Molecular Detection of Occult Disease in Non-Small Cell Lung Cancer

This is a collaborative Study with Dr. James D. Luketich at the University of Pittsburgh Cancer institute.

This project is funded through 2007 by an R01 grant to Dr. Godfrey from the NIH.

Summary: In patients with non-small cell lung cancer (NSCLC), tumor stage is the strongest determinant of prognosis. Stratification of patients into stages facilitates individual treatment decisions based on the survival statistics of a population. Within these staged populations however, subsets of patients with apparent early disease will still suffer cancer recurrence. This is due to the inability of current staging methods to detect small numbers of disseminated tumor cells (occult metastases or micrometastases) in these patients (reviewed in Coello et al. 2004 link to pdf). Reverse transcription-PCR (RT-PCR) has been shown to detect the presence of micrometastases in histologically negative specimens, and these findings correlate with poor outcome. Unfortunately, routine clinical application of this technique has been limited by “false positive” results in control tissues, a low specificity for predicting disease recurrence, and the lack of simple, standardized assays for multi-center trials. We have recently shown that quantitative RT-PCR (QRT-PCR) can discriminate between true and false positives, and that this results in an improved ability to predict recurrence (link to godfrey et al Clin Ca Res 2001 pdf). Furthermore, in our study of 30 histologically node-negative esophageal cancer patients, a positive QRT-PCR result was the strongest, independent risk factor for recurrence. In the lung cancer study, we will use QRT-PCR to detect micrometastases in the lymph nodes, blood and bone marrow of NSCLC patients. Recurrence and survival data will be collected and will enable us to determine the prognostic value of QRT-PCR for detecting both lymphatic and hematogenous tumor spread. Our goal is to improve the accuracy of NSCLC staging so that prognosis and treatment options are more closely correlated with individual patient outcomes.

Background: Lung cancer is currently the most common cause of cancer-related death in the USA and represents the second most common malignancy in both men and women ¹ . The incidence of lung cancer remains high despite anti-smoking campaigns and health education efforts and it is anticipated that lung cancer will continue to be a significant health-care problem for the foreseeable future.

Lung cancer is classified into two major sub-types, small cell lung cancer and non-small cell lung cancer (NSCLC).  Small cell lung cancer is usually disseminated at the time of diagnosis and systemic chemotherapy is the treatment of choice. Although NSCLC does respond to chemotherapy, surgical resection is the preferred treatment for patients with localized disease. Factors such as histologic sub-type, tumor differentiation and genetic alterations have not consistently been shown to influence the outcome of patients with NSCLC compared to the stage of the disease. The revised international staging system for NSCLC ² (shown in Table 1) uses Tumor (T), lymph node (N) and distant metastases (M) status to group patients into stages. The importance of accurate staging of NCSLC patients cannot be overemphasized. Development and testing of new treatment strategies depend heavily on knowledge of the end results achieved for carefully staged patient groups.  In addition, inaccurate staging results in less than optimal treatment of patients, and impedes progress in the evaluation of new, stage-specific treatment options for NSCLC patients.  For these reasons it is vital that we continue to pursue improvements in the current staging system, including the application of molecular research and technology ² .

In the absence of distant metastases, the most important factor in NSCLC staging is the extent of tumor spread to lymph nodes ⁴ (Figure 1).  For staging purposes, lymph nodes are divided into three levels (N1, N2 and N3 (Figure 2)) with each increasing level being associated with a higher stage, and worse prognosis.  Stage I patients have no pathologic evidence of metastatic tumor spread (pN0) and the standard of care for these patients is surgical resection with no adjuvant therapy.  Stage II patients have metastases to local lymph nodes (pN1), or a T3 tumor with no lymph node involvement. The primary treatment for stage II disease is surgery, although there are current clinical trials investigating the role of neoadjuvant and adjuvant chemotherapy or chemoradiotherapy in these patients ^5,6 .  Stage IIIA patients are those with tumor spread to ipsilateral mediastinal lymph nodes (pN2), or a T3 tumor with pN1 lymph node involvement. Three randomized clinical trials have shown that these patients benefit from neoadjuvant therapy followed by surgery ^7-9 . Patients with stage IIIB include tumors with significant local extension (T4), or contralateral mediastinal lymph node involvement (N3) and stage IV includes patients with distant metastases. Both stage IIIB and IV are generally considered inoperable with occasional exceptions ^10,11  and treatments are confined to radiation and chemotherapy. 

Stratification of patients into stages I-IV using the TNM system allows individual treatment decisions based on the survival statistics of a population, but within these populations, subsets of individual patients recur even in early stage disease.  For example, up to 40% of stage I NSCLC patients will suffer disease recurrence within 24 months despite potentially curative resection ⁴ .  A resected stage I patient has theoretically had all tumor removed based on current staging. There is accumulating evidence that this primary treatment failure is due to the inability of current staging methods to detect small numbers of disseminated tumor cells (micrometastases).  Thus, the ability to detect micrometastases would result in more accurate staging, and population statistics that more closely resemble individual outcomes. This is the overall goal of our research proposal.

Improved staging of NSCLC patients is important since it not only allows more accurate prognosis for the clinician and patient, but also may lead to effective treatment recommendations for subsets of upstaged patients. For example, if the standard pathologic staging of a patient is stage I, but molecular techniques identify this as a true IIIA stage, entry into a neoadjuvant protocol could lead to better survival. Postoperative chemotherapy and/or radiotherapy have found a less than 5% improvement in 5-year survival ¹² but ¹³ several trials have demonstrated a survival benefit using preoperative (induction) chemotherapy followed by surgery in NSCLC patients. Two such prospective randomized trials compared primary surgery with induction chemotherapy and surgery in stage IIIA NSCLC ^8,14 . Both trials showed a significant improvement in survival for patients treated with induction therapy and this advantage persisted with multivariate analysis and long-term follow-up ^9,15 . Trials in early stage (I and II) NSCLC have recently been completed and also suggest a survival benefit ^5,6 .  Success of these trials in stage I patients is based on the hypothesis that chemotherapy, or chemoradiotherapy, will eradicate radiographic and histologically occult loco-regional or systemic micrometastases in subsets of the patients who would otherwise suffer recurrence. Treating all stage I patients is not ideal, since 60% or more will remain disease-free with surgery alone and need not be subjected to the side effects of chemotherapy. Identification of occult metastasis prior to resection would therefore allow a more accurate selection of patients at greatest risk of recurrence, and allow assessment of chemotherapy in patients who are most likely to benefit.

Limitations of current staging.  Current imaging techniques (CT scan, PET scan, bone scan) cannot accurately identify microscopic tumor spread and require up to a square centimeter of tumor for consistent identification ^16,17 . Patients with no suspicious mediastinal lymph nodes (> 1 cm by CT, or no identification of increased glucose uptake by PET) and with no distant metastases are candidates for surgical resection without further work-up. Only patients with a suspicious preoperative imaging study undergo invasive biopsy and pathologic examination of lymph nodes or distant sites. At the present time, no guidelines or recommendations exist for the detection of micrometastasis to lymph nodes, bone marrow, blood, or other sites in NSCLC patients.

