May 2008
Feature Story

William Check, PhD

Evaluating a new molecular assay, like anything else, depends on how you look at it. In the simplest view, we can look at sensitivity and specificity compared with the existing gold standard. In the case of GeneSearch BLN (Veridex LLC), the recently approved intraoperative reverse transcriptase-PCR assay for detecting metastases in axillary sentinel lymph nodes of patients with breast cancer, the reference standard is H&E-stained permanent section (plus immunostaining for cytokeratin when indicated). In the validation study, the GeneSearch Breast Lymph Node Assay had a sensitivity and specificity of 87.6 percent and 94.2 percent, respectively, compared with permanent section (Blumencranz P, et al. Am J Surg. 2007; 194: 426– 432). Another perspective would be to compare GeneSearch to the most widely used intraoperative procedure for evaluating sentinel lymph nodes, frozen section. In the same study, sensitivity was higher for GeneSearch than for frozen section, 95.6 percent versus 85.6 percent, even at the macromet level, but specificity was lower, 94.3 percent versus 97.8 percent.

However, just as GeneSearch seeks to replace a binocular technique with a more searching inspection of sentinel lymph node samples, we can replace the typical two-question analysis with a more extensive interrogation. If the American poet Wallace Stevens could imagine 13 ways of looking at a blackbird, we can surely devise 13 ways of looking at a molecular assay:

Some of these questions are specific to GeneSearch; others are endemic to breast cancer pathology but have been raised in a more urgent manner by the introduction of the molecular assay. Several are being addressed in postmarketing studies.

Detecting cancer intraoperatively in a sentinel lymph node of a patient with breast cancer is important because it makes it possible for the surgeon to do an axillary lymph node dissection immediately. Based on data presented to the Food and Drug Administration, says Robert McCorm­ack, PhD, vice president of medical affairs at Veridex, “GeneSearch BLN was approved to be used as a stand-alone test for intraoperative testing, or as an adjunct to permanent section in assessing sentinel lymph nodes of patients with breast cancer.” It is meant to replace current intraoperative testing methods, he says, or to provide an intraoperative test for hospitals not currently doing one.

Peter W. Blumencranz, MD, member of Surgical Asso­ciates of West Flori­da, Clearwater, and medical director of breast health services for Morton Plant Mease Health Care, participated in the GeneSearch validation study. “All that data looked very good,” he says. “FDA approved the assay in July of 2007 as an adjunct to permanent section and as a stand-alone intraoperative procedure. We are using it here at my hospital to make decisions about axillary node dissection, but we are still doing frozen section.” His advice for any institution that might want to adopt the assay is to try it first without using it for clinical decisionmaking: “They need to get into a comfort zone with the assay compared with using histology.”

In the GeneSearch validation study, Juan P. Palazzo, MD, professor of pathology at Thomas Jefferson University Hospital, directed the central pathology laboratory. Surgical pathologists at the central site looked at cases blinded and compared their readings to the diagnosis provided in the original hospital. Pathological diagnosis was then compared with the result of the molecular assay. “Correlations turned out to be excellent,” Dr. Palazzo says. Agreement between local and central sites for H&E was 95 percent, and there was good agreement between diagnosis from permanent section and assay results.

Dr. Palazzo attributes differences between permanent section and GeneSearch largely to sampling effects. “The nice thing with this test is that we study the entire lymph node not used for histology,” he says. When H&E and the molecular assay were both positive or negative, that signaled a firm result. However, there was a “handful” of discordant cases, Dr. Palazzo says. “We think that has to do with how the lymph node was sampled. As surgical pathologists we know this happens in a small number of cases. I think there are no false-positives [with GeneSearch],” Dr. Palazzo asserts. “I think there is tumor in the lymph node but it was sectioned; it didn’t show up on the slide.” (For further discussion, see “Findings of an evidence-based assessment,” page 81.) Dr. Palazzo sees GeneSearch as a “nice complement” to what they are doing now with sentinel nodes. “Instead of leaving tissue behind, we are sampling 80 percent to 90 percent of the tissue, which gives us more surety,” he says.

