ORIGINAL ARTICLE

Malignant ileocaecal serotonin-producing carcinoid tumours: Thepresence of a solid growth pattern and/or Ki67 index above 1% identifiespatients with a poorer prognosis

JANET L. CUNNINGHAM1, LARS GRIMELIUS2, ANDERS SUNDIN3,

SMRITI AGARWAL4 & EVA T. JANSON1

1Department of Medical Sciences, Section of Endocrine Oncology, Uppsala University, Sweden, 2Department of Genetics and

Pathology, Uppsala University, Sweden, 3Department of Oncology, Radiology and Clinical Immunology, Uppsala University,

Sweden and 4Christian Medical College, Vellore, India

AbstractPatients with malignant serotonin-producing carcinoid tumours in the jejunum, ileum and caecum generally have longsurvival expectancy. In some patients, however, tumour progression is more rapid and there is a need to identify them at anearly stage. The purpose of this study was to determine if histopathological characteristics and/or Ki67 and apoptotic indicesare of prognostic value in cases of metastatic disease. Eighty-one patients with this tumour were included in the study; allhad metastases and their survival range was 1�223 months. Five growth patterns were identified and described. For 57patients whose tumour material was available, the Ki67 and apoptotic indices were calculated for ten randomly selectedtumour areas and ‘hot spots’. A Cox regression analysis was used to test if histopathology and/or Ki67 index]1% couldidentify patients whose survival might be shorter than anticipated. One of the histopathological growth patterns-the solid(non-organoid) cell pattern-was correlated to shorter survival in both primary tumours and metastases, when comparedwith the organoid growth patterns (hazard ratio 2.9 and 2.3, p50.01). In 75% of primary tumours and 67% of metastases,the average Ki67 index wasB0.5%. Ki67 index in ‘hot spots’ ranged from 0.1 to 14%. Ki67 index]1%, in both primarytumour and metastases, identified patients at increased risk of shorter survival (hazard ratio 5.4 and 2.5, p50.01). Theapoptotic index was very low in all cases. We conclude that in patients with metastazising serotonin-producing carcinoids,two independent criteria, a solid growth pattern and Ki67 index]1%, can be used to identify patients with a poorerprognosis. This study also showed that Ki67 indexB2% cannot, as previously suggested, be used to indicate a benignprogression for this tumour category.

The term carcinoid has been used to describe

neuroendocrine tumours (NETs) in the gastrointest-

inal tract (GI-tract), lung, thymus and ovary. Malig-

nant serotonin-producing carcinoids (MSPC) arise

from enterochromaffin cells (EC cells) in the jeju-

num, ileum, caecum and ascending colon [1]. These

tumours secrete, in addition to serotonin, tachyki-

nins, and in cases with development of liver metas-

tases, these hormones cause clinical symptoms,

i.e. the carcinoid syndrome. The WHO classification

of endocrine tumours and another more recent

classification calls these tumours ‘‘well differentia-

ted (neuro-)endocrine carcinomas (malignant carci-

noids)-low-grade malignant serotonin-producing

carcinoids/carcinomas’’. Both classifications suggest

that a Ki67 index B2% is associated with benign

tumour progression [2,3].

Median duration of survival from diagnosis is

around 6 years depending on the spread of disease,

though the range can be from months to decades

[4�6]. Clinically, there is a need to identify patients

at risk of shorter survival at an early stage in order to

facilitate the choice of more radical surgical and

medical treatments. The histopathological structure

and proliferation rate are both used for this purpose

in other tumour entities [7,8]; however, MSPCs

have hitherto been described as having an insular

growth pattern and a low proliferation rate in

common. Possible variation within these parameters

has not been revealed in earlier studies, largely due

Correspondence: Janet L. Cunningham, Department of Medical Sciences, Section of Endocrine Oncology, Lab 14, Research Department 2, Uppsala

University Hospital, SE 751 85, Uppsala, Sweden. Tel: �46 18 611 4913. E-mail: janet.cunningham@medsci.uu.se.

Acta Oncologica, 2007; 46: 747�756

(Received 28 June 2006; accepted 10 January 2007)

ISSN 0284-186X print/ISSN 1651-226X online # 2007 Taylor & Francis

DOI: 10.1080/02841860701218659

Act

a O

ncol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

118.

127.

75.2

25 o

n 05

/20/

14Fo

r pe

rson

al u

se o

nly.

to small sample sizes and/or heterogeneous tumour

material [2,9�20].

A histopathological classification of NETs was

first suggested in 1971 by Soga and Tazawa [9].

Their study included 62 fore-, mid- and hindgut

NETs. Only three of the included midgut tumours

were located in the ileum and three in the right

colon, while the remaining 14 originated in either

the appendix or caecum (exact location not speci-

fied). The authors concluded that their growth

pattern classification correlates to the site of the

primary tumour. Of the 20 midgut NETs included

in their study, 19 had an insular growth pattern (type

A) in either a pure form or mixed with trabecular

growth pattern (type B) or acinar and rosette growth

pattern (type C) while one was trabecular and acinar

(type B�C). In a similar study [10], Jones and

Dawson included 15 argentaffin carcinoids from the

small intestine and one from the proximal colon and

categorized all of these tumours as either insular

(type A1) or insular-acinar/glandular (type A1/A2).

Johnson et al. examined growth patterns in NETs

from different localizations and related them to

survival but did not report the outcome specifically

for MSPCs [11].

