June 2010
Feature Story

William Check, PhD

A couple awaits the birth of a child, following their usual routines during the leisurely pregnancy. One day—whomp!—the baby arrives. Everything is disrupted for a while, until they develop new, adapted approaches to daily life.

A similar process is taking place in pathology with regard to the diagnosis of non-small-cell lung cancer, or NSCLC. For many years pathologists followed routine histologic and cytologic procedures for this diagnosis. Meanwhile, gradual changes occurred: Understanding of molecular events in this neoplasm advanced, biomarkers were defined, and targeted molecular therapies were approved. Still, basic diagnostic protocols were not radically altered. Then, relatively abruptly over the past few years, clinical trials revealed critical inter­actions between some of the new thera­peutic agents and histologic subtypes of NSCLC, making the differentiation between squamous and ad­enocarcinoma man­datory in many cases. Now a period of adaptation is taking place while the best techniques are identified to make this distinction in routine daily practice.

During this transition there will inevitably be uncertainty about the best approach. With regard to one commercially offered immunohistochemistry panel, there is even strong objection. However, there is no uncertainty or disagreement about two aspects of the subclassification of NSCLC, says Philip T. Cagle, MD, director of pulmo­nary pathology at The Methodist Hospital, Houston, and professor of pathology at Weill Medical School. First, simply diagnosing “non-small-cell carcinoma” is no longer adequate when a specific cell type can be determined. Oncologists increasingly need a specific cell type diagnosis whenever possible. Second, Dr. Cagle says, “Community pathologists sitting at their microscopes doing the work they typically do have a chance to make a significant contribution to selecting patients who will receive the new therapies.” He calls these new treatments “the greatest breakthrough in lung cancer in decades.” In addition to standard histology, in some cases it will be necessary to do special stains by IHC. “Traditional histology and implementation of molecular findings are combining to move this field forward,” Dr. Cagle says.

“Historically, making that distinction [between squamous and adenocarcinoma] has been meaningless because of the poor prognosis of both subtypes,” says Saul Suster, MD, chair of the Department of Pathology at the Medical College of Wisconsin. “What has changed is the introduction of targeted therapies, such as tyrosine kinase inhibitors [TKIs] and monoclonal antibodies against epidermal growth factor receptor [EGFR]. Some types of tumors are more prone to having these receptors expressed and more likely to respond to some of these drugs.” For example, adenocarcinomas are more frequently EGFR positive, and EGFR-positive advanced lung cancers respond strongly to TKIs. “In addition,” Dr. Suster says, “some oncologists consider that certain drugs are contraindicated in squamous tumors because of the risk of severe pulmonary bleeding.”

Henry D. Tazelaar, MD, consultant in pathology at Mayo Clinic Scottsdale and professor of pathology at Mayo Clinic College of Medicine, says some chemo­therapeutic protocols include drugs that have been shown to result in life-threatening hemorrhage if used in patients with squamous histology. The chief culprit is the antiangiogenesis agent bevacizumab (Avastin), a monoclonal antibody directed against vascular endothelial growth factor, or VEGF. “Historically, distinguishing squamous from nonsquamous histology was not of great significance because the treatment for all metastatic non-small-cell lung cancers was pretty much the same,” Dr. Tazelaar says. “Now with this complication of hemorrhage in patients with squamous histology, practice has shifted. Frequently, clinicians want to know if a tumor is squamous or nonsquamous.”

Making the required subclassification is often difficult because of the small amount of tissue available in patients with advanced, inoperable lung cancer, who make up the great majority of those with lung cancer diagnoses and who are treated with chemotherapy. About 70 percent of lung cancers are diagnosed on small biopsies or cytology specimens alone. “This work is fraught with difficulty,” Dr. Tazelaar says, “partly because of sampling problems. It is known that lung cancer can be quite heterogeneous. It can be squamous in one area and adeno in another.”

In addition, no one stain is perfect at predicting squamous versus nonsquamous histology, he says. “We are learning more about how best to use ancillary techniques to subclassify non-small-cell lung cancer. The jury is still out on which tests to use and in which patients.” Establishing definitive histology might not be necessary in patients who are going to surgery, in Dr. Tazelaar’s view. “It may not be important for an immediate treatment decision. If the patient presents later with metastases, you might want to do it then.”