Dr. Palazzo plans to use the molecular assay in conjunction with frozen section. He is more cautious about using it to replace histology. “I am a surgical pathologist trained in morphology,” he says. “There are very few things in clinical pathology that have completely replaced morphology. First there will be multi-institutional trials to see if the initial results with GeneSearch are confirmed. Then we may say whether to use it alone. It will not replace frozen section in the immediate future.”

Maura Pieretti, PhD, HCLD, was also involved in the clinical trials of Gene­Search. She is scientific director of molecular diagnostics at BayCare Laboratories of Morton Plant Hospital in Clearwater, and she may have been the first to offer the assay clinically under fresh frozen conditions. Baycare has been doing GeneSearch testing since August 2007. “We are doing it intraoperatively simultaneously with frozen section,” says Dr. Pieretti, who has no plans at this time to commit the whole sentinel lymph node to GeneSearch.

She, too, thinks of GeneSearch’s lower specificity as a sampling issue. “The major component of the lymph node goes into the molecular assay,” she says. “Although it is controversial, one school of thought says that the more you sample a node, the more mets you will find. Some centers try to do dozens of serial sections for H&E to have a representation of the entire node, which is time-consuming and tedious. GeneSearch allows us to look at a large portion of the node in one go.” Dr. Pieretti acknowledges that the molecular assay doesn’t provide information about the size of the metastasis. But, she suggests, “We hope this assay becomes quantitative. Real-time PCR lends itself to quantitative measurement.”

At Dartmouth-Hitchcock Medical Center, Gregory Tsongalis, PhD, HCLD, FACB, director of molecular pathology and associate professor of pathology, has worked with GeneSearch and calls its performance “pretty robust.” While many people will use it as an intraoperative assay, he says, “I’m not sure I’m ready to use it that way yet.” He wonders whether getting an intraoperative result is more important than waiting a few days to get traditional pathology results. “This question will only be settled by having more data,” he says.

Wendy A. Wells, MD, associate professor of pathology and lead breast surgical pathologist at Dartmouth, says intraoperative frozen sections are not routinely performed on breast sentinel nodes at her institution. “We have a busy breast service, and we don’t have staffing to do intraoperative molecular diagnostic testing,” she says, “so we were never going to offer GeneSearch in that way.” Dartmouth’s surgeons have been going back and doing a second procedure for completion axillary clearance, at the time of margin re-excision or mastectomy, if permanent sections of the sentinel node or nodes are positive. However, at Dartmouth, only 30 percent of the sentinel nodes are positive, and of those cases only 21 percent have further positive axillary nodes after completion dissection. “That’s a small percentage to justify doing full axillary dissection on every case with a positive sentinel node,” she says. “Like others nationwide, our surgeons are now beginning to question the current standard of care that they need to do full axillary dissections on all positive sentinel node cases.”

Though data are still coming in, Dr. Wells says, nomograms have been devised for patients with a positive sentinel node that look at the size of the metastasis, the presence or absence of extranodal extension, characteristics of the main tumor such as grade and size, and lymphatic space invasion. These nomograms, she says, provide a risk score showing the likelihood that the patient will have more positive nodes in the axilla. “If the risk is low enough, some surgeons are opting to irradiate the axilla and not do a full axillary dissection. These issues are still being assessed in prospective clinical trials.”

At Dartmouth, GeneSearch will be used to complement permanent node sectioning. “We are definitely using the assay off-label,” Dr. Wells says. “I’m trying to use it to become more labor efficient in our processing of sentinel nodes and rigorously test the assay against other current tools for assessing sentinel nodes,” such as serial sectioning and immunohistochemistry. “And I’m really keen to extend that evaluation by developing an RNA bank of excess high-quality RNA in these node-positive patients.” Having GeneSearch RNA extracts will allow her to do that.