Pathologic examination of lymph nodes is carried out both intraoperatively, on frozen tissue, and then again postoperatively following formalin-fixation and paraffin-embedding of the lymph nodes. In both cases, one or two 4-micron sections are placed on slides, stained with hematoxylin and eosin (H&E) and examined for the presence of tumor cells. Fixed tissue examination is used for the final pathologic staging since it provides better tissue architecture and more time for examination by the pathologist compared to intraoperative examination. Even so, final pathology still suffers from low sensitivity for detecting individual tumor cells, or small foci of tumor. Pathologic examination only samples a very small percentage of each lymph node and it has been calculated that a pathologist has only a 1% chance of detecting a micrometastatic focus of three tumor cell diameter ¹⁸ . Indeed, there is evidence that micrometastases are missed by routine histological examination in a significant number of all cancer cases. In one study, lymph nodes from routine histological analysis of breast cancer patients were re-examined with extensive serial sectioning and 9% of patients were upstaged ¹⁹ .  Survival correlated with the micrometastases detected by serial sectioning in this study, demonstrating that these findings were clinically significant. Unfortunately, serial sectioning of every lymph node is too time-consuming to be used except in specific cases, such as sentinel lymph node biopsy, where only 1-2 nodes require examination. Thus, it appears that low sensitivity and sampling error are major factors in the current histopathologic analysis of lymph nodes. Methods are required that allow greater lymph node sampling and sensitive detection of otherwise occult metastases.

Importance of micrometastases in NSCLC.  Two methods are commonly used for detection of micrometastases, IHC and RT-PCR. Detection of micrometastases in the bone marrow and lymph nodes of NSCLC patients, however, has focused almost exclusively on the use of IHC (reviewed in ^20,21 and Coello et al. 2004 ). The targets of these studies are usually epithelial cell cytokeratins such as CK19, and CK18, or the surface glycoprotein EGP40.

In four studies on lymph nodes, micrometastases detection resulted in the upstaging (N0 to N1 or N1 to N2) of 22-70% patients ^22-25 . Three of these studies included clinical follow-up and reported a strong correlation between micrometastasis and disease free or overall survival ^22-24 .  The largest and most recent of these studies, by Kubuschok et al. ²² , found that IHC positive cells in the lymph nodes were associated with reduced disease free survival (p < 0.0001), overall survival (p = 0.0001) and, using multivariate analysis, a 2.7 fold increased risk for tumor recurrence. Furthermore, histologically occult N2 metastases detected by IHC have been shown to result in survival similar to patients with histologic N2 disease ²⁴ . This agrees with our findings in esophagus cancer using QRT-PCR to detect micrometastases in lymph nodes ²⁶ . RT-PCR has been used to detect occult lymph node metastases in NSCLC in only one study ²⁷ . This group used RT-PCR for the cell surface glycoprotein MUC-1 mRNA, and found that 38% of pN0 lymph nodes were positive for MUC-1 expression. The prognostic importance of this finding is unknown since no follow-up data have been published.

Hematogenously disseminated cancer cells have been found by IHC analysis in the bone marrow samples of 18-60% of NSCLC patients with no overt metastases ^22,28-31 . In all of these studies, the presence of IHC positive cells significantly correlated with disease-free survival.  Furthermore, in lymph node negative patients, multivariate analysis showed that bone marrow status was an independent predictor of overall survival, with an increased risk for shorter survival ³¹ . The prognosis of patients with pathologic N1 or N2 disease however was independent of bone marrow status, indicating that early hematogenous and lymphatic spread may be differentially regulated.  Detection of micrometastases in bone marrow could be extremely important clinically given the recent finding that biphosphonates reduce the risk of skeletal complications and visceral metastases in breast cancer patients with IHC positive cells in the bone marrow ³² .

Bone marrow aspiration is an invasive procedure and in most of the studies cited was obtained at the time of surgery with the patient anesthetized. Blood collection by venipuncture on the other hand is a simple and routine practice carried out in clinic.  In breast and prostate cancer, many studies have tried to detect tumor cells in blood and correlate findings with stage of disease. Results have been variable but several studies have shown prognostic significance ^33-36 . Only one similar study has been carried out in NSCLC ³⁷ . Using RT-PCR for CEA mRNA, Kurusu et al. found that 60% of preoperative blood samples from NSCLC patients were positive. Postoperatively this number fell to 27%. In both cases, CEA positivity by RT-PCR was correlated with pathologic TNM stage.

No studies have included micrometastatic analysis of lymph nodes, blood and bone marrow on the same patients. Thus, the relative prognostic value of each is unknown. In the proposed study we will analyze all three tissues on all patients, and use multivariate analysis to determine the significance of each variable.

Detection of micrometastases by RT-PCR.  Identification of micrometastases by RT-PCR relies on the detection of epithelial cell specific marker mRNAs in tissue that does not usually express that mRNA species. Theoretically, RT-PCR should be a very sensitive technique due to the power of PCR amplification. Indeed, some authors have reported detection of 1 cancer cell in a background of 10⁷ normal cells ³⁸ . The majority of groups however report sensitivities of 1 in 10⁵ or 10⁶. This variability is probably due to differences in methods, and the expression level of the marker mRNA in the cancer cell type being used for the spiking experiments. 

RT-PCR has been used extensively to analyze lymph nodes, blood and bone marrow in many tumor types and all studies report frequent RT-PCR positivity in histologically negative samples. Recent studies in melanoma, prostate, esophageal, colorectal, and breast cancer have also shown a correlation between RT-PCR positivity and disease recurrence ^39-43 . In these studies, up to one half of the lymph node was used for RNA isolation and RT-PCR.  Since sampling error is the biggest limitation of pathologic examination (by IHC or H&E), the ability to assess large amounts of tissue in a single assay is probably the most significant advantage of RT-PCR.  There are however problems that need to be addressed before RT-PCR can become a clinically useful tool.

With few recent exceptions all RT-PCR studies to date have relied on gel-based assays and simple positive/negative detection of marker mRNAs as the criterion for the existence of occult micrometastases. While results from these studies show that sensitivity of the assay is high, the specificity is low. In a study by Liefers et al. ⁴⁰ , 14 of 26 histologically N0 colon cancer patients had evidence of N1 disease using RT-PCR, and the remaining 12 were N0 by both analyses. Of the 12 N0 patients, only one recurred during the 6-year follow-up period, while 7 of the 14 RT-PCR N1 patients suffered recurrence. Using recurrence as an end point, a sensitivity of 88% was achieved. The specificity however was only 61%, since 7 patients with RT-PCR positive nodes did not recur. Other studies have shown a similar low specificity for standard RT-PCR. In melanoma patients, Shivers et al. achieved 86% sensitivity and 51% specificity ⁴⁴ , and Bostick et al. reported 100% sensitivity and 67% specificity ³⁹ . These assays would therefore result in overstaging of many patients. One likely reason for this poor specificity, is the presence of background (often referred to as ectopic or illegitimate) expression of marker mRNAs seen in some control lymph nodes.  For example, Schoenfeld used a nested RT-PCR assay for CK19 and found positive results in lymph nodes from patients without cancer ⁴⁵ . More recent reports concur ^45-47  but this remains an area of controversy in that some investigators continue to report that all control samples from patients with benign disorders are invariably negative ^27,48,49 . As noted above, the most likely reason for this variability is that no two studies use the same RT-PCR methods, and thus sensitivity levels differ. This background expression in normal samples has led some authors to question the utility of RT-PCR for micrometastasis detection ⁴⁷ . Indeed, without quantitative studies, the extreme sensitivity of RT-PCR may actually interfere with its ability to discriminate between normal nodes and those with occult metastases. Our data however shows that quantitative RT-PCR can easily discriminate between background expression, and expression due to the presence of micrometastases. Furthermore, quantitation allows sensitivity and specificity to be optimized since the assay results are no longer a simple positive or negative, but a continuous variable for expression level. In Specific Aims 1 and 2 of this proposal we intend to use the advantages of quantitative RT-PCR to comprehensively determine the micrometastasis status of NSCLC patients and its effect on survival.