General recommendations for NET management

suggest that a cut-off Ki67 index of 2% can be used

to distinguish benign tumours from MSPCs

[2,3,21]. In our experience, however, very few low

grade malignant carcinoids display a Ki67 index over

2% at diagnosis despite massive metastatic disease.

Supporting this, Canavese et al. [22] reported that in

a set of 17 MSPCs, Ki67 expression ranged from 0.0

to 1.76%. A pilot study by Chaudhry et al. [19],

showed that Ki67 expression of more than 1 Ki67

labelled cell/mm2 correlated to a shorter survival

when calculated in biopsies from liver metastases in

14 untreated patients with MSPCs. A cut-off of 2

mitoses/mm2, i.e. higher than in Chaudhry’s study,

showed a non-significant tendency to identify pa-

tients with a poorer prognosis in MSPCs [18]. These

three studies indicate that the Ki67 index is often

lower than 2% in malignant tumours and it is

therefore possible that the Ki67 index cannot be

used to indicate malignancy in these tumours.

Within the category of patients with MSPCs, how-

ever, criteria are needed to identify patients with

more aggressive tumour forms.

In this study, we examined variations in histo-

pathological growth pattern, proliferation index and

apoptotic index in patients with MSPCs. Our inten-

tion was to test the possibility of using growth

pattern, a Ki67 index cut-off of 1%, and apoptotic

index as prognostic markers to identify patients at

risk of a clinically significant shortened survival.

Materials and methods

Patients and material

The present retrospective study included 81 patients

with MSPCs. All primary tumours originated from

the distal ileum, caecum or proximal colon. Appen-

diceal carcinoids were excluded from this study.

Histopathological structure was assessed in tumour

specimens from 56 primary tumours, 38 mesentery

metastases and 27 liver metastases of which 18 were

from needle biopsies. Additional tumour tissue was

available from 57 of these patients for analysis of

Ki67 immunoreactivity. The Ki67 index was calcu-

lated in 36 primary tumours, 29 mesentery metas-

tases and 20 liver metastases of which 11 were

needle biopsies. In four cases, only biopsy material

was available for Ki67 assessment. The tumours

were diagnosed histopathologically at the Laboratory

for Pathology and Cytology and the patients were

treated at the Department of Endocrine Oncology at

University Hospital, Uppsala. Tumour tissue was

fixed in 10% buffered neutral formalin for 1�2 days,

dehydrated, embedded in paraffin wax and routinely

stained using haematoxylin/eosin and Van Gieson’s

solution. All tumours included displayed chromo-

granin A (CgA) and serotonin immunoreactivity,

and/or argyrophil [23] and argentaffin reactions

[24].

This population consisted of all patients with

MSPCs operated on their tumour at Uppsala Uni-

versity Hospital between 1980 and 2004 and from

whom tumour tissue was available and event-free

survival could be calculated. After diagnosis, all

patients underwent assessment of urinary 5-hydro-

xyindoleacetic acid levels and tumour size by com-

puterized tomography and ultrasound investigations

2�4 times per year. Plasma-CgA was also measured

routinely after 1990.

At the time of tumour tissue collection (surgical

tissue specimens or needle biopsy), 51 patients were

untreated; six had been treated with alpha-interferon

for less than 12 weeks and were considered as

‘untreated’ in data processing. Interferon treatment

has been shown to induce significant fibrosis after 6

months, which can affect the growth pattern [25]

and also the Ki67 index [19]. The 24 patients who

had received treatment for longer than 12 weeks

before tumour tissue collection were considered

‘treated’. All patients were treated postoperatively

according to clinical praxis at our Department

[26,27]. Seventy-nine received alpha interferon and

68 were treated with a somatostatin analogue. Liver

embolization was performed in 18 patients and 12

were treated with chemotherapy for a short period.

The clinical profile of the patients included in this

study is presented in Table I.

748 J. L. Cunningham et al.

Act

a O

ncol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

118.

127.

75.2

25 o

n 05

/20/

14Fo

r pe

rson

al u

se o

nly.

The study was reviewed and approved by the local

Medical Ethics Committee at Uppsala University

Hospital.

Immunohistochemistry

Sections were cut 4 mm thick and attached to

positively charged glass slides. The sections were

deparaffinized in xylene and rehydrated to distilled

water. They were then microwave treated at 700 W

for 15 min and 350 W for 10 min, except when

staining for CD31 and CD34 when microwave

treatment was 10 min at 700 W. The DAKO

EnVisionTM plus system or ZYMED NBATM kit

was used according to the manufacturer’s instruc-

tions. Diaminobenzidine was used as a chromogen

and sections were counterstained with Mayer’s

haematoxylin. The following primary antibodies

were used, with antigen retrieval buffer and antibody

dilution given in parentheses: serotonin, clone

5HT-H209 (citrate pH 6, 1:10); CD31 and CD34

clones JC70A and QBEnd10 (TRS pH 6, 1:50);

Ki67, clone MIB-1 (citrate pH 6, 1:100) all from

DAKO, Glostrup, Denmark; chromogranin A clone

LK2H10, (Tris pH 8, 1:2000) from Boehringer-

Mannheim, Mannheim, Germany; and finally, poly-

clonal cleaved caspase-3 (Tris pH 8, 1:200) from

Cell Signalling Technologies, Danvers, Mass., USA.

