College of American Pathologists
Printable Version

  Groping for ground zero
  in cancers


cap today

April 2005
Cover Story

In cancer, as in politics, everything funnels into the primary. Both, it should be noted, have about equal rates of success.

Oncologists are direct. Theirs is a ready-aim-fire sort of existence. They want to know what kind of cancer their patient has—breast, lung, prostate—so theyll know how to treat it. Had they been in charge of the last presidential election, it’s easy to imagine them ducking New Hampshire, the Iowa caucus, and Super Tuesday in favor of something more streamlined: one primary, one candidate.

Pathologists, on the other hand, could be mistaken for DNC roadies, sharing a cup of coffee with all the candidates and weighing the merits of Joe and Carol, the two Johns, Al and Dennis and Howard. Each primary is a way to sift through issues and possibilities, with the best bet—though not necessarily the winner—emerging at the end of a long, crisscrossing march.

Even if pathologists prefer to take the direct route, they can’t always do so, particularly when it comes to identifying the primary site of a cancer. In fact, what sounds like a simple question—What kind of cancer is it?—may never even be answered.

Carcinomas of unknown primary, as they’re commonly called, are troubling. But not always for the obvious reasons.

One problem is obvious: Some carcinomas have standard and efficacious therapies, and oncologists want to take advantage of them whenever possible. From a patient’s perspective, the problem is even more transparent: The specter of an untraceable cancer has a nightmarish quality to it.

But those problems ignore other, more subtle difficulties. One is age-old: Oncologists want definitive answers, while pathologists don’t always have the tools to provide them. And occasionally oncologists may have their answer without even realizing it, when they fail to obtain a thorough history and to pass it along to pathologists.

It’s also no secret that even the best chemotherapies are less than perfect. "If you find an honest oncologist—not that there aren’t any, but to find one without a particular bias—they’ll tell you there aren’t that many super effective chemotherapeutic agents," says Andrew L. Folpe, MD, associate professor, Department of Pathology and Laboratory Medicine, Emory University Hospital. Oncologists are fond of talking about the various sensitivities of specific tumors to specific treatments, he notes, "but they never, in my experience, have much trouble coming up with a regimen to give a patient when you say, ’I’m not that sure what it is.’"

Speaking by phone from his office in Atlanta, Dr. Folpe says, "I’m sitting here staring at 150 papers about treating carcinoma of unknown primary with three specific regimens, and they all seem to combine basically the same three drugs, as far as I can tell. In fact, as I look, I can see that they’re overlapping the way they treat lung and the way they treat urothelial cancers, and a whole bunch of other tumors."

It’s possible, then, that an unknown primary may not be the incubus it first appears to be. Identifying the primary "often is not hugely important," says Dr. Folpe. But while the unknown may not be cataclysmic, it continues to draw physicians forward, like agnostics on a pilgrimage.

The best place to start, as any married woman knows, is by asking for directions.

Cases that start out looking like tumors of unknown origin—another common sobriquet—may have an obvious if overlooked source.

"Pathologists need to make every effort to ensure that there is not a known history of malignancy," says Richard W. Brown, MD, medical director of the laboratory at Memorial Hermann Southwest Hospital, Houston, and chair of the CAP Histotechnology Committee.

Rod Miller, MD, director of immunohistochemistry, ProPath Laboratory Inc., Dallas, receives his fair share of cases marked as tumors of unknown origin or unknown primary, on which he then runs a battery of immuno stains. "But sometimes when the smoke clears, we find out that the clinician just didn’t tell the pathologist that his patient had a prior history of a particular tumor. If we know that history ahead of time, we will approach these cases quite differently." That’s why he encourages pathologists to make a simple phone call before they start working up these tumors. "It’s worth a few minutes of your time to ask the clinician whether the patient has any prior pertinent history that might guide us toward the most likely primary site," says Dr. Miller, who is also a member of the CAP Cell Markers Committee.

Those are the easy cases—terra incognita has, in fact, been traveled before, and the unknown primary is actually hiding in plain sight. The number of cases initially labeled as "unknown primary" can drop further as parallel tracks of detection unfold.

