Claude Monet (French, 1840-1926), Pourville (detail), 1882.
Natural
History Studies and the Evolution of Regional Treatment Modalities for Patients
Withy Isolated Liver Metastases From Primary Colon and Rectum Carcinoma
Glenn
Steele, Jr, MD, PhD
Estimates for 1995 indicate
that 150,000 cases of colon and rectum adenocarcinoma occurred in North America,
58,000 patients died of colorectal cancer, and 18,000 patients died with liver
as the predominant or only site of metastatic disease from recurrent colon and
rectum adenocarcinoma. Results from several trials can now better define the
therapeutic and survival benefit of surgical resection, cryosurgical ablation,
and intrahepatic arterial therapy in the treatment of patients with metastatic
colorectal cancer in the liver.
Introduction
Of the approximately 150,000
cases of colon and rectum adenocarcinoma estimated in 1995 in North America,[1]
nearly 58,000 patients died of colorectal cancer, 30,000 patients of colorectal
cancer from liver metastases, and 18,000 patients with liver as the predominant
or only site of metastatic disease from recurrent colon and rectum adenocarcinoma.
Most patients with isolated liver metastases from colon and rectum adenocarcinomas
who are amenable to consideration for surgical resection of isolated metastases
have no symptoms. Evidence for tumor recurrence usually comes from a rise in
serial plasma carcinoembryonic antigen (CEA) levels or elevated serum alkaline
phosphatase levels noted months or even years after primary tumor resection.
More often than not, such patients are confirmed to have metastatic liver disease
by follow-up computed tomography (CT) scan, magnetic resonance imaging, ultrasonography,
or other diagnostic procedures initiated by the CEA abnormality.
Natural History
This rise
in postoperative serial CEA as an effective marker for site-specific colon and
rectum carcinoma recurrence may be related to the fact that Kupffer's cells
of the liver and lung and fixed alveolar macrophages, respectively, have receptors
for transmembrane CEA or CEA-like molecules.[2] More important, clinical adjuvant
treatment studies such as trial 6175 of the Gastrointestinal Tumor Study Group
(Table 1) defined that patients with significant slopes of serial CEA rise had
their greatest predisposition for recurrence in the liver or in the liver and
the lungs. This stimulated our interest in the possibility of regional treatment
for such patients. 
Initially, we evaluated
the natural history of patients at our institutions who presented with isolated
metastatic disease to define whether our subset of patients was similar to those
with an expected survival of four to six months after diagnosis of synchronous
or metachronous liver metastases from colon and rectum adenocarcinoma. The vast
majority of deaths in patients with liver metastases are caused by liver failure.
We and other institutions treating secondary and tertiary referral patients
may be dealing with a patient population that has a biologically advantaged
natural history. In a consecutive group of 125 patients who were seen at Brigham
and Women's Hospital and at Dana-Farber Cancer Institute with liver-only metastatic
disease (all with at least five-year follow-up), median survival was 12.5 months.
Survival correlated with histology as defined by conventional morphologic criteria.
The majority of patients had moderate differentiated histology with a median
survival of 17 months. The small subset of patients with poorly differentiated
or undifferentiated carcinoma had a distinctly poorer median survival of six
months. Equally important, these patients did not die of disease that remained
isolated to the liver. This was the first indication in our own studies of colon
and rectum carcinoma that poorly differentiated disease presented a characteristic
natural history. Subsequent to this study and confirming studies at other institutions,
patients found to have poorly differentiated morphology as a major component
of their metastatic disease, even if initially isolated to a single site, are
treated as systemic disease failures since the initial presentation quickly
foreshadows diffuse extrahepatic disease manifestation.
