Introduction
Barrett’s esophagus (BE) is a known precursor of esophageal adenocarcinoma (EAC),
which has a rapidly rising incidence and a 5-year survival rate of less than 20 %
[1]
[2]. Endoscopic surveillance with biopsies and expert histologic review is recommended
by gastrointestinal societies to identify dysplasia and early EAC when endoscopic
eradication therapy (EET) can effectively prevent progression [3]
[4]
[5]. The effectiveness of this approach is limited by the random nature of endoscopic
biopsy sampling, which may miss dysplastic areas, and by significant interobserver
variation in histologic evaluation of biopsies [6]
[7]. A diagnosis of low-grade dysplasia (LGD) confirmed by a gastrointestinal subspecialist
pathologist is a significant predictor of progression to high-grade dysplasia (HGD)
and EAC, and current society guidelines recommend EET or increased surveillance for
patients with BE who have confirmed LGD [4]
[8]. However, the reported progression rates for confirmed LGD vary widely and overdiagnosis
of LGD is common [6]
[7]
[9]
[10]
[11], making it unclear whether therapeutic intervention is warranted for all patients
with LGD. Pathologic analysis also can be confounded by cases that are difficult to
diagnose, for example, if there is a background of inflammation, which can result
in a diagnosis of “indefinite” for dysplasia (IND). These IND cases also present a
management challenge because there is substantial heterogeneity between studies reporting
on the progression risk associated with IND [12].
The rate of malignant progression from non-dysplastic (ND) BE is very low [13]
[14], and clinical factors such as segment length, sex, and age may help guide decisions
regarding surveillance intervals [15]
[16]
[17]. However, a subset of patients with ND BE will progress to HGD/EAC during the recommended
3–to 5-year surveillance interval, resulting in significant uncertainty and anxiety
regarding management, as well as over-utilization of endoscopic surveillance that
has not been effective in reducing the incidence and mortality of EAC [18]
[19]. There is a need for objective tests to identify BE patients with ND, IND or LGD
who are at increased risk for progression. These patients could benefit from early
EET to prevent progression, or increased surveillance to detect dysplasia earlier.
Furthermore, a test that could identify patients who are at very low risk for progression
could prevent unnecessary EET and over-surveillance.
The TissueCypher Barrett’s Esophagus Assay has been developed and validated in five
multi-institutional studies to predict progression from ND, IND, and LGD to HGD and
EAC, and also to detect prevalence of HGD/EAC in BE that can be missed on surveillance
endoscopy [20]
[21]
[22]
[23]
[24]. TissueCypher utilizes a multiplexed fluorescence imaging platform that objectively
extracts quantitative data on multiple epithelial, stromal, and morphometric features
in intact tissue specimens [25]
[26]. TissueCypher integrates quantitative image analysis data and provides a risk score
ranging from 0 to 10 that is used to classify patients as low-, intermediate- and
high-risk for progression to HGD/EAC within 5 years. This three-tier risk stratification
approach may provide adjunctive evidence supporting EET in for BE patients with BE
with high-risk TissueCypher scores, while also enabling a surveillance-only approach
or extension of surveillance intervals in patients with low-risk scores. The predictive
performance of TissueCypher in BE has been well described; however, the extent to
which the test influences the management of patients with BE has not yet been studied.
The objectives of this study were to determine whether TissueCypher results impact
physician decision-making in the management of BE and to characterize these changes
in terms of altering endoscopic surveillance intervals and use of EET.
