Dubious Diagnoses?

The messages are everywhere around us – campaigns for breast cancer screening, for prostate cancer testing, for gynecological exams, for colonoscopies. Patients are warned to stay ahead of the potential risks by making sure they have regular checks. Though there’s plenty of dissent in the medical community about cancer screening – who should get it, when it should be done, which methods are most reliable – it’s hard to deny that early testing can save lives. But how reliable is this screening, and how much does success depend on the pathologist behind the test?

Diagnostic disagreements

A recent report in JAMA has drawn a lot of attention for its investigation of diagnostic concordance between pathologists interpreting breast biopsy specimens (1). The authors of the paper attempted to quantify the degree of disagreement between diagnoses provided by different pathologists for the same specimens. To do so, they generated a set of 240 excisional or core needle breast biopsy specimens randomly selected from pathology registries affiliated with the Breast Cancer Surveillance Consortium. From each biopsy, new slides were prepared in a single laboratory for consistency, and the best of those slides was selected by consensus panel for inclusion in the set of test cases. Specimens exhibiting atypia and ductal carcinoma in situ (DCIS) were oversampled, as were cases from women either in the 40-49 age category or with mammographically dense breast tissue. These types of samples were emphasized because age and breast density are key risk factors for both benign breast disease and cancer, and because atypia and DCIS are often more difficult to diagnose or appear “borderline” between multiple diagnostic categories – so the researchers predicted that there would be more discordance between different pathologists’ conclusions.

The cases were first reviewed by a panel of three experienced, internationally recognized pathologists who were blinded both to previous interpretations and to one another’s conclusions; at this point, the pathologists were in unanimous agreement on the diagnosis of 75 percent of cases. After resolving the remainder of cases by consensus, the 240 slides were randomly divided into four test sets and distributed to pathologists in the United States (all of whom had at least one year of experience interpreting breast specimens and intended to continue for at least one additional year) for interpretation. Slide examples for each diagnostic category can be seen in Figure 1. The invited participants received one hematoxylin and eosin-stained slide for each case, as well as information on the type of biopsy and the patient’s age. They were not given any diagnostic definitions or specific instructions; rather, they were asked to evaluate the cases as they would in their standard laboratory practice.

Of the 6,900 total interpretations in the study (115 pathologists reviewing 60 cases each), participants agreed with the consensus reference diagnosis 75.3 percent of the time – but the agreement varied widely depending on the type of specimen. Concordance for invasive breast cancer cases, for instance, was high (96 percent), but low (48 percent) for atypia (Figure 2). Discordance was higher when pathologists reported that a case was “difficult” or “borderline,” requested second opinions, or lacked confidence in their own assessments.

A key finding in the JAMA study is the variability of pathologist interpretations. Though the danger of underinterpreting a case is clear, it’s less obvious how overinterpretation (which, in the study, occurred in 3 percent of DCIS cases, 17 percent of atypia cases, and 13 percent of benign breast tissue specimens) can cause harm. Unnecessary medical intervention, additional incursion of costs and the possible detrimental psychological effects of a cancer diagnosis are all potential consequences. But knowing the level of discordance between pathologists and the risks associated with overinterpretation may cause anxiety and uncertainty among women who undergo breast cancer screening. It may even discourage them from receiving the appropriate monitoring (see “Prompting patient participation”). But what exactly is “appropriate” monitoring – and who determines how much is too much?

Testing: who, what, and when?

Some suggest it should be the physicians and researchers with the greatest knowledge and understanding who determine “appropriate” levels of testing – but even when offered the chance to reduce the amount of unnecessary screening, practitioners may not take it. Acknowledging estimates that as much as 30 percent of healthcare spending may not go toward improving patient health, a study published in JAMA Oncology (2) investigated regional imaging rates for both breast and prostate cancers. The researchers conducted a retrospective cohort study to look at how closely the American Society of Clinical Oncology (ASCO) recommendations against imaging to stage cancer in patients with low-risk disease were followed.

Patients were considered to have low-risk breast cancer if their disease was in situ, stage I, or stage II; low-risk prostate cancer included patients with stage T1c or T2a disease, a Gleason score of six or less, and a PSA level below 10 ng/mL (see The Pathologist’s previous feature on the value of PSA testing, (3)). ASCO’s Choosing Wisely guidelines (4) recommend against positron emission tomography (PET), computed tomography (CT), or radionuclide bone scans for patients meeting these criteria – but physicians continue to order them anyway. In the JAMA Oncology study, the overall rate of inappropriate imaging for breast cancer was 41.8 percent, while for prostate cancer it was 44.4 percent. There are a number of factors that might cause a clinical practitioner to order a test that isn’t recommended; for instance, a doctor might take a “better safe than sorry” approach by preference, a patient might request imaging, or fears of malpractice might prompt an unnecessarily high level of caution.

