|Dr. Kurtin and colleagues worked with hematologists to improve test use. “We wanted to have a data-driven algorithm,” he says, “rather than something that was emotional on our part or emotional on the clinician’s part.”
Not that he’s applying to be an understudy for Ted Allen on “Chopped.” But if Paul J. Kurtin, MD, won’t be tossing chefs from the kitchen anytime soon, he shows a keen eye when ridding another domain—this one at Mayo Clinic—of what doesn’t work.
Dr. Kurtin, of Mayo’s Division of Hematopathology, talked about diagnostically efficient, cost-effective test use for hematologic malignancies from blood and bone marrow specimens, at the AACC meeting in Los Angeles in July. As it turns out, pathologists have plenty of ways to reduce unnecessary testing, get the right answers, and save money. When improving which tests are ordered, he says, working closely with clinicians is best.
You start, Dr. Kurtin suggested, by finding what doesn’t work. “Look at your practice—where are the tests not effective? We had a sense of what wasn’t diagnostically efficient.”
The tests can be complicated, so clinicians may inadvertently order the wrong tests. Clinicians also tend, like Tolstoy, to be quite comprehensive. They generate a vast differential diagnosis, Dr. Kurtin said, then reasonably feel they have to address all the entities with a test. “If a bone marrow is one of those things, and if a pathologist looks at the morphology of the bone marrow, that differential diagnosis goes from huge to small, very, very fast.” The problem comes when ancillary studies are done reflexively.
These tests don’t always add useful information, and they can be misleading. If a clinician orders a T cell receptor gene rearrangement on a patient who ends up having a myeloid disorder, for example, and that T cell receptor gene rearrangement suggests that there was clonal T cell population in the patient, “You’ve got another problem now,” said Dr. Kurtin. Is something amiss in the patient’s T cells? Or is this an elderly patient with a false-positive due to immune senescence? “These are the kinds of issues that get involved when you have too many tests brought to bear on the clinical problems.”
Then there’s cost—there’s nothing cheap about these tests. “We have a fiduciary responsibility to our patients to make sure that we don’t bankrupt them by ordering too many tests.” Dr. Kurtin made the same argument on behalf of labs. Tests generate revenue, but they also create costs, in terms of space, equipment, and personnel. And if reimbursement comes up short—and it does—those costs are borne by the lab.
Mulling over these issues, Dr. Kurtin and his colleagues in the hematopathology division wondered how they might create a more collaborative relationship with the hematologists who order these tests. “Could we develop test algorithms that the pathologists would help to apply in the context of evaluations of bone marrows and blood specimens from hematology disease patients?” he asked.
Their vision for what they wanted was clear: The hematologist would evaluate the patient, generate a differential diagnosis, and order tests; the pathologist would look at the morphology of the blood and bone marrow specimen and revise the differential diagnosis based on what the morphology showed. “Reasonably so. There is value in morphology,” said Dr. Kurtin.
Then, guided by an algorithm, the pathologist would refine the clinician’s orders. You heard that correctly—&38220;change the orders that the clinician had originally made to get to the heart of the diagnostic problem,” Dr. Kurtin said. The “extra stuff,” as he called it, would, in essence, be chopped.
There was no talk of cutting out the clinicians. “We had to connect with our clinicians. They knew stuff,” Dr. Kurtin said. “We needed their expertise to help us develop these algorithms. They had experience with taking care of patients. They knew what patients demanded.” If pathologists were going to change physician test orders, “We certainly needed their endorsement to make that intervention,” he said.
To develop new testing algorithms, Dr. Kurtin and his colleagues followed a method that he said can be applied to any kind of testing.
First, they looked at their practice generally, and, based on their experience, identified possible targets for better test use.
Next, they reviewed their practice data as well as the literature to refine those hunches and make recommendations. “We wanted to have a data-driven algorithm rather than something that was emotional on our part or emotional on the clinician’s part.” That made the next step easier: taking their recommendations to the clinicians and garnering consensus.
Once final guidelines were approved, they designed ways to implement the algorithms—an important step, said Dr. Kurtin. “We didn’t want them to be based on individual practitioners; we wanted them to be automatic as much as possible.”
After the algorithms were put in place, the lab audited them to make sure they were right—and adjusted them if they weren’t.
