Cancer and Metastasis Reviews 22: 465–472, 2003.

# 2003 Kluwer Academic Publishers. Manufactured in The Netherlands.

Intraperitoneal cancer dissemination: Mechanisms of the patterns of spread

C. Pablo Carmignani1, Tessa A. Sugarbaker2, Christina M. Bromley3, and Paul H. Sugarbaker*

1
Surgical Oncology, Washington Cancer Institute, Washington, DC; 2 Marin County General Hospital,
Marin County, CA; 3 Biostat Solutions Inc., 24311 Preakness Dr, Damascus, MD 20872

Key words: peritoneal carcinomatosis, mucinous carcinoma, appendix cancer, ovarian cancer, sarcoma

Summary

Background

Well known patterns govern the distribution of hematogenous and lymphatic metastasis of cancer. In the
past the distribution of cancer cells free within abdominal cavity received little attention and was thought to
be a random event. However, surgical observation led the authors to generate and test hypotheses regarding
patterns of spread that vary with tumor type, with the intraperitoneal environment, and with the
physiology of the peritoneal surface tissues.

Methods

The distribution and volume of peritoneal surface malignancy was prospectively recorded in 129 patients
with 5 different types of tumors at the time of cytoreductive surgery. The malignancies studied included
pseudomyxoma peritonei (PMP) of appendiceal origin, colonic mucinous adenocarcinoma (MA), non-
mucinous colonic adenocarcinoma (NMA), ovarian carcinoma (OV) and sarcoma (SA). The abdominal
and pelvic cavity was divided into 3 horizontal sectors, 9 regions and 25 sites. The incidences of tumor
implants in these designated areas were statistically analyzed for each tumor type and comparisons between
tumor types studied.

Results

The magnitude of intraperitoneal cancer dissemination was similar for mucinous tumors, including PMP
and MA and significantly higher than for non-mucinous tumors. Also the mucinous cancers were more likely
to be present in the upper horizontal sector than were non-mucinous. When NMA was compared to PMP
and MA the epigastric region was significantly less likely to contain tumor. For all cancer diagnoses the
colon, greater omentum and cul-de-sac of Douglas were most often affected. The ileocecal valve region was
more likely to have large tumor masses on its surface than small bowel surface or small bowel mesentery.

Conclusions

Peritoneal carcinomatosis had a wider distribution when mucinous fluid was present; this cancer
distribution by intraperitoneal fluid hydrodynamics occurred regardless of histologic aggressiveness. The
organs that have peritoneal fluid resorption (omentum and omental appendages) have a high incidence of

* Corresponding author.
E-mail: Paul.Sugarbaker@Medstar.net
466 Carmignani et al.

implants. Small bowel and its mesentery free to move by peristalsis had a reduced incidence of implants as
compared to the ileocecal area, which is fixed to the retroperitoneum. These observations may facilitate
efforts to treat peritoneal surface malignancy.

1. Introduction study. Data prospectively recorded included age,

sex, histopathologic diagnosis and grade (accord-
Hematogenous and lymphatic metastases of gas- ing to the Manual for Staging of Cancer), interval
trointestinal cancer follow a well defined pattern of between first surgery and cytoreduction, and
distribution [1,2]. The regional lymph nodes and tumor status in abdominopelvic sectors, regions
the liver are the initial sites for dissemination in a and sites (see below) for each patient [6]. Data
majority of patients. This knowledge is applied on were recorded in an electronic database.
a daily basis in the clinical management of patients
with abdominal or pelvic malignancy. Intraper-
itoneal cancer spread is a third mechanism of 2.B. Description of sectors, regions and sites
dissemination affecting a large number of patients
with gastrointestinal cancer, ovarian cancer and The peritoneal cavity was divided into three
abdominal or pelvic sarcoma. Peritoneal surface horizontal abdominopelvic sectors (upper, middle
dissemination can be a consequence of serosal and lower) and three vertical sectors (right, axial
invasion, perforation of a viscus by cancer, or the and left) (Figure 1). Also nine abdominopelvic
result of radiologic or surgical interventions for regions (umbilical, right upper, epigastrium, left
diagnosis or treatment [3–5]. upper, left flank, left lower, pelvis, right lower and
In the past it has been assumed that intraper- right flank) were identified. Similar methodology
itoneal cancer dissemination is a random process. has been used to assist in the selection of patients
However, observations of surgeons interested in for treatment of peritoneal surface malignancy [7].
the treatment of peritoneal surface malignancy In addition, 25 abdominopelvic sites were
have suggested patterns governing the dissemina- identified. These included the abdominal incision,
tion of cancer cells and that these patterns differ
with the primary site of the malignancy [3].
Based on prospective data gathered from 129
operative procedures for peritoneal carcinomatosis
and sarcomatosis, this study statistically searched
for patterns of intraperitoneal cancer dissemina-
tion and variations in these patterns. These data
were used in an attempt to accept or reject
hypotheses regarding mechanisms of intraperito-
neal cancer dissemination.