When there were no clinical implications, attention was not paid to subclassification of NSCLC, agrees William D. Travis, MD, attending thoracic pathologist at Memorial Sloan Kettering Cancer Center. “Now we have new drugs. Basically that is what all the excitement is about,” he says. In addition to bevacizumab, another important new agent is pemetrexed (Alimta), a novel antifolate drug that is effective in adenocarcinoma and NSCLC NOS, but not in squamous carcinoma. Add to this the fact that bevacizumab is not given to patients with squamous carcinoma because of the risk of hemorrhage. “So, for the first time in the history of non-small-cell lung cancer, we need to subclassify,” Dr. Travis says. “First, because we have effective drugs and, second, because of important histological distinctions that define specific therapies” (Gridelli C. Curr Opin Oncol. 2009; 21:97–98; Scagliotti G, et al. Oncologist. 2009;14: 253–263; Tiseo M, et al. Expert Rev Anticancer Ther. 2009;9:425–435).

Dr. Cagle emphasizes the significance of this new role for pathologists: “This is an amazing change to those of us who have been practicing pathology all these years.” He describes the pathologist’s role in this as “very real” and says that specifying cell type and subtype depends on conventional histology, sometimes supplemented with IHC stains.

Dr. Travis points out the impact of small tissue specimens in specifying tissue subtype in NSCLC (Travis WD, et al. J Thorac Oncol. 2010;5:411–414). About 120,000 cases of unresectable NSCLC each year in the U.S. are diagnosed on small biopsies and cytologies, he notes. “That is a huge number. We pathologists have not historically paid attention to classification of non-small-cell lung cancer in small biopsies and cytologies.” Dr. Travis, who has been a leader in developing the last two WHO lung tumor classification schemes, points out that these classifications do not address this issue and there are no historical guidelines for diagnosing lung cancer in small biopsies. “As a result, there are not standardized criteria,” he says.

This situation will soon change, however, thanks to a new classification scheme devised by a panel of the International Association for the Study of Lung Cancer, which Dr. Travis chairs. “For the first time in the history of lung cancer,” he says, “we will have an internationally developed, standardized set of criteria for the classification of small biopsies and cytology with nomenclature developed in a multidisciplinary fashion in conjunction with oncologists who will be reading our diagnoses. This is going to be a big deal in the lung cancer field and should have a dramatic impact on how these tumors are classified.”

A more recent implication of small biopsies arises from the advent of molecular analyses. “One of our big goals is to preserve tissue for molecular studies,” Dr. Travis says. “When we get these tiny biopsies, we must be careful to reserve some tissue to be analyzed for EGFR expression. Every slice of tissue that you use for an antibody stain is one less opportunity to do a molecular study.”

Nowhere in oncology is improved treatment needed more urgently than in lung cancer. Worldwide five-year survival among lung cancer patients ranges from six percent to 14 percent for men and seven percent to 18 percent for women, Dr. Cagle says. More people die from lung cancer each year than they do from the next three common cancers combined. “Worldwide, lung cancer is a growing problem and will not go away simply because people quit smoking,” he says. “After stopping smoking people still have an increased risk of lung cancer for many years. Even if every smoker in the world stopped today, we would still have many cases for decades to come.”

In the forthcoming volume Advances in Surgical Pathology: Lung Cancer, of which he is one of several editors/authors, Dr. Cagle describes contemporary chemotherapy protocols for NSCLC. Improvement in treatment outcomes was achieved with platinum-based doublets containing third-generation drugs such as gem­citabine, docetaxel, paclitaxel, or vinorelbine. In two randomized, placebo-controlled phase III trials, addition of bevacizumab to a platinum-based doublet produced superior efficacy compared with chemotherapy alone. As already noted, this benefit is, unfortunately, limited to nonsquamous histology, because of the risk of severe toxicity in those with squamous carcinoma. Pemetrexed in combination with cisplatin was superior to gemcitabine/cisplatin in advanced NSCLC patients with nonsquamous histology, but gemcitabine/cisplatin was superior in squamous carcinoma. Thus, for both of these new targeted agents, pathologists need to differentiate squamous from nonsquamous tumors in NSCLC specimens.