It’s true that best practices for care in staging cancers are different to those in screening for them, but in both cases, clinicians may be ordering unnecessary tests, and neither patients nor practitioners benefit from the results. Tests that aren’t needed for medical care cost doctors’ time, healthcare systems money, and patients their peace of mind – especially when the people responsible for reading the scans might disagree in their interpretations.

Prompting patient participation

Not everyone wants to put the decision-making power in the doctors’ hands. Some advocate for patients’ rights to choose whether or not they want to be screened for cancer. And in fact, a first-of-its-kind study in The Lancet recently found that women who are better educated about the risks of breast cancer screening are less likely to want to take part in screening programs (5). The randomized controlled trial involved 879 Australian women aged 48 to 50 (the age at which screening commonly begins in Australia and in other countries with similar programs). Those given information on overdetection were found to have a less favorable attitude toward breast screening and, as a result, significantly fewer intended to be screened when compared with controls. Nonetheless, overall attitudes toward screening remained positive. The majority of the women also reported that they had not been aware of the facts surrounding overinterpretation before the study.

“Recent international reviews have called for better, more balanced information to be provided to women when they are invited to breast screening”, said Jolyn Hersch, lead author of the Lancet study. She added, “Overdetection is considered one of the most important downsides, but most women are unaware of the risk. Screening can detect inconsequential breast cancers, leading to overdiagnosis and overtreatment. And this treatment can include surgery, radiotherapy, hormone therapy, and chemotherapy, all of which have side effects.” An article published in JAMA Internal Medicine shows that most patients overestimate the benefit of interventions, but underestimate the potential risk (6) – but studies like the JAMA and Lancet ones regarding breast cancer screening seek to change that.

And it seems that the message is being heard. Recent statistics from the UK’s National Health Service show that the number of women attending breast screening in the country has fallen for the third year in a row. Media attention given to the hazards of screening (7,8) could be partly to blame. The Lancet study’s authors believe that patients should be provided with all of the information, so that they can make their own decisions about screening. Hersch explains that the current breast screening programs aren’t perfect, and can do both harm and good. “The national breast cancer screening program in Australia states that its aim is to reduce illness and death from breast cancer. The evidence suggests that breast screening does lower the number of women who die of breast cancer, but whether it reduces illness overall is questionable, because of the effects of overdetection. This is why it is so important to give women evidence-based, accessible information, so that they can decide what is best for them personally.”

Educating for empowerment

There’s no question that patients should have the final word on their own health care. But unfortunately, it isn’t as straightforward as educating them about the benefits and risks so that they can make informed choices. As every physician – and especially pathologist – knows, medicine is as much an art as it is a science, and evaluating the pros and cons of breast cancer screening isn’t as simple as adding and subtracting the evidence.

Given the media attention the new JAMA study on diagnostic discordance has received, it’s reasonable to worry about the effect on patients. In an editorial, Nancy Davidson (University of Pittsburgh Cancer Center, PA, USA) and David Rimm (Yale University School of Medicine, New Haven, CT, USA) wrote, “An undesirable short-term outcome from the study by Elmore et al. will undoubtedly be heightened anxiety among women who undergo breast biopsy and concern among their physicians about the accuracy of the pathologic diagnosis (9).” Patients may opt out of screening even when it would be advisable, or may place less faith in the diagnoses reached by pathologists, prompting second opinions and additional testing that burden the healthcare system even more.

That isn’t the way Davidson and Rimm think the results should be interpreted, though. Their editorial continues, “This study confirms that the majority of diagnoses […] are readily and accurately made by practicing pathologists.” They agree that cancer screening isn’t a perfect solution – the JAMA study identifies areas where process improvements are needed, and that there are ambiguous cases in which a second opinion would be valuable. But patient anxiety, and decision-making based on that anxiety, may not be necessary. After all, the study was conducted under conditions quite different to those in a pathologist’s daily work – only one slide per specimen, no second opinions or outside consultation, no requests for additional tissue, and no clinical information or imaging findings other than the patient’s age. Even the caseload was unrealistic, with large numbers of slides showing atypical hyperplasia and DCIS – borderline situations that comprise only a small fraction of those seen in day-to-day practice (see “Not Reflective of Clinical Practice”).

Michael Misialek, associate chair of pathology at Newton-Wellesley Hospital (Newton, MA, USA) writes, “While the study’s findings may not be surprising to physicians who understand the challenges of diagnosing complex breast cases, news of the article could lead to unnecessarily heightened anxiety for patients and the public as breast cancer is a highly publicized and pervasive disease (10).” He adds, “The study confirmed that the majority of breast pathology diagnoses, especially at either end of the spectrum (benign without atypia and invasive breast cancer) are readily and accurately made by practicing pathologists regardless of practice setting.” (See “Collaborating on Cancer Care.”) Much like in the Lancet paper about overdiagnosis, it’s most important to ensure that patients not only have, but understand, all of the relevant information about diagnostic concordance before making treatment decisions that can significantly affect their lives.

