• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of annsurgLink to Publisher's site
Ann Surg. Jun 2004; 239(6): 828–840.
PMCID: PMC1356291

Molecular Detection of Micrometastatic Breast Cancer in Histopathology-Negative Axillary Lymph Nodes Correlates With Traditional Predictors of Prognosis

An Interim Analysis of a Prospective Multi-Institutional Cohort Study

Abstract

Objective:

We sought to establish the clinical relevance of micrometastatic disease detected by reverse transcription polymerase chain reaction (RT-PCR) in axillary lymph nodes (ALN) of breast cancer patients.

Background:

The presence of ALN metastases remains one of the most valuable prognostic indicators in women with breast cancer. However, the clinical relevance of molecular detection of micrometastatic breast cancer in sentinel lymph nodes (SLN) and nonsentinel ALN has not been established.

Methods:

Four hundred eighty-nine patients with T1–T3 primary breast cancers were analyzed in a prospective, multi-institutional cohort study. ALN were analyzed by standard histopathology (H&E staining) and by multimarker, real-time RT-PCR analysis (mam, mamB, muc1, CEA, PSE, CK19, and PIP) designed to detect breast cancer micrometastases.

Results:

A positive marker signal was observed in 126 (87%) of 145 subjects with pathology-positive ALN, and in 112 (33%) of 344 subjects with pathology-negative ALN. In subjects with pathology-negative ALN, a positive marker signal was significantly associated with traditional indicators of prognosis, such as histologic grade (P = 0.0255) and St. Gallen risk category (P = 0.022). Mammaglobin was the most informative marker in the panel.

Conclusion:

This is the first report to show that overexpression of breast cancer–associated genes in breast cancer subjects with pathology-negative ALN correlates with traditional indicators of disease prognosis. These interim results provide strong evidence that molecular markers could serve as valid surrogates for the detection of occult micrometastases in ALN. Correlation of real-time RT-PCR analyses with disease-free survival in this patient cohort will help to define the clinical relevance of micrometastatic disease in this patient population.

The primary objective of cancer staging is to be able to classify patients by the extent of disease into groups with similar clinical outcomes and so facilitate patient management. In the setting of breast cancer, one of the most important prognostic indicators is the presence of axillary lymph node (ALN) metastases. Frequently, ALN disease status is the critical parameter for determining whether adjuvant systemic chemo or hormonal therapy is recommended.1–3 As a result, staging for newly diagnosed clinical stage I and II breast cancer patients has traditionally included an ipsilateral ALN dissection (ALND). Unfortunately, standard H&E histopathologic analysis of ALN has limitations. A number of studies have shown that performing additional tissue sections and/or immunohistochemical staining (IHC) of ALN increases metastases detection by up to 25%.4–6 Furthermore, these retrospective studies suggest that the prognosis for patients with occult disease is similar to patients with pathology-positive ALN.4,5,7 These findings imply that the development of more sensitive methods to detect micrometastatic disease in ALN could significantly improve breast cancer staging.

The recent identification of genes overexpressed in breast cancer combined with advances in molecular biology provide such an opportunity for improving breast cancer staging.8–16 We and others have shown that the reverse transcription polymerase chain reaction (RT-PCR) is capable of detecting metastatic disease in ALN of breast cancer patients,15,17 with a sensitivity of up to one cancer cell per 107 normal cells.18–20 Ironically, the exquisite sensitivity of RT-PCR has hindered its clinical application because the majority of potential markers have some baseline expression in normal tissues.21,22 Due to the fact that conventional RT-PCR techniques are at best semiquantitative, it has been difficult to differentiate between baseline gene expression in normal tissues and increased gene expression associated with breast cancer.8,21,23–29 As a result, some investigators consider PCR technology to be problematic for clinical application with false positive and/or clinically irrelevant results a concern.8,21,23,25–27,29–31

Real-time RT-PCR solves these limitations through the use of an online fluorescence detection system that precisely quantifies the amount of PCR product. We have previously shown that real-time RT-PCR can differentiate between baseline gene expression in normal tissues and cancer-associated gene overexpression.32,33 For example, CEA, CK19, and muc1 have detectable baseline expression in normal lymph nodes, but expression levels in ALN with metastatic breast cancer is 5-fold to 3500-fold higher.32 Our data indicate that a combination of multimarker analysis and quantitative real-time RT-PCR can be a precise and powerful tool for the detection of breast cancer ALN metastases. Furthermore, the genes mam, PIP, PDEF, CK19, CEA, muc1, and mamB have particular promise for breast cancer detection.15,32,33

Although these results suggest that molecular markers could serve as valid surrogates for metastatic and micrometastatic breast cancer, their clinical relevance is unproven. To address this, the Minimally Invasive Molecular Staging of Breast Cancer (MIMS) trial was initiated. This trial represents the first prospective cohort study in which a multimarker, real-time RT-PCR analysis was applied to the detection of breast cancer micrometastases in ALN. Sentinel and/or nonsentinel ALN from 489 breast cancer subjects with T1–T3 primary tumors were analyzed by standard histopathology and multimarker, real-time RT-PCR analysis. The study was designed with sufficient statistical power to correlate molecular analyses with clinical outcome at 5 years. Although the clinical outcome data are not yet available, we show in this interim report that real-time RT-PCR is able to sensitively detect metastatic breast cancer in ALN and that overexpression of breast cancer–associated genes in subjects with pathology-negative ALN is correlated with traditional indicators of poor prognosis.

MATERIALS AND METHODS

Study Design

A prospective cohort study design was adopted. Upon recruitment, eligible participants with stage I, IIa, or IIb breast cancer were requested to consent to tissue sampling from ALN, SLN, bone marrow, and whole blood. Tissue sampling was accomplished during the planned surgical procedure while the subject was under anesthesia. Real-time RT-PCR analyses were performed on all tissue specimens submitted to the Central Molecular Diagnostics Laboratory at the Medical University of South Carolina (MUSC). Staging was performed according to American Joint Committee on Cancer (AJCC) guidelines and included the results of routine histopathology of ALN. Patients received the current standard of care without their clinicians’ knowledge of, or reference to, the molecular analyses. The Clinical Innovation Group (TCIG, Charleston, SC) served as the coordinating center, and all study data were collected, processed, and analyzed at this central facility.

