Digital Pathology Story Appears in Wall Street Journal

| June 9, 2015

Facebook and a number of folks last night during a bad time for the Chicago Blackhawks made me aware of a piece by Ms. Landro at the Wall Street Journal that touches on a number of issues including mention of use of digital pathology for image analysis as illustrated by work being done at UPMC with GE and Omnyx.  There is an image in the original story showing microcalcifications in a breast section and another image showing Dr. Dabbs at UPMC using digital pathology.

I think this piece is a good news-bad news story well done by WSJ and Ms. Landro. The good news is that it highlights digital pathology, the regulatory issue that remains with FDA and manufacturers and the work that Dr. Nakhleh at Mayo Jacksonville and others working with CAP and ADASP to establish guidelines for necessary internal reviews and the like.

The bad news I think is the story starts out with “Pathologists are the most important doctors that patients never meet.”

Whose fault is that?  Probably pathologists.  Huddled over microscopes for 100 years being locked doors and behind LIS screens will make that happen pretty quickly.  As mentioned before — it is time for us to stop being the “doctor’s doctor” and be the “patient’s doctor”.

The story also mentions the study that keeps on giving – the “JAMA Breast Biopsy Study” with the lead author again being quoted as saying the right diagnosis was about as accurate as “the flip of a coin” referring to the statistical probability of 50% when flipping a coin and the concordance between ADH and DCIS in an enriched study set reviewed by many different pathologists.  The limitations of the study have been described previously here and elsewhere and we recognize there are “gray” areas in pathology, this being one of them but do we have to refer to concordance to “ground truth” as accurate as tossing a coin once or 50 or 100 times?

Lastly, Dr. Jean Simpson, chair of the CAP’s Cancer committee and a pathologist in Nashville, TN says that digital pathology may play a role, but it can’t replicate the pathologist’s trained eye in recognizing subtle cues and patterns to diagnose disease. “There is a mysterious aspect to what we do,” Dr. Simpson says.

Perhaps she was misquoted in the story or perhaps this was really said but I am curious as to what is “mysterious” about what we do when referring to making a diagnosis.  Digital pathology is not replicating what we as pathologists do, it is to replicate and improve upon what the microscope we have been huddled over for more than 100 years can do.

Whole slide images need to be exact replicates of the glass slide – no more, no less.  Perfect “Xerox” copies if you are old enough to remember when photocopies were all “Xerox”.  The whole issue with FDA, the null hypothesis to reject is “Is whole slide imaging inferior to the light microscope?”  Based on numerous studies including the largest and most comprehensive review by Dr. Bauer and colleagues at Cleveland Clinic and several other smaller studies looking at specific diseases, organs, etc…I think we are realizing that whole slide images are equal to, if not better than light microscope images when reflected on the back of the same pathologists’ retinas.

I respectfully disagree with Dr. Simpson and would argue the technology will not replace what we do as practioners but will in some part replace how we do what we do it – perhaps for real-time concordance reviews on say, challenging breast cases with folks from many different locales.

What seems “mysterious” about what we do is the lack of well-defined diagnostic criteria that makes resolving the right diagnosis against a “gold standard” that experts could only agree with 75% of the time.

I bet digital pathology will remove much of the mystery and we as pathologists will not be the most important doctors patients never meet.

New Ways Doctors Reach Agreement on Patient Diagnoses 

Studies show many breast biopsies are misdiagnosed; some hospitals use digital images

Pathologists are the most important doctors that patients never meet. Their expertise is essential to help diagnose disease, figure out how far it has spread, and determine the best treatment options.

They huddle over microscopes much as they have for the last 100 years, peering at slivers of human tissue, cells and bodily fluids on glass slides to unlock the medical mysteries in cells.

There can be significant differences in how pathologists interpret what they see under the microscope or with the naked eye, a growing number of studies show. That can lead to serious consequences for patients if pathologists miss cancer, or conversely, if they interpret a benign tumor as malignant.