Our studies showed that
the more extensive the disease, the sooner the patients died. Most of our patients
had limited disease, and those who had three or fewer liver nodules (as defined
by the most frequently used study at that time - technetium liver/spleen scan)
had a median survival of 24 months - a better rate than that reported in various
surgical and medical oncologic survey textbooks.[3,4] Most were asymptomatic
and had no disability, and they were classified as so-called "zero,"
according to the Eastern Cooperative Oncology Group performance status, with
a median survival of 18 months. Finally, none of the 125 patients in this natural
history series had any effective therapy for liver metastases, and most had
no therapy attempted whatsoever. Despite the fact that they survived longer
than was commonly assumed, thereby representing a biologically select subset
of patients, they all died within five years of follow-up, mostly of consequences
of inexorable growth of tumor in the liver.
The dominant message from
this natural history study is obvious. Historical controls (particularly those
that advertise four to six-month survival after diagnosis of synchronous or
metachronous liver metastases) are not legitimate in judging the efficacy of
regional treatment, whether it be surgery, regional chemotherapy, or a combination
of the two. A prospective nontreatment control group should be defined or, at
the very least, the historical controls to be used should represent a patient
subset that is equivalent to that defined above.
Hepatic Arterial Fluorodeoxyuridine
Treatment
Our first therapeutic involvement
in patients with liver-dominant or liver-only metastatic disease from colon
and rectum adenocarcinoma was a reinitiation of an earlier investigation[5]
in which percutaneous polyethylene catheters were placed via the brachial artery
into the hepatic artery and fluorodeoxyuridine (FUdR) infused for approximately
three weeks. The rationale for the three-week treatment period was not biologic;
rather, it was based on the fact that complications (eg, thrombosis of the vessel,
intimal disruption, extravasation of the catheter through the hepatic artery,
and/or dislodgment migration of the catheter or infection at the percutaneous
puncture site) dictated the three-week delivery as the maximum tolerable duration.
The clinical
utility of the Infusaid pump (Metal Bellows Corporation, Sharon, Mass) allowed
reexamination of the continuous infusion concept because of the technical elegance
of the actual infusion system. The technical advantages of the pump (the fact
that it could be fully implanted, the availability of a silastic egress catheter
that could be surgically placed for long periods of time in precisely the right
location in the proper hepatic artery via a gastroduodenal arteriotomy in most
patients, and the definition of flow into the tumor itself by means of macroaggregated
albumin isotope scans) allowed a longer duration of treatment. Needless to say,
individual institutional studies varied in their initial enthusiasm and in their
range of response and survival data. For example, anatomic responses were cited
in 55 or 76 patients at the University of Michigan compared with 7 of 17 at
the Swedish Hospital in Seattle, 5 of 17 at the Dana-Farber Cancer Institute,
and 9 of 31 at the University of Chicago. Our own experience was pessimistic.
However, at present, numerous national and international multi-institutional
studies have evolved to common endpoints (Table 2). Response of metastatic disease
in the liver is significantly better when intra-arterial infusion was compared
with systemic infusion. However, median survival in the two groups was essentially
the same. Since treatment crossover was allowed in several studies when patients
failed either of the treatment modalities, survival could not be accurately
assessed. However, increased extrahepatic disease occurred in the patients who
had increased response probability for liver metastases with hepatic arterial
treatment. In addition, the need for surgical implantation (ie, an operation)
and regional infusion-related toxicities were significant additional detriments.
The sclerosing, cholangitis-like, dose-related toxicity was not always reversible
after cessation of treatment. 
Several points can be made.
First, hepatic arterial FUdR treatment should be considered in patients who
have symptomatic liver-only or liver-predominant metastatic disease from colon
and rectum cancer. The goal, however, is symptomatic relief rather than extension
of life. Second, since most of these patients are asymptomatic, standard application
of hepatic arterial infusions using FUdR or FUdR and mitomycin is not justified
at the present time outside of formal clinical trials. Our group supports regional
treatment trials, particularly as an adjuvant to liver metastasis surgery because
of the high incidence and rapid development of tumor failure in the residual
liver, and systemically in most patients, even after successful resection or
cryosurgical ablation of isolated liver metastases. In addition, there is reasonable
intellectual rationale for comparisons of regional infusion vs systemic infusion
of cytokines with and without chemotherapy. Such applications should be performed
as part of formal phase I, II, or III clinical protocols.