Among the pathologists consulted for this article, there is unanimity on the path to making this distinction. “We start with standard published histologic criteria,” Dr. Suster says. “Sometimes we use IHC stains to better resolve poorly defined neoplasms. Squamous tumors tend to express cytokeratin 5/6 [CK5/6] and nuclear p63, whereas carcinoembryonic antigen and thyroid transcription factor 1 [TTF-1] are more commonly expressed in adenocarcinoma.”

Dr. Tazelaar adds desmoglein 3 (DG3) as a good marker of squamous differentiation, though the antibody is not widely available. “We are now beginning to use small panels of antibodies to help us find signs of squamous or nonsquamous differentiation if we do not see obvious features by light microscopy,” he says. Obvious features include intracytoplasmic mucin, which indicates ade­nocarcinoma, and keratinization, suggest­ing squamous subtype. “If you see either of these features you don’t need to stain,” Dr. Tazelaar says. “But in many cases there is not much material or the tumor is rather poorly differentiated.”

“In the old days,” Dr. Suster recalls, “mucicarmine or PAS was used to demonstrate mucin.” They are simpler and cheaper than IHC, he says, but IHC is done routinely in most labs and most pathologists go directly to IHC. “It is more specific,” he acknowledges, “but histochemistry is still a valuable tool to demonstrate mucin.”

Dr. Cagle notes that the IHC markers are usually mutually exclusive. “We expect that if p63 or CK5/6 is positive, then TTF-1 will be negative and vice versa.”

Dr. Travis limits the number of stains he does to preserve tissue. “At most I try to do two stains, TTF-1 and p63, and maybe mucin,” he says. “If you start doing larger panels of immunostains, you have nothing left for molecular testing. That’s a disaster for the patient. Unfortunately,” he continues, “some people perform panels of multiple antibodies trying to make this distinction. I don’t know any antibody better than TTF-1 for adenocarcinoma or p63 for squamous.” TTF-1 has the added value, Dr. Travis says, of being a pneumocyte marker that can favor a lung primary when it is positive or raise the question of a possible metastasis when it is negative.

It is over this last point that disagreement arises. In February of this year, Clarient launched its five-antigen IHC panel called PulmoType to help subclassify lung cancer. “PulmoType has a very high accuracy rate with a very low uninterpretable rate,” says Ken Bloom, MD, Clarient’s chief medical officer. PulmoType was compared head-to-head against p63 and TTF-1, neither of which is on the panel, in 1,100 patients, the largest study of p63/TTF-1 ever published (Ring BZ, et al. Mod Pathol. 2009;22:1032–1043). Sensitivity was higher for PulmoType (88.6 percent versus 74.1 percent) as was negative predictive value (84.8 percent versus 60.3 percent). The two methods had equal positive predictive value and specificity. PulmoType’s equivocal rate was 11 percent, while the p63/TTF-1 combination was uncallable 22 percent of the time (both markers positive or negative). Overall accuracy was 90.7 percent for PulmoType and 80.2 percent for p63/TTF-1.

PulmoType consists of five immunostains—CK5/6, MUC-1, CEACAM5, TRIM29, and SLC7A5. How did an “optimal” panel come to contain several antigens that no one has ever heard of? Dr. Bloom explains that the ultimate origin of several of these markers was expression-array analysis done by a research group at Stanford that looked at unique mRNA clustering in various kinds of tumors. Applied Genomics Inc. was formed to identify robust and reproducible antibodies that could identify these clusters. Ultimately, 80 satisfactory antibodies were created.