Experimental design and data analysis. The overall experimental design is shown as a flow diagram in Figure 3, and described in the following text. Based on our current clinical volume, we estimate that the UPMC Thoracic Surgery Department will see approximately 500 new cases of NSCLC for each year of the proposed study. Of the 500 patients seen each year, approximately 235 will be either stage IIIb or IV following clinical staging and will not be candidates for surgery. Approximately 265 remaining patients will undergo surgical resection and/or staging (approximately 500 procedures), and a pathologic stage will be determined. With this volume of patients, we will easily be able to enroll enough patients for this study. Patient accrual will occur for years 1-3 of the study, or until accrual goals, determined by statistical power analyses below, are met for each specific aim.

Specific Aims.

Specific Aim 1: To determine the prognostic importance of occult lymph node metastases detected by quantitative RT-PCR. In this aim, we will compare histological and IHC analysis of lymph nodes with QRT-PCR analysis for expression of CEA and CK19 in (a) patients with benign disorders (controls) and (b) NSCLC patients. The patients with benign disorders will provide our baseline levels of CEA and CK19 expression. These will define the lower limit of expression that can be used as a cut-off in further analyses. Control lymph nodes will be obtained from patients undergoing laparoscopic surgery for gastroesophageal reflux disease and hiatal hernias. At the University of Pittsburgh Medical Center we perform approximately 100 such cases per year and routine resection of the hernia sac and fat pad yield an abundance of benign lymph nodes for this part of the study.  For NSCLC patients, we will collect portions of all lymph nodes sent by the surgeon for intraoperative pathology consultation.  This will include nodes taken at the time of mediastinoscopy (N2 and N3), thoracoscopy (N1) and resection (wedge, lobectomy or pnemonectomy (N1 and N2)). Results of the intraoperative pathology analyses will be recorded for comparison with QRT-PCR, but final pathology will be used to stratify the patients into the appropriate groups for sub-aims 1a-1c.  Since the tissue used for the research project will be different from that used for pathologic analysis, we will also cut sections to perform our own microscopic examination of lymph node tissues used for QRT-PCR.  All research lymph node tissue will be mounted in optimal cutting temperature compound (OCT) to enable sectioning on a cryostat. Nodes will then be serially sectioned at 4mM thickness. The first three, and then every eighth, ninth and tenth section will be placed on a microscope slide while all intervening sections will be used for RNA isolation. A total of 53 sections will be cut initially resulting in 18 slides and 35 sections for RNA isolation (yields 2-7mg total RNA depending on size of lymph node). Thus, 3 slides will be available for H&E or IHC staining and examination every 32-44mM, or approximately every 2-cell widths. Aliquots of RNA from the 35 remaining sections will be analyzed for CEA and CK19 expression using our TaqMan QRT-PCR procedure, and data analysis will be performed on three subsets of patients as described in sub-aims 1a-c.  When QRT-PCR results and final pathology results are discordant, we will first review the original H&E slides to confirm the original diagnosis and then we will use the sections cut from the research tissue for a separate H&E and IHC analysis. Thus we will be able to determine if the discordance is due to sampling error, epithelial cell contamination or other unanticipated characteristics of the lymph nodes. While this part of the study will be extremely labor intensive, we believe that the ability to correlate QRT-PCR findings with the best possible pathologic examination using IHC is one of the strongest aspects of this project. If QRT-PCR is ever to be used in a clinical setting, comparisons such as this will be essential for FDA approval.

Sub-Aim 1a:  To determine the prognostic importance of QRT-PCR positive N1 and N2 lymph nodes in Pathologic-N0 (pN0) patients. This group will consist of patients diagnosed with early stage cancer. This includes patients with no evidence of lymph node involvement on scans and also patients with mediastinoscopy results negative for pN2 disease. Many patients staged by mediastinoscopy however will have pN1 disease diagnosed at the time of surgery. Therefore, for entry into this part of the study we will require that N1 nodes be sampled at the time of resection, and that the final pathologic stage is pN0.  In this group 30-40% of patients are expected to recur, thus this is the group in which QRT-PCR staging could have the greatest impact. We will first determine if the absolute CEA and CK19 expression levels measured by QRT-PCR correlate with risk of recurrence, as expected from our observations in esophageal cancer. Next we will define CEA and/or  CK19 expression level cut-off values for stratifying patients as QRT-PCR positive or negative. In order to obtain the best evaluation of our assay, we need to determine the expression cut-off values on a subset of patients (training set) and then apply this cut-off to the remaining patients (assessment set). Therefore the cut-off values will be calculated using a subset of the pN0 patients as described in the statistical analysis section below. This cut-off will then be used to stratify the remainder of patients in sub-aim 1a. Patients will be classified as qN0, qN1 or qN2 based on QRT-PCR for  CEA, CK19 or a combination of both.  Since these patients are all pN0, we anticipate that most of them will fall into the qN0 or qN1 categories although a small number may also have qN2 invovement. For any samples that are positive by QRT-PCR, we will carry out H&E and immunostaining to try and visually identify cells responsible for the QRT-PCR signal. We will then calculate concordance rates for QRT-PCR, H&E and IHC positivity.  Next, we will use recurrence information on all patients to determine which QRT-PCR marker, or combination of markers, provides the best sensitivity and specificity for predicting disease recurrence.  Finally, we will plot Kaplan-Meier disease free and overall survival curves for qN0, qN1 and qN2 patients.  Relative risk estimates will be calculated. With our current volume of NSCLC patients, we estimate that we will see approximately 180 pN0 patients per year, split  65:115 between stage IA (T1N0M0) and IB (T2N0M0) respectively. With this number of patients we will be able to determine the risk of recurrence associated with micrometastases in all stage I patients as well as in stage IA and IB separately.  Our preliminary data suggests that micrometastatic nodal disease is as clinically important as pathologic nodal disease. Therefore, we expect survival in these groups to drop to levels similar to stages IIA (T1N1M0) and IIB (T2N1M0) respectively.  Based on the power analysis below,  the acrual goal for this part of the study is 400 patients and should be achieved in the first three years of the study. Since we expect to obtain 12-17 mediastinal and hilar lymph nodes from each patient, a total of 4800-6800 lymph nodes will be studied in this part of the proposal.  Carried out over 4 years, this will require analysis of 24-34 lymph nodes per week, a number which is easily achievable by the technical personnel on this project.

Sub-Aim 1b: To determine the clinical significance of QRT-PCR positive N2 nodes in pN1 patients.

This aim is identical to aim 1a except that the patients will be that relatively small group with pN1 (T1N1M0 and T2N1M0) disease but without N2 disease seen on pathology.  Lymph nodes will be collected at the time of resection and, if applicable, at mediastinoscopy. Cut-off limits for QRT-PCR will be determined as in Aim 1a using a subset of the pN1 patients.  We will then use recurrence information to see if micrometastatic qN2 disease is associated with poor prognosis in these patients. Given the relatively low number of patients presenting with stage II disease, and the even lower number in the stage IIA group,  we will not have sufficient statistical power to subdivide our survival analysis into T1 and T2 , QRT-PCR positive or negative. Thus, in this aim, all stage II patients will be analyzed as a single group. The accrual goal for Sub-Aim 1B is 105 patients and we anticipate that this will be reached in the first three years of the project. This will add a further 6-9 lymph nodes per week to be analyzed.

Patients with micrometastatic N2 disease should probably be considered for neoadjuvant chemotherapy protocols. However, the diagnosis of pN1 is typically obtained post-resection, and therefore this is not possible. In the future, it is possible that N1 node staging could be accomplished via thoracoscopy and that this could become a standard part of NSCLC patient staging. If so, intraoperative QRT-PCR staging (Specific Aim 3) could play an important role in the care of these patients.