Dr. Wells is not willing to report an H&E-negative, GeneSearch-positive node as positive. “We are not there yet,” she says. “We may get there, but we need a lot more data.” She and others have talked to Dartmouth clinicians about how they will clinically interpret such scenarios as pN0 (mol+) using the current AJCC Cancer Staging Manual (6th edition). Moreover, she sees continuing value in permanent section. “It is still imperative to do H&E for the architectural information that is still considered clinically relevant,” she says, “such as where the tumor extends out of the node into the perinodal fat.” Her oncologists still want to know about extranodal extension of tumor because it qualifies patients automatically for radiation to that area. Size also still matters to them.

Dr. Wells concedes that pathologists have hit a “morphologic wall” in some diagnostic areas. “We do need something better. First it was immunohistochemistry and next will be molecular assays. I believe that we, as pathologists, must embrace validating these assays rigorously and painstakingly, but we should not hold them to higher standards than some of our current histologic methods where interpretive reproducibility and clinical relevance are also often less than perfect.” If after validation and outcomes data they turn out to be comparable, “that’s fine,” she says. But it’s premature, in her view, to say that histology may at some point be a thing of the past.

At Thomas Jefferson University in Philadelphia, GeneSearch is still being validated and not yet being used clinically, says Tina Bocker Edmonston, MD, director of molecular diagnostics. “Based on the size of my lab and how our hospital is set up, we were thinking of running it as a batch assay once a week. Right now, I don’t have the staffing to have one of my techs ready to do the molecular assay under frozen section conditions. We never know when a case will come.” Because extracts would not be assayed immediately, if the assay does find metastases, the surgeon will have to do a second procedure for axillary lymph node dissection. Another consequence of delayed testing is that Dr. Edmonston has been validating the assay with extract preserved in RNA Later solution. “We have to demonstrate that preserved RNA is equal to if not better than frozen tissue because the assay is only approved with fresh frozen tissue,” she explains.

At the University of Texas Health Science Center at San Antonio, breast pathologist I-Tien Yeh, MD, says, “We have looked at the GeneSearch assay and decided not to set it up.” Dr. Yeh, who is professor of pathology, cites many reasons for this decision. “Other pathologists on my staff felt it would be of minimal utility,” she says. “And I spoke to some breast surgeons. From a clinical point of view, they didn’t feel that the small increase in sensitivity was sufficient for them to advocate setting this assay up.” As for GeneSearch’s increased sensitivity for micrometastases (0.2 to 2.0 mm), which are most difficult to detect on frozen section, those can be detected on H&E and immunostain without much problem, she points out. “I’m not sure we need PCR to detect those.”

That GeneSearch consumes about half of the nodal tissue is sometimes a disadvantage, in Dr. Yeh’s view. “It uses more tissue than we use for frozen section examinations. Greater sampling will, of course, lead to greater detection; this is true no matter what type of assay you perform, whether you cut additional H&Es, perform immunostains or PCR. However, we sometimes go back to tissue blocks from surgeries that were performed in prior years to analyze a new parameter that was unavailable at the time of the initial surgery. This would not be possible if all the tissue is consumed,” she says. Her bottom line is that, for most small or medium hospitals, “I’m not sure how practical it is.”

Each of the laboratories that is adopting GeneSearch is at a different stage in integrating the assay. “What we are doing now is essentially validating controls and making sure the assay works well here,” Dr. Palazzo of Thomas Jefferson says. “At that point we are planning to use it clinically.” Dr. Palazzo has discussed extensively with surgeons how they would handle situations in which frozen section is negative and GeneSearch is positive. “In the right context I have no doubt they would do the dissection,” he says. “They have told me they would proceed with axillary dissection depending on the characteristics of the primary cancer—size, degree of differentiation, presence of vascular invasion—all of which one can get from core biopsy.”

Based on the validation study, Dr. Palazzo estimates that approximately 7.5 percent of cases will be negative on frozen section and positive on GeneSearch. In his view, this increment justifies the additional cost of the molecular assay. “Any practicing breast pathologist will tell you about cases that, for many reasons, he went back to the block of a sentinel node that was called negative, cut deeper, and found mets. That situation can be avoided using GeneSearch.”