All these principles have been applied to creating algorithms for lymphoma staging bone marrows, chronic myeloid leukemia, myeloproliferative neoplasms, myelodysplastic syndromes, plasma cell disorders, and acute promyelocytic leukemia. Coming soon (possibly by year’s end): revised approaches for acute leukemia, chronic lymphocytic leukemia, and eosinophil/mast cell disorders. (The algorithms are at mayomedicallaboratories.com, under “test utilization” and “algorithms.”)
They began with bone marrow cytogenetic evaluation for lymphoma staging.
The typical workup in these bone marrows has been to look at the morphology, to do flow cytometry, and to do cytogenetics with or without FISH testing, Dr. Kurtin said. Because the literature lacked data about the utility of cytogenetics or FISH testing in this context, he and his colleagues did their own study.
The assumption was that lymphoma type was established on the basis of a tissue other than bone marrow and that bone marrow was performed for staging, not for a primary diagnosis. The questions they wanted to answer were: Does routine karyotyping of lymphoma staging bone marrows add information about presence or absence of lymphoma? About lymphoma classification? About other diseases that might concurrently affect the bone marrow?
They started with 574 patients who had a lymphoma diagnosis during a three-month period at Mayo Clinic; 298 of them had a contemporaneous bone marrow done for staging purposes. Of those 298, 112 specimens had cytogenetics performed on them.
In the cases where cytogenetics was performed on bone marrow, 41 of the specimens were involved by malignant lymphoma by morphology and phenotyping. In 32 of them, the genetics results were normal. “So the genetic studies did not have good sensitivity to pick up the presence of lymphoma in the bone marrow,” Dr. Kurtin said. In the remaining nine cases, the genetics were abnormal—but they didn’t provide new information about the lymphomas. The abnormalities were expected, based on the lymphoma type. “So, for example, if the patient had follicular lymphoma, there was a 14;18 translocation in the bone marrow. If the patient had mantle cell lymphoma, there was an 11;14 translocation in the bone marrow.”
There were 71 cases that were negative for malignant lymphoma, eight of which had abnormal genetics. None of them had lymphoma involvement, and none of them had a karyotype that suggested a lymphoma.
In three of the patients, the bone marrows were normal. They had single metaphase abnormalities, which, Dr. Kurtin reminded his audience, do not meet the genetic definition of a clone, and thus cannot be used diagnostically.
Three other patients were elderly males who had lost chromosome Y in their bone marrows. It’s a common finding related to aging in males and has no pathogenic significance.
One patient, with a follicular lymphoma, did not have lymphoma in the bone marrow, but had been heavily treated for follicular lymphoma in the past and had developed acute myeloid leukemia—the abnormal karyotype, add(5)(q13) and -7, was related to that, Dr. Kurtin said. The last patient, with diffuse large B-cell lymphoma, had a morphologically normal bone marrow. The cytogenetic abnormality was a constitutional abnormality unrelated to the disease.
There were eight other patients who had other abnormalities noted by morphology that suggested an underlying myeloid disorderâ€”not a lymphocytic disorder. In five of the patients the karyotypes were normal.
Mayo’s practice data showed:
- cytogenetics in staging bone marrows does not improve sensitivity for lymphoma diagnosis over and above morphology and phenotyping.
- cytogenetics does not add useful additional data in lymphoma positive bone marrows.
- normal bone marrow morphology is sometimes associated with abnormal cytogenetics, but the abnormalities are not clinically significant.
- if there is a specific myeloid abnormality (acute myeloid leukemia, myelodysplasia, myeloproliferative neoplasms), suggested by morphology, cytogenetics can help determine the significance.
- in heavily treated patients, those who have genetic abnormalities in their bone marrows prior to auto-bone marrow transplant “have an awful clinical course after bone marrow transplant, because their genetically abnormal bone marrows don’t engraft. And so pretransplant we routinely do cytogenetics to exclude morphologically occult therapy-related myelodysplasias prior to reinfusing those bone marrows from these patients,” said Dr. Kurtin.
Based on their study, “We’re only going to do cytogenetics in those cases where there seems to be a good indication for them. And the indications are myeloid problems, not lymphoid problems, except for bone marrow transplant.”
Spurred by this success, the lab decided to do the same thing for lymphoma FISH panels. In their practice data, Dr. Kurtin and colleagues drew conclusions similar to those for cytogenetics.