Sub-Aim 1c: To determine the clinical significance of residual micrometastatic disease in pN2 patients following neoadjuvant chemotherapy. The goal of Sub-Aim 1c is to determine the clinical significance of QRT-PCR positivity in patients who have been histologically downstaged following chemotherapy.  Patients in this arm of the study have histologically positive N2 nodes documented during their initial evaluation by mediastinoscopy and then go on to receive neoadjuvant chemotherapy or chemo-radiotherapy. Patients who choose to receive their neoadjuvant therapy in our center will enter a Phase II chemotherapy trial (IRB #97-086). The therapy generally consists of three cycles of platinum-based chemotherapy and up to 5400 rads of loco-regional radiotherapy. This trial is conducted by two of our co-investigators, Dr. James D. Luketich and Dr. Chandra Belani.  The analyses to be carried out on these patients are essentially the same as described in sub-aim 1c above.  Histologically downstaged patients will be classified as QRT-PCR N0,N1 or N2 and we will plot Kaplan-Meier disease-free survival curves for these three sets of patients. Two randomized trials have shown that neoadjuvant therapy in this setting improves survival ^8,14   but the subsets with the best survival include those with a complete response to N0 or from N2 to N1 ⁶³ . However, even patients who are rendered histologically node negative have a greater than 50% chance of developing recurrent disease. The goals of this subaim will be to assess the QRT-PCR status of histologically negative nodes following neoadjuvant therapy and determine if this information will predict recurrence in some patients. Important clinical information could be obtained which might impact on who should undergo aggressive surgical resection, and may help guide post-resectional chemotherapy and radiotherapy in patients found to be at higher risk of recurrence.  We estimate that a total of 50 patients per year will be diagnosed with N2 disease, and receive neoadjuvant therapy and that approximately 40-50% will be downstaged. Thus, during the four years of patient accrual, we anticipate that a total of 80-100 downstaged patients will be studied. Approximately 15 lymph nodes will be resected at the time of surgery yielding a total of 1200-1500 lymph nodes for study.

Specific Aim 1 - Statistical Analysis Approach:  The investigation of the usefulness of quantitative RT-PCR for biomarkers CEA and CK19 mRNA levels  will be assessed in three stages.  In the first stage the quantitative levels of both markers as determined by QRT-PCR will be tested for their significance in a proportionate hazards regression model for disease-free survival.  Predictors will include quantitative mRNA levels of CEA, CK19 and their cross product, to test for interaction.  Once either or both markers are shown to affect disease-free survival, the second stage of the investigation will proceed.  In the second stage, the usefulness of expressing QRT-PCR as a binary value (+/-) for diagnosing micrometastasis and predicting recurrence will be evaluated. The predictive ability of CEA and CK19 mRNA levels will then be compared by ROC curve analysis in a random sample of one half of the cohort of pNegative patients.   If one ROC curve (CEA or CK19) is uniformly superior, the associated biomarker will have uniformly better classification and will be selected for restaging pathologic negative patients.  If the two ROC curves cross they will be compared by the permutation test method of Venkatraman and Begg ⁶⁴ for paired data.  To assist with a definition of recurrence that is relatively independent of survival time, a proportionate hazards cure  model ^65,66  will be fit to find evidence of a diminished hazard rate. If the cure rate model is appropriate, recurrence at the time of the change point will be used for classifying patients by recurrence status.  If the cure model is not appropriate a clinically relevant time will be chosen or a time-dependent ROC analysis method will be used ⁶⁷ .

The optimum cutoff value of the superior marker will be determined by classification accuracy, i.e., the value that correctly classifies the recurrence status of the greatest proportion of patients.   Once the optimum cutoff is set, the third stage of the data analysis will proceed in which the cutoff determined in the first half of the patient cohort (training set), will be evaluated in the second half of patients (assessment set).  Diagnostic statistics of sensitivity, specificity, predictive value of positive and negative diagnoses and classification accuracy will be calculated in the assessment group. Since the definition of recurrence may be time dependent, the most powerful evaluation of classification by QRT-PCR will be achieved by restaging pNegative patients and testing for differences in disease-free survival by the log-rank test between QRT-PCR positive and QRT-PCR negative patients.  If there are insufficient patients to divide into two groups, one for development of the cutoff and one for evaluation of the selected cutoff, then the same cohort will be used for both steps with K-fold cross-validation ⁶⁸ .

Power and Sample Size:   The most important potential use of QRT-PCR in a clinical setting may be the ability to  restage pathologically negative patients and successfully predict disease recurrence.  It is likely that if there are enough patients for a predictive model for QRT-PCR outcome as a binary variable, QRT-PCR positive (q+) or QRT-PCR negative (q-), then sample size will be more than adequate for evaluating QRT-PCR as a quantitative variable.   Accordingly, the predictive ability of restaging as either QRT-PCR positive or negative, as determined by a log rank test  for three year disease free survival, was chosen for conservatively computing desired sample sizes. 

Sample size calculations applied the following assumptions.  Two groups of  patients, all pathologically negative (p-), will be restaged as  q+ or q-.  P-/q+ patients will recur at the historical rate of p+ patients.  For example, pN0 patients who are restaged as  pN0/qN1 will recur at the rate of pN1 patients.  Since we are unsure what difference to expect between q+ and q- patients we assume that if q+ patients have at least double the recurrence hazard  of q- patients, enough patients should be accrued to allow a statistically significant difference with a one tailed log rank test at a = .05.  It is further assumed that 25% of  p- patients will be restaged at q+ and that the 3 year disease free survival, which is not readily available, is approximately the same as the 4 year overall survival for patients in various lung cancer stages. 

In the table below the required number of patients are shown for each sub-aim in Specific Aim 1, in which a log rank test will be conducted to assess the disease free survival between q+ and q- patients.  Calculations are shown for selected combinations of accrual and follow-up, all designed to accommodate a five year study.

The entries show the number of patients needed per year of accrual, for either a two or three year accrual plan that will give either 80% or 90% power to detect a two group difference in disease free survival.  These numbers can be compared to the expected accrual rates for UPMC lung cancer patients by disease stage on page 42 (Table 2).  Sample size requirements for each of the sub-aims are discussed below.

Sub-Aim 1a sample size: We assume that if QRT-PCR determination of CEA or CK19 is to have prognostic importance, then  pN0 patients who are QRT-PCR negative should differ significantly in disease free survival from pN0 patients who are restaged as QRT-PCR positive.  A guide for power and sample size calculations is the ability to detect a hazard ratio of two, that is, when the hazard rate of QRT-PCR positive patients is double that of QRT-PCR negative patients. 

It is assumed that the stage 1A pN0/q+ patients who recur will do so at the historical rate of stage 2A pN1 patients, that is a disease-free survival of 58% at 3 years.  With 3 years of accrual and 2 years of follow up 64  patients per year are needed to detect a significant difference of double the hazard ratio with a log rank test with alpha = .05 and power = .80. This is  within the  65 stage 1A patients expected per year at UPMC.  Similarly, stage 1B patients studied in aim 1a are assumed to exhibit disease free survival of 43%, identical to historical Stage 2B disease free 3 year survival.   Since 115 stage 1B patients are expected to be eligible per year,  either 2 years of accrual with 2 years of follow-up or 3 years of accrual and 1 year of follow will be sufficient to provide 90% power to detect a significant difference between survival proportions of 43% in the QRT-PCR positive group and 66% in the QRT-PCR negative group (ratio of hazard rates = 2.0).