It’s a delicate matter, involving the cleft between community and academic pathology. Dr. Folpe describes cases that arrive from smaller hospitals, in which "you get the sense that the patient has not been worked up at all before the pathology was sent out, and the referring pathologist and referring clinician just haven’t thought it through." As the consulting pathologist works up the case, the referring physicians may begin to look into matters more deeply. "Often what happens is when you get back to them, they’ll say, "Oh yeah, it turns out she’s got a tumor in the ovary." And your own findings will be consistent with that."

"It appears to be a much bigger problem in community-based practices; that is, community-based practices tend to send them to reference labs or academic centers for big immunohistochemical workups," says Dennis Sgroi, MD, director of breast pathology at Massachusetts General Hospital and associate professor of pathology at Harvard Medical School. Larger institutions have extensive IHC facilities; by using a broad panel of antibodies, pathologists often can reduce the number of "unknown" cases.

But not always. "Often what happens even when they send them out to these reference labs, reference labs can just say it’s a poorly differentiated carcinoma," says Dr. Sgroi. "And so you’re back to ground zero. You’d be surprised at how many they can’t figure out, even after big workups."

The flip side is that community-based pathologists have ample opportunity to engage with other physicians, a practice that can erode when pathologists work with multiple hospitals or are completely off-site. "As people get divorced from having that interaction, between the surgeon, the oncologist, the pathologist, the radiologist, then we end up doing broader and broader workups, which are, in my opinion, less effective," says Dr. Folpe. It may be that finding the primary does indeed require a complex workup, just as sometimes the best way to reach a destination is to fire up a GPS device. But sometimes it’s enough to roll down the window and ask a passerby for directions.

Though estimates vary, it’s generally agreed that cases involving unknown primaries account for no more than 10 percent of all newly diagnosed malignancies. Some say that figure may drop to four percent; others say it may be closer to two or three percent, and still others suggest four to six percent is a reasonable range.

Allen M. Gown, MD, says these tumors are a staple of the consultation service at PhenoPath Laboratories, Seattle, where he is medical director and chief pathologist. His lab sees six to 12 new cases a day of carcinoma of unknown primary.

Dr. Sgroi considers the problem to be significant, noting that even if the lower percentages are the more accurate figures, the total numbers will still be large, given the overall number of cancer diagnoses made each year. "It’s more common than non-Hodgkin’s lymphoma. It’s one of the top 10 most common cancer diagnoses," he says.

It’s also the fourth most common cause of cancer mortality, he says. Why are they so aggressive? "That’s the $64 billion question," Dr. Sgroi says. "We don’t know. We wish we did. But we don’t." The biology of these tumors is not well studied. They’re likely not a homogenous group, and for whatever reason, they have metastasized widely in the absence of a clinically definable tumor in their site of origin.

Drawing on his own area of expertise, breast cancer, Dr. Sgroi notes that the tumors that tend to respond to chemotherapy—at least initially—turn out to be poorly differentiated. That’s not necessarily a surprise, since these tumors often proliferate rapidly, and certain chemotherapeutic agents tend to work better when cells are dividing. But overall the outcome is poor. "You can reduce tumor load, but what’s left is still dividing at a great rate—that’s the simple way of looking at it," Dr. Sgroi submits.

Who are these artful dodgers?

Often they’re hepatomas, which can present as either a single mass or multiple masses in the liver. They can also have a morphologic appearance that simulates a variety of other tumors, says Dr. Miller. Ovarian tumors may also have deceptive morphologic appearances that push pathologists to consider metastatic disease from a number of other, nonovarian primary sites. Differentiating lung and colon tumors on morphology alone can also be ticklish—Dr. Miller points to the example of a patient with colon cancer who presents two years later with a lung nodule. A biopsy reveals an adenocarcinoma, but the morphology doesn’t reveal its origin.

Small tumors can easily lie in purdah. So can tumors that regress. Some are tucked away in areas that are difficult to image. Occasionally a primary tumor site is removed along with the organ—a gallbladder removed for what is presumed to be benign disease, for example, may contain a small carcinoma that later shows up in the liver. The primary is now gone—and was never found in the first place.

Dr. Gown divides cases of unknown primary into two particular carcinomas. As an example of one category, he describes a woman in her 60s or 70s who appeared to be disease-free following a breast cancer five or 10 years earlier but now has a tumor in her lung. The differential is clear and limited: Is this a new lung primary, or is this a metastasis from the previous cancer?

