Volume 210, Number 2
In General Surgery
Resection of Hepatic Metastases
from Colorectal Cancer
GLENN STEELE, JR., M.D., and T.S. RAVIKUMAR, M.D.
During the past decade the results of slightly fewer than 1000
resections of liver metastases from colorectal carcinoma have
been analyzed, retrospectively reanalyzed, and reviewed. The
following are confirmed conclusions: major liver resection can
be performed safely (less than a 5% operative mortality rate);
20% to 25% ofthese patients are cured; no other regional therapy
options have any curative potential. The following caveats are
also obvious: most patients who are operated on are not cured;
although predictors have been proposed to select patients most
likely to benefit from surgery, none is discriminating in and of
itself; most therapy questions in this group of patients have not
been addressed in any formal way, surgery for isolated regionally
recurrent colon and rectum carcinoma remains an important
stopgap only until effective systemic therapy is discovered. This
review of our own and other single and multi-institutional pro-
spective and retrospective data will be framed by the following
questions. (1) Does resection of liver metastases cure patients
or simply select those who would have survived in the long-term
without any therapy? (2) In the absence ofany formalized, prop-
erly designed trial, how can one judge the benefit of resection?
(3) Why do metastases recur only in the liver? (4) What new
therapies should focus on the predominant secondary failure sites
in the majority of patients who do not benefit from hepatic me-
From the Department of Surgery, New England Deaconess
Hospital, Harvard Medical School, Boston, Massachusetts
Selection of Patients
Before examining the outcome of resection of hepatic
metastasis from colorectal carcinoma, one must determine
the degree of selection for patients with isolated, poten-
tially resectable liver metastases. Pestana et al.' in an au-
topsy series, reported that the liver was involved in 70%
ofall patients who died from disseminated colon and rec-
tum carcinoma. Because 147,000 new cases ofcolon and
rectum carcinoma will be diagnosed this year and roughly
50% ofpatients will experience recurrence within 5 years,
liver metastasis remains a prevalent mode offailure. Fur-
thermore, numerous studies220 of the natural history of
colorectal carcinoma, including our own, have shown that
liver metastasis is the primary determinant ofpatient sur-
vival. However, this large number of patients with the
liver as one site of disseminated disease is irrelevant to
the surgeon. For effective regional therapy, there must be
no extraregional disease. As shown in recent prospective
adjuvant treatment trials,21'22 the liver is the first or only
site offailure in fewer than 20% ofpatientswho experience
recurrence after primary colon and rectum cancer resec-
Correspondence and reprint requests to: Glenn Steele, Jr., M.D., De-
partment ofSurgery, New England Deaconess Hospital, Harvard Medical
School, 110 Francis Street, Boston, MA 02215.
Accepted for publication: February 8, 1989.
STEELE AND RAVIKUMAR
tion. Some of these patients who present initially with
regional disease (for instance, those with poorly differ-
entiated adenocarcinoma) will show rapid proliferation
of disease outside the liver. In addition, one third of the
patients who have liver only or liver-predominant colo-
rectal cancer metastases staged by any of the presently
available noninvasive diagnostic tests will be found at
surgery to have extrahepatic disease or metastases in the
liver not amenable to surgical resection.23 New staging
techniques, such as intraoperative ultrasound (IOUS),24
will exclude even more patients found to have additional
deep liver metastases as small as 2 mm to 4 mm in di-
ameter. As the surgeon becomes stricter in excluding pa-
tients with multiple liver lesions or with disease outside
the liver, resection results will seem better. Furthermore,
as the surgeon selects patients with more indolent disease
(i.e., metastases isolated to the liver or lung that grow
slowly and do not spread to other organs), results from
resection will also seem better. The varying selection bias
may explain differences in outcome among the many sin-
gle institution hepatic metastasis resection series. Com-
posite series will offer a more realistic survival estimate
but are impossible to analyze because individual surgeon's
criteria for patient exclusion is not possible to extract.
Adson et al.'6 and Fortner25 have attempted to estimate
the maximum potential therapeutic benefit ifhepatic me-
tastasis resection were applied to all appropriately selected
patients with colon and rectum carcinoma recurrence.