2. Methods

2.A. Study design

Peritoneal carcinomatosis was studied in 129

consecutive patients undergoing cytoreductive
surgery at a tertiary care urban teaching hospital
by the senior author. These patients with perito-
neal carcinomatosis and sarcomatosis came from
different areas of the US and other countries.
Observations were made during the first cytor- Figure 1. Abdominopelvic sectors and abdominopelvic
eductive surgery and confirmed by histologic regions.

anterior parietal peritoneum, body of stomach,
gastric antrum, ligament of Treitz, lesser omen-
tum, liver surface, right hemidiaphragm, left
hemidiaphragm, greater omentum, spleen surface,
pancreas, small bowel wall and mesentery, ileoce-
cal area, colon, right and left paracolic gutter,
right and left pelvic side wall, bladder, vagina, cul-
de-sac of Douglas, retroperitoneum and kidney.
During the surgical procedure each region and site
was evaluated to determine the presence or
absence of tumor. Following an all-or-none rule,
presence vs. absence of tumor was defined as
macroscopically detectable implants of any size vs.
no visible evidence of disease. When tumor
nodules were present, the lesion size was also
recorded as less than or equal to 5 cm and greater
than 5 cm. The number of patients with involve-
ment of each region and site for each type of
tumor were determined. Involvement of sectors
was determined by an addition of abdominopelvic
regions.
2.C. Incidences
The overall incidence for each tumor type was
calculated by dividing the total number of
abdominopelvic regions or sites involved with
disease by the total possible number of regions
or sites. For example, nine abdominopelvic
regions 6 46 patients with pseudomyxoma perito-
nei resulted in a total of 414 possible regions.
Three hundred and fifty-nine abdominopelvic
regions contained tumor masses for an incidence
of 87%. These data allow the numerical expression
of the extent of intraperitoneal distribution for a
type of cancer.
The incidence of a tumor type within a
particular abdominopelvic region or site was
defined as the ratio between the number of patients
that were positive for presence of tumor in this
region or site over the total number of patients
with this tumor type. These data allowed a
comparison of incidence of implants among
regions or sites within a disease.
2.D. Statistics
Descriptive statistics and cross-tabulations were
used to summarize characteristics of sectors,
regions and sites. w2 and p-values were calculated
Intraperitoneal cancer dissemination 467
for overall incidence of tumor distribution, tumor
distribution within abdominopelvic sectors and the
difference between overall distribution in the
epigastric region. The z-scores were used to
evaluate the incidence of tumor implants at a
particular abdominopelvic site as compared to the
overall incidence. The overall incidence was 51.3%
(1,653 positive abdominopelvic sites divided by
3,225 possible sites [129 patients 6 25 sites]). The
positive or negative standard deviation from the
mean was determined; the z-scores between
0.99
and 0.99 were not considered meaningful [8].
Tumor masses on mobile vs. fixed segments of
small intestine were compared by w2 based on
relative risk and 95% confidence intervals. A two-
sided p-value of less than 0.05 was considered as
significant.
3. Results
