|
SLIDES
& TRANSCRIPTS
Tuesday,
March 6
Loco-Regional
Control: Staging
Scott A.
Hundahl, MD, FACS
|
| Slide
1: Introduction |
Good morning. For this session, I will be focusing on loco-regional
control, staging.
TOP
|
| Slide
2: Outline |
|
 After
discussing the survival impact of completeness of resection and
lymphadenectomy, I will go on to staging issues and I will finish
with what I see as opportunities for improvement.
TOP
|
| Slide
3: Definitions |
|
I show this slide just to clarify a confusing point. Prior to the
mid-1990s, the Japanese used an R factor to describe lymphadenectomy.
To avoid confusing with the UICC R factor, which refers to completeness
of resection, the Japanese coined the term, D factor, to describe
lymphadenectomy.
D factor designations are based on the Japanese general rules, not
the AJCC categories.
TOP
|
| Slide
4: Completeness of Resection |
|
The first issue, the completeness of resection. Echoing what Marty
showed us yesterday, in this 1,600 patient series from Germany,
achieving a complete negative margin resection was, on multivariate
analysis, the single most important prognostic variable. This is
something the surgeon has impact on.
TOP
|
| Slide
5: Antral Tumors |
|
Now, the Japanese have documented substantial five-year survival
with surgical resection of disease in celiac-based lymph nodes,
seen here in the blue bars. They have also shown, in the orange
bars, a fair incidence of disease in at least some of these node
groups. This slide is for antral tumors,
TOP
|
| Slide
6: Middle-Third Tumors |
|
 but
one sees the same thing for upper third tumors
TOP
|
| Slide
7: Upper-Third Tumors |
|
and middle third tumors.
TOP
|
| Slide
8: Justification |
|
 Now,
despite these data justifying D-2 type nodal resections, as Marty
summarized yesterday, two large, prospective randomized trials of
D-2 versus D-1 lymphadenectomy have failed to document improved
survival with the D-2 procedure.
TOP
|
| Slide
9: Dutch Lymphadenectomy Results |
|
 In
the Dutch trial, mortality was high, particularly for the D-2 group
and there was no difference in overall survival.
TOP
|
| Slide
10: MRC Lymphadenectomy Results
|
|
 In
the MRC trial, mortality was even higher, and there was again no
difference in overall survival. A confounding factor in both trials
was pancreatic and splenic resection. In general, this was a procedure
only required in the D-2 group and, in both trials, this portion
of the D-2 resection generated considerably mortality and morbidity.
In the early 1990s, after both these trials had begun, most groups
in Japan and elsewhere had switched to pancreas and spleen-preserving
D-2 operations.
TOP
|
| Slide
11: D-2 + Lymphadenectomy
|
|
In expert centers, 30-day mortality for D-2 operations is significantly
better than what we saw in these trials. Some have used this to
advocate concentration of gastric cases in expert centers.
TOP
|
| Slide
12: Randomized Trials Conclusions
|
|
 As
Marty emphasized yesterday, the conclusion of these trials is straightforward.
Bottom line, D-1 probably is okay as a minimum procedure.
TOP
|
| Slide
13: Impact of Under-Treatment
|
|
 Now,
I want to share with you some results from a surgical analysis of
Intergroup 0116/SWOG 9008, which will further refine how we view
lymphadenectomy.
TOP
|
| Slide
14: SWOG 9008 (INT 0116) |
|
You are already familiar with the scheme for that trial, basically
post-op 5FU-based chemoradiation versus post-op observation.
TOP
|
| Slide
15: Surgical Information Sources |
|
In addition to operative and pathology reports, we collected some
important information using a surgical checklist and data form.
In particular, we captured information on how each lymph node station
around the stomach was handled by the surgeon, prior to any survival
analysis.