Silver stains

Grimelius silver nitrate stain [24,28] was used to

reveal argyrophil reaction and a modified Masson

Fontana method for argentaffin reaction [29]. The

former stain identifies most NE cells, while the latter

displays EC cells (serotonin producing) in the GI-

tract.

Microscopical assessments

The immunostained sections were coded and exam-

ined microscopically and representative areas in CgA

or Grimelius stained sections were photographed

digitally at low magnification and printed in colour.

The photos were coded and sorted by two observers

(L.G. and J.L.C.) into groups according to their

main growth pattern. All sections were further

evaluated, this time by microscopy using both low

and high magnification. To distinguish acinar struc-

tures from vascular lumen within tumour cell

groups, tumour material was immunostained with

markers for vascular endothelium (CD31/CD34). A

growth pattern was deemed to be predominant if it

was present in more than half of the tumour. When

three or more patterns were present the predominat-

ing pattern could be present in less than 50% of the

tumour section.

Ki67 index and apoptotic index

The tumour cell Ki67 index was calculated by two

evaluators (J.L.C. and S.A.) in freshly cut sections

immunostained with antibodies to Ki67 (MIB-1).

Tissue from tonsil, and proliferating cells in intest-

inal mucosa and stroma surrounding the tumour

served as positive control material. The immunor-

eactive (IR) and the non-IR cell nuclei were counted

at 400� magnification using a square graticule

(10�10 mm) in one of the oculars. Nuclear staining

intensity varied between cells. IR tumour cells with a

delimited nuclear contour were counted excluding

IR stromal cells. Serial sections, stained with anti-

bodies recognizing CgA, were used to help identify

tumour cells. At least ten randomly selected tumour

areas (average) as well as the area with the greatest

number of IR tumour nuclei (‘hot spot’) in relation

to tumour cells were analysed. The Ki67 index was

expressed as a percentage of IR tumour nuclei. In

cases where Ki67 index group differed between

evaluators or between several sections from the

same tumour, an average of the indices was used as

the average Ki67 index and the highest index was

used as the ‘hotspot’ Ki67 index. To estimate

interpersonal variation in the calculation of Ki67,

ten sections with a Ki67 index in the range 0.5 to

1.5% were evaluated on three separate occasions.

Apoptotic cells were identified immunohistochemi-

cally using antibodies to cleaved caspase-3 and the

apoptotic index was calculated in the same manner

as the Ki67 index.

Survival and statistical analysis

Data were stored in an Access database and analysed

using the SAS statistical program package. The end-

point for all analyses was event-free survival, defined

as months from surgery, or biopsy to death. For

patients still living or who died of other causes, the

interval from surgery to the last documented follow-

up was used as a censored observation. Other causes

of death were death from other malignancies,

cerebral ischaemia, myocardial infarction and sui-

cide. Few patients were autopsied as most died at a

Table I. An overview of the clinical background.

Sex (male/female) 28/53

Median age (years) at operation [range] 62 [34�82]

Median survival (months) after operation

[range]

60 [1�223]

Diarrhoea and/or flush at operation 66

Metastases at operation 78

Metastases at clinical follow-up 81

Ileocaecal carcinoids 749

Act

a O

ncol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

118.

127.

75.2

25 o

n 05

/20/

14Fo

r pe

rson

al u

se o

nly.

local hospital. Medical records were reviewed in

order to determine the cause of death in all cases.

Three age categories were defined from patients’

ages at operation: B65 years, 65�75 years, and�75

years. Tumour burden in liver and mesentery was

estimated using radiological examinations prior to

operation. Three grades of tumour burden were

defined: minimal (B5 lesionsB5 cm in diameter),

intermediate (5�10 lesions or one�5 cm in dia-

meter), massive (]10 lesions or one�10 cm).

Statistical analyses were performed separately for

primary tumours and metastases. Before pooling the

treated and untreated patients, the distribution of

gender, age categories, growth pattern and Ki67

index was compared using Fisher’s exact test and

possible interaction between treatment/no treatment

of the tumours and Ki67 index on survival was

tested with a Cox regression model. Patients who

were still alive (n�3) as well those who had died

from causes not related to the tumour (n�7) were

censored in the analysis (n�10). Correlations be-

tween the average Ki67 index and ‘hot spot’ areas as

well as between Ki67 index and tumour burden were

studied using a Spearman correlation coefficient.

Analysis of the association between predominant

growth pattern and survival was performed using a

Cox regression analysis, adjusted for age and for sex

to calculate a hazard ratio. To identify the way to

calculate the Ki67 index having the best predictive

value, a Kaplan-Meier graph and a Mantel-Cox log-

rank test were used, as well as analysis with a Cox

hazard regression model adjusted for age category

and sex. All p-values are two-sided and a p-value

50.05 was considered a statistically significant

result.

Methodological error

The methods used in this study are limited in the

following ways. The ability to assess adequately the

type of cell growth pattern of the carcinoid and to

calculate the Ki67 index accurately might have been

impaired by the inclusion of tissue obtained from

1.2 mm needle biopsy specimens from the liver.

Evaluation of tumour growth patterns and Ki67

index can be marred by inter/intrapersonal variation.