We have modified their estimates as follows. If one as-
sumes that 50% ofthe 147,000 patients who will be newly
diagnosed with colon and rectum carcinoma this year will
fail within 5 years, approximately 70,000 patients will be
at risk for liver metastases as part of their recurrence.
Eventually, 60% to 70% of these patients will have the
liver as one involved organ. However, the liver will be the
first or predominant site of failure in only 20% of these
patients. Thus at most, 9000 patients could be considered
for liver resection. However, if one assumes some prob-
ability ofcomorbid disease that would exclude any major
operation, and if one assumes that some patients would
rapidly show progressive extrahepatic disease, then per-
haps another 1000 to 1500 patients would be excluded
from surgery. At the time of surgery, one third of these
7500 patients would be ineligible for resection.226 Thus,
approximately 5000 patients would be potential candi-
dates for resection each year. Ifone accepts the cumulative
operative mortality rate of 5%, 250 patients would not
survive the perioperative period. With the theoretical 4750
patients operated on and alive after their operations, and
a 20% to 25% 5- and 10-year disease-free survival, it is
apparent that using the most optimistic estimates no more
than 1000 patients per year could be cured by surgical
resection. This puts the clinical discussion into proper
The Natural History of Hepatic Metastases
At the time of surgery, no surgeon who finds that his
or her patient has a resectable liver metastasis will ran-
domize that patient to resection versus sham laparotomy.
Analysis ofpatient benefit from hepatic metastasis resec-
tion, therefore, must include comparison to outcome in
natural history series that include comparably selected
With the single exception of Adson et al.,27'28 all sur-
geons who report their hepatic metastasis resection series
have stated that their only goal was to cure the patient.
Although Adson's palliative approach for some ofhis pa-
tients and his attempt to define a postoperative quality-
of-life index is interesting, most patientswho present with
pain and deterioration of liver function chemistries or
synthetic function have, in general, far-advanced liver tu-
mors that should not be considered for major surgical
resection. In our own series,23'29 and in all single institu-
tional series with the exception of Dr. Adson's, patients
are operated on with curative intent. By and large these
patients are asymptomatic or only mildly symptomatic
at the time their liver metastases are diagnosed and when
they come to surgery. In addition, most patients have no
comorbid disease. Such selected surgical patients, then,
cannot be compared to the general population ofpatients
with synchronous or metachronous liver metastases from
colon and rectum cancer who have been predicted to have
a mean survival rate ofonly 6 to 9 months after diagnosis.'
Among the natural history studies listed in Table 1,
several have characterized survival among selected pa-
tients similar to those included in hepatic resection series.
These include 1970 and 1983 reports by Cady et al.,6"14
the Goslin et al.2 study published in 1982, and the recent
Finley and McArdle study20 ofoccult hepatic metastases.
The patients with limited liver involvement described in
these natural history series are undoubtedly more anal-
ogous to those on whom we operate. The benefits of our
surgical therapy should be evaluated by a comparison with
these historical nontreated controls.
Our investigation of 125 patients who were referred to
a tertiary cancer center during an era when hepatic me-
tastasis resection was not in vogue is representative ofthe
other natural history series in which most patients were
asymptomatic at the time of diagnosis. The following
conclusions were evident. First, the simplest way to es-
tablish prognosis was to ask the patient how he or she felt.
In the majority of our patients there was no deficit in
work or life style, so called Eastern Cooperative Oncology
Group (ECOG) functional status 0. As can be seen in
Figure 1A, the patients with no dysfunction had a median
survival of 18 months, regardless of the extent of their
liver disease. Second, patients who were found to have
more disease died more quickly than those with less dis-
Ann.Surg* August 1989
RESECTION OF HEPATIC METASTASES
TABLE 1. Liver Metastasesfrom Colorectal Cancer: Natural History Series
Pestana C, Reitemeier RJ, Moertel CG, et al.
Jaffe BM, Donegan WL, Watson F, et al.
Bengmark S, Hafstrom L
Oxley FM, Ellis H
Cady B, Monson DO, Swinton NW
Abrams MS, Lerner HJ
Neilson J, Balslev I, Jensen H-E
Baden H, Anderson B
Wood CB, Gillis CR, Blumgart LH
Welch JP, Donaldson GA
Bengsson G, Carlsson G, Hafstrom L, et al.