New guidelines aim to help pathologists resolve disagreements. Pathology specialists are trying to better define characteristics that determine diagnoses of cancer. Digital technology could help doctors share slides and interpret them more accurately. The goal: to help ensure patients get the most accurate diagnoses and follow-up care.

A study published in March in the Journal of the American Medical Association indicated as many as one in four breast biopsies are initially misdiagnosed. Another study published last year in the Annals of Surgical Oncology of more than 5,000 patients found diagnoses changed in 5% of cases after a report on the same specimen was reviewed by a specialist in melanoma pathology at a melanoma center in Australia. Even when pathologists agreed on a diagnosis, the stage of cancer was changed in 22% of cases after reviews.

Typically, doctors who are evaluating patients for cancer or other diseases send tissue samples from a biopsy to a pathology lab, either in a hospital or at a commercial lab that contracts with health-care providers. Pathologists, who complete an additional three to four years of training after medical school, have different areas of expertise in particular diseases. They are often present during surgery to evaluate specimens removed during procedures. After preparing slides and examining them microscopically, pathologists render a diagnosis, often in consultation with others in their lab.

For many years there have been no formal policies to resolve how to address varying pathologists’ opinions about a patient’s diagnosis. A panel of experts convened by the College of American Pathologists and the Association of Directors of Anatomic and Surgical Pathology issued new guidelines earlier this year, citing studies that show disagreements among pathologists and errors can happen in more than 10% of cases. They ask pathology labs to develop standard procedures to review cases and resolve disagreements, ideally before sending results back to referring doctors.

Pathology labs do conduct internal reviews when cases “fall into the gray zone and are not clearly benign or malignant,” and pathologists often call on colleagues who may have more expertise in a certain type of cancer, says Raouf Nakhleh, a pathologist at the Mayo Clinic in Jacksonville, Fla., and co-author of the guidelines. They suggest doctors hold conferences in labs and standardize criteria for cases where pathologists often have differing interpretations such as some lesions detected in the thyroid and esophagus.

Hardeep Singh, a patient safety researcher at the Michael E. DeBakey VA Medical Center and Baylor College of Medicine in Houston, says the entire system of specimen collection, processing and interpretation needs to be improved. He says pathologists should be part of the medical team and “not siloed as much from the rest of patient care.”

Breast cancer pathology is emerging as a concern. For example, for the March study in JAMA 115 pathologists in eight states reviewed 60 breast biopsy test sets, out of a total of 240 cases. They disagreed with expert opinions in 25% of cases. The highest rate of disagreement was for a precancerous condition known as ductal carcinoma in situ, or DCIS, and for abnormal cells known as atypia, which don’t always progress to DCIS or cancer. In the case of abnormal cells, pathologists correctly diagnosed less than half of cases, which is “like the flip of a coin,” says Joann G. Elmore, lead author of the study and a professor in the Department of Medicine at the University of Washington in Seattle.

Dr. Elmore says over-interpreting biopsies can lead women whose condition isn’t serious to have unnecessary follow-up surgery, radiation and hormone therapy, while missing a more serious case can lead to the development of invasive cancer down the road.

Marisa Weiss, founder of the nonprofit Breastcancer.org patient education site says patients should request a copy of their pathology report, ask whether it has been reviewed by a second doctor, and seek a second opinion if they have concerns about its accuracy. Dr. Weiss is also a radiation oncologist who relies on pathology reports to make treatment decisions.

To help patients understand their pathology reports, the College of American Pathologists offers a primer on how to read them and a website, MyBiopsy.org with information by cancer type on how pathologists make a diagnosis and what they are looking for in a biopsy sample.

Erika Fleck, 73, had a biopsy after a mammogram last December that was read by a pathologist as ductal carcinoma in situ. Because of concerns about the high rate of misdiagnosis, the surgeon recommended that she seek a second opinion at another hospital, Magee-Womens Hospital of UPMC, part of the University of Pittsburgh Medical Center. Pathologists who reviewed the results there diagnosed a condition known as atypical papilloma, a benign lesion in the breast that can be associated with an increased risk for breast cancer.