When the evaluators disagreed on the growth pattern

classification, the tumours were re-evaluated, when

possible together with new sections from other areas

in the tumour. In cases where average Ki67 index

group differed between evaluators or between several

sections from the same tumour, the average of the

calculations was used in the analysis. To estimate

interpersonal variation in the calculation of Ki67, ten

sections with a Ki67 index in the range of 0.5 to

1.5% were evaluated on three separate occasions.

Results

When using Fisher’s exact test, no significant differ-

ences were observed in the distribution of gender,

age categories, growth pattern or Ki67 index be-

tween the untreated and treated patient groups. An

interaction test showed no significant interaction

between treatment and growth pattern or Ki67

index. The statistical analysis was therefore con-

ducted on the pooled results from treated and

untreated patients.

Growth pattern and survival

We identified five growth patterns, as shown in

Figure 1. The insular, trabecular and acinar (also

called glandular or rosette) growth patterns have

been described earlier [9]. Two additional patterns

were seen: solid and ‘small cell nest’. Areas with

insular cell arrangement were found in most tu-

mours studied. The tumour cells in the mucosa and

submucosa were arranged in one or more of the

growth patterns described, sometimes with periph-

eral palisading arrangement. In muscularis propria, a

less regular trabecular pattern usually predominated,

whereas in subserosa, the tumour cells showed a

tendency to resume the predominating pattern

evident in the upper layer.

The term solid was used to describe tumour cells

with a non-organoid growth pattern. Tumour cells

with the solid growth pattern were densely packed,

expressing CgA and serotonin and were sometimes

arranged in broad bands without regular stromal

divisions. Vascular structures penetrated the solid

tumour bands, in contrast to the other growth

patterns where vessels lay in the stroma surrounding

the tumour cell aggregations. The ‘small cell-nest’

Figure 1. A�E. The microphotographs illustrate the five growth patterns in the mucosa/submucosa in MSPCs (Grimelius silver stain).

(A) Insular growth pattern: some of the insular nodules show peripheral palisading. Scale bar�100 mm. (B) Insular-trabecular

growth pattern. (C) Insular-acinar growth pattern: insular cell nests with glandular or rosette formations. Inset�endothelial

immunostaining showing vessles in the stroma surrounding tumour cell nests. (D) Small cell nest growth pattern: irregular small cell

nests together with slender trabeculae and abundance of fibrous stroma. (E) Solid growth pattern: solid aggregations of tumour cells, a

pattern deviating from that seen in most MSCPs. Inset�endothelial immunostaining showing vessels within tumour cell aggregations.

(F & G) Primary tumour and metastases with solid growth pattern showing chromogranin A IR tumour cells. Scale bar�200 mm. Inset:

haematoxylin-eosin stained tumour section showing uniform nuclei without mitosis (H) A tumour area without Ki67 IR cells and an

abundance of Ki67 IR tumour cells in the intestinal mucosa (lower left corner). Scale bar�50 mm.

750 J. L. Cunningham et al.

Act

a O

ncol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

118.

127.

75.2

25 o

n 05

/20/

14Fo

r pe

rson

al u

se o

nly.

Figure 1 (Continued)

Ileocaecal carcinoids 751

Act

a O

ncol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

118.

127.

75.2

25 o

n 05

/20/

14Fo

r pe

rson

al u

se o

nly.

growth pattern describes tumour cell aggregations

containing few and even lone tumour cells, distrib-

uted in a fibrotic stroma. The ‘small cell-nest’

pattern was seen in untreated patients and was

therefore not a result of treatment. Almost all

tumours displayed more than one growth pattern.

The distribution of the predominant growth patterns

among the tumours is listed in Table II.

The Cox proportional hazard model of overall

survival adjusted for age and sex was used to test if

any growth pattern was associated with an increased

relative hazard ratio. When each growth pattern was

tested against the other patterns pooled into one

group, only the solid pattern was associated with a

greater risk of death per time fraction (Table III).

This result was the same for primary tumours as for

metastases, but was especially prominent for the

mesentery metastases (hazard ratio 3.6, pB0.01).

Ki67 and apoptotic index versus survival

An average Ki67B0.2% was found in 23/36 primary

tumours and 21/49 metastases. Median survival in

patients with an average Ki67 index ]1% in this

material was less than half that of patients with

average Ki67 indexB1% (Table IV). The average

Ki67 index was closely correlated to the ‘hot spot’

index and the Spearman Correlation coefficient was

]0.93 (pB0.01).

Patients at risk of shorter survival were identified

by using a cut-off of 1% for both average and ‘hot

spot’ Ki67 indices. This was found for both primary

tumours and metastases, using a Cox proportional

hazard model of overall survival adjusted for age and

sex (Table III). The closest correlation between Ki67

index and survival was seen in primary tumours. The

most significant prognostic marker was an average

Ki67 index ]1% in primary tumours (Table III).

Plotting patient survival, in relation to Ki67 index

above and below 1% in the primary tumour, on a

Kaplan-Meier graph demonstrated this significant

difference in survival (Mantel-Cox log rank test

p�0.02; Figure 2). Explorative testing was also

performed on the material using Ki67 index cut-

offs of 0.5% and 2% and significantly shortened

survival was seen even in patients whose primary

tumour had a Ki67 index �0.5%.