Dr. Pieretti is able to do GeneSearch as an intraoperative procedure because BayCare Laboratories has a setup that makes it possible to do the test efficiently. “We have three medical technologists trained in the test. They take turns to be on call for the assay.” One of them goes to set up when she gets a notice at the onset of anesthesia. A small area adjacent to the frozen section room has a window through which the surgical pathologist gives the technologist half of the sentinel node. “Frozen results come out a lot earlier,” Dr. Pieretti says. “We call in the assay result to the surgeon.” While the GeneSearch work area is not divided into separate areas, as is typical for PCR work areas, the work follows a directional flow. Turnaround time—from the time the node leaves the operating room two floors away to when the surgeon receives the result—is 37 minutes. Dr. Pieretti works primarily with one surgeon, “which makes the logistics much easier to coordinate,” she says.

Reimbursement, it seems, is trickier. “It has been very hard to find out what are we getting paid for the test,” Dr. Pieretti says. “I am just now learning how complex it is to look at payment. We are billing with appropriate CPT codes, but reimbursement is based on the whole treatment of the patient. It’s hard to separate out payment for one component.”

The molecular pathology laboratory at Dartmouth-Hitchcock is doing something “a bit different” than the claim for the FDA-cleared product, Dr. Tsongalis says. “We are sectioning the lymph node differently. I’m not sure we are quite ready to sacrifice the whole node as they recommend and cut as they recommend, so we validated a different way of cutting the node that still picks up the majority of metastases and we can confirm by traditional pathology. We are not looking to replace what we were doing, but to complement and enhance it.”

To validate their approach, Dr. Wells did a detailed retrospective analysis of the processing, interpretation, and clinical outcome data for the sentinel nodes performed at Dartmouth over several years, Dr. Tsongalis says. “She showed very nicely that metastatic tumor cells are scattered throughout a huge portion of the lymph node, not localized in one region,” he says. “Traditional histology may miss small clusters of very few tumor cells—micromets—but right now no one knows what they mean anyway. So we found that we can sacrifice a smaller portion of the lymph node for the molecular assay and still be able to detect the tumor cells.”

Dr. Wells has conducted two studies, one exploring the distribution of metastases in positive nodes and one comparing GeneSearch to her histologic procedure. She explains that the first study was instigated “when we were approached by Veridex to validate their assay. We didn’t quite fit their profile, since we do not routinely perform intraoperative frozen sections on our sentinel nodes.” Another difference was how Dr. Wells sections sentinel nodes. “When our institution first started evaluating sentinel nodes, there were no CAP or ADASP [Association of Directors of Anatomic and Surgical Pathology] guidelines. We opted to perform H&E-stained serial sections and cytokeratin immunohistochemistry, as needed,” she says. “We know that this methodology is more than some national guidelines now recommend, but in order to validate the GeneSearch assay as optimally as possible, we needed to maintain the sectioning technique from which we had already generated all of our histologic and outcome data. So we weren’t too keen to take the center of the node and give that tissue up for the assay.”

She asked if, instead of taking the middle of the node, they could take the outer ends of the nodes. “Veridex was not happy with that; it was not according to the FDA label.” So Dr. Wells identified the node blocks from 30 or 40 prior consecutive cases with positive sentinel nodes. “These nodes had been cut in half and sectioned in the middle per our usual protocol,” she says. She de-paraffinized them, flipped them over, and re-embedded them so that the previously cut surface was now face down in the block and the ends of the node were face up and cut into. She then re-sectioned them. Her question was: In these known positive nodes, would there also be tumor on the outer edges of the node? “For every case, I was able to find metastatic disease in the outside portions of the node for deposit sizes ranging from 0.12 mm to 10 mm,” Dr. Wells reports. “So in a positive node there is not always a single meta­static deposit in the middle, but rath­er tumor cells appear to extend around the entire node capsule.”

Dr. Wells drew two conclusions. First, “Tumor deposit size in a node depends on where and how that deposit is measured, and the rigid size cut-offs [micro versus macro] of the current AJCC staging system may not be as clinically meaningful as hoped.” Second, “We could consider using the outside of the node for the GeneSearch assay and continue to cut the node in the middle for our usual protocol so that we could use our prior histologic and clinical outcome data for optimal validation.”