In the other category might be a man in his mid-50s who presents with multiple lesions in the liver. This gives rise to a much more open-ended question: Where is the primary site?

One subset of "unknown" tumors is somewhat easier to place: primary peritoneal serous carcinomas in women. These tumors, though akin to ovarian cancer biologically, spread primarily over the peritoneal surface; they are a well-recognized category of tumor that scatters widely without having a so-called tumor mass in a particular organ. "It’s possible that there are other tumors like this that we have not identified," says Dr. Brown. "But generally speaking, one assumes in these tumors of unknown primary that there is a site of origin that you simply can’t identify."

Faced with these varying shades of the unknown, what’s the best way for pathologists to search for answers?

Pathologists are most effective when there is a defined differential diagnosis, says Dr. Gown—for example, a patient with a history of colon cancer presenting with a mass in the ovary. Given that focused differential, "We have a panel of immunohistochemical markers that can be very effective." Ditto for cases in which a patient has a tumor in, say, the vicinity of kidney, adrenal, and liver, prompting the question, Is it adrenal cortical, renal cell carcinoma, or hepatocellular carcinoma? In these scenarios, looking for the primary resembles the subtle, skillful motions of fly-casting.

Of course, you can also land a fish by dropping grenades in the water. "Where we’re least effective is in the setting of, a random lymph node pops up, it’s an undifferentiated carcinoma histologically, and then we start to throw in markers," says Dr. Folpe. "My experience has been that we very, very often end up with a fairly broad differential diagnosis, even after we’ve done an extensive workup on it. And then it ends up getting bounced back to the clinicians."

Histology should be the major guide at the beginning of any workup, Dr. Gown says, but it should not be used to override other evidence, in part because histological appearances can change as tumors progress. ProPath’s Dr. Miller, who says nearly all his work is referral from other laboratories, begins with a thorough review of a standard H&E section. Then, depending on the tumor’s appearance, he and his colleagues will turn to a panel of immunostains. Sometimes just a few stains will suffice to let Dr. Miller narrow down the likely primary site, if not identify it. In other cases, it may take a dozen or so stains.

The most promising markers to come along in recent years are the so-called homeobox proteins, or transcription factors, says Dr. Brown. These include thyroid transcription factor-1, or TTF-1, which is unique to lung and thyroid, and CDX-2, which is unique to gastrointestinal sites. These proteins regulate the development of the cell line in normal tissues and are frequently overexpressed in tumors from those particular sites.

Dr. Folpe reiterates the value of IHC in the context of focused panels. "In so many settings, we have only a limited number of magic bullets in terms of the immunos," that is, markers that truly are fairly lineage-specific, he says. He and his colleagues at Emory usually begin their workups by looking at a combination of cytokeratin subsets, then move through their lineage-directed markers. Dr. Folpe catalogs their choices: "We have a couple of lung markers—TTF-1 and surfactant A. We have markers directed against mesothelium, but also sort of work for ovarian cancer, such as WT1. We have a relatively recent marker of enteric tumors, CDX-2. We have a marker of hepatocellular carcinoma, Hep Par 1. For breast, we have a marker called gross cystic disease fluid protein-15." Though there may be one or two others, he says, "the truth is, there aren’t that many more. There really aren’t good markers for gastric cancers, pancreatic cancers, biliary cancers, small intestinal carcinomas; breast cancers that are negative for hormone markers and for gross cystic disease fluid protein-15; lung cancers that are negative for TTF-1 and surfactant A—and about 20 percent of them will be."

That leaves a "pretty good number of tumors that we can’t do a whole lot with," Dr. Folpe continues. That’s why clinical correlation is critical. "That lets you say, ’It’s got a nonspecific phenotype, but that’s OK for this type of tumor.’ In other words, I’m not seeing anything that’s inconsistent with a lung primary, for example. So we’re not raising the possibility of another tumor."

IHC has other limits. Dr. Miller warns of the more common pitfalls. One is relying on so-called predilute, ready-to-use primary antibodies, "which I personally think are very dangerous reagents to use in a diagnostic setting, because in that situation you’re relying on manufacturers to determine the sensitivity and specificity of the diagnostic laboratory’s immunostains, a task which is the laboratory’s responsibility, not that of the manufacturer," he says.