3.A. Patients characteristics
Of the 129 patients in this cohort 67 (51.9%) were
male and 62 (48.1%) were female. Median age was
48 years (range 53). Five different types of tumors
were encountered including 46 cases of pseudo-
myxoma peritonei of appendiceal origin (PMP,
35.6%). The histology of all the PMP tumors was
grade 1 mucinous type. There were 33 mucinous
adenocarcinomas of the colon (MA, 25.6%). The
histology was grade 2 mucinous adenocarcinoma
in 24 patients (72.7%) and grade 3 mucinous
adenocarcinoma in nine patients (27.3%). In 22
patients the histology was non-mucinous adeno-
carcinoma of the colon (NMA, 17%). In 12 of
them the histology was grade 2 (54.5%) and in 10
grade 3 (45.5%). In 12 patients surgery was for
ovarian carcinoma (OV, 9.3%). In three patients
the histology was grade 1, in two grade 2 and in
seven grade 3. In 16 patients the diagnosis was
sarcoma (SA, 12.4%). Five patients were grade 1,
six grade 2 and five grade 3. Prior surgery, with
diagnostic and/or therapeutic intent, had been
performed in 123 of the 129 patients (95.3%).
There was a median of 12 (range 316) and a mean
of 29.3 + 45.3 months between the initial surgery
and the cytoreduction.
468 Carmignani et al.
Figure 2. Comparison of overall incidences of cancer implants among the five different tumor types.
3.B. Overall incidence of tumor distribution was
increased in PMP and MA
The overall incidence of mucus-producing tumors
(PMP and MA) within the nine abdominopelvic
regions were similar and not significantly different
from each other (w2 ¼ 4.35, p ¼ 0.820) (Figure 2).
The distribution of mucinous tumors implants for
PMP and MA combined was significantly higher
than that of non-mucus producing tumors (NMA,
OV and SA; w2 ¼ 30.34, p ¼ 0.001, data not
shown).
The overall incidence of tumor distribution for
12 ovarian tumors, three of which produced
mucus, was significantly different when compared
to PMP ðw2 ¼ 21.15, p ¼ 0.007Þ and not-signifi-
cantly different when compared to MA
ðw2 ¼ 12.40, p ¼ 0.134Þ.
3.C. Tumor distribution within the upper horizontal
abdominopelvic sector was increased for mucinous
tumors
Differences in the distribution of tumor implants
to upper abdomen, middle abdomen and lower
abdomen and pelvis were assessed by comparing
incidences in the upper vs. middle vs. lower
horizontal sectors. No significant differences were
seen between mucus-producing tumors (PMP vs.
MA, w2 ¼ 3.28, p ¼ 0.194). When they were
compared to non-mucinous tumors, the presence
of mucinous cancer was significantly higher in the
upper sector (PMP vs. NMA, w2 ¼ 20.21,
p ¼ 0.001; MA vs. NMA, w2 ¼ 8.35, p ¼ 0.015;
MA and OV, w2 ¼ 6.67, p ¼ 0.036). The only
mucinous vs. non-mucinous tumor distribution
within the horizontal sectors that was not sig-
nificant was the comparison between MA and SA
ðw2 ¼ 4.68, p ¼ 0.097Þ. When NMA, OV and SA
incidences were compared among themselves,
differences were not significant (Table 1).
We also made a comparison between the upper
sector and the mean of middle and lower sectors.
Results were the same in terms of predominance of
upper sector involvement for PMP and MA (data
not shown).
3.D. Failure of NMA to disseminate to the
epigastric region
The overall incidence of implants throughout the
peritoneal cavity was compared to the incidence in
 Intraperitoneal cancer dissemination 469