Goslin R, Steele G Jr, Zamcheck N, et al.
Levine AN, Donegan WL, Irwin M
Lahr CJ, Scong S-J, Cloud G, et al.
Wagner JS, Adson MA, Van Heerden JA, et al.
Wood CB. In: Van de Velde CJH, Sugarbaker P (eds.)
Finan PJ, Marshall RJ, Cooper EH, et al.
Fujimoto S, Miyazaki M, Kitsukawa Y, et al.
Finlay IG, McArdle CS
Am J Surg 1964; 108:826-9
Surg Gynecol Obstet 1968; 127:1-11
Cancer 1969; 23:198-202
Br J Surg 1969; 56:149-152
Surg Gynecol Obstet 1970; 131:697-700
Dis Colon Rectum 1971; 14:431-4
Acta Chir Scand 1971; 137:463-5
Scand J Gastroenterol 1975; 10:221-3
Clin Oncol 1976; 2:285-8
Am J Surg 1978; 135:505-11
Am J Surg 1981; 141:586-9
Dis Colon Rectum 1982; 25:749-54
JAMA 1982; 247:2809-10
Semin Oncol 1983; 10:127-33
J Clin Oncol 1983; 1:720-6
Ann Surg 1984; 199:502-7
Dordrecht: MartinusNijhoff 1984; 47-54
Br J Surg 1985; 72:373-7
Dis Colon Rectum 1985; 28:588-91
Br J Surg 1986; 73:732-5
ease in their liver (Fig. IB). Individuals who were asymp-
tomatic at presentation and who had fewer than three
metastases had a median survival of24 months, which is
considerably longer than the 4- to 6-month survival pe-
riods reported in unselected populations ofpatients with
synchronous or metachronous liver metastases. This and
other natural history series have shown that a select group
ofpatients with good biology liver metastases from colon
and rectum carcinoma will live a long time. However,
death is the uniform outcome, and in ourexperience death
is always related to the inexorable tumor growth in the
Finley and McArdle20 have proposed that perhaps as
many as 20% to 25% of patients who have no known
metastatic disease at the time oftheir primary colon and
rectum cancer surgery may have occult liver metastases.
These investigators have attempted to define the doubling
time of liver tumors in patients with purported occult
disease compared to the doubling time ofliver tumors in
patients who have known metastases at the time of pri-
mary tumor resection.30 The biologic approach is good,
even if impractical. Theoretically, the surgeon could do
with liver metastases what has already been done with
pulmonary metastases from sarcoma, i.e., calculate the
# of Liver Nodules
FiGS. IA and B. (A) Survival rate and performance status at presentation.
(B) Survival rate and number of liver nodules.2
Vol. 210- No. 2
STEELE AND RAVIKUMAR
doubling time of a given liver metastasis to estimate the
proliferative thrust or biology ofthe disease before making
the decision to resect. Unfortunately, however, the pre-
cision ofestimating doubling time even ofknown metas-
tases is hampered by inadequate resolution capacity of
most noninvasive diagnostic liver imaging techniques.
Doubling time estimation is further limited by imprecise
estimates ofwhen the actual metastasis occurred. Varia-
tions in tumor growth, dependence on tumor volume (so-
called Gompertzian growth kinetics), variations in tumor
spread to other organs not related to tumor volume, and
variations in multifactorial host antitumor response all
make the biology of a given tumor impossible to define.
Thus, the surgeon is left with clinical criteria for selecting
his or her patients and will most often operate on the
asymptomatic patient with less liver tumor who is likely
to have a reasonable outcome.
Two aspects in this comparison between the natural
history and resected groups will be particularly important:
(1) patterns offailure or death with and without successful
surgical resection, and (2) examination ofplateaus in sur-
vival curves among the "best biology patients" selected
in nonsurgical "historical control" and resection series.
Changes in patterns of failure after surgery imply that
surgery is, in fact, accomplishing something. A plateau in
the survival curve indicates that some patients are cured.