Sub-Aim 1b sample size:   As in aim 1a, sample size is dictated by attempting to detect a significant difference between pN1 patients who are restaged as N2 positive or negative by QRT-PCR.  The historical disease-free survival for pN2 positive patients is about 14% at 3years. If the QRT-PCR positive pN1 patients have a similar disease free survival as pN2 patients and their survival is double the hazard of QRT-PCR negative patients (three year survival = .37), then  35 patients per year  are required for 80% power  for 3 years of accrual and 1 year of follow up.  This number is within the 35 – 40 stage II lung cancer patients expected per year at  UPMC. 

Sub-Aim 1c sample size:  This sub aim will attempt to restage patients who are initially pN2, who are treated with pre-operative chemotherapy, and who have a complete pathologic response to their chemotherapy.   We assume that any such patient who is q+ will have a poor prognosis and for the purpose of sample size determination, will exhibit 3 year disease free survival of only 27%, approximately the same as Stage IIIA pN2 patients.  Patients who are q- will have one half the hazard of recurrence or 52% 3 year disease-free survival.  For a  log rank test to claim a significant survival difference, 35 patients per year would be required for 3 years of accrual and two years of follow-up.  It is unlikely that we will be able to accrue 35 patients per year to this study. Therefore, accrual will continue into year 4, until sufficient numbers are reached.

Specific Aim 2:  To determine the prognostic importance of hematogenous spread of tumor cells detected by TaqMan quantitative RT-PCR.  In this aim we will analyze both peripheral blood (Sub-Aim 2a) and bone marrow samples (Sub-Aim 2b) from controls and from stage I-IIIA NSCLC patients. Samples will be obtained at two time points, preoperatively and again several weeks post-operatively. Statistical analyses indicate that only 300 patients are required for this specific aim but we will  approach all patients in Specific Aim 1 for blood and bone marrow collection. Control blood samples will be obtained from patients with gastro-esophageal reflux disease but with no evidence of Barrett’s esophagus or esophageal dysplasia. These patients will be consented to our esophagus cancer risk registry research protocol and blood will be drawn in clinic. Using these samples, we will determine if simple QRT-PCR positivity and/or absolute CEA and CK19 expression levels in blood and bone marrow correlate with pathologic-tumor stage or with risk for disease recurrence.  Also, in combination with data from Specific Aim 1, we will be able to determine the prognostic value of QRT-PCR staging for lymphatic and hematogenous spread both independently and combined. There is evidence that bone marrow micrometastasis and pathologic lymph node status are independent  prognostic variables in NSCLC. However, none of these studies have assessed micrometastatic disease in lymph nodes of the pN0 patients and therefore it is still not known if pN0 patients with bone marrow micrometastases actually have micrometastatic lymph node involvement also. Specific Aims 1 and 2 combined will be able to address this important question.

Sub-Aim 2a. Detection of tumor cells in blood of NSCLC patients.  Many studies have shown the presence of tumor cells in peripheral blood of cancer patients although the clinical relevance of these findings is questionable. At this point, it seems unlikely that  detection of tumor cells in the blood will have sufficient sensitivity or specificity to be useful on its own. However, the use of a simple, quantitative assay may still have some prognostic value. In particular, detection of tumor cells remaining several weeks after surgery may indicate the need for systemic treatment despite an apparently curative resection. Furthermore, this small part of the project is simple and straightforward in terms of blood collection, RNA isolation and analysis. We therefore feel that these data merit inclusion in the current study.

Blood samples will be collected from all consenting patients seen by the Thoracic Surgery Section at UPMC at two different times. Pre-operative blood samples will be collected in heparinized vaccutainers either in the clinic or at the time of surgery, before the first incision is made. To avoid potential contamination with epithelial cells from the skin puncture, the first tube of blood will not be used for QRT-PCR analysis. Post-operative samples will be collected at the second follow-up clinic visit (typically 2-4 months). Blood will be processed immediately by centrifugation through a Ficoll gradient and isolation of mononuclear cells. RNA will be isolated from the pelleted cells using Qiagen RNA-easy columns (Qiagen, Valencia CA), and quantitated by spectrophotometry.  All bloods will be analyzed for CEA and CK19 expression and we will test for correlation between QRT-PCR and (a) tumor stage and (b) recurrence of disease.

Sub-Aim 2b. Detection of tumor cells in bone marrow of NSCLC patients.  Several groups have shown that cancer cells can be detected in bone marrow of cancer patients and that this correlates with poor prognosis. QRT-PCR could provide a simple and fast method for detection of micrometastatic disease but, to date, no one has reported the use of this approach in NSCLC or any other solid tumor.  In this study we will take bone marrow, from all consenting patients, at the time of surgery.  Bone marrow (3-5ml) will be aspirated from one iliac crest prior to surgical incision  and also from the rib in cases where thoracotomy is carried out by the surgeon. Thoracotomy requires removal of a small section of rib and this excess tissue is then available for study. Bone marrow will be obtained by immediately opening the rib and curetting the bone marrow into heparinized media. This process results in high quality bone marrow yielding approximately 40-60 million cells per patient ²⁸ .  Bone marrow from the iliac crest will be aspirated into a syringe containing heparin to prevent clotting. All bone  marrow samples will be sent for immediate processing (Ficoll purification of mononuclear and epithelial cells) in the pathology tissue bank laboratory.  After isolation of mononuclear cells, the bone marrow samples will be split into two aliquots, one for RNA isolation and QRT-PCR, and one for formalin-fixation and embedding in paraffin. The fixed cells will be stored and sections will be cut for immunostaining of any samples found to be positive by QRT-PCR.  Thus, as in Specific Aim 1, we will attempt to visually identify cells responsible for QRT-PCR positivity.

Aim 2 - Statistical Analysis Approach:  Aim 2 will evaluate the prognostic ability of two novel indicators of  micrometastasis, mRNA expression levels by QRT-PCR in bone marrow and peripheral blood.  Patients will be asked to provide blood and bone marrow samples prior to, or during surgery and again at their first post-operative visit.  Patients will be accrued for three years and followed for disease-free survival.  The endpoint will be time to recurrence following resection of their primary tumor.  Prognostic importance of biomarkers in blood and bone marrow will be evaluated by Cox proportionate hazards models.  Should CEA or CK19 mRNA levels significantly affect disease free survival, other predictors will be added to the Cox models including pathologic stage, quantitative levels of CEA/CK19 by QRT-PCR in lymph nodes, and any other clinical or pathologic parameters found to predict disease-free survival. If proportionate hazards models are successful in predicting disease-free survival from mRNA as a quantitative predictor, then expression levels will be converted to a binary variable (positive or negative).   The binary classification will then be tested for ability to diagnose micrometastasis using disease recurrence as the standard.  As in aim 1,  various cutpoints over the range of quantitative expression will be evaluated for sensitivity, specificity and classification accuracy.

In the proposed Cox models, values of mRNA at the two different times will be considered as candidate predictors.  However, due to the invasive nature of obtaining bone marrow aspirates it is likely that many patients who undergo surgery and consent to removal of bone marrow at surgery will not consent to post operative bone marrow aspiration.  Therefore we anticipate that a subset analysis will be conducted by rebuilding Cox models of disease-free survival in the subset agreeing to a second bone marrow sample.  To assess the validity of inference from this anticipated small subset of patients,  patients who do submit to post operative  bone marrow sampling will be compared to patients who do not agree.  The comparison between the two subgroups will consist of comparison of variables measured for both subgroups such as age, gender, clinical stage, geographic residence etc.  If the two groups of patients are similar with respect to other variables then results on the subset of patients with a second bone marrow sample can be extrapolated to the larger sample.  If the two groups differ, inference about the utility of obtaining the post-operative bone marrow sample will be limited to the subgroup of patients agreeing to the second procedure.