Patients and methods
A prospective study was conducted that collected information on decisions made by
two physicians regarding management of BE prior to and after receiving results from
the TissueCypher Barrett’s Esophagus Assay. TissueCypher is a commercially available
test for physicians managing patients with BE. The test is performed on slide blanks
recut from existing tissue blocks of biopsies or mucosal resections for BE, which
are sent to the Cernostics clinical laboratory (CLIA#: 39D2110302, Pittsburgh, Pennsylvania,
United States) for testing. TissueCypher integrates quantitative image analysis data
from nuclear morphology and nine protein-based biomarkers (p16, alpha-methylacyl-coA
racemase [AMACR], p53, CD68, cyclooxygenase [COX]-2, CD45RO, hypoxia inducible factor
[HIF]-1alpha, HER2 /neu and cytokeratin-20 [K20]). A clinical report is generated
that includes a risk score ranging from 0 to 10, with 0 indicating lowest risk and
10 indicating highest risk. This score is then stratified into three risk classes
(low 0 to < 5.5, intermediate 5.5 to < 6.4, or high 6.4 to 10) that predict progression
to HGD/EAC within 5 years [20]. The results provided are adjunctive to the ordering physician’s workup for patients
with BE.
TissueCypher tests ordered at Geisinger Medical Center between June 2016 and April
2020 were considered for this study. A pre-TissueCypher management plan survey was
sent to physicians after receipt of the test order. The survey included the following
endoscopic surveillance interval options: 3 and 6 months and 1, 3, or 5 years, or
other and the following treatment options: no treatment, mucosal resection, radiofrequency
ablation, cryotherapy, esophagectomy, referral to oncologist, and other (free text
describing the treatment plan). A post-TissueCypher management plan survey was then
sent after the test results were reported.
Additional data collected were age at the time the test was ordered, sex, pathologic
diagnosis of biopsies for which the test was ordered, worst historical pathologic
diagnosis, Prague classification, segment length (short < 3 cm, long ≥ 3 cm), sex,
and indication/reason for ordering the test. Indications for ordering the test could
be grouped into categories of support for a do-not-treat strategy in LGD, support
for a do-not-treat strategy in other situations, risk stratification for IND, evaluation
of persistent intestinal metaplasia after EET, and clinical/endoscopic concern. “Clinical/endoscopic
concern” included patients with worrisome endoscopic features (such as polypoid lesions,
or abnormal appearing but non-nodular mucosa), a history of intramucosal carcinoma,
or very discordant pathological results from initial outside biopsy compared to repeat
biopsy at referral center.
The pretest management plans (endoscopic surveillance interval and treatment plan)
were compared to the management plans after receiving the test results to determine
whether management decisions were changed by the test results, and to characterize
the changes. Pretest and post-test management recommendations were recorded by two
gastroenterologists (DLD and HSK) at a single center. Only patients with both pre-
and post-TissueCypher management recommendations were included in the analysis. Management
changes were characterized as upstaged if the plan changed from surveillance-only
to therapeutic intervention, additional therapy was recommended, ongoing therapy or
surveillance was recommended where discontinuation was being considered, and/or increased
surveillance frequency was recommended. Management changes were characterized as downstaged
if the plan was changed from therapeutic intervention to a surveillance-only approach,
or the surveillance frequency was reduced after review of test results. A chi-squared
test was used to evaluate associations between categorical variables. The study protocol
was reviewed and approved by the Geisinger Health System Institutional Review Board
(IRB), Danville, Pennsylvania, United States. All authors had access to the study
data and reviewed and approved the final manuscript.
