Reviewer 1 report (DE Jarbol)
1. Some key baseline
assumptions were included in
the model. It is hard to
understand the rationale
behind all the assumptions.
Assumption no. 4: If there is
no relief of symptoms after
initial management, or if
symptoms recur, an endoscopy
will be offered; However,
relief of symptoms will not
be expected in all patients
with uninvestigated
dyspepsia, treated for H
pylori infection. Further,
the fluctuation of dyspepsia
symptoms is well-known and
severity as well as symptoms
may vary in the same patient
over time. The evidence to
support the assumption
concerning endoscopy to all
people < 55 years with
recurrence of symptoms or no
relief of symptoms after
initial management is not
fully reliable. The
assumption needs further
explanation.
2. Assumption no. 5: The
rationale for long-term PPI
therapy here is not
understandable and needs
further explanation.
3. Probabilities employed in
the model were based on
published literature where
available and where
necessary, the numbers were
supplemented with expert
opinion based on experience
treating dyspepsia patients.
Limitations of this approach
Responses to
reviewer
We agree that there is not
evidence to support the
practice of endoscopy for
all patients in the setting
of non-relief or recurrent
symptoms. Yet this
reflects the common
gastroenterology practice
in our own community,
and we believe it is likely
to be common across
much of the U.S. Our
decision model was
designed to assess costeffectiveness of initial
diagnostic tests, not costeffectiveness of the full
algorithm, but the only
way to determine this is
by modeling the full care
process. Thus, we started
with what we believe to
be a common practice
pattern, rather than with
an idealized practice
pattern. And then we
determined the impact of
the choice of diagnostic
test given that practice
pattern.
Same
Agree that this is an
important limitation, and
may not have been clearly
enough acknowledged in
the Discussion section.
New or changed text in the
manuscript
Changed title to include the
phrase “Assuming a High
Resource Intensity Practice
Pattern”. Also, at the
beginning of the Methods
section, added "In order to
calculate the impact (both cost
and benefit) of the choice of
diagnostic test we had to first
create a model of the expected
care process for these patients.
We intentionally did not use an
idealized care process, but
rather modeled it to reflect
typical local practice for
managing dyspepsia (Figure 1).
We also believe it to be a
reasonable representation of
practice patterns for dyspepsia
across much of the US." Under
Assumption 4, added "This
assumption was followed not
in an attempt to reflect
evidence-based practice, but
rather in an attempt to reflect
a “typical” approach to
dyspepsia management."
Under Assumption 5, added
"As above, this assumption
was followed in an attempt
to reflect typical practice
patterns, rather than
evidence-based practice per
se."
Modified paragraph in
Limitations section to read
"The primary limitation of
our analysis was our set of
clinical assumptions. These
results should not be
considered to have validity
outside of those
should be discussed.
4. The incremental costeffectiveness ratios for each
strategy are presented in
table 5. However, the results
are not reflected in the
discussion
5. The discussion and
conclusion is not quit well
balanced and supported by the
data in the way that the
authors conclude that ?in the
initial choice of noninvasive
testing strategy does not
have a significant influence
on the overall quality and
cost for care for patients
presenting with previously
uninvestigated dyspepsia. The
data does not support much
about the quality of care for
patients.
assumptions. This includes
both the assumed practice
pattern (Figure 1) and the
numbers (costs and
probabilities). In clinical
settings that do not fit these
assumptions our results may
not apply."
On further reflection, the
ICER numbers as originally
presented here may
confuse rather than
enlighten readers. For
one thing, we used as our
denominator symptomfree years rather than
QALYs, yet most published
ICER analyses are
presented as incremental
cost per QALY. Second,
our ICER analysis was
based on the mean costeffectiveness results, but
our discussion emphasizes
the spread of the results,
i.e. the high degree of
overlap in the probabilistic
sensitivity analysis.
Finally, the results are
described entirely by
Table 3 and Figure 2.
Accordingly, we deleted
the ICER table.
The discussion section was
using "quality and cost of
care" as being
synonymous with the cost
effectiveness measure
used in our study, namely
cost per symptom free
year. This is potentially
ambiguous, and the
reviewer may have had a
different definition of
"quality of care" in mind.
We adjusted our language
accordingly.
Deleted Table 5 and the
references to Table 5.