Table 1. Comparison of abdominal horizontal sectors incidence 3.F. Tumor masses on mobile vs. fixed segments of
of peritoneal surface spread between 5 tumor types small intestine
Comparison Upper Middle Lower w2
(%) (%) (%) (p-value) The incidence of tumor masses greater than 5 cm
was compared in the ileocecal area, small bowel
PMP vs 89.9 89.1 81.2
MA 71.7 83.8 80.8 3.28 (0.194)
surface and small bowel mesentery. The ileocecal
PMP vs 89.9 89.1 81.2 area was 2.27 ðp ¼ 0.005Þ times more likely to
NMA 39.4 68.2 63.6 20.21 (0.001) have tumor masses present than small bowel
PMP vs 89.9 89.1 81.2 surface (95% Confidence interval: 1.28–4.02). It
Ov 50 72.2 80.6 18.72 (0.001) was 7.75 ðp ¼ 0.001Þ times more likely to develop
PMP vs 89.9 89.1 81.2
Sa 45.8 72.9 60.4 12.74 (0.002)
large tumor masses than small bowel mesentery
MA vs 71.7 83.8 80.8 (95% Confidence interval: 3.57–16.75) (Figure 4).
NMA 39.4 68.2 63.6 8.35 (0.015)
MA vs 71.7 83.8 80.8
Ov 50 72.2 80.6 6.67 (0.036) 4. Discussion
MA vs 71.7 83.8 80.8
SA 45.8 72.9 60.4 4.68 (0.097)
NMA vs 39.4 68.2 63.6 The presence of mucus ascites produced by all
Ov 50 72.2 80.6 2.15 (0.342) grades of mucinous adenocarcinoma resulted in a
NMA vs 39.4 68.2 63.6 wider distribution of cancer cells throughout the
Sa 45.8 72.9 60.4 1.65 (0.438) abdomen and pelvis. Mucinous tumors reliably
Ov vs 50 72.2 80.6
Sa 45.8 72.9 60.4 4.75 (0.093)
occupied spaces within abdominopelvic cavity that
are rarely involved when non-mucinous cancer
occurs, for example the epigastric region.
the epigastric region for each disease. The com- Previous publications have shown that intraper-
parisons obtained for PMP, MA and NMA are itoneal hydrodynamics directs peritoneal fluid up
shown in Figure 3. There was no sparing of to subdiaphragmatic spaces, explaining the well
epigastric cancer implantation for PMP or for MA known phenomena of subphrenic abscess after
ðw2 ¼ 0.21, p ¼ 0.646Þ. For NMA the overall acute appendicitis and Fitz-Hugh-Curtis Syn-
incidence of cancer implantation was 57% and drome [9,10]. It may be reasonable to assume
the incidence in the epigastric region only 27%. that the same fluid movement causes cancer cells
There was a significantly decreased ratio for NMA within mucinous ascites to frequently involve the
implantation in the epigastric region when com- upper abdominal spaces.
pared to PMP ðw2 ¼ 7.81, p ¼ 0.005Þ or to MA Data in Tables 1 and 3 show that all grades of
ðw2 ¼ 5.59, p ¼ 0.018Þ. cancer cells within ascites have similar patterns of
distribution and that this distribution is dependent
3.E. Deviation of implant incidence at a specific on physical rather than biologic (cancer aggres-
abdominopelvic site from the overall incidence – siveness) factors. Although MA and PMP have
z-scores very different invasive character and different
capabilities of hematogenous and lymphatic
The overall incidence of implants at 3,225 abdo- metastasis, they present similar patterns of perito-