To consider the possibility that the results ob-

tained reflect differences in survival due to tumour

burden, the spread of disease in the patients was

assessed. No difference was observed in the tumour

burden at diagnosis between patients with Ki67

index over vs. under 1%, or between those exhibiting

differing growth patterns. Figure 3 shows radiologi-

cal images from a representative patient with a

MSPC shortly before operation. Multiple metastases

are evident but the calculated Ki67 index is low in

the surgical specimens obtained from both primary

tumour and metastases.

The apoptotic index in the primary tumours and

metastases was under 0.01% in both treated and

untreated patients regardless of their Ki67 index,

and necrosis was seldom seen in the tumour sections

studied.

Growth pattern and Ki67 index in primary tumour and

metastases from the same patient

Analysis of the prognostic value of growth pattern

and Ki67 was performed separately for primary

tumours and metastases. Factors associated with

high risk, when seen in primary tumours, were

most often also seen in the metastases from the

same patient.

The predominating growth pattern in primary

tumours was often seen, at least in some parts of

the metastases. Sections for histopathological eva-

luation of primary tumours and metastases were

available from 37 patients; of these, 17 had the same

predominant growth pattern in both the primary

tumour and metastases. In 15 other patients, the

same growth pattern was seen in both primary

tumour and metastases, but the predominant pattern

differed. In the remaining seven cases, the metas-

tases displayed a growth pattern not present in the

primary tumour. Generally, patients with a solid

growth pattern in the primary tumour also had this

pattern in the metastases. In four cases, a solid

growth pattern was seen in the metastases only.

In 16/23 patients, the Ki67 index group (over/

under 1%) was the same in both primary tumour

and metastases. In six of the remaining patients, the

Ki67 index was above 1% only in metastatic tissue.

The last patient was found to have an average Ki67

Table II. Tumour cell growth pattern varied and five growth patterns could be identified. Three patterns correspond to patterns, type A, B

and C, described earlier (9). The table shows the distribution of tumours studied in relation to their dominating growth patterns. Areas with

an insular growth pattern were found in nearly all tumours studied.

Insular (A) Trabecular (AB) Acinar (AC) Solid ‘Small cell-nests’

Primary tumour 11 11 16 10 8

Local metastases 5 11 4 12 6

Liver metastases 9 4 5 5 4

752 J. L. Cunningham et al.

Act

a O

ncol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

118.

127.

75.2

25 o

n 05

/20/

14Fo

r pe

rson

al u

se o

nly.

index greater than 1% in the primary tumour and

only a ‘hot spot’ Ki67 index exceeding 1% in the

metastases which, in this case, was a needle biopsy.

Methodological error

The correlations between survival and the solid

growth pattern or Ki67 index ]1% were lower in

needle biopsy material than in surgical material. The

inter/intrapersonal variation in growth pattern eva-

luation was marginal regarding solid growth, but the

predominant growth pattern in a minor subset of

tumours with organoid growth pattern was difficult

to assign.

In the majority of tumours, the average Ki67 index

wasB0.5%. In these cases, interpersonal variation

between the two evaluators was minimal (90.1).

Variation in Ki67 index calculation increased with

the number of proliferating cells and with the

amount of tumour material available for examina-

tion. To assess the intrapersonal variation, ten

tumour sections each with an average Ki67 index

between 0.5 and 1.5% were evaluated on three

additional occasions by one evaluator. The standard

deviation of the calculated index was 0.2. The two

individuals calculating the Ki67 index placed the

tumours in the same average Ki67 index group

(over/under 1%) in all but two cases. The inter-

personal variation in ‘hot spot’ evaluation was

greater; however, the same ‘hot spot’ Ki67 index

group (over/under 1%) was assigned in all but three

cases.

Discussion

Most studies published on the histopathological

characteristics of malignant carcinoid tumours in

relation to clinical data have been based on collec-

tions of tumours from various regions of the GI-tract

and some studies also included NETs from other

organs, e.g., pancreas and lung. In the present study,

only primary tumours and metastases from well

Table III. Ki67 index ]1% and a solid growth pattern can be used to identify patients with a poorer prognosis. The relative hazard ratio for

each risk factor obtained in Cox proportional hazard models of overall survival in patients with MSPCs is shown with 95% confidence

intervals (95% CI). Models are adjusted for age and sex. Factors presented in italics were not included in the original hypothesis and were

tested exploratively.

n N in high risk group Hazard ratio (95% CI) p-values

Primary tumour

‘Hot spot’ Ki67 index ]1% 36 11 2.4 (1.1�5.5) 0.03

Average Ki67 index ]1% 36 5 5.4 (1.7�17.5) B0.01

Solid growth pattern 56 10 2.9 (1.3�6.3) B0.01

‘Hot spot’ Ki67 index ]0.5% 36 14 2.3 (1�1.9) 0.04

Average Ki67 index ]0.5% 36 9 3.3 (1.3�8.3) 0.01

‘Hot spot’ Ki67 index ]2% 36 7 3.6 (1.3�9.5) 0.01

Average Ki67 index ]2% 36 2 6.6 (1.2�35) 0.03

Metastases

‘Hot spot’ Ki67 index ]1% 49 21 2.0 (1.0�3.7) 0.03

Average Ki67 index ]1% 49 13 2.5 (1.2�5.1) 0.01

Solid growth pattern 62 17 2.3 (1.2�4.6) 0.01

‘Hot spot’ Ki67 index ]0.5% 49 33 1.8 (0.9�3.5) 0.09

Average Ki67 index ]0.5% 49 15 2.0 (1�4.2) 0.07

‘Hot spot’ Ki67 index ]2% 49 12 2.2 (1�4.6) 0.03

Average Ki67 index ]2% 49 9 1.7 (0.8�3.8) 0.18

Table IV. Ki67 index group in MSPCs is shown in relation to median survival regardless of cause of death (range shown in brackets).