Surgical Resection
Our initial consideration
of surgical resection of hepatic metastases was a consequence of the disappointing
results from hepatic arterial FUdR infusion, as well as the fact that we had
built up a considerable momentum in patient referral of the patient subset represented
by the natural history previously outlined. In addition, the technical and anatomic
considerations spun off from experience with liver transplant in the mid to
late 1970s were beginning to impact on lobar, extralobar, and segmental surgical
resection approaches to the liver. We and other single institutions reproduced
what Foster et al[6] summarized and what Wilson et al[7] prospectively proved
in their seminal publications. Our first surgical resection experience with
40 patients who were operated on at the Brigham and Women's Hospital was straightforward.
For the patients who were selected appropriately (ie, all patients were asymptomatic
at presentation, had three or fewer liver metastases, and had either moderate
or well-differentiated morphology), the expected five-year survival was no different
than in a highly selected group of patients with resectable hepatocellular carcinomas
in noncirrhotic livers. The key question in this and all single-institution
studies was what extent of survival was based on the effectiveness of the surgical
therapy vs surgery as a selection process - particularly since we had already
shown that such patients comprised a relatively biologically advantaged, highly
selected group, even without any effective therapy.
As one indication of the
effects of selection, our exclusion criteria weakened as we increased our accrual
of patients with liver-only or liver-predominant metastatic disease for 
liver resection. This was followed by a decrease in expected disease-free and
overall survival for the expanded resection group of over 100 patients. Subsets
were analyzed retrospectively, but the real proof of the effects of selection
is the application of prognostic rules to prospective study. Until the Gastrointestinal
Tumor Study Group 6584,[8] no such prospective studies were available in which
any purported prognostic variables could be tested.
One of the best attempts
to define a large (ie, unselected) patient denominator was the retrospective
collection of metastatic resection patients assembled in the mid 1980s.[9] This
retrospective collection is no better or worse in credibility of accrual or
follow-up than the individual institutional retrospective series; however, the
aggregation of 800 to 900 patients allows for a more realistic overall postresection
natural-history experience. Although the disease-free and overall survival plateaus
were lower compared with those in many of the individual series, the survival
curves plateaued, thus providing evidence of cure for some patients after successful
resection (Fig 1). The problem with concluding that these data showed evidence
of cure was the adequacy of the follow-up, particularly since all data were
defined retrospectively. How confident can one be about the plateauing survival
curves, even in this retrospectively collected, multi-institutional series?
In an attempt
to better define the therapeutic benefit of surgery, the Gastrointestinal Tumor
Study Group in 1984 defined protocol 6584 (Figs 2 through 4). The objectives
were to determine (1) the proportion of explored patients with resectable lesions
(ie, how good were the preoperative staging studies), (2) the important prognostic
variables, (3) the surgical
morbidity and mortality
in this multi-institutional group in which all patients were followed from the
time of accrual on the basis of their preoperative staging, and (4) survival
among the unresected, curatively resected, and noncuratively resected patients.
Unresected patients were
defined as those in whom the intraoperative staging showed unsuspected additional
liver disease or 
liver disease that could not be encompassed by any conventional surgical resection,
as well as patients with previously unsuspected (by the preoperative staging
studies) extrahepatic disease. Patients defined as noncuratively resected were
those who the surgeon felt had tumor-free resection margins but in whom the
pathologist defined tumor present at less than 1 cm from the cut edge of the
liver.
This study remains the single
prospective, multi-institutional approach to the above questions. Admittedly,
in the absence of a randomized, prospective study in which surgically staged
resectable patients would be simply followed vs resected and followed (a study
that would be neither practical nor ethical), this prospective treatment plan
will not conclusively differentiate the effect of surgery as a selection agent
vs a therapeutic modality. Nevertheless, it was as good as could be done at
the time or, in fact, subsequently.