minopelvic sites was 51.3%. Among the 25 neal surface distribution. In contrast MA and
different abdominopelvic sites studied the colon NMA have similar patterns of hematogenous and
was found to have the greatest positive deviation lymphatic metastasis but their peritoneal surface
from the mean with z-scores of 1.5 to 2.5. Greater distribution is different. The common feature for
omentum also had high z-scores except for ovarian PMP and MA is the presence of mucinous ascites,
tumors. The cul-de-sac of Douglas had high suggesting that mucus is a prominent facilitator of
z-scores except for NMA. The Treitz ligament, widespread intraperitoneal cancer distribution.
retroperitoneum and kidneys had lower z-scores. The motion of intraperitoneal fluid has a major
Complete information on the z-scores at the 25 role in the dissemination of malignant cells only
abdominopelvic sites is given in Table 2. when excess fluid is present. When tumors do not
470 Carmignani et al.
Figure 3. Overall and epigastric incidences compared between pseudomyxoma peritonei (PMP), mucinous adenocarcinoma (MA), and
non-mucinous adenocarcinoma (NMA).
produce fluid, their cells are restricted in motion
with implants near the primary site. Peritoneal
dissemination of NMA frequently involves colon,
greater omentum and small bowel; these are tissues
close to the primary site of cancer. Distant sites,
such as Treitz ligament and lesser omentum are
often involved when a fluid vehicle is present but
are spared when fluid is absent. The same concept
is supported by the analysis of peritoneal sarco-
matosis. Other gastrointestinal malignancies, like
pancreatic cancer and gastric cancer, should have
a seeding directly adjacent to the primary cancer
when ascites is absent but generalized dissemina-
tion when ascites is present.
These data may be interpreted to suggest that
cancer cells within a mucinous fluid are redis-
tributed on the abdominopelvic surfaces. In this
model free intraperitoneal cancer cells do not
immediately adhere to a peritoneal surface after
detachment from the primary malignancy. They
move with the peritoneal fluid vehicle to distant
sites by physical forces such as intraperitoneal
fluid hydrodynamics and gravity. They are rela-
tively excluded from peritoneal surfaces by sheer-
ing forces caused by peristaltic activity. Cancer cell
adherence and implantation is more likely to occur
in recesses within the peritoneal spaces such as cul-
de-sac of Douglas and subphrenic spaces. In
contrast cancer cells in the absence of a fluid
vehicle are more likely to adhere, implant and
progress proximal to the primary site [3].
Intraperitoneal fluid, especially mucus ascites,
may contribute to the large incidence of cancer
implants on tissues known to absorb peritoneal
fluid. The lymphatic lacunae on the peritoneal
surface of the diaphragm and the lymphoid
 Intraperitoneal cancer dissemination 471