Results
Patient characteristics
Pretest and post-test management recommendations were obtained for 60 patients with
BE and a diagnosis of ND (n = 18), IND (n = 25) or LGD (n = 17) diagnosed by a gastrointestinal
subspecialist pathologist at Geisinger Medical Laboratories. The flowchart in [Fig. 1] shows the tests that were included or excluded from the analysis. The mean age of
the included patients was 65.2±11.8 (range 35.1 to 85.0), 52 of 60 patients (86.7 %)
were male, and 43 of 60 patients (71.7 %) had long segment Barrett’s esophagus ([Table 1]). There were no significant differences in age, sex, histologic diagnoses, or the
ordering physician between the included patients and the patients who were excluded
due to lack of pre-management and post-management data, indicating that the exclusions
did not introduce bias. Indications for ordering TissueCypher were clinical/endoscopic
concern (n = 13), support of a do-not-treat strategy in LGD (n = 11), support of a
do-not-treat strategy in other situations (n = 8), risk stratification of IND (n = 25),
and evaluation of persistent intestinal metaplasia after EET (n = 3, patients had
complete eradication of dysplasia but not complete eradication of intestinal metaplasia
after at least one round of ablation). Repeat endoscopy with biopsies for patients
with IND was not done. All patients with diagnosis of IND were on chronic proton pump
inhibitor (PPI) treatment, and the TissueCypher test was ordered at the first diagnosis
of IND or after repetitive diagnoses of IND. The TissueCypher test reported a result
of low-risk for 39 patients (65 %), intermediate-risk for seven patients (11.7 %),
and high-risk for 14 patients (23.3 %) ([Table 2]).
Fig. 1 Flowchart of included/excluded patients.
Table 1
Patient characteristics.
|
Age
|
|
Mean ± SD
|
65.2 ± 11.8
|
|
Range
|
35.1–85
|
|
Sex, n (%)
|
|
|
52 (86.7)
|
|
|
8 (13.4)
|
|
Segment length, n (%)
|
|
|
17 (28.4)
|
|
|
43 (71.7)
|
|
Prague classification
|
|
|
4.7 (0, 16)
|
|
|
6.3 (1, 16)
|
|
Expert Dx, n (%)
|
|
|
18 (30)
|
|
|
25 (41.7)
|
|
|
17 (28.4)
|
|
TissueCypher results, n (%)
|
|
|
39 (65)
|
|
|
7 (11.7)
|
|
|
14 (23.3)
|
SD, standard deviation; Prague Classification, C: circumferential length and M: maximal
length; Dx, pathologic diagnosis; ND, non-dysplastic; IND, indefinite for dysplasia;
LGD, low-grade dysplasia.
Table 2
TissueCypher Results in Diagnostic Classes
|
All patients
|
ND patients
|
IND patients
|
LGD patients
|
|
n = 60
|
n = 18
|
n = 25
|
n = 17
|
|
Low-risk
|
39 (65)
|
15 (83.3)
|
17 (68)
|
7 (41.2)
|
|
Intermediate-risk
|
7 (11.7)
|
2 (11.1)
|
4 (16)
|
1 (5.9)
|
|
High-risk
|
14 (23.3)
|
1 (5.6)
|
4 (16)
|
9 (52.9)
|
BE, Barrett’s esophagus; ND, non-dysplastic; IND, indefinite for dysplasia; LGD, low-grade
dysplasia.
To determine the change in physician behavior due to the test results, the post-test
management recommendations were compared to the pretest management recommendations.
TissueCypher results had a statistically significant impact on decision-making regarding
management of BE. Following review of TissueCypher test results, the management plan
changed in 33 of 60 patients (56.7 %) ([Fig. 2]). The management plan was upstaged in eight of 14 patients (57.1 %) who had a high-risk
TissueCypher score, and the management was downstaged in 19 of 39 patients (48.7 %)
with a low-risk score (P < 0.0001). Eighteen of 39 patients who scored low-risk were not downstaged due to
other clinical/endoscopic concerns (n = 7), diagnosis of IND (n = 8) or because the
test was ordered in support of a do-not-treat strategy (n = 3). In the two upstaged
patients who scored low-risk, the recommended surveillance interval was reduced from
3 years to 1 year upon review of test results. In the subset of 33 patients whose
management plan was changed after receipt of test results, upstaging was significantly
associated with a high-risk test result, and downstaging was associated with a low-risk
result (P < 0.0001).