“Next, they applied these findings to the clinical problem of separating squamous carcinoma from adenocarcinoma,” Dr. Bloom says. To the 80 antibodies, Applied Genomics investigators added every antibody in the literature that was being used for subtyping NSCLC. They wound up testing 105 antibodies on a 551-sample discovery set. Analyzing the results with a variety of statistical tools, including “logistic regression” modeling, identified five antibodies as the best combination to classify adenocarcinoma and squamous carcinoma accurately. The result was validated on an independent sample of more than 1,100 specimens from three separate lung cancer cohorts.

When this research was undertaken, Dr. Bloom notes, “there was not really a good standardized way to distinguish adenocarcinoma and squamous subtypes in lung cancer. Everybody was using their own antibodies with their own titers and their own combinations.” In contrast, he says, “PulmoType is not a single institutional experience; it is not anecdotal. It was validated in over 1,100 patients.”

Clarient is often asked why p63 and TTF-1 are not in the panel. “They were among the 105 markers evaluated, but they didn’t make the cut,” Dr. Bloom says, “not because they aren’t great antibodies, but because the information they provide is redundant to the information already provided by the other antibodies in PulmoType. They turned out not to be the best antibodies to answer the question.”

In December 2009, Clarient acquired Applied Genomics and its antibodies, three of which were not commercially available. After re-validating Applied Genomics’ results, Clarient offered PulmoType to the pathology profession. “To me this panel is for stage three and four patients who will not have excision, who are having fine-needle or core biopsies, and who will be treated on the basis of a very small sample,” Dr. Bloom says. “PulmoType can help if the diagnosis is not obvious, which it isn’t in many small samples.”

PulmoType is offered on a “tech-only” or global basis. A pathologist sends in a tissue block, Clarient does the immunostains, and the pathologist can do the interpretation and bill for the professional fee. “If the case is sent to us on Monday,” Dr. Bloom says, “we receive it on Tuesday, and the client receives either the slides or the result on Wednesday, depending on whether Clarient or the client is interpreting the assay.” Dr. Bloom says four of the markers are billed as 88342 and one is billed as 88360. At current Medicare rates, reimbursement is about $550.

Dr. Travis is not convinced of PulmoType’s validity or superiority. Of the 2009 publication in Modern Pathology, he says, “The authors fail to address several major issues that make me concerned.”

First, he says, “All current data driving the need to distinguish squamous carcinoma from adenocarcinoma and non-small-cell carcinoma-NOS are based solely on light microscopy. The basic premise of this paper is that immunohistochemistry using the five-antibody panel makes for more accurate diagnosis than either light microscopy and/or TTF-1/p63 and that patient therapy should be decided based on these results. However, there is no clinical trial that confirms that diagnosis based on immunohistochemistry is more predictive of bevacizumab toxicity or pemetrexed efficacy.”

His second objection is that the only diagnosis that excludes a patient from pemetrexed or bevacizumab is squamous cell carcinoma. Yet the data were analyzed with regard to all subtypes. “They should have compared squamous cell carcinoma versus adenocarcinoma and unclassified carcinomas, which is the breakpoint that impacts clinical decisions,” Dr. Travis says.

Third, he points out that the authors do not provide the detailed results of staining for each of the individual antibodies, only the combination. “It really is not possible to compare the utility of the individual antibodies,” Dr. Travis says. “It is not surprising that a larger panel of antibodies might give you a smaller number of unclassifiable cases compared to a smaller panel.”

Finally, Dr. Travis reiterates his insistence that tissue must be saved for molecular testing. The drugs where histology matters—bevacizumab, pemetrexed, and TKIs—are used in patients who must be tested for EGFR mutations. Patients with advanced adenocarcinoma or NSCLC-NOS should be tested for EGFR mutations. EGFR-mutation-positive patients are candidates for TKI therapy; if the mutation status is negative, then they are eligible for pemetrexed or bevacizumab. “Using material for immunohistochemistry for five antibodies could waste tissue, as most patients with advanced lung cancer have diagnoses based on small biopsies or cytology,” he says.