Next Dr. Wells did a validation study on 59 consecutive patients in which both ends of the nodes were submitted for GeneSearch, and she did her routine histological evaluation plus cytokeratin immu­no­­studies for H&E-negative cases. She found that the molecular assay had a sensitivity of 88.9 percent (eight of nine positive samples) and a specificity of 93.5 percent. Based on these data, Dr. Wells plans to eliminate serial sectioning of all sentinel nodes (analysis showed that metastases present in the first tissue sections were identified in the last serial section 90 percent of the time) and incorporate GeneSearch into her diagnostic protocol. For every node, fresh tissue will be sampled from both ends of the node and processed in preparation for the GeneSearch assay. The number of tissue sections will be reduced to one H&E-stained slide. If the sentinel node is positive for malignancy by H&E stain, no further testing will be required. If the node is negative by H&E stain, then the GeneSearch assay will be performed. If the H&E-stained section is negative and the PCR result is positive, then further tissue levels and cytokeratin immunostudies will be performed. If the histology remains negative, then the nodes will be reported as negative, with the designation N0(mol+).

The data will be reevaluated in one to two years to assess correlation with the histologic gold standard, assay accuracy, cost-effectiveness, and clinical outcome.

At the University of Texas Health Science Center it was the cost-benefit ratio that kept the laboratory from setting it up, Dr. Yeh says. “Our surgeons only require intraoperative consult for sentinel lymph node examination during mastectomies, not during lumpectomies. And the majority of our breast cancer surgeries are lumpectomies.” Further, she estimates only about a 10 percent increase in sensitivity but no improvement in specificity. “It just wasn’t enough for us to feel that we could commit time for training, equipment setup, and space to this assay.”

Looking in more detail into the extent of benefit conferred by GeneSearch, Dr.Yeh notes that the positive lymph node rate overall for breast cancer “in our patient population is approximately 25 percent. About 75 percent of cases will not even need an axillary node dissection. So the potential 10 percent increase in sensitivity from 85 percent to 95 percent would only apply to one-fourth of patients whom we see. There are only a handful of patients who might benefit, even if every claim that Veridex makes is true.”

Financials didn’t look good either, Dr. Yeh says. “Based on our analysis of the cost to the laboratory for the assay versus current reimbursement rates from our insurance carriers, it was going to cost us more per assay than we would be reimbursed by our carriers. This may not be true in higher-volume labs, but in our lab we would lose money every time we performed the assay.”

Veridex’s Dr. McCormack explains the two postmarketing studies the FDA required. One is to substantiate the assay’s turnaround time when used intraoperatively. “We will be looking at the logistics of implementing the test in the lab at four sites,” he says. “We need to ensure that GeneSearch is rapid enough to be performed in the intraoperative setting.”

In the second, larger study, the assay’s accuracy will be evaluated. “This is a fairly extensive study,” Dr. McCormack says “It requires about 1,200 patients at clinical sites actually using GeneSearch as their test of record and where decisions about axillary dissection are being based on it.” The FDA required Veridex to follow patients who have a positive assay result and axillary dissection for outcomes. “We are asking whether in fact the test was correct in saying that the axillary nodes should have been removed,” Dr. McCormack says. “We are looking for a reduction in nodal or distal recurrence.”

Even while questions about the assay’s current application are being resolved, some people are looking to its wider use. Dr. Blumencranz, of Surgical Associates of West Florida, foresees a time when people will say, “Why not increase the efficiency of the pathologist’s valuable time? Let’s homogenize all of the sentinel node.” That is not approved, he cautions. “We are not doing that. But we may eventually homogenize the whole node intraoperatively.”

Dr. Tsongalis is thinking along the same line. He notes that permanent section provides information on the size and location of metastatic lesions, which is considered important for planning patient management and which cannot be obtained with a molecular assay.

“We will see whether the information we provide on the molecular level will trump that,” he says. “Have people used that information from H&E only because it is as good as it gets? With something new, will those parameters not be as important anymore?”

Related Links

• Findings of an evidence-based assessment