Presently available preoperative staging techniques lead
the surgeon to explore 30% or more patients who are
found not to be amenable to hepatic metastasis resec-
tion.2326 Contraindications include unsuspected extra-
hepatic disease (usually periportal
nodes) or additional disease in the liver that makes resec-
tion technically or biologically impractical.
In our own experience, ultrasound is the most effective
way to stage the liver. A caveat is the qualitative aspect
ofultrasound interpretation. Even with the best and most
interested ultrasonographer, however, resolution capacity
is limited to liver lesions equal to or greater than
diameter. CT scan, particularly with hepatocyte or Kupffer
cell dyes, may improve resolution capacity, but presently
CT is no better than ultrasound and may even be less
useful when attempting to define lesions in the left lateral
hepatic segment. Particular advantages of magnetic res-
onance imaging in defining liver lesions and extrahepatic
disease are still being investigated.3'
In a recent comparison ofpreoperative imaging studies
with intraoperative staging we added intraoperative ul-
trasonography (IOUS) to our repertoire.32 We prospec-
tively evaluated 117 patients; 167 lesions were detected
by IOUS. Of these, 61%, 76%, and 52% were detected by
preoperative CT scan, ultrasound, and angiography, re-
1 cm in
spectively. IOUS was superior to intraoperative visual ex-
amination or bimanual palpation of the liver, as well.
Parenchymal lesions as small as 2 mm to 4 mm could be
detected in the depths of the right and the left lobes. We
now routinely apply the IOUS transducer over the anterior
and posterior surfaces ofright and left lobes after the liver
is freed from its diaphragmatic attachments before any
commitment to resection. Use ofIOUS to stage the liver
at the time of primary colon or rectum cancer resection
might provide valuable natural history data, particularly
regarding occult liver metastases. Because, however, there
is no direct therapeutic benefit to the patient from this
added information, we reserve IOUS at the time of pri-
mary colon and rectum carcinoma resection for experi-
Additional routine studies performed before liver re-
section include lung tomograms or pulmonary CT scans.
In our early series,23 plain roentgenograms of the chest
missed occult pulmonary metastases found by one ofthe
other two imaging modalities in 10% to 15% of patients.
Our need to more effectively exclude occult pulmonary
disease was evident when analyzing patterns of failure.
Hepatic arteriography is performned before operation
by our group and by most surgeons who perform hepatic
resection. This is not done to find additional liver lesions,
although the combination of arteriography and CT scan
may be helpful to do so. It is simply to provide a road
map of the varied celiac arterial plexus. Particularly in
cases in which the major right hepatic arterial inflow orig-
inates from the superior mesenteric artery or when the
left hepatic artery originates from the left gastric, it is
helpful to know the anatomy before dissection ofthe porta
Finally, all patients are colonoscoped before surgery to
rule out the existence of polyps, suture line recurrence,
or metachronous cancers.
The major individual institutional series analyzing re-
sults from resecting colorectal cancer hepatic metastases
are listed in Table 2. Collected series, beginning with Fos-
ter's seminal work in 1977,33 are listed in Table 3. A com-
posite ofpatient numbers and factors predicting outcome
are summarized in Table 4.
All hepatic metastasis resection series, with the excep-
tion ofa single unpublished prospective registry from the
Gastrointestinal Tumor Study Group, are retrospective.
All, including the Gastrointestinal Tumor Study Group
prospective study, have no concomitant nonsurgically
treated controls. It is important to distinguish the collected
retrospective analyses, such as the Hughes et al.3436 he-
patic metastasis registry reports from the single institution
analyses.37175 Although Hughes and coworkers have ac-
Ann. Surg.-August 1989
RESECTION OF HEPATIC METASTASES
TABLE 2. Single Institutional Series ofColorectal Cancer Liver Metastasis Resection
Pack GT, Brasfield RD
Attiyeh F, Wanebo HJ, Sterns MW
Wanebo HJ, Semoglov C, Attiyeh F, et al.
Fortner JG, Silva JS, Golbey RB, et al.
Butler J, Attiyeh F, Daly JM
Flanagan L, Foster JH
Foster JH, Lundy J
Wilson SM, Adson MA
Adson MA, Van Heerden JA
Adson MA, Van Heerden JA, et al.
Logan SE, Meier SJ, Ramming KP, et al.
