Cytokine 27 (2004) 58e65

www.elsevier.com/locate/jnlabr/ycyto

Tumour necrosis factor-alpha, interleukin-6, and fasting seruminsulin correlate with clinical outcome in metastaticbreast cancer patients treated with chemotherapy

Hakan Bozcuka,), Gulcan Uslub, Mustafa Samura, Mustafa Yıldıza,Tomris Ozbenc, Mustafa Ozdog�ana, Mehmet Artacxa, Hasan Altunbasxd,

_IIlhan Akanc, Burhan SavasxaaAkdeniz University Medical Faculty, Department of Medical Oncology, Dumlupinar Bulvari, 07070 Antalya, TurkeybAkdeniz University Medical Faculty, Department of Internal Medicine, Dumlupinar Bulvari, 07070 Antalya, Turkey

cAkdeniz University Medical Faculty, Department of Biochemistry, Dumlupinar Bulvari, 07070 Antalya, TurkeydAkdeniz University Medical Faculty, Department of Endocrinology, Dumlupinar Bulvari, 07070 Antalya, Turkey

Received 19 January 2004; received in revised form 5 April 2004; accepted 6 April 2004

Abstract

Introduction: To assess the relationship of various growth factors and cytokines with the clinical outcome in metastatic breast cancer

patients receiving chemotherapy.Methods: Consecutive, metastatic breast cancer patients with measurable disease and receiving palliative chemotherapy wereprospectively evaluated for the predictors of progression free survival (PFS) and overall survival (OAS) in relation to serum insulin,

insulin resistance, interleukin-6 (IL-6), and tumour necrosis factor-alpha (TNF-a).Results: Estrogen receptor (ER) status, serum IL-6 and serum TNF-a were the independent determinants of PFS, with RR ¼ 0:28(0.13e0.60), P ¼ 0:001, RR ¼ 2:48 (1.24e5.61), P ¼ 0:012, and RR ¼ 0:48 (0.23e1.01), P ¼ 0:053, respectively. The factors relatedwith OAS in the multivariate analysis were histological grade (RR ¼ 7:88 (2.33e26.62), P ¼ 0:001), ER status (RR ¼ 0:18(0.06e0.57), P ¼ 0:003), serum insulin (RR ¼ 0:87 (0.77e0.97), P ¼ 0:016), and serum IL-6 (RR ¼ 5:99 (1.89e18.97), P ¼ 0:002).Conclusions: We show for the first time that fasting serum insulin and TNF-a levels are independent predictors for OAS and PFS,

respectively, in metastatic breast cancer patients. In addition, we also confirm that IL-6 is a poor prognosticator in this group. Theseresults suggest that insulin and TNF-a are important biomolecules that may be directly involved in vivo in the progression ofmetastatic breast cancer.� 2004 Elsevier Ltd. All rights reserved.

Keywords: Breast cancer; Chemotherapy; IL-6; Insulin; TNF-a

1. Introduction

Various mitogenic molecules and their receptors areincreasingly being explored to possess prognostic and/orpredictive value in breast cancer. Two better studiedmembers of the epidermal growth factor type 1 sub-family are: c-erb-B2 (HER-2), and epidermal growth

) Corresponding author. Liman mah., 39. Sokak, Adalya Sitesi, C

Blok, No: 4, Antalya 07070, Turkey. Tel.: C90-542-356-2027; fax:

C90-242-259-1454.

E-mail address: hbozcuk@antnet.net.tr (H. Bozcuk).

1043-4666/$ - see front matter � 2004 Elsevier Ltd. All rights reserved.

doi:10.1016/j.cyto.2004.04.002

factor receptor (EGFR, HER-1). C-erb-B2 overexpres-sion exists in 20e30% of infiltrating breast cancers [1]and has been shown to be associated with poorer diseasefree survival, resistance to cyclophosphamide, metho-trexate, fluorouracil, prednisone (CMFP) chemotherapyand to tamoxifen [2,3], and increased likelihood ofdistant metastasis at presentation [4]. EGFR is overex-pressed in 35e50% of breast cancers and, similarly,conveys resistance to tamoxifen [5,6].

Another group of mitogenic molecules consists ofinsulin and related compounds. Despite the fact thatinsulin like growth factor-1 (IGF-1), insulin and insulin

59H. Bozcuk et al. / Cytokine 27 (2004) 58e65

resistance have been shown to be associated with anincreased risk of developing breast cancer [7,8], and thata recent report showed that in early stage breast cancer,increased fasting serum insulin was linked with inferiordisease free survival [9], their roles in the clinical out-come of metastatic disease remain to be explored. Like-wise, solid data on the prognostic or predictive value ofother growth factors and mitogenic molecules in breastcancer is lacking.