Average Ki67 index in the majority of MSPCa is B0.5%. Patients with a Ki67 index ]1% have a significantly shorter median survival.

Average Ki67 index ‘Hot Spot’ Ki67 index

n

Survival in months

median (range) n

Survival in months

median (range)

Primary Tumours (n�36)

Ki67 index B1% 31 62 (1�216) 25 62 (1�216)

Ki67 index ]1% 5 30 (13�53) 11 35 (4�136)

Metastases (n�49)

Ki67 index B1% 36 62 (6�223) 28 64 (18�196)

Ki67 index ]1% 13 28 (1�136) 21 56 (1�223)

Ileocaecal carcinoids 753

Act

a O

ncol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

118.

127.

75.2

25 o

n 05

/20/

14Fo

r pe

rson

al u

se o

nly.

characterized, MSPCs were studied. To evaluate the

usefulness of histopathological characteristics, Ki67

index and apoptotic index as prognostic indicators,

these factors were studied in relation to survival. Our

study defines two factors that may help to identify

patients with a less favourable prognosis:

a solid growth pattern and a Ki67 index ]1%.

The apoptotic index was low in all the tumours and

gives no prognostic information. This report shows

that Ki67 index B2% is seemingly not useful as a

criterion for benign tumour behaviour.

This study demonstrates that a solid growth

pattern in primary tumours and/or metastases is a

risk factor for shorter survival. A solid growth pattern

describes solid, broad bands of tumour cells that have

maintained their neuroendocrine phenotype and

produce both CgA and serotonin. This growth

pattern has been included, in earlier studies, under

the definition of insular [11] or may have been, in

other studies, described as an undifferentiated growth

pattern. The latter term can, however, be misinter-

preted as a highly malignant tumour. The solid

growth pattern does not display nuclear atypia or

Ki67 indices above 15% and should not be confused

with an undifferentiated or high-grade malignant

tumour [3]. In the present solid growth pattern, the

vascular network penetrated the tumour cell aggrega-

tions; this is in contrast to the other growth patterns

where vascular structures were only found in the

stroma surrounding tumour nests. The solid growth

pattern was more frequently associated with a higher

Ki67 index than were the other growth patterns. In

these cases, however, proliferating cells tended to lie

in peripheral small cell aggregations and less often

within the solid tumour areas. One may hypothesize

that the solid growth pattern represents the loss/gain

of a biological control regulating the tumour cell’s

behaviour in relation to its surroundings.

The remaining patterns described in this study

represent variations of an organoid growth pattern.

Three patterns - insular, trabecular and acinar-are

consistent with those observed in midgut NETs by

Soga and Tazawa [9] but were more prevalent in our

study. These variations in organoid patterns may

represent differences in tumour biology but showed

no relation to clinical outcome in this study. The

growth pattern ‘small cell-nest’ was described in this

study and visually appears to be a fibrotic reaction

that is unconnected with treatment effects. Tumours

within the trabecular group often displayed features

from other growth pattern groups.

The ‘hot spot’ Ki67 index in patients with MSPCs

ranges from close to 0% to 14% and the average

Ki67 index from nearly 0% up to 4%. Despite

extensive tumour spread, the majority of tumours

displayed an average Ki67 index B0.2% in primary

tumours and B0.5% in metastases. Thus, a Ki67

index B2% as suggested in current classifications, or

even 0.1%, is not an indication of a non-metastasiz-

ing tumour.

Various methods for estimating tumour prolifera-

tion have been reported in the literature. Mitotic

activity was, in the pre-immunohistochemical era,

the only way to calculate cellular proliferation.

Mitosis represents a limited phase of the prolifera-

tion cycle. Mitotic and apoptotic cells are sometimes

difficult to differentiate. During recent decades,

immunohistochemical technique has been able to

demonstrate the presence of various nuclear proteins

during cellular proliferation; the Ki67/MIB-1 is,

currently, the most used as it is expressed in all

stages of cell proliferation. In the present study, we

calculated the number of proliferating cells in rela-

tion to the number of tumour cells in a unit area.

This method takes into consideration factors that

influence the counting results such as cell size, the

amount of fibrosis and compression artefacts. Cal-

culation of proliferation as a percentage of tumour

cells simplifies comparison between tumours.

In our study, the group of patients with an average

Ki67 index ]1% had a median survival only half

that of those with Ki67 indexB1%. Most tumours

in the study displayed a Ki67 index clearly under or

above 1%, which in clinical practice simplifies the

prognostic evaluation. In one and the same patient,

primary tumours and metastases were often placed

in the same KI67 index group (over/under 1%)

although Ki67 indices were generally higher in

metastases. In the Cox proportional hazard model,

Figure 2. The majority of patients with MSPCs display a Ki67

index B1%. Patients with a Ki67 index ]1% are at risk of

shortened survival. The figure shows a cumulative proportion of

patients surviving in relation to Ki67 index in ‘hot spots’ in

primary tumours. ***, Patients with a Ki67 indexB1%;

---------, Ki67 index ]1%. k, complete cases; �censored cases.

m, Patients with average Ki67 index ]1%. P-value is from a

Mantel-Cox log rank test.