Data Management

Case report forms (CRFs) indexed by a unique subject number were used to record the clinical data for research purposes. The site investigators maintained the key that linked subject number with subject name to ensure confidentiality of the data at the coordinating center. Once CRFs were completed, they were transmitted via express mail to the coordinating center for processing. The CRFs were independently double-key entered into a Clintrial 4.2 database. The database was programmed with consistency checks to ensure the data entered were within a valid range with logical sequences and that all required items were completed. Written data clarification requests were sent to the sites to update the information if deficiencies in the data were identified. Quality control of the data was conducted in 2 phases. Trained monitors verified reported data against source documents (ie, monitored) and ensured all applicable regulatory documents were current. An additional quality control step was conducted by randomly sampling 20% of CRFs in the database for a CRF-to-database audit. Both processes include written documentation of discrepancies identified.

Study Subjects

Five hundred fifty subjects with pathologically confirmed invasive breast carcinoma were enrolled in the MIMS Trial. Inclusion criteria were as follows: age 18 years and older, tumor size category of T1–T3; ECOG (Eastern Cooperative Oncology Group) performance status of 0 (normal), 1 (with symptoms but ambulatory), or 2 (in bed < 50% of the time); recent bilateral mammogram with normal contralateral breast; and chest x-ray with no evidence of metastatic breast cancer. Exclusion criteria were as follows: inflammatory breast cancer or Paget disease; prior ductal or lobular carcinoma in situ; clinical evidence of supraclavicular, infraclavicular, or ALN involvement; or known metastatic disease. Subjects with a history of previous treatment of breast cancer were also excluded. Informed consent was obtained in accordance with each participating center's institutional review board regulations.

ALN Specimens From Breast Cancer Subjects

The surgeon and surgical pathologist at the individual clinical centers selected the ALN for the study. In general, if a sentinel lymph node (SLN) biopsy was performed, the SLN was submitted, although this was not required. If an ALN dissection was performed, at least 3 ALN were submitted. Approximately one half of the selected SLN and ALN were sent to surgical pathology at the clinical center for routine histopathology. The other half of the selected ALN were snap-frozen in liquid nitrogen and shipped to the Central Molecular Diagnostics Laboratory on dry ice for real-time RT-PCR analyses. ALN were evaluated by standard hematoxylin and eosin (H&E) histopathology at the participating center. The pathology status for lymph nodes was based on the H&E staining only. If the ALN was positive only by IHC, the subject was considered to have pathology-negative ALN. Real-time RT-PCR analyses were performed in a blinded manner.

Lymph Node Specimens From Control Subjects Without Evidence of Malignancy

To define baseline expression levels for the molecular markers used in this study, normal lymph nodes from patients without evidence of malignancy were procured. Informed consent was obtained from 51 patients undergoing elective carotid endarterectomy at MUSC. None of the patients had a history or clinical evidence of malignancy. At the time of the procedure, a single cervical lymph node was removed, snap-frozen in liquid nitrogen, and sent to the Central Molecular Diagnostics Laboratory for real-time RT-PCR analyses.

RNA Isolation and cDNA Synthesis

Total cellular RNA was isolated from control lymph nodes and lymph nodes from breast cancer subjects using a guanidinium thiocyanate-phenol-chloroform solution (RNA STAT-60; TEL-TEST, Friendswood, TX). Lymph node specimens were removed from −70°C storage and immediately weighed without allowing the tissue to thaw. SLN were processed separately from ALN. A maximum of 3 ALN were pooled together into 1 sample, while SLN were analyzed individually. Tissue (≤0.15 g) was then homogenized in 1 mL of RNA STAT-60 using a model 395, type-5 polytron (Dremel, Racine, WI). Total RNA was isolated per the manufacturer's instructions, with the exception that 1 μL of a 50 mg/ml solution of glycogen (Sigma, St. Louis, MO) was added to the aqueous phase before addition of isopropanol. The RNA pellet was dissolved in 50 μL of 1× RNA secure buffer (Ambion, Austin, TX). RNA was quantified by spectrophotometry at 260 nm. cDNA was made from 5 μg of total RNA using 200 U of M-MLV reverse transcription (Promega, Madison, WI) and 0.5 μg Oligo (dT)12–16 in a reaction volume of 20 μL (10 minutes at 70°C; 50 minutes at 42°C; 15 minutes at 70°C).

Real-Time RT-PCR

The real-time RT-PCR primers have been previously reported:32,33 mglo: F 5′- TGAGTGCTGTCTCCATGTTTGA, R 5′- TCTGCTCCCCACCTCTAAGTTG; PDEF: F 5′-AGTGCTCAAGGACATCGAGACG, R 5′-AGCCACTTCTGCACATTGCTG; mam: F 5′-CGGATGAAACTCTGAGCAATGT, R 5′-CTGCAGTTCTGTGAGCCAAAG; CK19: F 5′-CATGAAAGCTGCCTTGGAAGA, R 5′-TGATTCTGCCGCTCACTATCAG; muc1: F 5′-ACCATCCTATGAGCGAGTACCC, R 5′- GCCACCATTACCTGCAGAAAC; PIP: F 5′-GCCAACAAAGCTCAGGACAAC, R 5′-GCAGTGACTTCGTCATTTGGAC; mamB: F 5′-AGCAGTGTTTCCTCAACCAGTC, R 5′-TCTGAGCCAAACGCCTTG; CEA: F 5′-TAAGTGTTGACCACAGCGACCC, R 5′-GTTCCCATCAATCAGCCAAGAA. Real-time RT-PCR analyses were performed on a PE Biosystems Gene Amp 5700 Sequence Detection System (Foster City, CA). All reaction components were purchased from PE Biosystems. The standard reaction volume was 10 μL and contained 1× SYBR Green PCR Buffer; 3.5 mM MgCl2; 0.2 mM dATP; 0.2 mM dCTP; 0.2 mM dGTP; 0.4 mM dUTP; 0.25 U of AmpliTaq Gold; 0.1 U of AmpErase UNG enzyme; 0.7 μL of cDNA template; 0.25 μM forward primer; and 0.25 μM reverse primer. The initial step of PCR was 2 minutes at 50°C for AmpErase UNG activation, followed by a 10-minute hold at 95°C. Cycles (n = 40) consisted of a 15 seconds denaturation step at 95°C, followed by a 1 minute annealing/extension step at 60°C. The final step was a 60°C incubation for 1 minute. All reactions were performed in triplicate. The threshold for cycle of threshold (Ct) analysis was set at 0.5 relative fluorescence units.