Fig. 2 Impact of TissueCypher results on management decisions. Upstaged management indicates
that the test results led to increased surveillance frequency and/or a change from
no treatment to recommendation of therapeutic intervention. Downstaged management
indicates that the test results led to decreased surveillance frequency and/or a change
from planned therapeutic intervention to recommendation of no treatment. *Treatment
of newly diagnosed co-morbidity took priority over BE management.
The management plan was upstaged after review of test results in 13 patients (21.7 %)
([Fig. 2]), resulting in a variety of changes in management plan: 1) a change from surveillance-only
to therapeutic intervention with ablation therapy (n = 7); 2) recommendation for endoscopic
mucosal resection (EMR) in addition to ablation (n = 1); 3) fundoplication and shorter
surveillance interval (from 1 year to 6 months) (n = 1); 4) shorter surveillance interval
(from 3 years to 1 year) (n = 2) 5) ongoing ablation where a surveillance-only plan
was being considered (n = 1); and 6) ongoing surveillance versus discontinuation of
surveillance (n = 1). In the group of patients whose management was upstaged, eight
scored TissueCypher high-risk, three intermediate-risk, and two low-risk, indicating
that the majority of patients were upstaged due to a high-risk test result. Three
of the 13 upstaged patients had a diagnosis of LGD, seven were IND and three had ND
BE.
The management plan was downstaged after test results were reviewed in 20 patients
(33.4 %), resulting in a change from therapeutic intervention with ablation to recommendation
of a surveillance-only plan in 19 patients, and extension of surveillance intervals
(from 1 year to 3 years) in one patient ([Fig. 2]). In the subset of patients whose management was downstaged after review of the
test results, 19 scored TissueCypher low-risk and one scored intermediate-risk, suggesting
that the majority of downstaged management decisions were due to a low-risk test result.
The management plan was not impacted for 25 patients (42.4 %) of whom 18 scored TissueCypher
low-risk, three scored intermediate-risk, and four scored high-risk. Two patients
were diagnosed with a comorbidity shortly after the test results were delivered, and
treatment of the comorbidity took priority over management of BE.
The impact of the test results was not significantly different between the three diagnostic
classes of ND, IND and LGD (P = 0.8904) ([Table 3]). In patients with ND BE, 50 % of management plan decisions were changed after review
of test results (16.7 % upstaged, 33.3 % downstaged), in IND 60.0 % of decisions were
changed (28.0 % upstaged, 32.0 % downstaged), and in patients with LGD 52.9 % decisions
were changed (17.6 % upstaged, 35.3 % downstaged) ([Table 3] and [Fig. 3a, ]
[Fig. 3b], and [Fig. 3c]). Downstaging of the management plan after review of test results was more frequent
in patients with LGD than with IND and ND. This is likely due to the indication for
ordering as the majority of test orders in patients with LGD were to support a do-not-treat
strategy. When an intermediate risk result was reported, other clinical factors were
taken into account to arrive at a treatment strategy, including comorbidities and
history of previous treatment such as refractoriness to endoscopic eradication therapy.
This approach led to upstaging of the management plan in three of seven patients who
scored TissueCypher intermediate-risk, downstaging in one of seven, and no change
in three of seven patients.
Table 3
Impact of TissueCypher on BE management decisions: Change in management[1]
|
All patients
|
ND patients
|
IND patients
|
LGD patients
|
|
n = 60
|
n = 18
|
n = 25
|
n = 17
|
|
Upstaged management
|
13 (21.7)
|
3 (16.7)
|
7 (28)
|
3 (17.6)
|
|
Downstaged management
|
20 (33.4)
|
6 (33.3)
|
8 (32)
|
6 (35.3)
|
|
No change
|
25 (41.7)
|
9 (50)
|
10 (40)
|
6 (35.3)
|
|
Other
|
2 (3.4)
|
0 (0)
|
0 (0)
|
2 (11.8)
|
|
Changed (up/down)
|
33 (55)
|
9 (50)
|
15 (60)
|
9 (52.9)
|
|
Not changed (none/other)
|
27 (45)
|
9 (50)
|
10 (40)
|
8 (47.1)
|
BE, Barrett’s esophagus; ND, non-dysplastic; IND, indefinite for dysplasia; LGD, low-grade
dysplasia.