One of such mitogenic molecules with a potential to berelated with the clinical outcome, interleukin-6 (IL-6),also a cytokine, is a multifunctional protein and hasgrowth promoting/mitogenic activity on T, hemopoieticstem, hybridoma and liver cells [10e13]. Although pre-vious reports suggest that it is a poor predictive andprognostic factor in breast cancer [14], and that its level iselevated in malignancies like myelomas, lymphomas andovarian cancer [15e17], conclusive evidence of clinicalsignificance in metastatic breast cancer does not exist atpresent. Tumour necrosis factor-alpha (TNF-a) is ano-ther cytokine that is extremely pleiotropic, and is alsoable to affect the expression of growth factors and othercytokines, via multiple signal transduction pathways [18].It is elevated in breast cancer patients undergoing mas-tectomy [19], but again whether this has any clinicalimplications is not known.

It was in an attempt to explore, in the setting ofmetastatic breast cancer treated with chemotherapy, thepredictive and prognostic roles of those less-studiedgrowth promoting and mitogenic molecules mentionedabove, that we conducted this study.

2. Results

2.1. General features

A total of 43 patients were enrolled. Three patientsdied due to progressive disease before completing twocourses of chemotherapy. In these three patients, al-though confirmation of progression was not possible, asimaging at the time of progression could not be per-formed, no major doubt existed about the presence ofprogression by clinical judgement. Thus, all 43 patientswere included accordingly in the survival analysis.

Median age was 50, 42% having an ECOG perfor-mance status of 1, and the majority (95%) had visceraldisease. Only two patients (5%) had measurable boneand soft tissue lesions. Forty-four percent had a taxanecontaining chemotherapeutic regimen, after a median oftwo previous different chemotherapy schedules. Of the43 patients, a total of 20 (47%, CR: 0 (0%), PR: 20(47%)) responded to chemotherapy (see Table 1). Of the13 cases with progressive disease, progression in 3 cases(those who died early in the study period) was decidedsolely by clinical judgement, whereas in the remaining 10

subjects, progression was confirmed by proper imagingtechniques.

The mean values in our cohort (and correspondingreference values in parentheses as defined in the litera-ture) for insulin, insulin resistance, IL-6, and TNF-a,respectively, were as follows: 10.4 Mu/l (4.7e11.1 Mu/l),2.6 (1.3e2.0), 14.6 pg/ml (6 pg/ml), and 6.5 pg/ml (2.4e3.6 pg/ml) [23,24]. The details of the patient, treatmentand laboratory characteristics are summarised in Table 1.

Body mass index prior to chemotherapy administra-tion was correlated in our cohort with insulin resistance(RHOMA) and serum insulin (r ¼ 0:41,P ¼ 0:006, and r ¼0:41,P ¼ 0:007, respectively). Two other important asso-ciations were between RHOMA and serum insulin (r ¼0:95,P! 0:001), and between IL-6 andTNF-a (r ¼ 0:39,P ¼ 0:010).

2.2. Progression free survival

During the study period, 29/43 (67%) of patientsprogressed. The median PFS was 141 days for the wholegroup. Factors with a Cox P value of !0.20 in theunivariate tests were; estrogen receptor status (negativevs. positive, RR ¼ 0:31, P ¼ 0:002), previous radiother-apy (No vs. Yes, RR ¼ 2:01, P ¼ 0:068), insulin re-sistance (%2.2 vs. O2.2, RR ¼ 0:56, P ¼ 0:125), serumTNF-a (%6.20 pg/ml vs. O6.20 pg/ml, RR ¼ 0:55, P ¼0:117), and serum IL-6 (%5 pg/ml vs. O5 pg/ml, RR ¼2:48, P ¼ 0:012). In the multivariate analysis (Coxmodel), three factors were significant; estrogen receptorstatus (RR ¼ 0:28 (CI: 0.13e0.60), P ¼ 0:001), serumIL-6 (RR ¼ 2:48 (CI: 1.24e5.61), P ¼ 0:012), and serumTNF-a (RR ¼ 0:48 (CI: 0.23e1.01), P ¼ 0:053, witha borderline significance). Refer to Table 2 for the asso-ciation of other factors with PFS and to Fig. 1a, b and cfor KaplaneMeier survival curves and, figure legend,for median survival times and associated P values.