Primary Data Analyses

Real-time RT-PCR data were quantified as Ct values that are inversely related to the amount of starting template; high Ct values correlate with low levels of gene expression, whereas low Ct values correlate with high levels of gene expression. Results were normalized to an internal control reference gene, β2-microglobin, by subtracting the mean Ct value of β2-microglobin from the mean Ct value of each respective gene (ΔCt value). Samples for which Ct values for β2-microglobin were equal to or higher than 22 were considered to contain inadequate RNA and were excluded from the analysis. To define baseline levels of gene expression and to define thresholds for marker positivity, 51 cervical lymph nodes from patients with no evidence of malignancy were analyzed. The size of this control group was based on 95% confidence interval data obtained from Geigy Scientific Tables.34 Threshold values for each individual marker were set at 3 standard deviations below the mean ΔCt value in the control group. A subject was considered to be positive for the molecular analysis if at least 1 marker in the panel was below the defined threshold. Data from real-time RT-PCR analyses were compiled in a Microsoft Access database and submitted to TCIG at MUSC for statistical analyses.

Statistical Analyses

Simple descriptive summary statistics (means ± standard deviations for continuous variables and proportions for categorical variables) were obtained to describe the demographic and clinical characteristics of the study sample. Degree of agreement between pathologic and molecular results was calculated as a Kappa statistic. Logistic regression analyses were performed using SAS version 8.0 software (SAS Institute Inc., Cary, NC) for the analysis of pathologic and molecular outcome adjusting for predefined baseline covariates that have been associated with pathologic outcome in prior studies. Predefined baseline covariates were tumor size, histologic grade, estrogen receptor status, progesterone receptor status, Her2neu status, and St. Gallen risk category (minimal/low risk: tumor size ≤ 1 cm, positive ER and/or PR status, grade I and age ≥ 35 years; intermediate risk: tumor size > 1 or 2 cm, positive ER and/or PR status, and grade I; high risk: tumor size > 2 cm, negative ER and/or PR status, grade II or III, or age < 35 years).1 Statistical significance was defined as P < 0.05.

RESULTS

Demographic and Clinical Characteristics

A total of 550 breast cancer patients were enrolled from 14 medical centers into the MIMS study. Of these, 489 subjects were analyzable. Sixty-one patients (11%) who initially consented were excluded from the final analysis due to 1 or more of the following reasons: subject withdrawal of consent (n = 6), subject ineligibility (n = 13), standard histopathology unavailable (n = 18), specimens lost or thawed during shipment (n = 17), or inability to obtain adequate mRNA from tissue specimens (n = 7). Reasons for subject ineligibility included presence of invasive cancer or lymphoma in the contralateral breast, previous diagnosis or treatment of cancer within the exclusionary timeframe, diagnosis of ALN metastases before subject registration, and final diagnosis of DCIS without invasive breast cancer. Demographic and clinical characteristics of the analyzable subjects are listed in Table 1.

Table thumbnail
TABLE 1. Patient Demographic and Clinicopathologic Characteristics

Precise Quantitation of Marker Gene Expression in Control Lymph Nodes

We have previously shown that the majority of known breast cancer–associated genes have some background expression in normal lymph nodes.32,33 For this study, we selected 7 breast cancer-associated genes (mam, mamB, PIP, CK19, muc1, PDEF, and CEA) known to be overexpressed in metastatic breast cancer compared with control lymph nodes.32,33 Baseline gene expression was precisely quantitated in 51 control lymph nodes by real-time RT-PCR. To obtain maximum specificity, a threshold value for marker positivity, ie, abnormal expression was set at 3 standard deviations beyond the mean ΔCt value for each gene (Fig. 1). At the defined threshold values, positivity was not observed for any marker in the control samples.

figure 10FF1
FIGURE 1. Expression of cancer-associated genes in normal control lymph nodes. Real-time RT-PCR analysis of lymph nodes from 51 negative control patients [32 males (+), 19 females (O)] was performed as described in Materials and ...

Routine Histopathology and Real Time Multimarker RT-PCR Analysis of the MIMS Study Cohort

Of the 489 analyzable patients enrolled on study, 145 (30%) had one or more standard histopathology pathology-positive ALN, and 344 (70%) had pathology-negative ALN. Using the above-defined threshold values, a multimarker RT-PCR analysis was then performed for the 489 subjects. Of this cohort, real-time RT-PCR analyses revealed a positive marker signal in 126 (87%) of 145 of subjects with pathology-positive ALN, and in 112 (33%) of 344 subjects with pathology-negative ALN. The frequency of overexpression for each marker is shown in Figure 2.

figure 10FF2
FIGURE 2. Frequency of gene expression in breast cancer subjects with pathology-positive [Path (+)] and pathology-negative [Path (-)] ALN. Real-time RT-PCR analysis of ALN from 145 subjects with pathology-positive ...

Multimarker Molecular Analysis Improves the Sensitivity of Metastatic Breast Cancer Detection

To assess the value of multimarker analysis for the detection of metastatic and micrometastatic breast cancer, we determined the frequency of gene overexpression for various marker combinations (Fig. 3). For patients with pathology-positive ALN (n = 145), mam was the most sensitive marker and was overexpressed in 114 (or 90.5%) of the 126 patients who were positive by molecular analyses. This result is consistent with previous studies demonstrating that mam has the highest diagnostic accuracy for the detection of metastatic breast cancer.32,35 Of the patients in the pathology and molecular-positive ALN subgroup who did not overexpress mam, CEA was the next most valuable marker. The combination of mam and/or CEA (ie, a 2-gene marker panel) was positive in 123 (97.6%) of 126 RT-PCR positive patients (Fig. 3). Of the patients in this subgroup who did not express mam or CEA, PIP was the next most valuable marker; the combination of mam and/or CEA and/or PIP (ie, a 3-gene marker panel) detected gene overexpression in 125 (99.2%) of 126 subjects.

figure 10FF3
FIGURE 3. Sensitivity of detection of metastatic disease by multimarker RT-PCR. This analysis includes lymph nodes from 126 patients with pathology-positive/marker-positive ALN (filled triangles) and 112 pathology-negative/marker-positive ALN (filled ...