Table 2. Involvement of sites based upon z-score calculated as standard deviation from the mean

Abdominopelvic site PMP MA NMA Ov Sa

Colon 1.5 1.5 2.5 2 2.5

Cul-de-sac of Douglas 1.3 1.4 2 1
Greater omentum 1.3 1.6 1.4 1
Small bowel 1.2 1.8 1.9
Right hemidiaphragm 1.2
Incision 1.2
Liver 1.1
Bladder 1.1
Left paracolic gutter 1.1
Left hemidiaphragm 1
Small bowel mesentery
1.1
Lesser omentum
1.1
Anterior parietal peritoneum
1.1
Stomach body
1
1.1
Gastric antrum
1
1.1
1.5
Vagina
1.4
1.4 1.4
1.2
Retroperitoneum
1.6
1.7
1.3 1
Treitz ligament
1.3
1.3
1.4
1.2
Kidney
2
2.1
1.1 1.7

Other sites assessed (spleen, pancreas, ileocecal area, right paracolic gutter, right side wall, left side wall) and blank spaces in this table
have a z-score between
0.99 and 0.99.

Figure 4. Comparison of the number of patients with tumor masses greater than 5 cm in ileocecal valve area, on small bowel surface
and on small bowel mesentery.
472 Carmignani et al.
aggregates on the greater omentum, lesser omen-
tum and omental appendages absorb peritoneal
fluid and thereby draw cancer cells to their surface.
The high z-scores of colon and greater omentum
and the high proportion of epigastric involvement
(lesser omentum) suggest that not only mucinous
fluid but also the resorption of this fluid are
important determinants of the distribution of
peritoneal surface malignancy.
The small bowel surface and the small bowel
mesentery have a large surface area as compared
to the ileocecal valve region. The small bowel and
its mesentery were theoretically exposed to a much
larger number of cancer cells than the ileocecal
area. However, the mobility of these portions of
the small bowel differ markedly. Large tumor
masses were more likely to develop in the ileocecal
valve region; this portion of the small bowel is
fixed on a short mesentery to the retroperitoneum
and has relatively less morbidity than other
portions of the small bowel. Malignant cells may
have difficulties to attach to the surfaces of the
intestine that are continuously moving with
peristalsis and rubbing against each other. Simi-
larly, movement of muscle may profoundly
decrease hematogenous metastasis to myocardial
and skeletal muscle [11,12].
A large proportion of these patients had a prior
surgical procedure before the collection of these
data with cytoreductive surgery. Although altera-
tion of the general pattern of carcinomatosis and
sarcomatosis was not noted, tumor cell entrap-
ment from prior surgery can be detected from
these data. The z-score for NMA within the
residual colon after prior resection was 2.5. Also
tumor progression at the operative site was a
prominent finding with ovarian cancer. The
vaginal cuff and pouch of Douglas has had z-
scores of 1.4 and 2. A high likelihood of local
implantation can be related to cancer cells
accumulating within fibrin near to the resection
of the primary tumor.
5. Conclusions
Directions taken by detached cancer cells within
the peritoneal cavity may be affected by a large
number of factors, including anatomic site of the
primary tumor, histologic type of tumor, changes
in intraabdominal pressure, gravity, presence of
peristalsis, resorption of peritoneal fluid, viscosity
and volume of fluid within the abdomen, perito-
neal adhesions, and fibrin entrapment from
trauma resulting from a surgical intervention.
Our data suggested that intraperitoneal fluid and
peritoneal surface motion (peristalsis) were pro-
minent mechanisms controlling the patterns of
spread of carcinomatosis.
References
1. Cole W: The mechanisms of spread of cancer. Surgery,
Gynecology & Obstetrics 137: 853–871, 1973
2. Viadana E, Bross ID, Pickren JW: The metastatic spread of
cancers of the digestive system in man. Oncology 35: 114–
126, 1978
3. Sugarbaker PH: Observations concerning cancer spread
within the peritoneal cavity and concepts supporting an
ordered pathophysiology. In: Sugarbaker PH (ed.) Perito-
neal Carcinomatosis: principles of management. Kluwer
Academic Publisher, Boston, MA, 1995, pp 80–99
4. Fortner JG: Inadvertent spread of cancer at surgery. J Surg
Oncol 53: 191–196, 1993
5. Solomon MJ, Egan M, Roberts RA, Philips J, Russell P:
Incidence of free colorectal cancer cells on the peritoneal
surface. Dis Colon Rectum 40: 1294–1298, 1997
6. AJCC. Manual for Staging of Cancer. Fourth ed. J.B.
Lipincott Company, Philadelphia, 1992
7. Jacquet P, Sugarbaker PH: Current methodologies for
clinical assessment of patients with peritoneal carcinoma-
tosis. J Exp Clin Cancer Res 15: 49–58, 1996
8. Weyn B, Jacob W, da Silva VD, Montironi R, Hamilton
PW, Thompson D, Bartels HG, Van Daele A, Dillon K:
Data representation and reduction for chromatin texture in
nuclei from premalignant prostatic, esophageal, and
colonic lesions. Cytometry 41: 133–138, 2000
9. Autio V: The spread of intraperitoneal infection. Acta
Chirurgica Scandinavica S(321): 5–31, 1964
10. Lopes Cardozo AM, Gupta A, Koppe MJ, Meijer S, van
Leeuwen PA, Beelen RJ, Bleichrodt RP: Metastatic pattern
of CC531 colon carcinoma cells in the abdominal cavity:
An experimental model of peritoneal carcinomatosis in
rats. Eur J Surg Oncol 27: 359–363, 2001
11. Weiss L: Biomechanical destruction of cancer cells in the
heart: A rate regulator for hematogenous metastasis.
Invasion Metastasis 8: 228–237, 1988
12. Weiss L: An analysis of the incidence of myocardial
metastasis from solid cancers. Br Heart J 68: 501–504,
1992
Address for offprints: Paul H. Sugarbaker, M.D., Washington
Cancer Institute, 110 Irving Street NW – suite CG-185,
Washington DC 20010.