1 In all patients and in subgroups of patients with pathologic diagnoses of ND, IND
and LGD (number and % of decisions impacted).
Fig. 3 Impact of TissueCypher results on management decisions in diagnostic subsets of ND,
IND, and LGD. Proportions of management plan decisions upstaged, downstaged or not
changed after review of TissueCypher results are shown in patients with a non-dysplastic BE; b indefinite for dysplasia, and c low-grade dysplasia. *17 patients had diagnosis of LGD, however, two patients were
excluded from this sub-analysis due to treatment of higher priority co-morbidity (see
Fig. 1 legend).
A subanalysis was conducted to evaluate the impact of TissueCypher results within
specific indications for ordering the test. The impact of test results on management
plan was significantly associated with the indication for ordering ([Fig. 4, ]
P = 0.0011). The largest impact was observed when the test was ordered to support a
do-not-treat (DNT) strategy for LGD (9/11 decisions impacted), DNT in other indications
(5/8 decisions impacted) and for risk stratification in patients with IND (15/25 decisions
impacted). Only one of 13 decisions was impacted when the reason for ordering the
test was “clinical/endoscopic concern.” The other indication for ordering had sample
size of three or less, which was insufficien for this subanalysis.
Fig. 4 Impact of TissueCypher results on management decisions for specific indications. The
number of patients whose management plan was not changed, upstaged or downstaged is
shown by indication for ordering the test, which included clinical/endoscopic concern,
support of a do-not-treat (DNT) strategy in LGD, support of DNT strategy in other
situations (DNT Other) and risk stratification in patients with indefinite for dysplasia
(IND). Indications with less than five test orders were excluded due to inadequate
sample size.
Discussion
In this prospective decision impact study, the TissueCypher Barrett’s Esophagus Assay
significantly influenced decision-making in management of patients with BE. Comparison
of post-test versus pretest management recommendations in 60 patients with BE showed
that the TissueCypher results impacted 55.0 % of management decisions regarding surveillance
intervals and therapeutic interventions. The management plan was upstaged after review
of TissueCypher results in 21.7 % of patients, which included decisions to intervene
with endoscopic eradication therapy (EET) instead of a surveillance only approach.
The upstaging of the BE management plan was associated with high-risk results from
the TissueCypher test, indicating that the high-risk result may influence physicians
to use therapeutic intervention to prevent potential progression to HGD/EAC, or to
increase surveillance in order to monitor these patients more closely. This is a key
finding as it suggests that the TissueCypher test has potential clinical utility to
improve outcomes by targeting early interventions and close surveillance to an “at-risk”
subset of patients, which may have been missed based on the current standard of care
The management plan was downstaged after review of TissueCypher results in 33.4 %
of patients, and 95 % of these decisions involved changing from EET to surveillance
only, with the other 5 % resulting in extension of the surveillance interval from
1 year to 3 years. This is also a critical finding as it indicates that TissueCypher
test results have the potential to reduce the overuse of both EET and surveillance
endoscopies, which may improve the efficiency of healthcare use in the management
of BE.
TissueCypher results impacted management decisions at similar rates in the diagnostic
subgroups of non-dysplastic BE (ND), indefinite for dysplasia (IND) and low-grade
dysplasia (LGD), suggesting that the test results may be used to guide management
decisions in all three of these clinical indications. However, the type of impact
differed between the diagnostic subgroups with downstaging occurring more frequently
in LGD than in IND and ND BE. This is likely due to the indication for ordering in
LGD, which was mostly for support of a do-not-treat strategy.