To determine the value of multimarker analysis for the detection of occult metastatic disease, we evaluated the frequency of gene overexpression in pathology-negative patients (n = 344) (Fig. 3). Of this cohort, 112 patients had at least 1 molecular marker positive (Fig. 2). Similar to the pathology-positive ALN results, we observed that the mam, CEA, and PIP marker combination yielded the highest apparent sensitivity (Fig. 3). However, in contrast to the 99.2% detection rate observed in pathology-positive patients, this marker combination only accounted for 105 (93%) of the 112 RT-PCR–positive/pathology-negative patients. These data suggest that multimarker analysis may be most important in the setting of minimal disease.

Real-Time RT-PCR Multimarker Positivity Significantly Correlates With Clinical Parameters That Predict a Poor Prognosis

To determine whether the results of the molecular analyses were correlated with traditional prognostic indicators, we performed 3 specific subgroup analyses: (1) the distribution of positive standard histopathology in ALN of all subjects (n = 489); (2) the distribution of positive molecular analyses in ALN of all subjects (n = 489); and (3) the distribution of positive molecular analyses in ALN of pathology-negative patients (n = 344) (Table 2). In each group, the distribution of positive histopathology or real-time RT-PCR increased with tumor size, clinical stage, histologic grade (grade I versus grades II & III), Her2neu status, and St. Gallen risk category (Table 2). These data strongly suggest that the distribution of real-time RT-PCR results is not random, but rather is associated with traditional prognostic indicators.

Table thumbnail
TABLE 2. Prognostic Characteristics, Histopathology, and Molecular Analysis of ALN

To determine the statistical significance of the apparent association between molecular analyses and traditional risk factors, individual logistic regression analyses were performed using standard histopathology or molecular status of ALN (positive versus negative) as the dependent variable and the following clinical parameters as independent variables: tumor size, histologic grade, ER status, PR status, Her2neu status, and St. Gallen risk category (Tables 3–5). For the distribution of positive standard histopathology in ALN of all patients (n = 489), standard histopathology was significantly associated with tumor size (>1 cm versus ≤1 cm; P < 0.0001, OR = 5.937, 95% CI [3.159 to 11.159]), histologic grade (II-III versus I: P = 0.0018, OR = 2.162, 95% CI [1.331 to 3.511]), and St. Gallen risk category (high versus low/intermediate: P = 0.0028, OR=2.316, 95% CI [1.335 to 4.018]) (Table 3). For the distribution of positive molecular analyses in ALN of all patients (n=489), significant associations were observed with tumor size (P < 0.0001, OR=2.608, 95% CI [1.711, 3.975]), histologic grade (P = 0.0003, OR=2.163, 95% CI [1.428, 3.275]), and St. Gallen risk category (P = 0.0009, OR=2.155, 95% CI [1.369, 3.393]) (Table 4). For the distribution of positive molecular analyses in ALN of subjects with pathology-negative ALN (n = 344), positive molecular analyses clearly increased with tumor size, although this did not reach statistical significance (P = 0.0889). However, molecular analyses were significantly associated with histologic grade (P < 0.0255, OR = 1.799, 95% CI [1.075 to 3.012]) and St. Gallen risk category (P < 0.0220, OR=1.946, 95% CI [1.101 to 3.441]) (Table 5). The collective findings from these logistic regression analyses confirm that the real-time RT-PCR analyses are significantly correlated with clinical parameters that predict a poor prognosis in the entire subject cohort (n = 489) and in the subset of subjects patients with pathology-negative ALN (n = 344). This rigorous statistical analysis suggests that the molecular markers described in this study might serve as surrogates for the presence of small amounts of cancer-like cells.

Table thumbnail
TABLE 3. Probability of Positive Pathology Relative to Clinical Variables
Table thumbnail
TABLE 4. Probability of Positive Molecular Analysis Relative to Clinical Variables
Table thumbnail
TABLE 5. Probability of Positive Molecular Analysis with Pathology-Negative ALN-Relative to Clinical Variables

DISCUSSION

To determine the clinical relevance of molecular detection of metastatic and micrometastatic breast cancer, we initiated a multi-institutional prospective cohort study (the MIMS Trial) using a panel of molecular markers designed to have a moderate to high diagnostic accuracy for the detection of breast cancer metastases.32,33 The MIMS trial as designed has the potential to provide a significant contribution to the field of breast cancer molecular diagnostics for several reasons. The analyses have been performed using for the first time a quantitative real-time RT-PCR and a multimarker gene panel with threshold levels for marker positivity based on quantitative analyses of a significant number of normal lymph nodes (n = 51). To provide validity, the study was prospectively designed, all real-time RT-PCR assays were performed in a blinded manner, and the results were analyzed by an independent statistical group. Finally, the number of subjects analyzed in the study (n = 489) should be large enough to provide the statistical power required to help define the clinical relevance of molecular detection of micrometastases.

The analyses of ALN from 489 breast cancer subjects by multimarker real-time RT-PCR has been completed. The results indicate that 33% of subjects with pathology-negative ALN have a positive marker signal in their lymph nodes by real-time RT-PCR. The primary hypothesis of the MIMS Trial is that in this subgroup of patients with pathology-negative ALN, the presence of a positive marker signal will be associated with a worsened clinical outcome (defined as the proportion of subjects having disease relapse within 5 years of diagnosis). In preliminary support of this hypothesis, Table 2 demonstrates that the rate of marker positivity increased with tumor size, histologic grade (grade I versus grades II & III), Her2neu expression, and St. Gallen risk category (Table 2). Even though the associations among marker positivity, tumor size, and Her2neu status did not reach statistical significance, the results of logistic regression analyses confirm that the associations among marker positivity, histologic grade, and St. Gallen risk category are statistically significant (Tables 3–5). These data provide evidence that marker positivity is associated with clinical parameters that predict poor prognosis, a finding that supports the concept of applying a molecular approach for the detection of micrometastatic breast cancer. Validation of this approach will require the correlation with the clinical outcomes at 5 years.

A major limitation of conventional RT-PCR is the inability to precisely quantify the amount of mRNA template, and we believe that this may explain the variable results that have been reported in the literature. Specifically, in the field of breast cancer molecular diagnostics, Noguchi et al18 reported no expression of muc1 and CK19 in control lymph nodes, while studies by Min et al21, Bostick et al,36 and Marchetti et al37 demonstrate that these 2 markers are expressed in normal lymph nodes and are therefore not suitable for molecular detection of micrometastatic breast cancer. Further, whereas the study by Bostick et al36 showed expression of CEA in normal lymph nodes, those investigations of Min et al21, Marchetti et al,37 and Ooka et al38 failed to detect expression of CEA in normal lymph nodes.