Thompson HH, Tompkins RK, Longmire WP
Taylor B, Langer B, Falk RE, et al.
Coburn CS, Makowka L, Langer B, et al.
Steele G Jr, Osteen RT, Wilson R, et al.
Cady B, McDermott WV
Cady B, Saenz N, Steele G Jr, et al.
Ekberg H, Tranberg K-G, Andersson R, et al.
Ekberg H, Tranberg K-G, Andersson R, et al.
Woodington GF, Waugh JM
Boey J, Choi TK, Wong T, et al.
Blumgart LH, Allison DJ
Morrow CE, Grage TB, Sutherland DER, et al.
Rajpal S, Dasmahapatra KS, Ledesma EJ, et al.
Iwatsuki S, Byers WS Jr, Starzl TE
Kortz WJ, Meyers WC, Hanks JB, et al.
Tomas de la Vega JE, Donahue JE, Doolas A, et al.
August DA, Sugarbaker PH, Ottow RT, et al.
Coppa GF, Eng K, Ranson JHC, et al.
Fegiz G, Bezzi M, DeAngelis R, et al.
Petrelli NJ, Nambisan RN, Herrera L
Gennari L, Doci R, Bozzetti F, et al.
Bozzetti F, Bignami P, Morabito A, et al.
Nordlinger B, Quilichini MA, Parc R, et al.
Sesto ME, Vogt DP, Hermann RE
Brown DA, Pommier RF, Woltering EA, et al.
Stephenson ICR, Steinberg SM, Hughes KS, et al.
Am J Surg 1955; 90:704-716
Dis Colon Rectum 1978; 21:160-2
Am J Surg 1978; 135:81-5
Ann Surg 1984; 198:306-16
Surg Gynecol Obstet 1986; 162:109-113
Am J Surg 1988; 155:378-82
Am J Surg 1967; 113:551-7
Cancer 1970; 26:493-502
Am J Surg 1978; 135:389-394
Curr Probl Surg 1981; 18:161-202
Arch Surg 1976; 111:330-4
Ann Surg 1980; 191:576-83
Arch Surg 1984; 119:647-51
World J Surg 1987; 11:511-20
Arch Surg 1982; 117:25-8
Ann Surg 1983; 197:375-88
Can J Surg 1983; 26:215-7
Surg Gynecol Obstet 1987; 165:239-46
Am J Surg 1984; 147:554-9
Ann Surg 1985; 201:204-9
Am J Surg In press
Br J Surg 1986; 73:727-31
World J Surg 1987; 11:541-7
Surg Clin North Am 1962; 42:507-18
Am J Surg 1963; 105:24-9
SugGynecolObstet 1981; 153:844-8
Worl J Surg 1982; 6:320-45
Surgery 1982; 92:610-4
Surg Gynecol Obstet 1982; 155:813-6
Ann Surg 1983; 197:247-53
Ann Surg 1984; 199:182-6
Surg Gynecol Obstet 1984; 158:46-8
Surg Gynecol Obstet 1984; 159:223-8
Ann Surg 1985; 201:210-8
Ann Surg 1985; 202:203-8
Ital J Surg Sci 1985; 15:259-65
Am J Surg 1985; 149:205-9
Ann Surg 1986; 203:49-54
Ann Surg 1987; 205:264-70
Ann Surg 1987; 205:256-63
Surgery 1987; 102:846-51
Arch Surg 1988; 123:1063-6
Ann Surg 1988; 203:679-87
cumulated significant data that provide a less-selected
picture ofoutcome, few ofthe data are new. Most ofthe
patients have also been included in individual series.
Composite data is particularly difficult to analyze because
specific selection criteria from individual participating
surgeons are no longer extractable from the larger data
pool. For this discussion we have focused on outcome
and predictors of outcome derived from the individual
series and we refer to the Hughes et al. collection only
where there are marked differences.
Since the watershed hepatic metastasis resection studies
ofFoster33'42-45 and Adson et al.,46-4' and the reemphasis
of liver anatomy in the technical prerequisites of liver
resection by Starzl and his colleagues7677in the mid to
later 1970s, the number ofseries reporting liver resection
has increased dramatically and may soon equal the num-
Vol. 210 * No. 2