Endogenous biotin artifact, which can occur in tumors with endogenous biotin, can also trip up unwary pathologists. If pathologists use an avidin/biotin-based detection system and fail to block the endogenous biotin, tumors will often light up with unsuspected markers—one imagines an electrifying grand finale capping what until then has been a docile backyard fireworks display.

Pathologists should bear in mind the spectrum of reactivity these markers possess. A new marker is often touted for its specificity, Dr. Miller notes, but "after it gets out in a number of laboratories and is studied for a period of time, it very frequently becomes clear that the antibody is not all that it’s cracked up to be." He sidesteps this problem by using multitumor tissue controls for all his staining. His routine preparations contain some 80 different tumors, he explains. When he brings a new antibody onboard, he optimizes it using those preparations, which lets him stain 80 tumors simultaneously. (He keeps a key to identify the tumors.) He’ll know within a day or two whether the claims for these new antibodies are valid.

While IHC remains the workhorse for identifying unknown primaries, there are times when it’s appropriate to keep it in the barn, so to speak. Patients with extremely short life expectancy may not benefit from having their primary identified, says Dr. Miller.

He’s less inclined to accept the objections of those who think a panel of immunostains is a wasted expense. "There are even pathologists who suffer from something that I call immunohistochemical guilt," he says, who are "inappropriately confident in their ability to interpret H&E sections." Some pathologists do perform like rodeo stars in this arena. "But most of us normal types of pathologists just can’t tell where these tumors are coming from, or oftentimes even what kind of tumor it is. And if you are overconfident at those types of cases, that can lead to problems and missed diagnoses."

Serum markers are often misleading in cases involving unknown primaries, cautions Dr. Brown, who adds that many oncologists nevertheless give them great weight. A woman with a widely metastatic tumor in the abdomen may have significantly elevated levels of CA-125. That doesn’t necessarily indicate a primary ovarian or peritoneal cancer; the rise could be due to irritation of the pelvic peritoneum. Oddly, in many cases there seems to be no correlation between expression of tumor markers on the tumor and the serum level. It’s not unusual to find a patient with an elevated CEA level despite a CEA-negative tumor, or vice versa, Dr. Brown says.

Markers such as PSA remain the exception rather than the rule, adds Dr. Gown. Indeed, PSA has remained a sort of sangraal for investigators hoping to find similar serum markers for breast and lung cancers. The search is far from over, and may still yield results; the advent of molecular techniques to identify gene products makes that prospect even more likely than in the past. On the other hand, these other tumors may be more complex than prostate, upending the promise of serum markers.

There is another possibility for identifying primary sites, though it’s hardly a rousing one: autopsy.

The difficulty with autopsy, says Dr. Brown, is that when the patient dies, he or she typically has a widely disseminated malignancy. "At that point, there is so much tumor burden all over the body, it becomes a guessing game as to which tumor mass in their body might be bigger than another. So in my personal experience, and in the literature as well, autopsy is rarely more successful than the antemortem studies in identifying a primary site."

Dr. Miller hazards that in 65 percent of the cases, he and his colleagues can provide clinicians with either a primary site or a short list. "We can often narrow the list of possibilities substantially," he says.

And when he can’t? Then he turns to face the wall. "I actually have a list there of all the various primary sites for cancers in the body," he says. "And I will specifically address each one of those primary sites in my comments." He’ll begin by noting he can’t identify the primary site, then provide a list of the most likely possibilities—say, pancreas, lung, or stomach. Next, he’s often able to say with conviction that other tumors are extremely unlikely or excluded—say, kidney, adrenal, hepatoma, bladder, or prostate. "Even when I can’t tell them precisely where it’s from, I think we can give them information that will help clinicians direct their search much more efficiently."

The best piece of advice may be, as it so often is, the easiest to follow: make better use of what’s available. No one is denying the importance of the markers developed in recent years. They’re more robust and easier to use, and they’ve allowed pathologists to improve their work considerably. "But we do our best job when we begin the process armed with some clinical information," Dr. Folpe says. "So I would encourage people to wait to do the big workup until after you’ve got some sense from the oncologist of what’s going on. Because then you can be a little bit focused."