Supplemental data on the
majority of patients in whom CT scan was obtained show the inadequacy of preoperative
staging. In 139 patients, CT scanning missed hepatic disease in 10 patients,
underestimated liver involvement in 46 patients, and overestimated liver involvement
in five patients. In addition, CT scanning missed extrahepatic disease in 17
of 144 patients. Although subsequent refinements in CT technology and application
of magnetic resonance imaging may have improved the resolution capacity for
liver metastases staging and for extrahepatic disease, our impression is that
with the exception of CT portography or 
translaparoscopic or intraoperative ultrasonography, preoperative staging remains
grossly inadequate.[10]
Several immediately apparent
outcomes were surprising. As shown in the initial report of this trial[8] (Table
3), the amount of blood used for curative or noncurative resection was remarkably
low, even in the mid to late 1980s. To put this in historical perspective, between
six and 12 units of bank blood replacement were commonly used during major hepatic
resection 10 to 15 years ago, while at present, particularly with the use of
t he cell saver and autologous blood, a good deal of liver surgery is performed
using no bank blood whatsoever. Length of stay seemed to be reasonable, given
the date of the trial (before much of the recent financial pressure to decrease
length of stay) and the extent of the surgery. We believe that this trial provided
the first evidence that major techniques of liver resection were "exportable"
to institutions that were not necessarily tertiary care facilities. At present,
our own group's median length of stay after major liver resection is 10 days.
The effects of the advent of intraoperative ultrasonography (IOUS), cryosurgical
ablation, and resections aided by the Cavitron ultrasonic surgical aspirator
also are apparent from the data.
Patterns of recurrence were similar to numerous individual institutional experiences.
The patients were at high risk for residual liver recurrence as well as extrahepatic
recurrence. Therefore, any adjuvant treatment protocols should be designed to
compare systemic therapy to surgery alone or systemic and regional therapy combined
with resection vs metastasis resection alone. Although these studies are ethically
compelling, they have been extraordinarily difficult to mount. The 4.6% 30-day
mortality and the 13% incidence of abscess or biloma were acceptable, given
the date of the trial and the multi-institutional accrual. Most of the bilomas
could be drained percutaneously, with many not even necessitating hospitalization.
The most important outcome
was survival. At the time of initial publication, the overall and disease-free
survival of the entire group had not plateaued. At median follow-up of approximately
24 months, no hint of any plateau was seen in either overall or disease-free
survival, even for the curatively resected patients (Fig 5).[9] The only signifi
cant information was that noncuratively resected and unresectable patients had
the same median survival of approximately 20 months, while the 36-month median
survival of curatively resected patients was significantly higher. Conclusions
at this early publication were straightforward: if it could be determined before
the parenchymal transection that adequate margins could not be obtained, then
noncurative or any debulking hepatic metastasis resection was not justified.
The introduction of IOUS has made the prediction of resection adequacy more
accurate.
More recently, follow-up on this initial (and only) prospective hepatic metastasis
resection surgery study has indicated that almost none of the patients are ultimately
cured of disease, although six patients remain alive (minimum follow-up is six
years from the inclusion in the study) (Fig 6).[11] The implication is that
although we are achieving a significant number of "complete responses"
when the medical oncologist's criteria for success in treatment of metastatic
or systemic cancer are used, we are probably curing very few of these patients.
The broader conclusion from this realization is that since we are not ultimately
curing these asymptomatic patients, we should emphasize the application of additional
systemic or regional therapy in a trial-type format to this group of patients
after surgery and continue to decrease the morbidity and mortality of surgical
techniques.