754 J. L. Cunningham et al.

Act

a O

ncol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

118.

127.

75.2

25 o

n 05

/20/

14Fo

r pe

rson

al u

se o

nly.

patients with a Ki67 index above 1% have a higher

risk of shorter survival compared to those with Ki67

index below 1%. This risk increased further when

several areas over 1% were identified (as is the case

for an average Ki67 index ]1%) and when areas

over 1% were identified in primary tumours. Scru-

tiny of our data revealed that this mortality risk

started to increase already at a Ki67 index of 0.5%.

Stronger correlations were usually seen in surgical

tissue specimens from both primary tumour and

metastases in contrast to needle biopsy tissue speci-

mens from liver metastases. This disparity can be

explained by the limited amount of tumour tissue

available for study in a biopsy. This is a general

problem, as in many cases needle biopsy is the only

tissue specimen available for analysis.

Tumour growth is the balance between cell pro-

liferation and cell death. MSPCs seldom demon-

strate necrosis. The apoptotic index was extremely

low in all cases and independent of the Ki67 index.

Recent genetic [30�32] and expression studies

[33] indicate that it is likely that MSPCs have a

distinct tumour biology when compared with other

NETs. It may therefore no longer be appropriate to

extrapolate results from heterogeneous populations

of NETs that originate from different regions in the

GI tract and other organs and assume them to be

valid for MSPCs [2]. This study serves two pur-

poses. First, it provides a careful description of

variations in tumour cell growth patterns and pro-

liferation rates in relation to survival in this category

of patients. Second, in addition to established

clinical prognostic factors [34], the identification of

areas with a solid growth pattern as well as Ki67

index ]1% in tumour tissue specimens charac-

terizes patients with MSPCs having a poorer prog-

nosis. These criteria are simple to determine and can

be applied in a clinical setting.

Acknowledgements

The authors thank Asa Forsberg, laboratory techni-

cian, for her excellent technical assistance, Mikael

Bjork, IT coordinator, Uppsala University Hospi-

Figure 3. Radiological images from a representative patient with a MSPC show that the Ki67 index cannot be used to indicate a non-

metastasizing tumour. Preoperative intravenous contrast-enhanced computerized tomography (CT) of the liver performed in the portal

venous inflow phase shows numerous, well-vascularized, contrast-enhanced metastases (see arrows). Ki67 index in surgical material from

the primary tumour and in the metastases was B0.2%.

Ileocaecal carcinoids 755

Act

a O

ncol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

118.

127.

75.2

25 o

n 05

/20/

14Fo

r pe

rson

al u

se o

nly.

tal for his computer support and Lars Berglund,

biostatistician and Head of the Department of

Biometrics at Uppsala Clinical Research Centre,

for his assistance in the statistical analysis of the

data. The Swedish Cancer Society (project number -

4880-B05-03XBB), Lion’s Fund for Cancer Re-

search at Uppsala University Hospital and the

Torsten and Ragnar Soderbergs Research Founda-

tion are acknowledged for supporting the study.

References

[1] Williams ED, Sandler M. The classification of carcinoid

tumours. Lancet 1963;/i:/238�9.

[2] Kloppel G, Perren A, Heitz PU. The gastroenteropancreatic

neuroendocrine cell system and its tumors: The WHO

classification. Ann NY Acad Sci 2004;/1014:/13�27.

[3] Solcia E, Kloppel G, Sobin LH. Histological typing of

endocrine tumours. WHO International Histological Classi-

fication of Tumours. 2nd ed. Berlin: Springer; 2000.

[4] Modlin IM, Lye KD, Kidd M. A 5-decade analysis of 13,715

carcinoid tumors. Cancer 2003;/97:/934�59.

[5] Zar N, Garmo H, Holmberg L, Rastad J, Hellman P. Long-

term survival of patients with small intestinal carcinoid

tumors. World J Surg 2004;/28:/1163�8.

[6] Janson ET, Holmberg L, Stridsberg M, Eriksson B, Theo-

dorsson E, Wilander E, et al. Carcinoid tumors: Analysis of

prognostic factors and survival in 301 patients from a referral

center. Ann Oncol 1997;/8:/685�90.

[7] Gleason DF. Histological grading and clinical staging of

prostatic carcinoma. In: Tannenbaum M, editor. Urologic

Pathology: The Prostate. Philadelphia: Lea and Febiger;

1977. p. 171�97.

[8] Elston CW, Ellis IO. The Breast, 3rd ed. Edinburgh:

Churchill Livingstone; 1998.

[9] Soga J, Tazawa K. Pathologic analysis of carcinoids. Histo-

logic reevaluation of 62 cases. Cancer 1971;/28:/990�8.

[10] Jones RA, Dawson IM. Morphology and staining patterns of

endocrine cell tumours in the gut, pancreas and bronchus

and their possible significance. Histopathology 1977;/1:/137�50.

[11] Johnson LA, Lavin P, Moertel CG, Weiland L, Dayal Y,

Doos WG, et al. Carcinoids: The association of histologic

growth pattern and survival. Cancer 1983;/51:/882�9.

[12] Van Eeden S, Quaedvlieg PF, Taal BG, Offerhaus GJ,

Lamers CB, Van Velthuysen ML. Classification of low-grade

neuroendocrine tumors of midgut and unknown origin.