Sample size for Aim 2:   The method of analysis for Aim 2 is to test for significant influence of a regression parameter in a proportionate hazards model.  Sample size can be determined by calculating the number of patients for a  log rank test since the log rank test is equivalent to the score test for the null hypothesis of equal hazards.  For the purpose of power analysis, patients will be divided into two groups with respect to the median of their quantitative mRNA levels and tested for equality of disease free survival between groups.  Table 4 below shows the number of patients required for a log rank test to detect a significant difference in disease free survival if the highest expression level subgroup had double the hazard of the lowest subgroup.  Calculations are based on 80% and 90% power, alpha of .05, three years of accrual and 1 year of additional observation.  Sample sizes in the table are somewhat conservative because the actual analysis of blood and bone marrow mRNA levels will use the continuous, quantitative expression levels rather than splitting patients into two groups.  According to the table, 100 patients each year for three years will be sufficient to adequately analyze mRNA levels in peripheral blood or bone marrow over a wide range of disease-free survival scenarios. 

Specific Aim 3:  To develop and test a rapid, quantitative RT-PCR system for intraoperative lymph node staging. Currently our intraoperative QRT-PCR  procedure requires extensive handling for RNA isolation and RT-PCR set up. Although RNA isolation and RT-PCR set-up can be done very quickly, neither is practical in the setting of a surgical pathology laboratory. For this reason we propose to develop a fully automated and integrated RNA isolation and QRT-PCR system that will provide a result from frozen sections of lymph nodes in less than 20 minutes.  We anticipate that frozen tissue sections will be placed in a single-use, disposable cartridge that will then integrate into a quantitative thermal cycler instrument. RNA isolation and addition of RT-PCR reagents will occur in the cartridge and the mixture will then be automatically transferred to an attached PCR tube for RT and QPCR. QPCR will be monitored in real time and software will automatically call the sample positive or negative. Since the input into this system will be frozen tissue sections, it will allow for routine histologic analysis of lymph nodes, and add the extra senstitivity of the QRT-PCR procedure without appreciably delaying the information reaching the surgeon.  Once this system is fully developed it will provide a simple, standardized assay for multi-center clinical trials of micrometastasis detection by QRT-PCR.  Under Specific Aim 3a we will continue to develop this rapid QRT-PCR system and then, in aim 3b, we will use rapid QRT-PCR to analyze lymph nodes obtained from  NSCLC patients at the time of surgical staging. Results of the rapid QRT-PCR assay will be compared with intraoperative frozen section, final pathology and TaqMan QRT-PCR results.

Sub-Aim 3a. To develop rapid QRT-PCR chemistry which is completely internally controlled.

This aim will be carried out in collaboration with Cepheid (Sunnyvale CA).  Cepheid develops and manufactures miniature, fully integrated bioanalytical test systems which combine advanced microfluidics, microelectronics and software. The goal of these products is to give healthcare providers automated, portable instruments which produce faster results with higher accuracy. Cepheid already has a DNA based, fully automated PCR instrument for detection of bacteria within 20 minutes. This automated system carries out both DNA isolation and PCR in a sealed, disposable cartridge which fits into an optical PCR instrument for cycling and quantitation. On the basis of our preliminary data, using Cepheids Smart Cycler system, we have persuaded the company to explore development of an RNA-based instrument for clinical use. This instrument will accept frozen section tissue slices and carry out RNA isolation, reverse transcription and QRT-PCR with no further handling by the pathologist. We believe that this can be done in under 30 minutes and that the cost can be kept reasonable with single use, disposable cartridges. While Cepheid has the necessary engineering and nucleic acid isolation expertise for this project, our role is to develop the reverse transcription and QRT-PCR chemistries that will eventually go into the cartridge.

Sub-Aim 3b. To test the reaction chemistry developed in aim 1 on lymph nodes collected from NSCLC patients undergoing mediastinoscopy.  The greatest impact of intraoperative staging of NSCLC will come from analysis of N2 nodes at the time of staging via mediastinoscopy. Patients who have positive N2 nodes are currently offered neoadjuvant chemotherapy prior to resection and this has been shown to impart a survival benefit.  Presumably a significant number of patients with N2 disease are missed by histologic analysis due to sampling error and the inability to detect micrometastatic disease.  We believe that intraoperative QRT-PCR will be able to more accurately stage these patients and, in addition, will allow for immediate surgical resection in patients wthout N2 disease. However, the first requirement of the intraoperative QRT-PCR assay is that it give a positive result in all cases where histology is positive. Thus the first goal of this aim is to compare intraoperative QRT-PCR results with frozen section and final pathology on nodes that are positive. Next we will determine if micrometastases detected by intraoperative QRT-PCR, in histologically negative nodes, are associated with a higher risk of recurrence. To do this, we will analyze lymph nodes taken at medistinoscopy and previously analyzed by TaqMan QRT-PCR in specific Aim IB. This will allow us to compare results using the two methods, determine concordance and correlate results with clinical follow up to determine the prognostic importance of micrometastases detected by intraoperative QRT-PCR. If successful, molecular information regarding micrometastases could, for the first time, be made available intraoperatively to the surgeon.

In the final cartridge system we envision that all reagents will be lyophilized and compartmentalized for addition to the reaction following RNA isolation in the cartridge. For aim 3b however we will continue to use standard laboratory methods for RNA isolation and RT-PCR setup until the cartridge system becomes available.

Sample size for Aim 3:  Sub aim 3a focuses initially on assay development  which does not require statistical analysis. In the second part of sub aim 3a we will carry out a reproducibility study by taking multiple measurements of the same specimen and computing the intra-class correlation coefficient and the standard error of the measurements.  Sub aim 3b will evaluate the concordance of rapid QRT-PCR determination with pathologic determination of N2 nodes.  While it is expected that QRT-PCR will discover micrometastasis in pathologically negative nodes, it will be possible in this aim to assess whether QRT-PCR will also be positive in pathologically positive nodes.  For widespread application of the technology, the two methods should agree, that is a p+ lymph node should also be q+ in a high proportion of cases.   We assume that 80% agreement (80% of p+ nodes are also q+) is the highest unacceptable agreement rate and that 90% agreement is the lowest rate that would be acceptable.  To test the hypothesis that a single binomial proportion is greater than 80% with 90% power to detect a proportion as high as 90%, will require 112 p+ nodes.  If 97 of the 112 p+ nodes are also q+ we will conclude that rapid QRT-PCR agrees with the pathologist’s positive determination at least 90% of the time.

Literature Cited:

1. Boring,C.C., Squires,T.S. & Tong,T. Cancer statistics, 1992 [published erratum appears in CA Cancer J Clin 1992 Mar-Apr;42(2):127-8]. CA Cancer J. Clin. 42, 19-38 (1992).

2.   Mountain,C.F. Revisions in the International System for Staging Lung Cancer. Chest 111, 1710-1717 (1997).

3.   Mountain,C.F. A new international staging system for lung cancer. Chest 89, 225S-233S (1986).

4.   Naruke,T., Goya,T., Tsuchiya,R. & Suemasu,K. Prognosis and survival in resected lung carcinoma based on the new international staging system [published erratum appears in J Thorac Cardiovasc Surg 1989 Mar;97(3):350]. J. Thorac. Cardiovasc. Surg. 96, 440-447 (1988).