Intraoperative Ultrasonography
The most interesting (and
most publicized) aspects of the development of IOUS transducers (Fig 7) are
(1) increased staging ability to more accurately define patients who should
be treated for liver-only metastatic disease and to achieve reasonable margins
by either resection or cryosurgical ablation and (2) an ability to define, in
real time, the adequacy of liquid nitrogen delivery in order to achieve destruction
of the tumor by freezing without sacrifice of "innocent bystander"
liver tissue in a segment or lobe, particularly when the tumor is relatively
small but deeply placed within the right or left lobe of the liver parenchyma.
The application of ultrasound
technology to surgical staging and to cryosurgical ablation of liver metastases
from colon and rectum carcinoma (and one should not extrapolate from the experience
in colon and rectum adenocarcinoma to any other histologic type of tumor) should
be
seen as a model for introduction of any new therapeutic or diagnostic technology.
The essential questions are: Does the new technology work? Does it replace something?
How does it benefit the patient?
Our initial postulates concerning
the introduction of cryosurgical ablation in patients with liver metastatic
disease have essentially all been disproven. Nevertheless, we have found that
IOUS is beneficial to all patients in terms of staging capability. This is obvious
in a number of trials[12] and is summarized in an initial prospective application
in comparing IOUS to preoperative imaging that was in competition at the time.
The resolution capacity of IOUS or the translaparoscopic ultrasound probes is
approximately three times that of all but CT portography, which is probably
comparable. We now can see lesions in the depths of the right and the left lobes
that are no larger than 3 to 4 mm in diameter. We also can predict when lesions
that might be otherwise clinically "insidious" - located at the take-off
of the right hepatic vein or next to the junction of the middle hepatic vein
and the cava - which might obviate obtaining clean margins after the technical
commitment for a parenchymal resection. Finally, with the advent of laparoscopic
staging and translaparoscopic ultrasound, we can define reasons for not moving
ahead in approximately 50% of such patients, thus obviating the need for a full
laparotomy.[13] It will be important that trocar-site recurrences in patients
with primary colon and rectum carcinoma are not experienced in patients with
advanced disease.
Cryosurgical Ablation
Other references have described
the efficacy of the cryosurgical ablation itself and the ability to monitor
real-time freeze margins that are as adequate as resection margins.[14] At our
last formal review of the Deaconess experience, 56 (80%) of 70 patients who
underwent cryosurgical ablation had isolated metastases from moderately to well-differentiated
colorectal carcinomas. Our survival experience parallels the follow-up data
from the Gastrointestinal Tumor Study Group. Among the patients who had no residual
disease (75% of our overall treatment group), the expected disease-free survival,
overall survival, and patterns of failure are comparable to patients who underwent
liver resection with adequate resection margins (defined by both the surgeon
and the pathologist). The patients with residual disease, after either resection
or cryosurgical ablation, fared considerably worse with a median survival of
approximately 19 months. In our series and in the series of Morris et al (personal
communication) and Ravikumar et al[14](in which inclusion criteria are comparable),
the morbidity and mortality of the cryosurgical ablation is considerably less
than those of resection. No mortalities have occurred in our patients, and the
median length of stay is five da ys compared with that for resection, with between
10 and 14 days' median survival and a mortality between 3% and 5%.
Cryosurgical ablation allows
us to try an option in cancer control that is no better, but certainly no worse,
than resection. Morbidity is decreased and mortality is eradicated when patients
are selected and treated appropriately. However, cryosurgery has not allowed
us to expand our treatment to more profuse liver metastases. This is a consequence
of our realization that the biology of patients who have more than a limited
number of metastases indicates additional diffuse disease in the liver or systemic
disease that is not altered by any regional approach. Cryosurgery does not eradicate
tumors that are contiguous to major portal or hepatic venous structures due
to the heat-sink effect from high-volume blood flow. Such tumors must be resected
with adequate sleeves taken from these vessels after appropriate vascular control.
Cryosurgical ablation cannot be substituted since it cannot be performed percutaneously,
at least in the patients referred to us. At the present time, percutaneous cryosurgical
ablation is a merchandising concept that we regard with caution. Finally, patients
who have undergone cryosurgical ablation or resection can be retreated with
either a second ablation or a second resection if they present with recurrent
isolated liver metastases in either the residual or regrown liver.