2.3. Overall survival

Eighteen out of 43 (42%) cases died during the courseof the study. The whole cohort had a median OAS of415 days. Seven factors had a Cox P value of less than0.20 in the univariate analysis. These were tumour his-tological grade (1,2 vs. 3, RR ¼ 2:35, P ¼ 0:080), estro-gen receptor status (negative vs. positive, RR ¼ 0:24,P ¼ 0:003), progesterone receptor status (negative vs.positive, RR ¼ 0:42, P ¼ 0:086), previous radiotherapy(No vs. Yes, RR ¼ 2:68, P ¼ 0:047), insulin resistance(%2.2 vs. O2.2, RR ¼ 0:34, P ¼ 0:036), serum insulin(continuous variable, RR ¼ 0:92, P ¼ 0:067), and serumIL-6 (%5 pg/ml vs. O5 pg/ml, RR ¼ 2:73, P ¼ 0:046).The final Cox model selected four of these variables;tumour histological grade (RR ¼ 7:88 (CI: 2.33e26.62),P ¼ 0:001), estrogen receptor status (RR ¼ 0:18(CI: 0.06e0.57), P ¼ 0:003), serum insulin (RR ¼ 0:87

60 H. Bozcuk et al. / Cytokine 27 (2004) 58e65

Table 1

Patient, treatment and laboratory characteristics

n (%) Mean Median Range

General patient characteristics

Total 43 (100)

Age 50 50 32e76

Performance status (ECOG)

0 3 (7)

1 18 (42)

2 17 (40)

3 5 (11)

Comorbidity

Present 29 (67)

Absent 14 (33)

Menopausal status

Premenopausal 22 (51)

Postmenopausal 21 (49)

Body mass index 27.50 27.64 17.09e39.92Baseline CA 15-3 140.84 70.00 9.23e969.50

c-erb-B2 expression

(by immunohistochemistry)

0 to ‘‘++’’ 23 (53)

‘‘+++’’ 20 (47)

c-erb-B2 expressiona 0.60 0.60 0.00e0.95

Estrogen receptor

Positive 26 (60)

Negative 15 (35)

Unknown 2 (5)

Progesteron receptor

Positive 18 (42)

Negative 23 (53)

Unknown 2 (5)

Pathology

Invasive ductal

carcinoma

32 (74)

Other 11 (26)

Grade

1 4 (9)

2 28 (65)

3 11 (26)

Total maximum

transverse

diameter (cm)

3.8 3.0 1.0e13.4

Number of organs

involved by tumour

2 or less 33 (77)

3 or more 10 (23)

Visceral organ

metastasis

Present 41 (95)

Absent 2 (5)

Treatment characteristics

Number of chemotherapy

regimens recieved

2 0e6

Previous radiotherapy

Yes 28 (65)

No 15 (35)

Previous hormonal treatment

Yes 18 (42)

No 25 (58)

Chemotherapy regimen

administered

Taxane containing 19 (44)

Vinorelbin containing 5 (12)

CA or CEb 18 (42)

CMF 1 (2)

(CI: 0.77e0.97), P ¼ 0:016), and serum IL-6 (RR ¼ 5:99(CI: 1.89e18.97), P ¼ 0:002). See Table 2 for the associ-ation of other factors with OAS, and Fig. 1d, e and f forKaplaneMeier survival curves, and figure legend, formedian survival times and associated P values.

Of note, among the treatment factors tested, numberof chemotherapy regimens previously received or type ofchemotherapy regimen administered failed to improvePFS (P ¼ 0:264 and 0.963, respectively), or OAS(P ¼ 0:581 and 0.553, respectively).

3. Discussion

The first novel contribution from this study is thedemonstration that patients living more progression freehave increased serum TNF-a with a borderline signifi-cance. To the best of our knowledge, no previous reportof a prognostic value of serum TNF-a in metastaticbreast cancer patients exists in the literature. It haspreviously been shown in vitro that TNF-a blocksgrowth of breast cancer cells by impairing signals fromIGF-1 receptor [25]. It has also been suggested in T-47Dbreast carcinoma cells that TNF related pathwaysmay augment chemotherapy effect [26]. These factorstogether may potentially explain why serum TNF-a O6.20 pg/ml is associated with 52% less risk of progres-sion in our cohort.