5.    Depierre,A., Milleron,B. & Moro,D. Phase III trial of neo-adjuvant chemotherapy in resectable stage I (except T1 N0), II and IIIA non-small cell lung cancer: The french experience. Proc Am Soc Clin Oncol 18, 465a. 1999.

6.   Pisters,K.M. et al. Induction chemotherapy before surgery for early-stage lung cancer: A novel approach. Bimodality Lung Oncology Team. J. Thorac. Cardiovasc. Surg. 119, 429-439 (2000).

7.   Martini,N. et al. Preoperative chemotherapy for stage IIIa (N2) lung cancer: the Sloan- Kettering experience with 136 patients. Ann. Thorac. Surg. 55, 1365-1373 (1993).

8.   Roth,J.A. et al. A randomized trial comparing perioperative chemotherapy and surgery with surgery alone in resectable stage IIIA non-small-cell lung cancer. J. Natl. Cancer Inst. 86, 673-680 (1994).

9.   Rosell,R. et al. Preresectional chemotherapy in stage IIIA non-small-cell lung cancer: a 7-year assessment of a randomized controlled trial. Lung Cancer 26, 7-14 (1999).

10.   Luketich,J.D., Martini,N., Ginsberg,R.J., Rigberg,D. & Burt,M.E. Successful treatment of solitary extracranial metastases from non-small cell lung cancer. Ann. Thorac. Surg. 60, 1609-1611 (1995).

11.   Luketich,J.D. & Burt,M.E. Does resection of adrenal metastases from non-small cell lung cancer improve survival? Ann. Thorac. Surg. 62, 1614-1616 (1996).

12.    Chemotherapy in non-small cell lung cancer: a meta-analysis using updated data on individual patients from 52 randomised clinical trials. Non-small Cell Lung Cancer Collaborative Group.  BMJ 311, 899-909 (1995).

13.    Postoperative radiotherapy in non-small-cell lung cancer: systematic review and meta-analysis of individual patient data from nine randomised controlled trials. PORT Meta-analysis Trialists Group. Lancet 352, 257-263 (1998).

14.   Rosell,R. et al. A randomized trial comparing preoperative chemotherapy plus surgery with surgery alone in patients with non-small-cell lung cancer. N. Engl. J. Med. 330, 153-158 (1994).

15.   Roth,J.A. et al. Long-term follow-up of patients enrolled in a randomized trial comparing perioperative chemotherapy and surgery with surgery alone in resectable stage IIIA non-small-cell lung cancer. Lung Cancer 21, 1-6 (1998).

16.   Luketich,J.D. et al. Role of positron emission tomography in staging esophageal cancer. Ann. Thorac. Surg. 64, 765-769 (1997).

17.   Gupta,N.C., Graeber,G.M., Rogers,J.S. & Bishop,H.A. Comparative efficacy of positron emission tomography with FDG and computed tomographic scanning in preoperative staging of non-small cell lung cancer. Ann. Surg. 229, 286-291 (1999).

18.   Gusterson,B.A. & Ott,R. Occult axillary lymph-node micrometastases in breast cancer [letter; comment]. Lancet 336, 434-435 (1990).

19.    Prognostic importance of occult axillary lymph node micrometastases from breast cancers. International (Ludwig) Breast Cancer Study Group [see comments]. Lancet 335, 1565-1568 (1990).

20.   Pantel,K., Cote,R.J. & Fodstad,O. Detection and clinical importance of micrometastatic disease. J. Natl. Cancer Inst. 91, 1113-1124 (1999).

21.   Cote,R.J., Hawes,D., Chaiwun,B. & Beattie,E.J., Jr. Detection of occult metastases in lung carcinomas: progress and implications for lung cancer staging. J. Surg. Oncol. 69, 265-274 (1998).

22.   Kubuschok,B., Passlick,B., Izbicki,J.R., Thetter,O. & Pantel,K. Disseminated tumor cells in lymph nodes as a determinant for survival in surgically resected non-small-cell lung cancer. J. Clin. Oncol. 17, 19-24 (1999).

23.   Maruyama,R. et al. Relationship between early recurrence and micrometastases in the lymph nodes of patients with stage I non-small-cell lung cancer. J. Thorac. Cardiovasc. Surg. 114, 535-543 (1997).

24.   Izbicki,J.R. et al. Mode of spread in the early phase of lymphatic metastasis in non-small- cell lung cancer: significance of nodal micrometastasis. J. Thorac. Cardiovasc. Surg. 112, 623-630 (1996).

25.   Chen,Z.L. et al. Frequency and distribution of occult micrometastases in lymph nodes of patients with non-small-cell lung carcinoma. J. Natl. Cancer Inst. 85, 493-498 (1993).

26.   Godfrey,T.E., Raja,S., Finkelstein,S.D., Kelly,L.A. & Luketich,J.D. Quantitative Reverse Transcription-Polymerase Chain Reaction Predicts Disease Recurrence in lymph Node-Negative Esophagus Cancer Patients. N.Engl.J.Med. In Review. 2001.

27.   Salerno,C.T. et al. Detection of occult micrometastases in non-small cell lung carcinoma by reverse transcriptase-polymerase chain reaction. Chest 113, 1526-1532 (1998).

28.   Cote,R.J. et al. Detection of occult bone marrow micrometastases in patients with operable lung carcinoma. Ann. Surg. 222, 415-423 (1995).

29.   Ohgami,A. et al. Micrometastatic tumor cells in the bone marrow of patients with non- small cell lung cancer. Ann. Thorac. Surg. 64, 363-367 (1997).

30.   Pantel,K. et al. Immunocytological detection of bone marrow micrometastasis in operable non-small cell lung cancer. Cancer Res 53, 1027-1031 (1993).

31.   Passlick,B., Kubuschok,B., Izbicki,J.R., Thetter,O. & Pantel,K. Isolated tumor cells in bone marrow predict reduced survival in node- negative non-small cell lung cancer. Ann. Thorac. Surg. 68, 2053-2058 (1999).

32.   Diel,I.J. et al. Reduction in new metastases in breast cancer with adjuvant clodronate treatment [see comments]. N. Engl. J. Med. 339, 357-363 (1998).

33.   Corey,E. et al. Detection of circulating prostate cells by reverse transcriptase- polymerase chain reaction of human glandular kallikrein (hK2) and prostate-specific antigen (PSA) messages. Urology 50, 184-188 (1997).

34.   Cama,C. et al. Molecular staging of prostate cancer. II. A comparison of the application of an enhanced reverse transcriptase polymerase chain reaction assay for prostate specific antigen versus prostate specific membrane antigen. J. Urol. 153, 1373-1378 (1995).

35.   Olsson,C.A. et al. The use of RT-PCR for prostate-specific antigen assay to predict potential surgical failures before radical prostatectomy: molecular staging of prostate cancer. Br. J. Urol. 77, 411-417 (1996).

36.   Ghossein,R.A., Carusone,L. & Bhattacharya,S. Review: polymerase chain reaction detection of micrometastases and circulating tumor cells: application to melanoma, prostate, and thyroid carcinomas. Diagn. Mol. Pathol. 8, 165-175 (1999).

37.   Kurusu,Y., Yamashita,J. & Ogawa,M. Detection of circulating tumor cells by reverse transcriptase- polymerase chain reaction in patients with resectable non-small-cell lung cancer [see comments]. Surgery 126, 820-826 (1999).

38.   Mori,M. et al. Molecular detection of circulating solid carcinoma cells in the peripheral blood: the concept of early systemic disease. Int. J. Cancer 68, 739-743 (1996).

39.   Bostick,P.J. et al. Prognostic significance of occult metastases detected by sentinel lymphadenectomy and reverse transcriptase-polymerase chain reaction in early-stage melanoma patients. J. Clin. Oncol. 17, 3238-3244 (1999).

