. 16(13): p. 3452-62.

Patel, K., et al., Enhancement of a multianalyte serum biomarker panel to identify lymph node metastases in non-small cell lung cancer with circulating autoantibody biomarkers. Int J Cancer, 2010. 129(1 ): p. 133-42. Shersher, D.D., et al., Biomarkers of the Insulin-Like Growth Factor Pathway Predict Progression and Outcome in Lung Cancer. Ann Thorac Surg, 201 1.

Rinewalt, D., et al., Development of a serum biomarker panel predicting recurrence in stage I non-small cell lung cancer patients. J Thorac

Cardiovasc Surg, 2012. 144(6): p. 1344-50; discussion 1350-1 .

Henschke, C.I. and D.F. Yankelevitz, CT screening for lung cancer: update 2007. Oncologist, 2008. 13(1 ): p. 65-78.

Field, J.K., et al., International Association for the Study of Lung Cancer Computed Tomography Screening Workshop 2011 report. J Thorac Oncol, 2012. 7(1 ): p. 10-9.

Farlow, E.C., et al., A multi-analyte serum test for the detection of non- small cell lung cancer. Br J Cancer, 2010. 103(8): p. 1221 -8.

Hanahan, D. and R.A. Weinberg, Hallmarks of cancer: the next generation. Cell, 201 1. 144(5): p. 646-74.

Grimberg, A. and P. Cohen, Role of insulin-like growth factors and their binding proteins in growth control and carcinogenesis. J Cell Physiol, 2000. 183(1 ): p. 1 -9.

Bredin, C.G., Z. Liu, and J. Klominek, Growth factor-enhanced expression and activity of matrix metalloproteases in human non-small cell lung cancer cell lines. Anticancer Res, 2003. 23(6C): p. 4877-84.

Aberle DR, Adams AM, Berg CD, et al. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med 201 1 ;365:395- 409.

Breiman L. Random Forests. Mach Learn 2001 ;45:5-32.

Brieman L, Friedman J, Olshen R, Stone C. Classification and Regression Trees. Belmont, California: Wadsworth Co, 1984. Team RDC. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing, 2006. Table IV. Multianalyte classification panel development of 28 biomarkers (levels 3-10 provided).

Claims

1. A method of assessing risk of lung cancer versus benign disease in a subject, the method comprising: obtaining a biological sample from the subject; determining a measurement for a panel of biomarkers in the biological sample, the panel comprising at least one biomarker selected from the group consisting of IGFBP-1 , IGFBP-2, IGFBP-3, IGFBP-4, IGFBP-5, IGFBP-6, IGFBP-7, IGF-1 , and IGF-2 and at least one biomarker selected from the group consisting of IL-6, IL-1 ra, IL-10, SDF-Ια+β, TNF-a, MIP-1a, slL-2Ra, CA-125, Eotaxin, OPN, sEGFR, and sE-Selectin; comparing the measurement to a reference profile for the panel of biomarkers and sorting the subject into a stratification group; and determining whether the subject is at risk for lung cancer based on the stratification group.

2. The method according to claim 1 , comprising determining the measurement for the panel of biomarkers wherein the panel comprises IGFBP-5 and IL-6.

3. The method according to claim 1 , comprising determining the measurement for the panel of biomarkers wherein the panel comprises IGFBP-5 and IL-6, IL-1 ra and IL-10.

4. The method according to claim 1 , comprising determining the measurement for the panel of biomarkers wherein the panel comprises IGFBP-5, IGFBP-4 and IL-6.

5. The method according to claim 1 , comprising determining the measurement for the panel of biomarkers wherein the panel comprises IGFBP-5, IGFBP-4, IGF-II and IL-6.

6. The method according to claim 1 , comprising determining the measurement for the panel of biomarkers wherein the panel comprises IGFBP-5, IGFBP-4, IGF-II and IL-6, IL-1 ra, IL-10, SDF-Ι α+β.

7. The method according to claim 1 , wherein the measurement is obtained after detection of an indeterminate nodule is found by LDCT detection.

8. The method according to claim 1 , wherein the measurement is a primary screen.

9. The method according to claim 1 , wherein the lung cancer is non- small cell lung cancer.

10. The method according to claim 1 , wherein the group comprises non- small cell lung cancer and benign disease.

1 1 . The method according to claim 1 , wherein a learning statistical classifier system is used to determine the reference profile and sort the subject into the group.

12. The method according to claim 1 1 , wherein the learning statistical classifier system is a Random Forest system.

13. The method according to claim 12, wherein the panel has a negative predictive value of about 95% or greater.

14. The method according to claim 1 , wherein the biological sample comprises plasma sample or serum sample.

15. The method according to claim 12, wherein the group determination is made with an accuracy of at least about 85%.

16. The method according to claim 1 , further comprising stratifying the subject into a treatment group.

17. The method according to claim 17 wherein the treatment group comprises additional testing, treatment for lung cancer or follow up at a future time.

18. The method according to claim 1 , wherein the measurement for the biomarker panel is determined after surgery to monitor the subject for postsurgical disease recurrence.

19. A kit for performing the measurement of the panel of biomarkers of the subject in claim 1 , wherein the kit comprises reagents for measuring at least two of the panel of biomarkers.

20. The kit according to claim 19, wherein the kit comprises reagents for measuring serum or plasma.

21. The kit according to claim 19, wherein the kit comprises reagents for measuring IGFBP-5 and IL-6.

22. The kit according to claim 19, wherein the kit comprises reagents for measuring IGFBP-5, IGFBP-4 and IL-6.

23. The kit according to claim 19, wherein the kit comprises reagents for measuring IGFBP-5 and IL-6, IL-1 ra and IL-10.

24. The kit according to claim 19, wherein the kit comprises reagents for measuring IGFBP-5, IGFBP-4 and IL-6, IL-1 ra, IL-10 and SDF-Ι α+β.

25. The kit according to claim 19, wherein the kit comprises reagents for measuring IGFBP-5, IGFBP-4, IGF-II and IL-6, IL-1 ra, IL-10, SDF-Ι α+β.