Coupled with morphology and phenotyping on blood specimens, he said, “FISH sometimes helps define lymphoma types in these specimens when you’re not going to do a tissue biopsy other than bone marrow.” Other than that, however, it’s insensitive for detecting lymphoma in staging bone marrow specimens, and it rarely adds information about lymphoma classification in staging specimens. So, the lab does not do FISH routinely on staging bone marrows for patients with malignant lymphomas.
Other common tests for patients with malignant lymphoma are immune receptor gene rearrangements, either T cell receptor gene rearrangements or immunoglobulin gene rearrangements.
The former are usually done by PCR on blood or bone marrow specimens. A clonal pattern is present in 90 percent of patients with T cell lineage malignant lymphomas and leukemias; however, it’s also present in a fairly high percentage of patients with reactive conditions, “particularly if you study the blood, the bone marrow, or the skin,” Dr. Kurtin said. In the blood and the bone marrow, patients with acute viral infections, such as infectious mononucleosis, adenovirus, hepatitis B, and HIV, can all have clonal T cell receptor gene rearrangements.
Moreover, age-related immunosuppression renders individuals chimeras for subsets of T cells. “You lose T cell repertory,” he said. When labs do T cell receptor gene rearrangements by PCR in those patients, it’s not uncommon to pick up clones that look like malignant clones, but are actually related to the normal, aforementioned aging process. As if the ongoing Medicare debate weren’t scary enough, “Once you get past 65, you have a 10 percent incidence of having clonal T cell receptor gene rearrangements detected in your peripheral blood,” he said, “even though you’re completely normal.”
Nonclonal pattern is, most of the time, associated with nonmalignant lymphoid hyperplasia. There is a small rate (10 percent) of false-negatives for T cell lymphomas, he cautioned.
With immunoglobulin gene rearrangements, he said, “You see a clonal pattern in 90 percent of B cell lineage malignant lymphomas and leukemia.” A clonal pattern is quite rare in reactive conditions, and nonclonal patterns are associated with lymphoid hyperplasias.
While T cell receptor gene rearrangements suffer from high sensitivity and low specificity, he said, immunoglobulin gene rearrangements have fairly high specificity. Alas, in many instances they are—like Florida in this year’s presidential election—irrelevant, because there are faster, cheaper methods of looking for clonal B cell populations, including flow cytometry and immunohistochemistry.
This has been borne out in Mayo’s practice data. In immunoglobulin gene rearrangements ordered on 54 specimens (seven on peripheral blood, 47 on bone marrow), “It appeared to us that all of them were unnecessary.” In some cases, the tests were ordered in a context where there did not seem to be an underlying B cell disorder; in others, the lab already knew the B lymphocytes were clonal, thanks to flow cytometry.
For T cell receptor gene rearrangements in the peripheral blood, Dr. Kurtin and colleagues looked at 249 cases. This time, 171 appeared unnecessary. There were also 25 probable false-positives—exactly the 10 percent incidence that would be expected in elderly patients, he reminded his audience.
What about in bone marrow specimens? Of 172 cases, 136 T cell receptor gene rearrangement studies were unnecessary, he said, with 19 probable false-positives and 12 probable false-negatives.
No use in squirreling away these data. Dr. Kurtin and his lab colleagues shared their findings with their clinicians, and the response of one of the senior hematologists was striking:
“These tests are dangerous for patient care,” he told the lab. “Why do you let us order them?”
After a telling pause in his AACC talk, Dr. Kurtin went on to say, “So our recommendation for T cell and immunoglobulin gene rearrangement is that they’re poor screening modalities.” The lab uses them only to resolve specific diagnostic problems posed by morphology and phenotyping.
The algorithm puts the reviewing hematopathologist in the driver’s seat. Mayo’s lymphoma disease oriented group approved several principles regarding test use:
- The reviewing hematopathologist needs to approve cytogenetic test requests for bone marrow evaluations for lymphoma.
- Ditto for lymphoma FISH tests (which can also be approved by a geneticist).
- A hematopathologist needs to approve all immunoglobulin and T cell receptor gene rearrangement requests on blood and bone marrow.
Next up: improving test use for chronic myeloproliferative neoplasms. In this case, there’s plenty of literature on the topic, Dr. Kurtin said.
With numerous genetic tests available, what’s the best way for labs to “string them together,” as Dr. Kurtin put it, to arrive at a diagnosis?
In response, he offered the case of a 71-year-old male with fatigue, and with no hepatosplenomegaly. He was anemic, with normal MCV, WBC, and platelet count, and elevated LDH.