Ultimately, however, pathologists, like religious leaders, often face the disquieting prospect of having to respond to unanswerable queries. "We can assign a tumor, in most cases, to the broad category of a carcinoma versus, say, metastatic melanoma or lymphoma," Dr. Brown says. "But within the category of carcinoma, identifying the organ of origin is in many cases not successful." In many cases, he concedes, "We do not definitively answer the question."

That doesn’t sit well with clinicians. "I think the biggest misconception among oncologists is that we can tell in every case where the tumor’s coming from," Dr. Brown says. "It’s a great source of frustration on both sides."

No matter how hard oncologists continue to press for a specific answer, pathologists will continue to fall short. "We’re probably at the limits of what we can do with IHC," says Dr. Brown. Molecular profiling, in his opinion, is the Next Big Thing—the focus of much current research and the likely choice to pull pathologists past the IHC border. But depending on your point of view, microarrays may be both as significant and as everlasting as the Peace of Westphalia.

Dr. Sgroi, who’s done extensive research in gene expression profiling, predicts the method will surpass IHC, if for no other reason than it allows pathologists to look at a much larger number of variables than with IHC. While microarrays are not a shoo-in, many of the problems associated with them have been worked out in recent years, Dr. Sgroi says. And their use in identifying primary tumor sites appears to be a fairly straightforward application. "It’s not completely pie-in-the-sky," he says. Gene expression platforms may also be more cost-effective than IHC workups, he adds.

For all his enthusiasm, Dr. Sgroi is careful to temper his predictions by stating his strong support of IHC, calling it "very, very useful" and noting that he uses it "all the time." But, he says, "I think there are cases in which the gene expression platform may give you a higher probability of figuring out where the tumor comes from."

Dr. Folpe is a bit skeptical. He suggests that the genes being identified with microarray techniques are no different from the proteins that pathologists currently identify using IHC. "There’s not any real magic in what they’re doing. I think they make it sound sexy because it’s ’microarray’—it’s got a certain sort of sexy quality to it."

The premise is certainly reasonable, he says—it makes infinite sense that a lung tumor would have a molecular signature that would differ from the molecular signature of a colon tumor. "What I’m unclear about," says Dr. Folpe, "is whether we can sort through all the noise and all the genes that are up and down and sideways in a met. Can we sort out the important pieces of information to say, ’This is lung and not colon’? My guess is there’s an immense amount of overlap in the gene expression profiles."

Dr. Gown suggests the best bet may be their selective use to identify primary sets of tumors. Even more likely, he says, gene expression array data may be used to identify target genes; from there, appropriate antibodies can then be found. Because IHC is such a powerful technique, blending phenotypic and morphologic information may supplant molecular techniques, as has happened in the case of diffuse large B-cell lymphoma.

No one, least of all patients, is comfortable with unknown primaries. Dr. Folpe is the first to admit, "You’d like to find these things."

But there’s an edge to the obvious. As Dr. Folpe puts it, "In a perfect world, and maybe 30 years from now, when we have chemotherapeutic agents that are perhaps more tissue-specific, it will matter more. But I would say for the most part, at this point, it doesn’t matter much."

He doesn’t stop there. One could make the argument, Dr. Folpe says, that finding the primary may become less, rather than more, important. Chemotherapy is becoming less concerned with tumor type and more focused on protein expression—witness c-Kit and gastrointestinal stromal tumors. There’s also plenty of interest in epidermal growth factor receptor and HER2/neu. "We’re increasingly being asked to look at those in a wide variety of different primary sites, and there again, there’s more interest in whether the tumor has EGFR expression and less interest in whether it’s gastro or lung. Once we’re given the tumor in the lymph node, I can do an immunohistochemical test to see whether the tumor’s positive for EGFR. And if it is, they’ll know how to treat the patient," says Dr. Folpe. "They don’t care where it came from"—not unlike music buffs who download tunes to their iPods without ever owning the disc.

Given all that, it’s not unreasonable to wonder whether making fine primary distinctions will always matter. "I think it could go either way," Dr. Folpe says. Just like many elections.

Karen Titus is CAP TODAY contributing editor and co-managing editor.