We also proved in our cohort that IL-6 is negativelycorrelated with survival. Our observation that higherserum IL-6 is associated with inferior progression freeand overall survival is in accordance with the previousfinding of Zhang and Adachi [14], where they showed in46 metastatic breast cancer patients that serum IL-6value greater than a cut off value of 4 pg/ml wasassociated with 3.86 times increased risk of death asopposed to 5.99 times increased risk of death in ourcohort, when the cut off value was 5 pg/ml. Our findingon IL-6 together with that of Zhang et al., propose IL-6

Table 1 (continued)

n (%) Mean Median Range

Best objective response

Complete response 0 (0)

Partial response 20 (47)

Stable disease 10 (23)

Progressive disease 13 (30)

Cytokine/mitogenic markers

Insulin (Mu/l) 10.42 8.66 0.41e26.33

Insulin resistance 2.59 2.18 0.07e7.10

IL-6 (pg/ml) 14.56 5.00 2.30e100.00

TNF-a (pg/ml) 6.46 6.20 3.90e15.60

a Over a scale of 300, see text for details.b Cyclophosphamide, adriamycin or cyclophosphamide, epirubicin.

61H. Bozcuk et al. / Cytokine 27 (2004) 58e65

Table 2

Determinants of clinical outcome (factors associated with progression-free and overall survival)

Parameter Univariate tests Multivariate tests

RRa P RRa P

PFS OAS PFS OAS PFS OAS PFS OAS

General patient/tumor characteristics

Age (continuous variable) 1.00 0.98 0.996 0.349

Menopausal status

(premenopausal vs. postmenopausal)

0.85 0.70 0.673 0.482

Number of organs

involved by tumour (1,2 vs. 3,4)

0.91 0.84 0.832 0.758

Pathology

(invasive ductal carcinoma vs. other)

1.01 1.79 0.986 0.357

Grade (1,2 vs. 3) 1.55 2.35 0.299 0.080 7.88 0.001

Comorbidity (absent vs. present) 1.22 1.61 0.621 0.352

Estrogen receptor

status (negative vs. positive)

0.31 0.24 0.002 0.003 0.28 0.18 0.001 0.003

Progesteron receptor

status (negative vs. positive)

0.86 0.42 0.698 0.086 0.950

c-erb-B2 expression

(%2 vs. 3 by immunohistochemistry)

1.17 1.14 0.688 0.790

(as a continuous variable)b 3.67 3.28 0.225 0.367

Body mass

index (continuous variable)

1.03 1.06 0.384 0.239

Performance status (0,1 vs. 2,3) 1.56 1.45 0.238 0.446

Total maximum

transverse diameter

(logarithmic transformation)

0.48 0.81 0.285 0.808

Baseline CA 15-3

(logarithmic transformation)

1.44 1.53 0.388 0.429

Duration of disease

(%1 year vs. O1 year)

1.55 1.36 0.355 0.600

Treatment characteristics

Previous radiotherapy (No vs. Yes) 2.01 2.68 0.068 0.047 0.417 0.863

Previous hormonal

treatment (No vs. Yes)

0.93 0.53 0.851 0.225

Number of previous

chemotherapy regimens received

(%2 vs. O2)

1.54 1.02 0.264 0.963

Chemotherapy regimen

administered

(‘‘Older’’ vs. ‘‘Newer’’ regimen)c

1.24 1.34 0.581 0.553

Cytokine/mitogenic markers

Insulin resistance (%2.2 vs. O2.2) 0.56 0.34 0.125 0.036 0.407 0.724

Insulin (continuous variable) 0.97 0.92 0.407 0.067 0.87 0.016

TNF-a (%6.20 pg/ml vs. O6.20 pg/ml) 0.55 0.90 0.117 0.830 0.48 0.053

IL-6 (%5 pg/ml vs. O5 pg/ml) 2.48 2.73 0.018 0.046 2.64 5.99 0.012 0.002

Abbreviation: RR, relative risk; PFS, progression-free survival; OAS, overall survival.a RR for a categorical variable is defined as the risk for the last category with reference to the first category.b Over a scale of 300, see text details.c ‘‘Older’’ regimen: CA, CE, or CMF, ‘‘Newer’’ regimen: Taxane or Vinorelbin containing regimens.

as an independent negative prognosticator in breastcancer patients with metastatic disease.