TOP
|
| Slide
16: Surgical Data
|
|
In addition to the usual clinical and pathologic variables, we coded
three surgical variables: the type of gastrectomy, the D level of
lymphadenectomy, and a third variable which we termed the Maruyama
index of unresected disease.
TOP
|
| Slide
17: Maruyama Computer Program
|
|
 I
need to share some background concerning this latter variable. The
Maruyama computer program basically matches seven case characteristics
with a 3,843-patient data base from the National Cancer Center Hospital
in Tokyo.
Based on actual experience, the program predicts, among other things,
the percentage likelihood of disease at each of the 16 defined lymph
node stations around the stomach.
TOP
|
| Slide
18: Maruyama
Output |
|
 This
is an example of the output. As you can see, it is not a very user
friendly program. It is, however, very useful.
TOP
|
| Slide
19: Maruyama Computer Program Evaluations |
|
The performance characteristics of this program have been evaluated
in both Japan and Germany. In the German series, the accuracy of
predictions for perigastric nodes and for celiac-based nodes is
actually very good, 83 and 89 percent respectively.
TOP
|
|
Slide
20: Definition--Maruyama
Index
|
For the patients in the Intergroup 0116 trial, we used the Maruyama
program to predict the percentage likelihood of disease in the
regional nodes left behind by the surgeon, and we termed this
the Maruyama index of unresected disease.
TOP
|
| Slide
21: Methods--Statistical Analysis |
|
 We
used conventional statistical techniques, including Cox multivariate
regression, to analyze survival. Cases in the two arms of this trial
were well balanced in terms of all the variables listed here.
TOP
|
| Slide
22: SWOG 9008/INT 0116 Group Balance |
|
should point out that the patients entered in this trial generally
had more advanced disease. Here we see the stage distribution for
the Intergroup 0116/SWOG 9008 patients, in orange, compared to what
you would expect based on contemporaneous data from the national
cancer data base. In the SWOG group we see a lot more III-A and
III-B cases.
TOP
|
| Slide
23: Stage Distribution |
|
As you have already seen, the treatment group enjoyed an advantage,
both in terms of overall survival,
TOP
|
| Slide
24: SWOG
9008/INT 0116 Overall Survival |
|
and relapse-free survival.
This was a post-op trial with post-op registration. As such, we
probably captured standard of care as far as U.S. surgeons and lymphadenectomy
is concerned.
TOP
|
| Slide
25: D-level |
|
 Fifty-four
percent of the patients in this trial underwent D-0 lymphadenectomy,
inadequate in the opinion of most experts.
Interestingly, perhaps because lymphadenectomy is for early stage
cases, they tended to be the less extensive D-0 lymphadenectomies.
Because the difference between D-0 and D-1 was generally just one
node station, we observed no significant relationship between D
level and survival.
TOP
|
| Slide
26: Maruyama Index Summary |
|
We observed a predictable relationship between D level and Maruyama
index of unresected disease. We also observed that median Maruyama
index tended to increase with stage but, as we will see, basically
within the same Maruyama index category.
TOP
|
| Slide
27: Maruyama Index Prediction |
|
 Prior
to any survival analysis, I predicted that cases with a Maruyama
index less than five would have better survival.
TOP
|
| Slide
28: Maruyama Index Independent Predictor |
|
 Using
this as a cut off, we see that Maruyama index was an independent
predictor of both overall survival and relapse free survival. This
was the case even when we corrected for not just treatment but N
status and T status.
TOP
|
| Slide
29: Overall Survival Curve |
|
 The
survival curves for overall survival and relapse-free survival are
compelling.
TOP
|
| Slide
30: Relapse-Free Survival |
|
A curve for Maruyama index less than five appears to have plateaued
at almost 60 percent survival.
TOP
|
| Slide
31: MI vs. Survival |
|
 Using
a recursive partitioning technique which partitions the cases so
as to maximize the difference in log rank test, it looks like the
optimal cut points for Maruyama index are about the same, whether
you look at overall survival or relapse-free survival. The first
category is Maruyama index less than five. So, I guessed right.