To date, only 1 prospective study designed to assess the value of molecular detection of micrometastatic breast cancer has been reported.39 In this study, Sakaguchi et al39 reported that 11% of patients with pathology-negative ALN had evidence of CK19 overexpression by conventional RT-PCR. However, the authors concluded that these results were not clinically relevant because outcomes for patients who overexpressed CK19 were not significantly different than the outcomes for patients who did not overexpress this marker. In the present study, overexpression of CK19 was detected in only 3.8% of the subjects with pathology-negative ALN, a value that is significantly lower than that of Sakaguchi et al.39 These results suggest that Sakaguchi et al39 may not have been able to differentiate between baseline CK19 expression and CK19 overexpression using conventional RT-PCR, leading to a significant number of false positive results. This potential limitation of conventional PCR is addressed directly in the MIMS trial by the use of quantitative real-time RT-PCR.

In terms of the individual molecular markers, mammaglobin clearly appears to be the most valuable molecular marker for the detection of metastatic breast cancer in ALN, and is overexpressed in 79% of subjects with pathology-positive ALN (or 90% of subjects with pathology-positive ALN who were marker-positive). This result is concordant with our previous work in which mammaglobin was overexpressed in 94% of pathology-positive lymph nodes.32 It is also concordant with the study by Zehentner et al, in which mammaglobin was detected by real-time RT-PCR in 80% of pathology-positive ALN.35 Studies using conventional RT-PCR have also validated the high diagnostic accuracy of mammaglobin, with sensitivities ranging from 78% to 100% and specificities ranging from 86% to 100%.31,36,37,40,41

The use of multimarker gene panels for the detection of metastatic breast cancer has been proposed by many investigators.17,21,31,32,35,36 The results of this study confirm that a combination of markers significantly improves the potential value of the molecular assay as the sensitivity of the assay is significantly improved using the multimarker panel. However, the present data also suggest that a multimarker gene panel is most important for the detection of micrometastatic disease (low tumor burden). Our data indicate that to obtain a sensitivity of 97% in subjects with marker-positive/pathology-positive ALN, a 2-marker gene panel is sufficient (mam, CEA). However, to obtain a similar level of sensitivity in subjects with marker-positive/pathology-negative ALN a 4-marker gene panel is required (mam, CEA, PIP, CK19) (Fig. 3). These results suggest that in pathology-negative ALN, where the disease burden is presumed to be low, the diagnostic accuracy of the individual molecular markers is decreased, and an extended maker panel is required to achieve maximum disease detection.

Overall, the sensitivity of multimarker real-time RT-PCR for the detection of metastatic breast cancer was 87%. Evaluation of the 19 subjects with pathology-positive ALN who were marker-negative (false negatives) revealed a significant potential for sampling error. In fact, 15 of the 19 patients in this group had only 1 pathology-positive ALN, and the median number of ALN analyzed by pathology was 12 per subject. We suspect that, in these 19 false-negative cases, the ALN containing metastatic breast cancer may not have been available for RT-PCR analysis. Another source of sampling error is the fact that the ALN specimens were divided at the time of collection, with half of the specimen sent for routine pathology and half for real-time RT-PCR. Thus, there is the potential that the portion of the ALN with metastatic breast cancer was sent for routine pathology but not real-time RT-PCR.42

In summary, the interim analysis from this prospective clinical trial provides the first report of a statistically significant association between molecular detection of breast cancer micrometastases and traditional predictors of poor prognosis in subjects with pathology-negative ALN. If the long-term results of the MIMS trial support the hypothesis that molecular detection of breast cancer micrometastases is clinically relevant, the staging of breast cancer could be approached with increased sensitivity and accuracy. This may be particularly relevant for the analysis of SLN, as the overall accuracy of SLN biopsy appears to be predicated on the sensitive pathologic evaluation of the SLN. Ultimately the development of a sensitive molecular diagnostic assay for the detection of breast cancer micrometastases is likely to translate into an improved ability to tailor adjuvant therapies by identifying high-risk patients who would most likely benefit from aggressive systemic therapy.

ACKNOWLEDGMENTS

We thank Loretta Hoover, Yuehua Zhou, Megan Baker, Lisa Sooy, and James Raman for help in processing lymph node specimens. We acknowledge the HCC tissue procurement/tumor bank and specifically thank Dr. Debra Hazen-Martin and Margaret Romano. We are grateful to Thomas Smith at the MUSC Writing Center for providing constructive counsel.

Discussions

Dr. Kirby I. Bland (Birmingham, Alabama): I wish to thank the authors for providing me this manuscript well in advance of the meeting. Dr. Gillanders and Dr. Cole of the Medical University of South Carolina are to be commended for organizing and completing a well-organized, meritorious study that was conducted prospectively with these 13 other institutions that are listed.

Obviously we at UAB have great interest in the outcomes. We contributed 112 patients to this study via Dr. Urist, the PI from our institution. As you have heard today, 489 well-analyzed subjects were observed to have a positive marker, signaling 87% of those with pathologically positive ALNs and a third of those with pathologically negative nodes. Of the RT-PCR molecules that were evaluated, mammoglobin was the most important predictor in this group. I think that was somewhat a surprise to us.

The authors are to be congratulated, though, as this is the first study to our knowledge that has actually confirmed overexpression of these specific related cancer-associated genes. I have 3 questions for you, Dr. Cole.

Could the authors explain further to the Fellows and guests of this Association how they selected the particular mutated proteins that they examined? As you know, many of us have had grave doubts, especially for those such as CEA and CK19 as a discriminator to correlate with prognosis and clinical outcomes in breast carcinoma. Did the authors consider others such as p53, ras, fos, myc, or other mutated proteins in the primary tumor and/or the ALNs? Again, what is really the rationale for your methodology relative to this DOD study and its funding to evaluate these particular molecules?

Secondly, it is well known that the pathological, the biochemical, as well as the molecular variables that have been most important in detecting prognosis in breast cancer include the size, the grade, the ER and PR status of the tumor, as well as Her2neu expression. However, in your analysis, if my interpretation of the manuscript is correct, the ER-negative and PR-negative as well as the Her2neu-positive status (which are the ultimate discriminators of a bad outcome in these cases) did not correlate statistically with your predicted clinical outcomes. In addition, positive molecular analysis relative to these clinical variables also did not validate these traditional markers to be important in the pathologically positive and negative outcomes. I think it is essential that this be explained, if I have interpreted your data correctly.