The second novel finding from this study is theindependent association of the higher fasting seruminsulin levels with increased overall survival. Thus,serum insulin level has come out from this study asa positive prognostic factor. The explanation of themechanism(s) of insulin and survival association is notpossible at this stage, as this study was not designed to

answer this. However, we feel that the following mech-anisms may explain why high serum insulin level isaccompanied by better survival: firstly, insulin resistanceis a block at the post receptor level, and is accompaniedby high insulin levels [20]. It has also been shown thatinsulin induces mdr1b expression in rat hepatoma cellsby NF-KappaB via Raf-1 kinase signalling pathway[27]. Thus, taking the factors above into consideration,at the presence of a block at the post insulin receptor

62 H. Bozcuk et al. / Cytokine 27 (2004) 58e65

Overall survival (whole cohort)

Time (days)

10008006004002000

Cum

ulat

ive

over

all s

urvi

val

1,0

,9

,8

,7

,6

,5

,4

,3

,2

,1

0,0

Survival Function

Censored

Progression free survival (whole cohort)

Time (days)

10008006004002000

Cum

ulat

ive

prog

ress

ion

free

surv

ival

1,0

,8

,6

,4

,2

0,0

Survival Function

Censored

Overall survival according to serum IL-6

Time (days)

10008006004002000

talumu

Ci

lavivrus llarevo ev

1,0

,8

,6

,4

,2

0,0

IL-6

>5 pg/ml

>5 pg/ml-censored

</=5 pg/ml

</=5 pg/ml-censored

Progression free survival according to

serum TNF-alpha

Time (days)

10008006004002000

lavivrus ee rf noisse rg or p evitalumu

C

1,0

,8

,6

,4

,2

0,0

TNF-alpha

>6.20 pg/ml

censored

</=6.20 pg/m

censored

Overall survival according to serum Insulin

Time (days)

10008006004002000

talumu

Ci

lavivru s llare vo ev

1,0

,8

,6

,4

,2

0,0

Insulin

>9 Mu/l

>9 Mu/l-censored

</=9 Mu/l

</=9 Mu/l-censored

Progression free survival according to serum IL-6

Time (days)

10008006004002000

lavivrus eerf noissergorp evitalumu

C

1,0

,8

,6

,4

,2

0,0

IL-6

>5 pg/ml

>5 pg/ml-censored

</=5 pg/ml

</=5 pg/ml-censored

a d

b e

c f

Fig. 1. Survival curves of metastatic breast cancer patients with respect to cytokine/mitogenic markers. Median survival (MS) times and log rank P

values are as follows: (a) MS=141 days, (b) MS (IL-6% 5 pg/ml)= 380 days, MS (O5 pg/ml)= 109 days, P ¼ 0:014, (c) MS (TNF-a% 6.20 pg/

ml)= 109 days, MS (O6.20 pg/ml)= 380 days, P ¼ 0:112, (d) MS=415 days, (e) MS (IL-6% 5 pg/ml)= not reached, MS (O5 pg/ml)= 371 days,

P ¼ 0:038, ( f) MS (insulin% 9 Mu/l)= 371 days, MS (O9 pg/ml)= 558 days, P ¼ 0:086.

63H. Bozcuk et al. / Cytokine 27 (2004) 58e65

level, fasting hyperinsulinemia would occur, but thatwould not lead to increased mdr1b expression andactually in turn might diminish P-glycoprotein expres-sion. Consequently, less chemo-resistant tumour clonesmight be better killed by chemotherapy, resulting insuperior survival.

Secondly, when there is a post receptor block, thefunction of other growth factors like IGF-1 may also beaffected (diminished), as they may share the samereceptor with insulin to exert their effects leading tobetter survival [28]. Different effects of these two mech-anisms may even superimpose. It is noteworthy that therelationship of high insulin with better survival is notlinked with body mass index, as this was controlled forin the analysis. Interestingly, for breast cancer patientsin the adjuvant setting, higher insulin levels were foundto be negatively associated with disease free survival asproposed by Goodwin et al. [9]. This, in a way may seemin contrast to what we have shown in the metastaticsetting, and this difference may theoretically be relatedwith different cellular pathways and biological consid-erations in early stages (adjuvant setting), as well asdifferent activities of the aberrant insulin pathways inlate and early stages.

Other independent associations in this study that arein accordance with the findings from other studies arethose of positive estrogen receptor-better overall andprogression free survival, higher histological grade-worse overall survival [29e32]. These factors point outthat our sample is a typical metastatic breast cancerpopulation possessing classical prognostic associationsas defined in the literature.

Although our cohort is a relatively small one, andvarious chemotherapy regimens with different numbersof previous chemotherapy schedules are administered,since none of the treatment factors in this study, like thetype of chemotherapy and the number of previousdifferent regimens used, were significantly associatedwith survival, we think that the correlation of TNF-aand insulin with survival in our cohort is independentfrom treatment factors. Obviously, our results need con-firmation in larger controlled studies, preferably in che-monaive patients treated with a standard chemotherapyregimen.