Douglas Ross, MD, PhD, addressed Dr. Travis’ objections. Dr. Ross was part of the Stanford research group, went with Applied Genomics when it was formed, and is now chief science officer at Clarient. “[Dr. Travis] is quite right to point out that there are no clinical trials that explore IHC surrogates for morphology as markers for either toxicity or therapeutic response,” Dr. Ross says. The best experiment would have been to compare PulmoType directly to pemetrexed or bevacizumab clinical trial samples; Clarient is working on getting access to the former samples. However, Dr. Ross continues, “The premise of the [Modern Pathology] paper is that resection morphology is the gold standard and that IHC tests serve as an aid when morphology is unclear, either due to sparse sample or poor differentiation.” The Ring, et al., article compares PulmoType with TTF-1/p63 using morphology as the gold standard. “When morphology is clear, no IHC aids are necessary,” Dr. Ross says. “When histology is unclear, IHC aids can offer a histologic surrogate that, in the case of PulmoType, has been validated as accurate and that offers information when TTF-1 and p63 fail to make a call.”

In response to Dr. Travis’ contention that squamous carcinoma should not have been lumped with NSCLC-NOS, Dr. Ross says, “In fact we do point out in table 3 that PulmoType and TTF-1/p63 make determinations on cases where the clinical record indicates that the pathologists deemed a case ‘NSCLC-NOS’ or large cell. When they both assert histotype, they invariably make the same call, suggesting that these two independent measures of biologic differentiation agree and identify biology not apparent even on surgical resection specimens to pathologists.”

With regard to evaluating each marker in the PulmoType panel, Dr. Ross says, “We did not select individual markers, but instead selected the best combination of markers to maximize sensitivity and specificity. Combining the five antibodies in a weighted algorithm allows the test to take advantage of, in part, redundancy and, in part, independence of the stains.”

Dr. Ross agrees that preservation of tissue is an issue, saying, “Clarient is careful to manage tissue in a manner that ensures that EGFR and other molecular testing can be done on samples after histotype determination.” For instance, the company uses allele-specific PCR rather than sequencing approaches, which require much more sample. “Additionally,” Dr. Ross says, “we have shown we can perform molecular testing on the IHC-stained slides used for PulmoType, if necessary.”

Dr. Bloom raises another issue—reproducibility. “Current methods are causing problems in clinical trials,” he says, citing an editorial titled “If histology matters ...” by a group of oncologists (Stinchcombe TE, et al. J Clin Oncol. 2010;28:1810–1812). The authors write:

“If histology, especially when differentiating between squamous and nonsquamous histology, is now essential to the treatment of NSCLC, the inevitable questions are, How accurate are we in assessing histology? and, How great is the variation in the histologic diagnosis among pathologists?”

They cite evidence that reproducibility among pathologists is not high. In their own study, they had 12 expert pulmonary pathologists and 12 community pathologists review the same digital images from 96 primary NSCLC tumors. “Even for the binary distinction of squamous versus nonsquamous carcinoma,” they report, agreement was only moderate, with kappa values of 0.55 among all pathologists, 0.64 among expert pathologists, and 0.41 for community pathologists.

“Overall, this is a big deal,” Dr. Bloom says. “Patients’ therapies are going to be decided by our classification. Unfortunately, we continue to believe we are better than we are and to ignore evidence-based medicine.”

CAP TODAY asked Dr. Cagle for his impartial view of PulmoType. At the time he was asked, he had not yet used PulmoType, so he was unable to give a firsthand appraisal. “From a practical point of view, the potential value of PulmoType will depend in large part on the practice one is in,” he begins.

To set the context for his remarks, Dr. Cagle says, “No one disagrees that the proper classification of lung cancer cell type is now crucial for treatment.” Nor does anyone disagree that “This vital diagnosis of lung cancer cell type is not a problem in most cases. Provided that a pathologist is reasonably trained, competent, and experienced, the cell type of most lung cancers can be diagnosed on the biopsy and the cytology specimen using our routine stains.” Dr. Cagle notes that the importance of cytology in rendering a cell type diagnosis should not be overlooked because it may be more revealing than a biopsy (Nicholson AG, et al. J Thorac Oncol. 2010;5:436– 441).