The third question I would have is that the rate of marker positivity was correlated to increase with the tumor size, histologic grade, Her2neu expression, and, finally, the St. Gallen risk category. However, the associations here, as you said earlier, did not reach statistical significance, except in logistic regression analyses. While this is a leap of faith, how can the authors then conclude that these data provide evidence that marker positivity is associated with hopefully those clinical parameters that we can predict a poor prognosis? Therein, you and I will know who not to treat if they have a terrible phenotype expression, and those are the ones we really want to isolate.

In summary, I think this is a very important Stark II as one of the first important prospective studies to look at specific molecular staging of traditional pathological markers. And I hope, Dr. Cole, you and your colleagues will bring this back to this Association in 5 years and give us the clinical outcomes of these data. It is a very important paper. I greatly enjoyed it.

Dr. R. Daniel Beauchamp (Nashville, Tennessee): I too congratulate Dr. Cole and his colleagues on an innovative study. He has been doggedly pursuing this area for several years now, and this is starting to pay off.

I think that this project has the potential to really revamp the way that we stage breast cancer. I don't think we know that potential yet from this report. I think the aspect of the study that has the most promise is the analysis of bone marrow and the analysis in the peripheral blood samples that were not presented, because they are not yet available, as part of the interim analysis in the study.

One of the disturbing aspects of the data—and I think you explained why this happened—but one of the disturbing aspects was the 13% false negative rate. And that is combined with the 33% rate of PCR-positive but pathologically negative nodes. So we really don't know what the false positive rate is. We know at least that there is a 13% false negative rate. And that is particularly disturbing.

That may be explained in part by the selection of the lymph nodes to be analyzed. And I wonder if you have gone back to actually analyze only the sentinel nodes comparing PCR and histology in only the sentinel nodes, excluding those patients who had standard lymph node dissection up front? In this regard, the sentinel nodes could be focused on, and you would have more precise pathological staging. In that case, what was the rate of positivity and what was the false negative rate?

Have you gone back and done serial sections in immunohistochemistry on the ALNs where the PCR was positive but the pathology was negative? In other words, go back and do a serial section of those nodes to determine whether you actually missed micrometastases? And what was the rate in that circumstance?

Finally, have you yet established a correlation between pathologic positive lymph nodes and either blood or bone marrow PCR results? We anxiously await your results and the maturation of this study.

Dr. Henry M. Kuerer (Houston, Texas): I enjoyed the presentation. A great number of sites participated in the study. I have 2 questions. It appears that 33% of patients had a pathology-negative but PCR-positive result. Have you performed serial sections and/or immunohistochemistry on these negative nodes? Concerning the potential false positive PCR rate for your study, I know it is difficult to determine this for your study, but based on previous studies with these markers can you give us an idea of what you would expect?

Dr. William C. Wood (Atlanta, Georgia): Dr. Cole, this is an exciting study. Have you stratified the node-negative, molecularly positive patients for core or FNA biopsy versus open? We know that we seed little crumbs of tumor cells into these nodes when we do an invasive biopsy. Is that what we are seeing? Or are these truly micromets?

Dr. Michael J. Edwards (Little Rock, Arkansas): I enjoyed the paper. I would remind the authors that we also need to consider the impact of the stage shift adjustment in the lymph node biopsy. And we need stress to the readers of Annals of Sugery, as this discussion comes out that we don't even know the biology of the SLN-negative patient. It may be that none of those patients actually benefit from adjuvant chemotherapy. We simply have not followed a group of SLN-staged patients who were node-negative long enough to know what the natural history of that group is; now we throw on top of that the PCR staging, which makes us look through a glass even less clearly. I think this is very important. These kinds of studies need to be done. The authors need to be commended.

I think the important group of these patients is the 232 patients who were not only node-negative and sentinel node–negative but also PCR negative. Bill Wood designed the NAFE trial, which I implemented. Based on the predictors and historical standards of axillary dissection, Dr. Wood predicted a certain number of events that we would see with 36 months of follow-up. And Dr. Wood, that number of events, as I remember, was on the order of 60 or 70 events of either recurrence or death. In fact, with that median follow-up, we only saw 21. What that means is with the SLN group who were staged and node-negative were enrolled in that trial, we just didn't see people recur the way we thought they would. I wonder how much of this is due to PCR and how much of it is just actually sentinel node.

Dr. Roger R. Perry (Norfolk, Virginia): I also enjoyed this paper. To me, the important use of molecular markers would be to identify one that could serve as a surrogate for the patient's nodal status. So my question is, have you looked at these markers in the primary tumor? If so, what is the correlation with your findings in the lymph nodes? And can you use these markers as a surrogate for lymph node status?

Dr. David J. Cole (Charleston, South Carolina): I would like to thank the discussants for their insightful comments and questions.

With respect to Dr. Bland's question in terms of marker selection, we actually went back when we initiated the real-time PCR analysis and looked at the literature to identify known markers that might have significant overexpression in the breast cancer setting. One real problem is that there is no uniquely expressed breast cancer marker, and so there will be normal expression. We actually tested a 15-marker set, and we looked at the relative expression of these markers in a normal lymph node cohort. This was then compared with the level of marker overexpression in a known metastatic lymph node cohort that we have available in the lab.

We found that there are actually some markers that were useless. For example, 1 marker in our pilot study, which I didn't highlight, C-myc was totally useless because there was no statistical difference between normal and cancer expression.

The topic of molecular markers has been a moving target. This study was initiated 4 years ago, and every year there are new markers and genes identified. We have tried to be current. So if a new marker comes up, the question that we always ask is: Does it add anything, or is this yet another marker that is similar to the rest?

So, to answer the question, with our initial studies we looked at relative overexpression levels. The markers that had the clearest delineation between normal and cancer in the lymph node setting were chosen. There were 6 markers that met this criterion.

Admittedly, this is a work in progress. There are going to be more markers. We have evaluated Fos. We have not evaluated Ras. We recently described some markers such as KS14, which have some promise. I am hoping that we can establish some basic principles. Probably the details are going to be markedly different when this all comes out.