A blood smear revealed abnormal teardrop-shaped red blood cells, which usually indicates fibrosis in the bone marrow, as well as circulating nucleated red blood cells. “You shouldn’t have the rookies in the blood,” Dr. Kurtin said. “You should just have the mature red cells in the blood.” The smear also reveals other “rookies”—circulating left-shifted granulocyte precursors—as well as small megakaryocytes.
What was going on? “It looks like the bone marrow is being replaced by something.”
The test menu is extensive:
- Bone marrow aspirate/biopsy
- Cytogenetic analysis
- JAK2 V617F mutation analysis
- JAK2 Exon 12 sequencing
- cMPL Exon 10 sequencing
- BCR-ABL FISH
- BCR-ABL PCR dx screen
- BCR-ABL p190 quant PCR
- BCR-ABL p210 quant PCR
- MDS FISH
- KIT D816V mutation analysis
- CHIC 2 (FIP1L1/PDGFR alpha deletion FISH
So was the physician’s order, which included the JAK2 V617F mutation analysis on both blood and bone marrow, and most of the other assays on this list.
This came with a hefty price tag. “That is about $5,000 worth of genetic tests,” Dr. Kurtin said.
He and his colleagues bore down. Their first question focused on the aforementioned JAK2 analyses on bone and blood. “If you go to the factory, do you get a higher rate of JAK2 mutations than if you look at the blood? And so we looked at that.”
As part of a wider utilization study, they found 267 patients with concurrent blood and bone marrows on which they did JAK2 V617F mutations by allele-specific PCR. “In all but four cases, we got concordant results. This doesn’t speak too strongly to test both.” In the four cases with discrepant results, the mutation burden was at the very lower limit of detection of the assay. The bone marrow diagnoses in these cases included one obvious myeloproliferative neoplasm, and three were morphologically normal. Finally, he says, “It’s well known that you can get very, very low level V617F mutations in normal individuals.”
The discordant results meant no change in diagnosis, management, or outcome. That left only one plausible action. “Cancel duplicate tests,” Dr. Kurtin intoned. The lab now strikes duplicate JAK2 tests if theyâ€™re ordered on both blood and bone marrow. Had they done that in these cases, he said, they would have saved some $100,000 in patient charges.
In polycythemia vera, V617F aren’t the only mutations in play; other mutations, in Exon 12 of JAK2, have a low rate (five percent) of positivity in these patients as well. Dr. Kurtin suggested labs might want to screen with allele specific PCR for V617F, and do Sanger sequencing for Exon 12 if PV is still suspected in cases that are negative for JAK2 V617F. Sanger sequencing is less sensitive than allele-specific PCR, he cautioned, and should not be used as a JAK2 V617F mutation screening assay.
As for MPL Exon 10 sequencing, mutations are present in less than 10 percent of patients with either essential thrombocythemia or primary myelofibrosis. Nor will you see them if JAK2 mutations are present. Thus, they’re not useful in straightforward cases of classic chronic myeloproliferative neoplasms, or CMPN, which, as Dr. Kurtin reminded his audience, are easily recognized by clinical findings and bone marrow morphology.
But don’t drop them completely, he warned. “They are useful when there’s a clinical suspicion of a myeloproliferative neoplasm and the bone marrow morphology is equivocal.”
After looking at the literature, Dr. Kurtin and his colleagues determined that for the first pass, the appropriate initial tests for patients with suspected CMPN were: bone marrow aspirate and biopsy; chromosome analysis; a BCR/ABL FISH or BCR/ABL PCR to quickly confirm or exclude CML; and JAK2 V617F (blood or bone marrow, but not both). “And then follow the algorithm.”
“Once you do this,” Dr. Kurtin promised, “you’ll save zillions of dollars.”
Dr. Kurtin’s third example covered FISH studies in evaluating myelodysplastic syndromes.
Why even bring this up? Answering his own question, Dr. Kurtin said it has become routine practice to do cytogenetic and FISH analysis to look for genetic abnormalities common in these syndromes. FISH is thought to be more sensitive, but Dr. Kurtin and his colleagues wanted to test that assumption, given FISH is costly and time-consuming. If a routine karyotype could give the same information, so much the better.