In summary, we propose for the first time in meta-static breast cancer, fasting serum insulin as a predictorof OAS, and serum TNF-a as a determinant of PFS. Weadditionally prove that IL-6 is a negative prognosticfactor in this setting. This information may help in theprognostication of breast cancer patients with metastaticdisease. Additionally, the clinical significance of thesegrowth modulating factors points out the need for morework to explore better the intracellular events related tothese factors in metastatic breast cancer. As our resultssuggest that both insulin and TNF-a are linked with theprogression of metastatic breast cancer, perhaps, we also

need to start testing the potential role of blocking thepost insulin receptor cascade, and exploit TNF-a relatednovel treatment strategies in this setting.

4. Materials and methods

4.1. Patients

Amongmetastatic breast cancer patients attending theoutpatient unit of the Medical Oncology Department atAkdeniz University Medical Faculty in the year 2001, 43consecutive subjects with measurable disease were in-cluded in the study. Those with brain metastasis wereexcluded, whereas patients with bone only disease wereallowed only if their primary lesion was measurable.Subjects with Diabetes Mellitus, or using medicationssuch as steroids prior to the start of chemotherapy (at thetime of blood collection for cytokines andmitogens) wereexcluded from the study, since these factors could alterblood glucose and insulin levels or may cause insulinresistance. Only patients embarking on a new palliativechemotherapy regimen were permitted. Written consentfor treatment was obtained from each participant. Thechoice of the chemotherapy regimen for each patient wasdictated by the previous regimens received, the patientand disease related factors, and along with the sugges-tions of the primary physician in charge.

4.2. Methods

4.2.1. Sample collection, storage, immunoassaysand insulin resistance

For the immunoassay tests used in this study, earlymorning blood samples after 10 h of overnight fastingwere obtained from each subject a day prior to the onsetof 1st chemotherapy cycle. Plasma was collected usingEDTA as an anticoagulant. After centrifuging for15 min at 1000!g within 30 min of collection, sampleswere aliquoted, frozen at �40 (C, and stored until studytime. All samples were analysed later, at the same time.

Insulin resistance was determined by using the ho-meostasis model assessment (HOMA) methodology;three blood samples, each 5 min apart are separatelydrawn for insulin and glucose levels, and their mean isused in the following formula [20]:

Insulin resistance ðRHOMAÞ ¼ insulin!glucose=22:5

In order to measure the serum levels of insulin,TNF-a, and IL-6, Elecsys system insulin assay fromRoche (catalogue number 12017547), Quantikine hu-man TNF-a immunoassay (catalogue number DTA50),and Quantikine human IL-6 immunoassay (cataloguenumber D6050), from R&D Systems, Inc., Minneapolis,were used, respectively.

64 H. Bozcuk et al. / Cytokine 27 (2004) 58e65

4.2.2. Clinical outcomeFirstly, progression free survival (PFS), calculated as

the time period between onset of the 1st cycle ofchemotherapy and the date of progression or date lastseen, and secondly, overall survival (OAS), defined asthe period of time between onset of the 1st cycle ofchemotherapy and the date of death or date last seen,were calculated as the indicators of clinical outcome.During the study period, patients were seen at theMedical Oncology Outpatient Clinics regularly, at leastmonthly.

Data on patient, tumour and treatment character-istics at the study entry for each subject was collectedfrom the medical records. Dimensions of the metastaticlesions before the 1st and after the 2nd cycles, and aftereach additional 2 cycles of chemotherapy, and also atthe time of disease progression were assessed, andchange in tumour diameters and progression were de-fined according to the Response Evaluation Criteria inSolid Tumours (RECIST) [21].

To assess relationship with the clinical outcome,c-erb-B2 expression was also defined as a continuousvariable on a 300 (3!100) scale, basically as a ratio,with the denominator always being 300, and the numer-ator as the product of the percentage of cells staining(from 0 to 100) and the intensity of staining (from 1C to3C).