Ten percent to 15% of patients
who present with liver-only metastatic disease from colon and rectum adenocarcinoma
can be treated with ablation rather than resection. To select this subset of
patients, conventional criteria can be used that are based on well-defined outcome
studies. We do not believe that a randomized trial is necessary or could be
done with adequate accrual), although there is still much discussion about this
among our surgical colleagues.
Conclusions
No longer is the diagnosis
of metastasis to the liver from colon and rectum adenocarcinoma a summary death
sentence. Disease that is localized to the liver usually is biologically more
favorable than when extensive metastasis is present, and a variety of local/regional
treatment techniques are available that can significantly extend survival. Continued
research is required to define effective adjuvants to such local/regional approaches
and to minimize the regrowth of tumor in the liver or at distant sites.
References
- Steele G Jr, Jessup LM,
Winchester DP, et al. Clinical highlights from the national Cancer Data Base:
1995. CA Cancer J Clin. 1995;45:102-111.
- Thomas P, Petrick AT,
Toth CA, et al. A peptide sequence on carcinoembryonic antigen binds to a
80kD protein on Kupffer cells. Biochem Biophys Res Comm. 1992;188:671-677.
- Schwartz SI, Shires GT,
Spencer FC, et al, eds. Principles of Surgery. 5th ed. New York, NY: McGraw-Hill
Book Co; 1989.
- DeVita VT Jr, Hellman
S, Rosenberg SA, eds. Cancer: Principles & Practice of Oncology. 3rd ed.
Philadelphia, Pa: JB Lippincott Co; 1989.
- Cady B, Monson DO, Swinton
NW. Survival of patients after colonic resection for carcinoma with simultaneous
liver metastases. Surg Gynecol Obstet. 1970;131:697-700.
- Foster JH, Berman MM.
Solid liver tumors. In: Ebert PA, ed. Major Problems in Clinical Surgery.
Philadelphia, Pa: WB Saunders Co; 1977;22:1-342.
- Wilson SM, Adson MA.
Surgical treatment of hepatic metastases from colorectal cancer. Arch Surg.
1976;111:330-334.
- Steele G Jr, Bleday R,
Mayer RJ, et al. A prospective evaluation of hepatic resection for colorectal
carcinoma metastases to the liver: Gastrointestinal Tumor Study Group Protocol
6584. J Clin Oncol. 1991;9:1105-1112.
- Hughes KS, Simon R, Songhorabodi
S, et al. Resection of the liver for colorectal carcinoma metastases: a multi-institutional
study of patterns of recurrence. Surgery. 1986;100:278-284.
- Kane RA, Hughes LA, Steele
G Jr, et al. Impact of intraoperative ultrasound of the liver on surgical
decision-making. IHPBA. 1995. In press.
- Steele G Jr, Mayer R,
Lindblad A. A prospective evaluation of hepatic resection for colorectal carcinoma
metastases to the liver: follow-up report. J Hepatic Biliary Pancreatic Surg.
1995;2:122-125.
- Stone MD, Kane R, Both
A Jr, et al. Intraoperative ultrasound imaging of the liver at the time of
colorectal cancer resection. Arch Surg. 1994;129:431-436.
- Babineau TJ, Lewis WD,
Jenkins RL, et al. Role of staging laparoscopy in the treatment of hepatic
malignancy. Am J Surg. 1994;167:151-155.
- Ravikumar TS, Kane R,
Cady B, et al. A five-year study of cryosurgery in the therapy of liver tumors.
Arch Surg. 1991;126:1520-1524.
From the Department of Surgery
at New England Deaconess Hospital, Harvard Medical School, Boston, Mass. Dr.
Steele is now the Dean of Biological Sciences at the University of Chicago.
Back
to Cancer Control Journal Volume 3 Number 1