There are recurring cytogenetic abnormalities in MDS that can readily be detected by FISH, but they have relatively low incidence. Nonetheless, because some have prognostic significance, theyâ€™re important to recognize. Mayo has a FISH panel (similar to those found at most academic medical centers, he said), which looks for the following abnormalities: 5q-, -7/7q-, +8, 20q-, 13q-, 11q-/11q rearr, t(3;21), and inv(3).
Dr. Kurtin’s colleagues did a study that looked at 498 blood and bone marrow samples, on which conventional chromosome as well as FISH analyses were done. Data were analyzed based on success of conventional chromosome analysis.
If there are 20 or more metaphases, they found, the MDS FISH panel adds little value. “There was excellent correlation between cytogenetics and FISH,” Dr. Kurtin reportedâ€”95 percent. In three percent of cases, FISH added potential prognostic information, but was unnecessary as a diagnostic test because in these three percent of FISH-positive, cytogenetic-negative cases, the bone marrow morphology indicated the presence of myelodysplasia.
When less than 20 metaphases are evaluated, on the other hand, the FISH panel does appear to be useful. It accounted for a small fraction (five to 10 percent) of the cases in the Mayo study, but in 20 percent of those, FISH added diagnostic or prognostic information. Their conclusions, Dr. Kurtin said, are similar to those reached in other recent articles in the literature.
To identify chromosome abnormalities in MDS, the lab now does morphologic assessments and conventional chromosome analysis, followed by the FISH panel if there are less than 20 metaphases. When there are 20 or more metaphases, they skip the FISH panel, though targeted FISH is an option if karyotype by morphology is difficult to resolve.
Six months after implementing the first algorithm, for cytogenetics in lymphoma staging, the lab audited its performance, essentially doing the same study on which the algorithm was based.
This time, they looked at samples from 486 patients. Of those, 179 had a bone marrow, 59 of which had cytogenetics and 120 of which did not. “That seems like a success,” Dr. Kurtin said, noting the drop in cytogenetics performed.
In 32 of the 59 cases, cytogenetics was appropriate (11 were pretransplant patients, and 21 had a myeloid abnormality).
But, 27 were not. Why were they inappropriate? In 25 cases, the answer was: mea culpa. “The pathologist failed to cancel the cytogenetics test when they were supposed to do it,” said Dr. Kurtin.
Not that he’s pointing fingers. “Guilty!” Dr. Kurtin said. “I was one of them.”
In the other two cases, the clinicians reordered the tests, and the lab succumbed. “So they failed, too,” Dr. Kurtin said. Nonetheless, the algorithm works “if you actually do it.” Had they stuck to the algorithm, they would have done cytogenetics in only 19 percent of the bone marrow cases; if they had done all the cytogenetics that were canceled, that figure would have risen to 60 percent.
Dr. Kurtin said the audit points to modest success with the algorithm. More resounding success might have occurred under different circumstances, he suggested. Maybe they did the audit too soon; maybe the guideline, or even the lab’s communication, was unclear; maybe everyone just needed more practice. Maybe, just maybe, practice data, no matter how compelling, won’t light a fire under every physician. And, said Dr. Kurtin, “I know that the junior faculty members were very, very reticent to change the orders of senior hematologists.
“We have to mentor our junior pathologists,” as well as work with senior hematologists, so that the latter don’t put the former in a “bad place.”
In a later audit, done in June 2011, the lab canceled 123 tests and saved $200,000 worth of charges for that month. If the rate of genetics test reductions were annualized, he said, “we would also save two FTEs of cytogenetics technologist work effort.”
Is this an only-at-Mayo scenario? Dr. Kurtin said no. When another hospital applied the algorithms, “they went from performing cytogenetic studies on 95 percent of their bone marrows to doing cytogenetic analysis on only 38 percent, with the same incidence of cytogenetic abnormalities.”
It can be done anywhere, he insisted, for anyone willing to tackle the tough culture issues. Even then, however, perfection will be elusive. Human behavior has the uncanny tendency of getting in the way. “But perfection is not necessary,” he said. “We aim for 80 percent. Culture change is tough.”
Karen Titus is CAP TODAY contributing editor and co-managing editor.
Dr. Kurtin worked with the following people on the collaboration reported here: Rhett Ketterling, MD, and Daniel VanDyke, PhD, of the Division of Laboratory Genetics, and Curtis Hanson, MD, and Rebecca King, MD, of the Division of Hematopathology. Dr. King is now at the Childrenâ€™s Hospital of Philadelphia, University of Pennsylvania.