At The Methodist Hospital where Dr. Cagle practices, cell type is diagnosed in most cases by routine stains on routine histology and cytology. “Cell type immunostains are needed on only about 15 percent of our lung cancer biopsies,” he says. “We have a large, robust, in-house IHC lab that provides 24-hour service, so I can get results on TTF-1, p63, and CK5/6 (our usual panel) on the same day. I end up with maybe five percent or less that cannot be classified.” The inability to classify might be for technical reasons or because the cancer is so poorly differentiated that no test identifies a cell type. “In both situations, neither PulmoType nor any other test would be able to further clarify the cell type,” Dr. Cagle says.

In other settings, he continues, “PulmoType may be a great convenience or a good option for identifying the cell type of a lung cancer for which cell type cannot be determined on routine histology or cytology. Labs that do not have in-house immunostains for this differential might benefit from sending their tissue for PulmoType, which may be able to offer fast turnaround and may be very cost-effective for some practices.” PulmoType may assist in situations where the pathologist does not diagnose lung cancer cell types with as much frequency as a larger practice, he says. In particular, it may reduce interobserver and intraobserver variability in determining the cell type of difficult cases. “Even at The Methodist Hospital, we might consider trying PulmoType for the occasional patient for whom we cannot determine a cell type based on failure of our in-house immunoprofile,” Dr. Cagle concludes.

If PulmoType, a conventional IHC panel, evokes resistance, one can imagine the reaction to a more esoteric method for differentiating squamous from nonsquamous lung carcinoma—Prometheus’ ProOnc SquamousDx, which is based on quantitative expression of the micro RNA miR-205. James N. Lowder, MD, senior medical director of Prometheus, says this real-time quantitative RT-PCR assay measures the amount of miR-205, which is expressed at about 35 times higher levels in squamous cell carcinoma versus nonsquamous carcinoma. Initial exploratory work and limited validation (72 samples) showed 96 percent sensitivity and 90 percent specificity (Lebanony D, et al. J Clin Oncol. 2009;27:2030–2037). Further work was done in poorly differentiated lung tumors (Bishop JA, et al. Clin Cancer Res. 2010;16:610–619). “The Bishop study demonstrates that the test works well with very small specimens,” Dr. Lowder says. “It works in fine-needle aspirates. However, it is currently only available for formalin-fixed, paraffin-embedded tissue specimens.” Samples are sent to Prometheus, which does the test with a 10- to 13-day turnaround time at a cost of approximately $3,000.

Dr. Travis and his colleagues wrote an editorial for the Journal of Thoracic Oncology (2010;5:411–414), in which they said: “Other molecular techniques, such as micro-RNA, have been reported to distinguish squamous cell carcinoma and adenocarcinoma. However, whether they provide any advantage to light microscopy and immunohistochemistry remains to be determined.”

Clinical validation studies with ProOnc SquamousDx continue.

While advanced molecular therapies in NSCLC are so far limited to adenocarcinoma, Cesare Gridelli, MD, says in his 2009 commentary that this situation could already be changing (Curr Opin Oncol. 2009;21:97–98). Cetuximab (Erbitux), an anti-EGFR monoclonal antibody, is approved for colorectal carcinoma and head and neck cancer. Dr. Gridelli, of the Division of Medical Oncology, S.G. Moscati Hospital, Avellino, Italy, writes:

“Very recently, in the FLEX (firstline in lung cancer with Erbitux) phase III randomized trial, the combination of cisplatin and vinorelbine along with cetuximab demonstrated superiority in terms of response rate and overall survival compared with chemotherapy alone in the first-line treatment of advanced EGFR-expressing NSCLC.... [C]etuximab-based chemotherapy could represent a treatment option for all EGFR-positive NSCLC.”

In that case, cetuximab would become the first molecular agent effective in squamous NSCLC as well.

Dr. Cagle foresees this as a general phenomenon. “Mutations related to cell types and subtypes on which molecular targeted therapies can potentially have an impact have not yet been fully exploited,” he says. “It turns out that the first few have been related to adenocarcinoma. I expect more in the near future.” As therapies accumulate that are subtype-specific, the role of the pathologist in guiding therapy for lung cancer patients can only increase.