The question concerning variables with respect to clinical predictors and prognosis, and specifically Her2 and ER and PR status not being statistically related to node status. I have to say I was surprised by that outcome in this study. I think 1 problem that we are faced with is that Her2-positive or -negative status is institutionally derived. Therefore, the tests that are used and how the primary tumor is evaluated are variable. We have not been able to yet evaluate the primary pathology, which needs to occur, because I suspect that at least some of that would change if we had a routine and standardized process. Although we screen the pathologists and ask that they adhere to fairly uniform standards, I think there is variability with these specific tests. This may contribute to the less than statistically significant Her2 positivity and ER/PR positivity outcome. I do not have any further insight beyond this for that question.

To answer Dr. Beauchamp's question about the bone marrow and peripheral blood data, we are working on a negative control bone marrow cohort currently. With respect to the 13% false negative rate, I certainly hoped that we would do no worse than a 5% rate for false negatives. I have to think that at least a significant contributing factor is as I described: sampling error in the non-SLN cohort. For example, a significant number of the nodes that were differentially positive, path positive-marker negative, had a single node positive by pathology. We are averaging 4 of 12 nodes available for the molecular analysis. So it is not inconceivable that we are just not getting the tissue for analysis. I can tell you that we have done very careful and discreet studies using SLNs where one half went for histologic staining and the other half for PCR analysis. In this setting without simpling error, we have a 100% correlation with pathology positivity.

I am confident that molecular diagnostics have the requisite capability. I am also certain that sampling error in our cohort is going to be a compounding factor in terms of interpretation of the data. I am not sure in retrospect how it could have been designed differently, given the changing SLN paradigm.

I agree with Dr. Edwards’ comment about the significance of the PCR negative-pathology negative node cohort. I think that there are 2 sides to this story. On one hand you would hope that PCR positivity means something. However, I am fairly certain that a negative PCR outcome is going to give you a clean slate, ie, by the most sensitive test you cannot determine evidence of cancer, and so this means a better prognosis. I will be quite excited to see what the outcomes are for this group.

Again, I would like to thank the discussants and apologize if I missed any individual questions.

Footnotes

Source of Support: Department of Defense N00014-99-1-0784.

Drs. Gillanders and Mikhitarian contributed equally to this manuscript.

Reprints: David J. Cole, MD, MUSC Department of Surgery, 171 Ashley Ave, Room 420Q CSB, Charleston, SC 29425. E-mail: ude.csum@jdeloc.