The finding that high-risk results from TissueCypher led to upstaging of BE management
and low-risk results led to downstaging suggests that TissueCypher has the potential
to shift the current clinical practice paradigm from surveillance and treatment based
on the subjective histopathologic finding of dysplasia to objective risk assessment,
with early interventions targeted to high-risk patients, and reduced interventions
and less frequent surveillance in low-risk patients. This paradigm shift could improve
patient outcomes by preventing EAC in high-risk patients while reducing expenditures
on unnecessary procedures and endoscopic surveillance in low-risk patients. A recent
cost-effectiveness study compared TissueCypher-guided care versus the current standard
of care from the perspective of a large US health care system [27]. Intervention with EET in patients with high-risk TissueCypher scores, in parallel
with extension of surveillance intervals to 5 years in patients with low-risk scores,
was predicted to be cost-effective within 5 years, while reducing the progression
to HGD, EAC and EAC-related deaths by 51.7 %, 47.1 %, and 37.6 %, respectively, and
also reducing use of endoscopy with biopsy and pathology review by 16.6 %.
Overdiagnosis of dysplasia and the resulting interventions and intensive surveillance
is costly to the healthcare system. The excess cost associated with overdiagnosis
of LGD ranges from $ 3,115 to $ 8,072 per patient in the United States [28], and while EET procedures are considered to be both effective and safe, adverse
events occur at a rate of 8.8 % [8]. TissueCypher could be used to target EET and intensive surveillance for at-risk
patients, while avoiding overtreatment and overuse of surveillance for LGD and IND.
While patients with ND BE are rarely treated with EET, overuse of endoscopic surveillance
is common [19]. TissueCypher could be used for ND BE to both reduce the overuse of endoscopic surveillance
in low-risk patients and identify the small subset of high-risk patients requiring
closer surveillance or EET.
Because the impact of other risk prediction assays on management decisions in BE has
not been reported, there is no other test to which we can directly compare our decision
results. However, the impact of precision medicine tests has been extensively studied
in other disease areas showing comparable results in terms of how objective risk stratification
data can influence decisions made by physicians to manage complex diseases. For example,
a systematic review and meta-analysis of 8 studies evaluating the impact of the Oncotype
Dx Breast Cancer Recurrence Score found that the test results changed the treatment
recommendation regarding adjuvant chemotherapy in 33.4 % of patients with early stage
breast cancer [29]. In non-small-cell lung cancer, the VeriStrat test has been shown to lead to a 28.2 %
change in treatment recommendations by physicians, leading to a reduction in use of
costly and ineffective treatments [30].
The main strengths of this study are the prospective design, and inclusion of ND,
IND, and LGD cases for which risk stratification is clinically needed. An additional
strength is the study setting, which included a gastrointestinal pathology subspecialty
group for interpretation of the biopsy results.
The study limitations are that this was a single-center experience with only two physicians
providing management plans before and after receipt of test results. Data on adherence
to management plan recommendations or outcomes were not collected. Our center is an
expert BE referral center, and the test was ordered selectively rather than consecutively,
which resulted in a higher proportion of patients with IND and LGD than would be seen
in a typical community practice setting. Additional studies are needed to determine
the impact of test results in the community setting where the majority of patients
have ND BE. Six of the 60 patients had previously undergone EET and had biopsies showing
residual BE when TissueCypher was ordered. While a small study on TissueCypher has
been conducted in this setting [21], the majority of clinical studies on the test have been completed in patients who
have not been previously treated. Because this was a prospective study assessing physician
decision-making, there is a risk of the Hawthorne effect where participants may alter
their behavior because they are being observed.
Conclusion
In summary, objective risk stratification provided by the TissueCypher test had a
significant impact on physician decision-making in management of patients with BE.
Physicians upstaged management of their patients when the test reported high-risk
scores and downstaged management recommendations in response to low-risk scores, indicating
that the TissueCypher test has clinical utility with the potential to both improve
patient outcomes and reduce overuse of procedures in management of BE.