Changed "quality and cost of
care" to "cost-effeciveness of
care"
6. The authors are aware of
the limitations of their
analysis in the set of
clinical assumptions but I am
not convinced about the
representatives of the
results.
Minor essential revisions:
1 Table. 2: Variable1:
Prevalence. I guess it is the
H pylori prevalence, but this
should be clarified.
Discretionary Revisions
The authors prepare the
ground for a debate of the
issue of which non-invasive H
pylori test strategy should
be used in the management of
uninvestigated dyspepsia, but
also include empiric proton
pump inhibitor trial in the
model. The reason for this is
not fully explained in the
text.
2. The authors include
serologic tests in the model,
and argue that serology is
more widely used than would
be expected under the
recommended approach.
However, the serologic tests
are not recommended due to
inferior sensitivity and
specificity, and are
therefore not interesting
from a clinical point of
view.
3 The choice of cost per
symptom free year rather than
cost per correct diagnosis as
primary outcome is discussed
in the text. The aspect
considering a correct
diagnosis of H pylori,
including confirmation of the
eradication treatment effect,
in the light of preventing
future ulcer disease, ulcer
complications and cancer
disease is, however, not
discussed.
See our response and
changes corresponding to
item 3 above.
Change made
Changed to "H. pylori
prevalence "
We felt that empiric PPI
therapy should be
included for
completeness, both
because it is a legitimate
management option and
because in a sense it
represents a type of
noninvasive diagnostic
strategy.
Added to first paragraph of
Methods: "The first step of
strategies 1 through 5 was a
different noninvasive test, and
empiric PPI therapy was
included for completeness as a
sixth strategy"
Agree that serology is not
recommended, and this is
stated in the introduction.
We disagree that they are
not clinically interesting,
specifically because they
are widely used (also as
described in the
introduction).
Although HP eradication
theoretically should
reduce the risk of cancer
as well as ulcer
complications, we felt that
there was insufficient
evidence regarding the
magnitude of this effect to
attempt to model longterm benefits of HP
eradication.
The purpose of providing
this information was
clarified as "Some
previously published costeffectiveness analyses have
reported results in the form
of mean cost per correct
diagnosis; to allow easier
comparison with these we
provide cost per correct
diagnosis in Table 4."
4 The cost perspective taken
was social, however, the
analysis considered only
direct costs which mean that
the indirect costs were not
included, which could have
been discussed.
Currently recommended
practice for costeffectiveness analysis is to
include only direct
medical costs, as opposed
to indirect costs such as
time off of work, etc. The
term "societal
perspective" is fairly
standard in costeffectiveness analysis and
is meant to reflect
aggregate medical cost
impact as opposed to the
portion paid by the
patient, the hospital, the
insurance company, etc.
No change.
This overlaps with
Reviewer 1's first item,
addressed above. Again,
our intent was to compare
cost-effectiveness just of
the diagnostic testing
portion of care, not
compare costeffectiveness of
therapeutic approaches,
and in order to do this we
had to pick a "typical"
model of care. With
specific regard to
eradication testing, we did
not test this specifically
but it probably would not
have impacted the results
signficantly given that in
most of these patients,
eradication does not
result in dyspepsia relief -thus, most would receive
endoscopy anyway under
our assumptions.
See changes made above in
conjunction with response to
Reviewer 1's first comment.
Reviewer 2 report (Dr. Xavier
Calvet)
1. Despite the study deals
with a low-risk for cancer
population (less than 55 y
old, no alarm symptoms), the
authors assumed that all
patients whose symptoms
relapsed either after triple
therapy or a PPI trial will
receive an expensive
endoscopy. By contrast, as
suggested by Spiegel et al.
(Gastro 2002), it seems more
reasonable to test for H.
pylori and treat the
infection in patients failing
empirical PPI therapy before
endoscopy. It would also be
preferable to test patients
who had received eradication
treatment for cure of H.
pylori infection with and UBT
before endoscopy. Did these
plausible different
approaches change the study
conclusions?
2. Prevalence of H. pylori
infection is given for the US
general population. However,
this results are applicable
neither to other countries
(for example, Central and
South-American,
Mediterranean, African or
Asian countries) nor to
specific populations in the
US (as Afro-American,
Hispanic or Asian), that have
an Hp prevalence around 60%.