4.2.3. Statistical analysisCorrelation among body mass index, insulin resis-

tance, and levels of cytokine and mitogenic compoundswere evaluated by Spearman’s nonparametric correla-tion tests. In order to explore the factors related withprogression-free and overall survival, univariate andmultivariate Cox regression analyses were performed.In the case of abnormal distribution and where trans-formation to normality was not possible, independentinterval variables like insulin resistance, TNF-a andIL-6 were collapsed into categorical variables aroundthe corresponding median values to be included in thesubsequent analyses. However, in evaluating the associ-ation of serum insulin with survival, we used a differentmethodology by including serum insulin as a continuousvariable in the Cox analysis, as it was normally distrib-uted, in parallel to a previously recommended strategy[22], where we have no prior idea of any associationbetween the dependent and independent variables. Var-iables with a P value of again %0.20 in the univariateanalyses were, in the next stage, evaluated by multivar-iate Cox regression analysis, using forward LR. Survivalcurves were constructed and median survival times werecalculated according to KaplaneMeier method. Statis-tical analysis was performed with SPSS for Windowsversion 10.0 (SPSS Inc., Chicago, IL).

Acknowledgements

We thank Levent Undar, MD, for his help in inter-preting the immunoassay data. Competing interests:A partial financial support was received from AVENTISfor the immunoassays used in this study.

References

[1] Slamon DJ, Clark GM, Wong SG, Levin WJ, Ullrich A, McGuire

WL. Human breast cancer: correlation of relapse and survival

with amplification of the HER-2/neu oncogene. Science 1987;235:

177e82.

[2] Gusterson BA, Gelber RD, Goldhirsch A, Price KN, Save-

Soderborgh J, Anbazhagan R, et al. Prognostic importance of

c-erbB-2 expression in breast cancer. International (Ludwig)

Breast Cancer Study Group. J Clin Oncol 1992;10:1049e56.

[3] Pinto AE, Andre S, Pereira T, Nobrega S, Soares J. C-erbB-2

overexpression identifies a subgroup of estrogen receptor positive

(ERC) breast cancer patients with poor prognosis. Ann Oncol

2001;12(4):525e33.

[4] Bozcuk H, Uslu G, Pestereli E, Samur M, Ozdogan M,

Karaveli S, et al. Predictors of distant metastasis at presentation

in breast cancer: a study also evaluating associations among

common biological indicators. Breast Cancer Res Treat 2001;68:

239e48.

[5] Nicholson S, Richard J, Sainsbury C, Halcrow P, Kelly P, Angus

B, et al. Epidermal growth factor receptor (EGFr): results of

a 6-year follow-up study in breast cancer with emphasis on the

node negative subgroup. Br J Cancer 1991;63:146e50.

[6] Nicholson S, Halcrow P, Sainsbury JR, Angus B, Chambers P,

Farndon JR, et al. Epidermal growth factor receptor (EGFr)

status associated with failure of primary endocrine therapy in

elderly postmenopausal patients with breast cancer. Br J Cancer

1988;58:810e4.

[7] Hankinson SE, Willett WC, Colditz GA, Hunter DJ, Michaud

DS, Deroo B, et al. Circulating concentrations of IGF1 and risk

of breast cancer. Lancet 1998;351:1393e6.

[8] Stoll BA. Upper abdominal obesity, insulin resistance and breast

cancer risk. Int J Obes Relat Metab Disord 2002;26:747e53.[9] Goodwin PJ, Ennis M, Pritchard KI, Trudeau ME, Koo J,

Madarnas Y, et al. Fasting insulin and outcome in early-stage

breast cancer: results of a prospective cohort study. J Clin Oncol

2002;20(1):42e51.[10] Lotz M, Jirik F, Kabouridis P, Tsoukas C, Hirano T, Kishimoto

T, et al. B cell stimulating factor 2/interleukin-6 is a costimulant

for human thymocytes and T lymphocytes. J Exp Med 1988;

167(3):1253e8.[11] Leary AG, Ikebuchi K, Hirai Y, Wong GG, Yang YC, Clark SC,

et al. Synergism between interleukin-6 and interleukin-3 in

supporting proliferation of human hematopoietic stem cells:

comparison with interleukin-1 alpha. Blood 1988;71(6):1759e63.

[12] Wong GG, Witek-Giannotti JS, Temple PA, Kriz R, Ferenz C,

Hewick RM, et al. Stimulation of murine hemopoietic colony

formation by human IL-6. J Immunol 1988;140(9):3040e4.[13] Van Damme J, Van Beeumen J, Decock B, Van Snick J, De Ley

M, Billiau A. Separation and comparison of two monokines with

lymphocyte-activating factor activity: IL-1 beta and hybridoma

growth factor (HGF). Identification of leukocyte-derived HGF as

IL-6. J Immunol 1988;140(5):1534e41.