Ms. Fleck, a retired physical therapist who works part time in a school for disabled children, says having the second opinion was “huge.” She had the lesion surgically removed, which confirmed it wasn’t the precancerous condition first diagnosed. She avoided having to consider radiation therapy, but she was glad to have the lesion removed because of the risk “it could turn out to be something.”

UPMC and other medical centers around the country are studying the use of new digital technology that scans slides into a computer to create images that can be shared and compared electronically. Software used by such systems can help quantify biomarkers that are important for diagnosing a tumor, such as the protein HER2 in breast cancer, according to David Dabbs, chief of the department of pathology at Magee-Womens Hospital.

Currently, digital pathology is approved for primary and secondary diagnosis in Europe, but not yet in the U.S., where the Food and Drug Administration is reviewing whether the images are as good as physical slides of pathology specimens. UPMC and GE Healthcare, which have a joint venture called Omnyx, are currently cleared by the FDA to use Omnyx’s digital pathology system to review digital images of breast cancer tissue of patients being considered for the treatment Herceptin.

Jean Simpson, chair of the College of American Pathology’s Cancer committee and a pathologist in Nashville, TN says that digital pathology may play a role, but it can’t replicate the pathologist’s trained eye in recognizing subtle cues and patterns to diagnose disease. “There is a mysterious aspect to what we do,” Dr. Simpson says.

Write to Laura Landro at laura.landro@wsj.com

Corrections & Amplifications

An earlier edition of this story incorrectly said Dr. Raouf Nakhleh is a pathologist at the Mayo Clinic in Rochester, Minn. He is a pathologist at the Mayo Clinic in Jacksonville, Fla.

Source: Wall Street Journal

Tags: , , , , , , , , , , ,

Category: Advocacy, Anatomic Pathology, Clinical Laboratories, Current Affairs, Digital Pathology News, Education, General Healthcare News, Image Analysis, Laboratory Informatics, Laboratory Management & Operations, Microscopy, Pathology News, Patient Advocacy, Standards and Guidelines, TeleHealth, Web/Tech, Whole slide

Comments (1)

Trackback URL | Comments RSS Feed

  1. Mark Clymer says:

    Keith,
    I was looking forward to reading your perspective on the WSJ article, and I’d like to add my two (OK, three) cents.

    Despite advances in technology, do WSI scanners really deliver a “perfect ‘Xerox’” of the specimen? If so, then why aren’t the images identical between scanners? For all of the published articles demonstrating the “equivalence” of digital images to “brass and glass” observation, there are plenty of examples illustrating that the microscope continues to be the benchmark. Journals tend to publish articles describing scientific advancements, not the shortcomings. Yes, significant progress has been made, we’re getting closer to the reality of digital pathology and, I believe, the pending FDA guidance on technical performance assessment in slide scanning will only strengthen the argument for digital pathology in the US.

    A second observation is that digital images must be displayed to the pathologist’s retina. For the moment, monitor quality, performance, calibration and even color profile remain topics for debate. A visit to a consumer electronics store illustrates the wide range of inconsistency in monitor performance, and how do we know which display is accurate? Until standards are identified and adopted, pathologists will continue to view images on displays that vary in color and tonal reproduction.

    Third, in regard to the patient mentioned in the WSJ article who sought a second opinion, it is wonderful that digital images enable consultation, but doesn’t this become a slippery slope? Patients may not “like” or accept a first opinion and may continue to seek alternatives until they obtain an opinion that is satisfactory to them. The same could apply to insurers who would prefer a diagnosis with a lower cost for treatment, thus assuming the risk that the first treatment may be unsuccessful and follow up intervention may be required. And who pays for the pathologists’ time? Where digital pathology should contribute to a virtuous circle for patient care, it could quickly spiral out of control.