Hum Pathol 2002;/33:/1126�32.

[13] von Herbay A, Sieg B, Schurmann G, Hofmann WJ, Betzler

M, Otto HF. Proliferative activity of neuroendocrine tu-

mours of the gastroenteropancreatic endocrine system: DNA

flow cytometric and immunohistological investigations. Gut

1991;/32:/949�53.

[14] Moyana TN, Xiang J, Senthilselvan A, Kulaga A. The

spectrum of neuroendocrine differentiation among gastro-

intestinal carcinoids: Importance of histologic grading, MIB-

1, p53, and bcl-2 immunoreactivity. Arch Pathol Lab Med

2000;/124:/570�6.

[15] Sokmensuer C, Gedikoglu G, Uzunalimoglu B. Importance

of proliferation markers in gastrointestinal carcinoid tumors:

A clinicopathologic study. Hepatogastroenterology 2001;/48:/

720�3.

[16] Amarapurkar AD, Davies A, Ramage JK, Stangou AJ, Wight

DG, Portmann BC. Proliferation of antigen MIB-1 in

metastatic carcinoid tumours removed at liver transplanta-

tion: Relevance to prognosis. Eur J Gastroenterol Hepatol

2003;/15:/139�43.

[17] Kimura N, Miura W, Noshiro T, Miura Y, Ookuma T,

Nagura H. Ki67- is an indicator of progression of neuroen-

docrine tumors. Endocr Pathol 1994;/5:/223�8.

[18] Van Eeden S, Nederlof PM, Taal BG, Offerhaus GJ, Van

Velthuysen ML. A tumour with a neuroendocrine and

papillary serous component: Two or a pair? J Clin Pathol

2002;/55:/710�4.

[19] Chaudhry A, .Oberg K, Wilander E. A study of biological

behavior based on the expression of a proliferating antigen in

neuroendocrine tumors of the digestive system. Tumour Biol

1992;/13:/27�35.

[20] Burke AP, Thomas RM, Elsayed AM, Sobin LH. Carcinoids

of the jejunum and ileum: An immunohistochemical and

clinicopathologic study of 167 cases. Cancer 1997;/79:/

1086�93.

[21] .Oberg K, Astrup L, Eriksson B, Falkmer SE, Falkmer UG,

Gustafsen J, et al. Guidelines for the management of

gastroenteropancreatic neuroendocrine tumours (including

bronchopulmonary and thymic neoplasms). Part I-general

overview. Acta Oncol 2004;/43:/617�25.

[22] Canavese G, Azzoni C, Pizzi S, Corleto VD, Pasquali C,

Davoli C, et al. p27: A potential main inhibitor of cell

proliferation in digestive endocrine tumors but not a marker

of benign behavior. Hum Pathol 2001;/32:/1094�101.

[23] Grimelius L. A silver nitrate stain for alpha-2 cells in human

pancreatic islets. Acta Soc Med Ups 1968;/73:/243�70.

[24] Grimelius L, Wilander E. Silver stains in the study of

endocrine cells of the gut and pancreas. Invest Cell Pathol

1980;/3:/3�12.

[25] Andersson T, Wilander E, Eriksson B, Lindgren PG, .Oberg

K. Effects of interferon on tumor tissue content in liver

metastases of human carcinoid tumors. Cancer Res 1990;/50:/

3413�5.

[26] .Oberg K. Diagnosis and treatment of carcinoid tumors.

Expert Rev Anticancer Ther 2003;/3:/863�77.

[27] .Oberg K. Management of neuroendocrine tumours. Ann

Oncol 2004;/15(Suppl 4):/293�8.

[28] Grimelius L. The argyrophil reaction in islet cells of adult

human pancreas studies with a new silver nitrate procedure.

Acta Soc Med Ups 1968;/73:/271�94.

[29] Grimelius L. Silver stains demonstrating neuroendocrine

cells. Biotech Histochem 2004;/79:/37�44.

[30] Lollgen RM, Hessman O, Szabo E, Westin G, .Akerstrom G.

Chromosome 18 deletions are common events in classical

midgut carcinoid tumors. Int J Cancer 2001;/92:/812�5.

[31] D’Adda T, Pizzi S, Azzoni C, Bottarelli L, Crafa P, Pasquali

C, et al. Different patterns of 11q allelic losses in digestive

endocrine tumors. Hum Pathol 2002;/33:/322�9.

[32] Kytola S, Hoog A, Nord B, Cedermark B, Frisk T, Larsson

C, et al. Comparative genomic hybridization identifies loss of

18q22-qter as an early and specific event in tumorigenesis of

midgut carcinoids. Am J Pathol 2001;/158:/1803�8.

[33] Duerr E, Mizukami Y, Warshaw A, Kulke M, Chung C.

Gene expression profiles of pancreatic neuroendocrine

tumors and gastrointestinal carcinoids. Gastroenterology

2006;/130(Suppl 2):/S1887.

[34] Rorstad O. Prognostic indicators for carcinoid neuroendo-

crine tumors of the gastrointestinal tract. J Surg Oncol 2005;/

89:/151�60.

756 J. L. Cunningham et al.

Act

a O

ncol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

118.

127.

75.2

25 o

n 05

/20/

14Fo

r pe

rson

al u

se o

nly.