The second index, five to 140, and a third category greater than
140. The survival difference for all these categories was quite
remarkable.
TOP
|
| Slide
32: Treatment Interaction Analysis |
|
 Interestingly,
the adjuvant treatment in this trial, post-op chemoradiation, appeared
to work in all surgical and pathological subgroup. Even when the
operation was good, the chemoradiation improved results.
TOP
|
| Slide
33: SWOG
9008/INT 0116 Conclusions |
|
Our conclusion reflects this fact and also stressed that suboptimal
lymphadenectomy is common in the United States. At least by the
Maruyama index, suboptimal lymphadenectomy compromises survival.
TOP
|
| Slide
34: D-0 Lymphadenectomy |
|
Now, there is some evidence that inadequate lymphadenectomy is not
just a United States problem. For power calculations, both the Dutch
group and the MRC group used a historical figure of 20 percent survival
for patients meeting the inclusion criteria for those trials.
TOP
|
| Slide
35: Relative Survival |
|
In fact, without any change in T stage distribution, the survival
rate observed for the D-1 group was substantially better, 35 percent
in the MRC trial and 45 percent in the Dutch trial. This, to me
anyway, suggests that less than D-1 resection may have been quite
common in the historical group that was used for the power calculations.
TOP
|
| Slide
36: 5th Ed. AJCC Staging |
|
 As
you can see, AJCC staging for gastric cancer, of course, changed
with the fifth edition. The change was made basically to address
the problem of lymphadenectomy-related stage migration.
The N in the fifth edition was based on number of positive nodes
rather than anatomic location. As you can see, this new system works
fairly well.
Unfortunately, based on NCBD data from over 50,000 cases treated
from 1985 to 1996, only 18 percent of U.S. cases have node counts
which comply with the minimum AJCC criteria of 15 nodes per specimen.
Obviously, if a pathologist has only counted six or less nodes in
the specimen, it is impossible to have N-2 disease using fifth edition
staging.
TOP
|
| Slide
37: 5-Year Survival Rates |
|
 This
is also from the NCDB. Here we see that, without any change in T
stage distribution, observed stage stratified survival varies according
to the number of lymph nodes analyzed by the pathologist. The survival
differences are particularly noteworthy for stages I-B, II and III-A.
In the United States, unfortunately, we still have a stage migration
problem. We certainly need to take this into account when we start
comparing survival rates.
TOP
|
| Slide
38: Summary |
|
So, we have shown that the majority of U.S. patients undergo inadequate
lymphadenectomy and that this can significantly compromise survival
and staging.
TOP
|
| Slide
39: Improving Local-Regional Control & Survival |
To improve our results, using the Maruyama program to plan surgical
reception might correct the D-0 problem yet, at the same time,
minimizing the dissection of irrelevant node groups, and thus,
decrease the mortality and the morbidity of the procedure.
As Marty stressed yesterday, of course, we should avoid pancreatic
and splenic resection, unless we need this in order to achieve
negative margins. Finally, this is very much a disease where proper
case selection and appropriate training and experience impact
on outcome. Tertiary referral may be a desirable strategy, the
same way we do for pancreas and other cancers.
TOP
|
| Slide
40: Future Plans |
|
 Currently,
I am proposing a single arm phase II trial of Maruyama-guided surgery.
The American College of Surgeons Oncology Group could probably do
this as a single group.
As far as generalizable results are concerned, doing this through
SWOG of the Intergroup may actually be preferable.
TOP
|
| Slide
41: Cady's Paradigm |
|
One final slide deals with philosophy. I think surgical treatment
for this disease probably follows Cady's paradigm. Cady's paradigm
goes like this: The therapeutic effect of cancer surgery is akin
to that of a drug with a threshold effect, dose response up to a
certain plateau, but no therapeutic effect beyond this, only more
complications. Thank you.
TOP
|