[14] Zhang GJ, Adachi I. Serum interleukin-6 levels correlate to

tumour progression and prognosis in metastatic breast carcinoma.

Anticancer Res 1999;19:1427e32.

65H. Bozcuk et al. / Cytokine 27 (2004) 58e65

[15] Hirano T, Akira S, Taga T, Kishimoto T. Biological and clinical

aspects of interleukin 6. Immunol Today 1990;11(12):443e9.

[16] Hirano T. Interleukin-6 and its relation to inflammation and

disease. Clin Immunol Immunopathol 1992;62(1 pt 2):60e5.

[17] Watson JM, Berek JS, Martinez-Maza O. Growth inhibition of

ovarian cancer cells induced by antisense IL-6 oligonucleotides.

Gynecol Oncol 1993;49(1):8e15.

[18] Vilcek J, Lee TH. Tumour necrosis factor. New insights into the

molecular mechanisms of its multiple actions. J Biol Chem 1991;

266(12):7313e6.

[19] Sheen-Chen SM, Chen WJ, Eng HL, Chou FF. Serum concen-

tration of tumour necrosis factor in patients with breast cancer.

Breast Cancer Res Treat 1997;43:211e5.

[20] Radziuk J. Insulin sensitivity and its measurement: structural

commonalties among the methods. J Clin Endocrinol Metab 2000;

85(12):4426e33.[21] James K, Eisenhauer E, Christian M, Terenziani M, Vena D,

Muldal A, et al. Measuring response in solid tumours: unidimen-

sional versus bidimensional measurement. J Natl Cancer 1999;

91(6):523e8.

[22] Parmar MKB, Machin D. Survival analysis: a practical approach.

Great Britain: John Wiley and Sons; 1999.

[23] Demirbas B, Guler S, Cakir B, Culha C, Aral Y. Plasma tumour

necrosis factor-alpha and insulin resistance in non-diabetic

hypertensive subjects. Horm Res 2002;58(6):283e6.

[24] Benoy I, Solgado R, Colpoert C, Weytjens R, Vermeulen PB,

Dirix CY. Serum interleukin-6, plasma VEGF, serum VEGF, and

VEGF platelet load in breast cancer patients. Clin Breast Cancer

2002;2(4):311e5.

[25] Shen WH, Zhou JH, Broussard SR, Freund GG, Dantzer R,

Kelley KW. Proinflammatory cytokines block growth of breast

cancer cells by impairing signals from a growth factor receptor.

Cancer Res 2002;62(16):4746e56.

[26] Morgan M, Williams BA, Blay J, Hoskin DW. Chemosensitiza-

tion of T-47D breast carcinoma cells to TRAIL and Fas

receptor-induced killing. Anticancer Res 2002;22(2A):673e6.[27] Zhou G, Kuo MT. NF-kappaB-mediated induction of mdr1b

expression by insulin in rat hepatoma cells. J Biol Chem 1997;

272(24):15174e83.[28] Pandini G, Vigneri R, Costantino A, Frasca F, Ippolito A,

Fujita-Yamaguchi Y, et al. Insulin and insulin-like growth factor-I

(IGF-1) receptor overexpression in breast cancers leads to insulin/

IGF-1 hybrid receptor overexpression: evidence for a secondmech-

anism of IGF-1 signalling. Clin Cancer Res 1999;5(7):1935e44.

[29] Ryberg M, Nielsen D, Osterlind K, Skovsgaard T, Dombernow-

sky P. Prognostic factors and long-term survival in 585 patients

with metastatic breast cancer treated with epirubicin-based

chemotherapy. Ann Oncol 2001;12(1):81e7.

[30] Schneeweiss A, Hensel M, Goerner R, Khbeis T, Hohaus S,

Egerer G, et al. Comparison of double and triple high-dose

chemotherapy with autologous blood stem cell transplantation in

patients with metastatic breast cancer. Stem Cells 2001;19(2):

151e60.

[31] Tsuda H, Tsugane S, Fukutomi T, Nanasawa T, Yamamoto H,

Hirohashi S. Prognostic factors for recurrent breast cancer:

univariate and multivariate analyses including histological grade

and amplification of the c-erbB-2 proto-oncogene. Jpn J Clin

Oncol 1992;22(4):244e9.[32] Niskanen E, Blomqvist C, Franssila K, Hietanen P, Wasenius

VM. Predictive value of c-erbB-2, p53, cathepsin-D and histology

of the primary tumour in metastatic breast cancer. Br J Cancer

1997;76(7):917e22.