Are the conclusions of the
study the same with such high
prevalence?
3. The same applies for
costs, which could markedly
change from country to
country. How did major
differences in the cost of
the tests influence the study
conclusions?
4. Values of sensitivity and
specificity are generous,
especially for serology and
stool tests. It should be
remembered that the different
kits show marked differences
in sensitivity and
specificity. Sensitivity and
specificity may be very low
depending on the stool kit or
even the methodology of the
urea breath test ( Calvet et
al., Clinical Infectious
Diseases). In addition, as
variability of serology
accuracy is even larger, its
use is not recommended
We tested across
prevalence ranging from
5-40%, which we agree
would not be sufficient to
extend these results to
other countries. However,
the main limitation in
attempting to apply these
results outside the US is
more likely to be the
practice pattern
assumptions. Countries
which are more resourcesensitive (e.g. UK) or
resource-deprived (e.g.
developing countries)
would have very different
practice patterns for
diagnosing and managing
dyspepsia, and our
findings would likely not
apply.
Given that the underlying
model of clinical practice
for dyspepsia would likely
not apply to most other
countries, we did not
attempt to model cost
differences other than
what would be expected
in a typical US practice
setting.
We performed
probabilistic sensitivity
analysis across a range of
sensitivity and specificity
for each test that we
considered reasonable
given our literature
review. Our ranges didn't
go quite as low as 70%,
nor did we separately
analyze what the costeffectiveness would be at
that low end of test
performance. However,
given our results, it is
unlikely that this would
See changes made above in
conjunction with response to
Reviewer 1's third comment.
Same as above.
Added the following sentence
to the Discussion: "To use a
more extreme illustration, a
consequence of our underlying
model is that the diagnostic
test could be replaced with a
random number generator
without significantly impacting
cost-effectiveness. "
provided local validation was
performed (Maastricht
guidelines). What happens in
the model when values of
sensitivity and specificity
for serology are decreased
to, for example, 70%?
5. Obtaining the values for
assumptions of different
studies instead from direct
comparative trials leads to
bizarre results. Values
coming from different studies
heavily depend on the
differences in the baseline
characteristics of the
populations evaluated instead
of reflecting true
differences of the efficacy
of the treatments. So, for
example, the study assumes
that, after eradication, the
probability of symptom
relapse at one year is higher
in patients with peptic ulcer
than in those without. By
contrast, in the few
available randomized trials,
the one-year probability of
symptoms relapse in patients
with uninvestigated dyspepsia
after a PPI trial is near
100% (Rabeneck, AJG, Marmo
BMJ) whereas probability of
relapse is far lower after
eradication treatment in H
pylori positive patients
(Marmo BMJ). What happens in
the model when these
different values are
incorporated?
have made any difference.
Stool antigen testing was
modelled as being less
expensive than serology
and much more accurate,
yet the model did not
predict it to be
significantly more costeffective. As pointed out
in the Discussion, this
appears to be a
consequence of our
underlying clinical practice
pattern assumptions.
Agree with the reviewer
about the problematic
nature of using
assumptions derived from
different comparative
trials. For our study, we
chose to heavily weight
the findings of the
following Cochrane
review: Moayyedi P,
Deeks J, Talley NJ, Delaney
B, Forman D. An update of
the Cochrane systematic
review of Helicobacter
pylori eradication therapy
in nonulcer dyspepsia:
resolving the discrepancy
between systematic
reviews. Am J
Gastroenterol 2003
Dec;98(12):2621-6. This
review found that for
patients with nonulcer
dyspepsia and confirmed
HP infection, eradication
therapy provided only a
small average benefit.
This review thus has quite
different conclusions from
the articles the reviewer
cites here. Nonetheless,
we added language in the
limitations section to
Added the following to the
Limitations section: "Finally,
our analysis relied heavily on
the findings of a Cochrane
systematic review in which H.
pylori eradication was found to
have only a very small clinical
benefit for the average patient
with nonulcer dyspepsia. Our
findings may thus not apply to
patient subsets for which
eradication therapy could be
shown to have a larger average
benefit. "
clarify that our findings
would not apply to patient
subsets where eradication
therapy could be shown to
have a large effect in
reducing symptoms.