40.   Liefers,G.J. et al. Micrometastases and survival in stage II colorectal cancer. N. Engl. J. Med. 339, 223-228 (1998).

41.   Luketich,J.D. et al. Detection of micrometastases in histologically negative lymph nodes in esophageal cancer. Ann. Thorac. Surg. 66, 1715-1718 (1998).

42.   Wood,D.P., Jr. & Banerjee,M. Presence of circulating prostate cells in the bone marrow of patients undergoing radical prostatectomy is predictive of disease-free survival. J. Clin Oncol 15, 3451-3457 (1997).

43.   Fields,K.K. et al. Clinical significance of bone marrow metastases as detected using the polymerase chain reaction in patients with breast cancer undergoing high-dose chemotherapy and autologous bone marrow transplantation. J. Clin Oncol 14, 1868-1876 (1996).

44.   Shivers,S.C. et al. Molecular staging of malignant melanoma: correlation with clinical outcome.  JAMA 280, 1410-1415 (1998).

45.   Schoenfeld,A. et al. Detection of breast cancer micrometastases in axillary lymph nodes by using polymerase chain reaction. Cancer Res. 54 , 2986-2990 (1994).

46.   Zippelius,A. et al. Limitations of reverse-transcriptase polymerase chain reaction analyses for detection of micrometastatic epithelial cancer cells in bone marrow. J. Clin. Oncol. 15, 2701-2708 (1997).

47.   Bostick,P.J. et al. Limitations of specific reverse-transcriptase polymerase chain reaction markers in the detection of metastases in the lymph nodes and blood of breast cancer patients. J. Clin. Oncol. 16, 2632-2640 (1998).

48.   Schoenfeld,A., Luqmani,Y., Sinnett,H.D., Shousha,S. & Coombes,R.C. Keratin 19 mRNA measurement to detect micrometastases in lymph nodes in breast cancer patients. Br. J. Cancer 74, 1639-1642 (1996).

49.   Noguchi,S. et al. Detection of breast cancer micrometastases in axillary lymph nodes by means of reverse transcriptase-polymerase chain reaction. Comparison between MUC1 mRNA and keratin 19 mRNA amplification. Am. J. Pathol. 148, 649-656 (1996).

50.   Zippelius,A. & Pantel,K. RT-PCR-based detection of occult disseminated tumor cells in peripheral blood and bone marrow of patients with solid tumors. An overview. Ann. N. Y. Acad. Sci. 906, 110-123 (2000).

51.   Ginzinger,D.G. et al. Measurement of DNA copy number at microsatellite loci using quantitative PCR analysis. Cancer Res 60, 5405-5409 (2000).

52.   Raja,S., Luketich,J.D., Kelly,L.A., Finkelstein,S.D. & Godfrey,T.E. Intraoperative, quantitative RT-PCR detects nodal micrometastases in patients with esophageal cancer. American Association For Cancer Research 2001 Meeting Submitted(Abstract # 1791). 2001.
Ref Type: Abstract

53.   Raja,S., Luketich,J.D., Ruff,D.W., Kelly,L.A. & Godfrey,T.E. A Method for Increased Sensitivity of One-Step, Quantitative RT-PCR. Biotechniques 29, 702-705 (2000).

54.   Tassone,F. et al. Elevated Levels of FMR1 mRNA in Carrier Males: A New Mechanism of Involvement in the Fragile-X Syndrome. Am. J. Hum. Genet. 66, 6-15 (2000).

55.   Godfrey,T.E. et al. Quantitative mRNA expression analysis from formalin-fixed, paraffin-embedded tissues using 5' nuclease quantitative RT-PCR. Journal of Molecular Diagnostics 2, 84-91 (2000).

56.   Collins,C. et al. Positional cloning of ZNF217 and NABC1: genes amplified at 20q13.2 and overexpressed in breast carcinoma. Proc. Natl. Acad. Sci. U. S. A 95, 8703-8708 (1998).

57.   Shayesteh,L. et al. PIK3CA is implicated as an oncogene in ovarian cancer [see comments]. Nat. Genet. 21, 99-102 (1999).

58.   Raja,S., Luketich,J.D., Kelly,L.A., Finkelstein,S.D. & Godfrey,T.E. Intraoperative, quantitative RT-PCR detects nodal micrometastases in patients with esophageal cancer. American Association of Thoracic Surgery 2001 Meeting.San Diego, May 2001 . 2001.
Ref Type: Abstract

59.   Luketich,J.D. et al. The Role of Positron Emission Tomography in Evaluating Mediastinal Lymph Node Metastases in Non-Small Cell Lung Cancer. Clinical Lung Cancer In Press. 2-1-2001.
Ref Type: Journal (Full)

60.   Luketich,J.D. et al. Minimally Invasive Surgical Staging For Esophageal Cancer. Surgical Endoscopy Accepted, (2000).

61.   Landreneau,R.J. et al. The role of thoracoscopy in lung cancer management. Chest 113, 6S-12S (1998).

62.   Landreneau,R.J. et al. Wedge resection versus lobectomy for stage I (T1 N0 M0) non-small-cell lung cancer. J. Thorac. Cardiovasc. Surg. 113, 691-698 (1997).

63.   Sugarbaker,D.J., Herndon,J.E. & DeCamp,M.M., Jr. N2 status at resection predicts long-term outcome following induction therapy for Stage IIIA non-small cell lung cancer. Proc Am Soc Clin Oncol (1998).

64.   Venkatraman,E.S. & Begg,C.B. A distribution-free procedure for comparing receiver operating characteristic curves from a paired experiment. Biometrika 83, 835-848. 1996.
Ref Type: Journal (Full)

65.   Sy,J.P. & Taylor,J.M. Estimation in a Cox proportional hazards cure model. Biometrics 56, 227-236 (2000).

66.   Peng,Y. & Dear,K.B. A nonparametric mixture model for cure rate estimation. Biometrics 56, 237-243 (2000).

67.   Heagerty,P.J., Lumley,T. & Pepe,M.S. Time-dependent ROC curves for censored survival data and a diagnostic marker. Biometrics 56, 337-344 (2000).

68.   Davison A.C. & Hinkley,D.V. Bootstrap Methods and their Applications. Cambridge University Press, (1997).

69.   Gibson,U.E., Heid,C.A. & Williams,P.M. A novel method for real time quantitative RT-PCR. Genome Res. 6, 995-1001 (1996).

70.   Heid,C.A., Stevens,J., Livak,K.J. & Williams,P.M. Real time quantitative PCR. Genome Res 6, 986-94 (1996).

71.   PE Applied Biosystems user bulletin #2. Relative quantitation of gene expression.  1997.
Ref Type: Pamphlet

72.   Gold,P. & Freedman,S.O. Specific carcinoembryonic antigens of the human digestive system. J. Exp. Med. 122, 467-481 (1965).

73.   Beauchemin,N. et al. Redefined nomenclature for members of the carcinoembryonic antigen family.  Experimental Cell Research 252, 243-249 (1999).

74.   Savtchenko,E.S., Schiff,T.A., Jiang,C.K., Freedberg,I.M. & Blumenberg,M. Embryonic expression of the human 40-kD keratin: evidence from a processed pseudogene sequence. Am. J. Hum. Genet. 43, 630-637 (1988).

75.   Ruud,P., Fodstad,O. & Hovig,E. Identification of a novel cytokeratin 19 pseudogene that may interfere with reverse transcriptase-polymerase chain reaction assays used to detect micrometastatic tumor cells. Int. J. Cancer 80, 119-125 (1999).