REFERENCES

1. Goldhirsch A, Glick JH, Gelber RD, et al. Meeting highlights: International Consensus Panel on the Treatment of Primary Breast Cancer. Seventh International Conference on Adjuvant Therapy of Primary Breast Cancer. J Clin Oncol. 2001;19:3817–3827. [PubMed]
2. Woo CS, Silberman H, Nakamura SK, et al. Lymph node status combined with lymphovascular invasion creates a more powerful tool for predicting outcome in patients with invasive breast cancer. Am J Surg. 2002;184:337–340. [PubMed]
3. Cummings MC, Walsh MD, Hohn BG, et al. Occult axillary lymph node metastases in breast cancer do matter: results of 10-year survival analysis. Am J Surg Pathol. 2002;26:1286–1295. [PubMed]
4. de Mascarel I, Bonichon F, Coindre JM, et al. Prognostic significance of breast cancer axillary lymph node micrometastases assessed by two special techniques: reevaluation with longer follow-up. Br J Cancer. 1992;66:523–527. [PMC free article] [PubMed]
5. McGuckin MA, Cummings MC, Walsh MD, et al. Occult axillary node metastases in breast cancer: their detection and prognostic significance. Br J Cancer. 1996;73:88–95. [PMC free article] [PubMed]
6. Sedmak DD, Meineke TA, Knechtges DS. Detection of metastatic breast carcinoma with monoclonal antibodies to cytokeratins. Arch Pathol Lab Med. 1989;113:786–789. [PubMed]
7. Gardner B, Feldman J. Are positive axillary nodes in breast cancer markers for incurable disease? Ann Surg. 1993;218:270–277. [PMC free article] [PubMed]
8. Datta YH, Adams PT, Drobyski WR, et al. Sensitive detection of occult breast cancer by the reverse-transcriptase polymerase chain reaction. J Clin Oncol. 1994;12:475–482. [PubMed]
9. Gasparini G, Toi M, Gion M, et al. Prognostic significance of vascular endothelial growth factor protein in node-negative breast carcinoma. J Natl Cancer Inst. 1997;89:139–147. [PubMed]
10. Ioachim E, Kamina S, Athanassiadou S, et al. The prognostic significance of EGFR, C-erbB-2, Ki-67 and PCNA expression in breast cancer. Anticancer Res. 1996;16:3141–3147. [PubMed]
11. Anan K, Morisaki T, Katano M, et al. Assessment of c-erbB2 and vascular endothelial growth factor mRNA expression in fine-needle aspirates from early breast carcinomas: pre-operative determination of malignant potential. Eur J Surg Oncol. 1998;24:28–33. [PubMed]
12. Bieche I, Lazar V, Nogues C, et al. Prognostic value of chorionic gonadotropin beta gene transcripts in human breast carcinoma. Clin Cancer Res. 1998;4:671–676. [PubMed]
13. Harbeck N, Dettmar P, Thomssen C, et al. Prognostic impact of tumor biological factors on survival in node-negative breast cancer. Anticancer Res. 1998;18:2187–2197. [PubMed]
14. Linderholm B, Tavelin B, Grankvist K, et al. Vascular endothelial growth factor is of high prognostic value in node-negative breast carcinoma. J Clin Oncol. 1998;16:3121–3128. [PubMed]
15. Lockett MA, Baron PL, O'Brien PH, et al. Detection of occult breast cancer micrometastases in axillary lymph nodes using a multi-marker reverse transcriptase-polymerase chain reaction panel. J Am Coll Surg. 1998;187:9–16. [PubMed]
16. Molina R, Jo J, Filella X, et al. c-erbB-2 oncoprotein, CEA, and CA 15.3 in patients with breast cancer: prognostic value. Breast Cancer Res Treat. 1998;51:109–119. [PubMed]
17. Lockett MA, Metcalf JS, Baron PL, et al. Efficacy of reverse transcriptase-polymerase chain reaction screening for micrometastatic disease in axillary lymph nodes of breast cancer patients. Am Surg. 1998;64:539–543; discussion 543–544. [PubMed]
18. Noguchi S, Aihara T, Motomura K, et al. Detection of breast cancer micrometastases in axillary lymph nodes by means of RT-PCR. Comparison between MUC1 mRNA and keratin 19 mRNA amplification. Am J Pathol. 1996;148:649–656. [PMC free article] [PubMed]
19. Mori M, Mimori K, Inoue H, et al. Detection of cancer micrometastases in lymph nodes by reverse transcriptase-polymerase chain reaction. Cancer Res. 1995;55:3417–3420. [PubMed]
20. Schoenfeld A, Luqmani Y, Smith D, et al. Detection of breast cancer micrometastases in axillary lymph nodes by using polymerase chain reaction. Cancer Res. 1994;54:2986–2990. [PubMed]
21. Bostick PJ, Chatterjee S, Chi DD, 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. 1998;16:2632–2640. [PubMed]
22. Zippelius A, Kufer P, Honold G, et al. Limitations of reverse-transcriptase polymerase chain reaction analyses for detection of micrometastatic epithelial cancer cells in bone marrow [see comments]. J Clin Oncol. 1997;15:2701–2708. [PubMed]
23. Eltahir EM, Mallinson DS, Birnie GD, et al. Putative markers for the detection of breast carcinoma cells in blood. Br J Cancer. 1998;77:1203–1207. [PMC free article] [PubMed]
24. Giesing M, Austrup F, Bockmann B, et al. Independent prognostication and therapy monitoring of breast cancer patients by DNA/RNA typing of minimal residual cancer cells. Int J Biol Markers. 2000;15:94–99. [PubMed]
25. Goeminne JC, Guillaume T, Salmon M, et al. Unreliability of carcinoembryonic antigen (CEA) reverse transcriptase-polymerase chain reaction (RT-PCR) in detecting contaminating breast cancer cells in peripheral blood stem cells due to induction of CEA by growth factors. Bone Marrow Transplant. 1999;24:769–775. [PubMed]
26. Lopez-Guerrero JA, Gilabert PB, Gonzalez EB, et al. Use of reverse-transcriptase polymerase chain reaction (RT-PCR) for carcinoembryonic antigen, cytokeratin 19, and maspin in the detection of tumor cells in leukapheresis products from patients with breast cancer: comparison with immunocytochemistry. J Hematother. 1999;8:53–61. [PubMed]
27. Merrie AE, Yun K, Gunn J, et al. Analysis of potential markers for detection of submicroscopic lymph node metastases in breast cancer. Br J Cancer. 1999;80:2019–2024. [PMC free article] [PubMed]
28. Zhong XY, Kaul S, Thompson J, et al. Evaluation of the reverse transcriptase/polymerase chain reaction for CEA for the detection of breast cancer dissemination in bone marrow and peripheral blood. J Cancer Res Clin Oncol. 1999;125:669–674. [PubMed]
29. Zhong XY, Kaul S, Diel I, et al. Analysis of sensitivity and specificity of cytokeratin 19 reverse transcriptase/polymerase chain reaction for detection of occult breast cancer in bone marrow and leukapheresis products. J Cancer Res Clin Oncol. 1999;125:286–291. [PubMed]
30. Zach O, Kasparu H, Krieger O, et al. Detection of circulating mammary carcinoma cells in the peripheral blood of breast cancer patients via a nested RT-PCR assay for mammaglobin mRNA. J Clin Oncol. 1999;17:2015–2019. [PubMed]
31. Manzotti M, Dell'Orto P, Maisonneuve P, et al. Reverse transcription-polymerase chain reaction assay for multiple mRNA markers in the detection of breast cancer metastases in sentinel lymph nodes. Int J Cancer. 2001;95:307–312. [PubMed]
32. Mitas M, Mikhitarian K, Walters C, et al. Quantitative Real-time RT-PCR detection of breast cancer micrometastasis using a multigene marker panel. Int J Cancer. 2001;93:162–171. [PubMed]
33. Mitas M, Mikhitarian K, Hoover L, et al. Prostate-specific ets (PSE) factor: a novel marker for detection of metastatic breast cancer in axillary lymph nodes. Br J Cancer. 2002;86:899–904. [PMC free article] [PubMed]
34. Henderson AR. Assessing test accuracy and its clinical consequences: a primer for receiver operating characteristic curve analysis. Ann Clin Biochem. 1993;30:521–539. [PubMed]
35. Zehentner BK, Dillon DC, Jiang Y, et al. Application of a multigene reverse transcription-PCR assay for detection of mammaglobin and complementary transcribed genes in breast cancer lymph nodes. Clin Chem. 2002;48:1225–1231. [PMC free article] [PubMed]
36. Min CJ, Tafra L, Verbanac KM. Identification of superior markers for PCR detection of breast cancer metastases in sentinel lymph nodes. Cancer Res. 1998;58:4581–4584. [PubMed]
37. Marchetti A, Buttitta F, Bertacca G, et al. mRNA markers of breast cancer nodal metastases: comparison between mammaglobin and carcinoembryonic antigen in 248 patients. J Pathol. 2001;195:186–190. [PubMed]
38. Ooka M, Sakita I, Fujiwara Y, et al. Selection of mRNA markers for detection of lymph node micrometastases in breast cancer patients. Oncol Rep. 2000;7:561–566. [PubMed]
39. Sakaguchi M, Virmani A, Dudak MW, et al. Clinical relevance of reverse transcriptase-polymerase chain reaction for the detection of axillary lymph node metastases in breast cancer. Ann Surg Oncol. 2003;10:117–125. [PubMed]
40. Watson MA, Dintzis S, Darrow CM, et al. Mammaglobin expression in primary, metastatic, and occult breast cancer. Cancer Res. 1999;59:3028–3031. [PubMed]
41. Leygue E, Snell L, Dotzlaw H, et al. Mammaglobin, a potential marker of breast cancer nodal metastasis. J Pathol. 1999;189:28–33. [PubMed]
42. Slade MJ, Smith BM, Sinnett HD, et al. Quantitative polymerase chain reaction for the detection of micrometastases in patients with breast cancer [published erratum appears in J Clin Oncol 1999 Apr;17:1330]. J Clin Oncol. 1999;17:870–879. [PubMed]

Articles from Annals of Surgery are provided here courtesy of Lippincott, Williams, and Wilkins
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...