The source of the confusion

Oscar Bronsther, CEO and CMO of MetaStat, Inc. (Boston, MA, USA), feels that we need to do better than we have so far to treat patients as precisely as possible – not just in terms of providing better information, but in diagnostic methodology itself. In response to the JAMA study, he says, “As the disparity in diagnoses reveal, relying on traditional approaches to diagnosing cancer can lead to clinical mistakes, especially in premalignant cases. And if experts can’t even agree on what cancer looks like under the microscope, they surely can’t understand the underlying biology—and whether a specific cancer will become invasive.”

Bronsther explains that, in his opinion, the differences in diagnoses can be traced to the “grey zone” that sometimes exists between normal and malignant results. Although both groups of individuals who examined the slides in the JAMA study – community and academic pathologists – have substantial experience analyzing biopsy results, there’s still a measure of subjective judgment involved. Though the definition and morphological criteria of cancer have not changed for decades, it’s inevitable that, in the absence of more detailed analysis, opinions among the two groups regarding specific slides will vary. Nevertheless, he says, “the disparity is unsettling.”

There is an understandably high degree of interest in the earlier diagnosis of breast cancer and other malignancies. After all, that’s why current screening programs were established – on the basis that cancers caught earlier can be addressed with cheaper, less invasive treatments. But it isn’t enough simply to detect malignant tumors. Increasing the number of cancers we are able to identify without increasing our understanding of their clinical significance has the potential to result in significant overtreatment. Bronsther cautions against delivering a life-changing diagnosis such as breast cancer without all of the facts. “It is more important that we understand the metastatic potential of a tumor rather than just labeling a lesion a cancer,” he says. “A tumor with little metastatic potential is very different from a lesion with significant metastatic potential.”

Next-generation solutions

The uncertainty surrounding breast cancer screening and biopsy highlights the potential for next-generation diagnostics, which can offer more precise and personalized diagnoses. Molecular and epigenetic tests not only maximize the amount of data obtained, but can also give patients more refined knowledge of the specific challenges they face based on their diagnoses – information Bronsther says “is crucial, because once you tell them they ‘have cancer,’ you’ve frightened them and changed their life.” He’s optimistic that new kinds of diagnostics can lead to more effective, targeted therapies for specific subsets of cancer patients, as well as to increased savings for hospitals and patients.

Next-generation diagnostic tests have already been developed for the prediction of metastasis in women with breast cancer. Immunohistochemistry-based tests performed on biopsy tissue can tag the active sites of metastasis development; one test, for example, targets the three-cell structures necessary for metastasis (endothelial cells, perivascular macrophages and tumor cells expressing a particular chemotaxis protein) with a triple stain, and has even been able to determine whether a woman with newly diagnosed breast cancer is among the 35 percent likely to experience metastatic cancer or among those with breast tumors for whom metastasis is unlikely (11). Other diagnostic platforms quantify metastatic risk by measuring levels of prognostic markers – for instance the Mena protein, which promotes the actin polymerization and protein interaction necessary for cell migration. Tests like those are applicable not only to breast cancer, but also to other epithelial cell tumors such as colorectal, prostate and lung cancer.

“Cancer is complicated and messy,” Bronsther says, “and relying on traditional criteria to establish a diagnosis potentially leads to incorrect approaches to the needs of individual patients.” Better diagnostic tests will provide patients and their doctors with detailed, personalized information to accelerate the delivery of tailored cancer therapy. Next-generation tests are moving away from a morphological diagnostic approach to a more molecular one, a change that researchers hope will result in better outcomes for patients and significant savings for the healthcare system.

There’s still work to be done

Regardless of the approach taken, it seems clear that pathologists and patients have some work to do. While the diagnostic discordance noted in the JAMA study is not, in fact, as concerning as the media hype would have us believe, there’s certainly a need for better definitions of the various categories, and for better quality management of borderline and high-risk cases. Over- and underinterpretation are more worrying, because they may lead to unnecessary treatment or to missed diagnoses, respectively – and, as the JAMA Oncology report attested, controversy still exists about the role of increased monitoring in women with atypical biopsy results. Ideally, pathologists will embrace the idea of improving definitions and processes, consulting second opinions where necessary, and making use of next-generation testing in situations where it can offer additional insight into disease. And in the meantime, patients should be given the information and education necessary to understand the treatment decisions they, and their doctors, are making.