Skeletal Muscle Depletion and Markers for Cancer Cachexia Are Strong Prognostic Factors in Epithelial Ovarian Cancer

Objective Tumor cachexia is an important prognostic parameter in epithelial ovarian cancer (EOC). Tumor cachexia is characterized by metabolic and inflammatory disturbances. These conditions might be reflected by body composition measurements (BCMs) ascertained by pre-operative computed tomography (CT). Thus, we aimed to identify the prognostically most relevant BCMs assessed by pre-operative CT in EOC patients. Methods We evaluated muscle BCMs and well established markers of nutritional and inflammatory status, as well as clinical-pathological parameters in 140 consecutive patients with EOC. Furthermore, a multiplexed inflammatory marker panel of 25 cytokines was used to determine the relationship of BCMs with inflammatory markers and patient’s outcome. All relevant parameters were evaluated in uni- and multivariate survival analysis. Results Muscle attenuation (MA)—a well established BCM parameter—is an independent prognostic factor for survival in multivariate analysis (HR 2.25; p = 0.028). Low MA—reflecting a state of cachexia—is also associated with residual tumor after cytoreductive surgery (p = 0.046) and with an unfavorable performance status (p = 0.015). Moreover, MA is associated with Eotaxin and IL-10 out of the 25 cytokine multiplex marker panel in multivariate linear regression analysis (p = 0.021 and p = 0.047, respectively). Conclusion MA—ascertained by routine pre-operative CT—is an independent prognostic parameter in EOC patients. Low MA is associated with the inflammatory, as well as the nutritional component of cachexia. Therefore, the clinical value of pre-operative CT could be enhanced by the assessment of MA.


Introduction
Cancer cachexia is a significant factor contributing to the poor performance status and high mortality rate of cancer patients. Cancer cachexia is a multifactorial and often irreversible syndrome, characterized by metabolic and inflammatory disturbances [1] and accounting for up to 20% of cancer deaths [2]. Thus, evaluating the prognostic value of markers of cachexia and understanding the underlying molecular mechanisms is essential.
In search of markers of cachexia, image analysis of computed tomography (CT) scans has the potential to reveal muscle depletion and to provide potential prognostic imaging biomarkers in cancer patients. In a recent report, Martin et al strengthen the possible clinical relevance of body composition measurements (BCMs) obtained from CT scans in cancer patients prior to therapy [3]. Modern image-based estimations of whole body skeletal muscle mass are made via measuring the mean muscle attenuation (MA)-inversely related to muscle fat contentand the skeletal muscle index (SMI) in abdominal cross-sectional CT images at the level of lumbar vertebra 3 (L3) [4] which can also reveal otherwise occult muscle depletion [3].
A progressive and generalized loss of skeletal muscle mass [5] is considered a pathological phenomenon reflecting the wasting process associated with changes in several metabolic pathways in cancer patients [6]. The catabolic process is associated with an alteration in inflammatory cytokines [7] and includes protein synthesis inhibition as well as an increase in protein degradation [8]. Cancer cachexia involves a variety of catabolic proinflammatory cytokines, such as tumor necrosis factor alpha (TNFα) and interleukin (IL)-6, but also anti-inflammatory cytokines such as IL-4 or IL-10 are involved in the pathogenesis of cancer associated metabolic disturbances [9,10], with a negative impact on prognosis [11].
It has been accepted, that cancer cachexia, associated with an ongoing tumor associated catabolism and systemic inflammatory condition, affects performance status, therapy tolerance, and survival [3,12,13]. Due to the lack of typical early symptoms, diagnosis of epithelial ovarian cancer (EOC) is frequently made at advanced disease, which in turn is associated with worsening nutritional and performance outcomes. Unfortunately, an optimal marker of cancer cachexia and an inclusion in clinical routine is still missing, although various studies highlight the importance to consider cachexia associated changes in cancer patients [2,14].
The present study aimed to investigate the prognostic value of BCMs assessed by pre-operative CT in patients with EOC. The relationship of BCMs with markers of nutritional and inflammatory status will be evaluated using a multiplexed inflammatory marker panel and common clinical-pathological parameters.

Study population
All consecutive patients with primary EOC treated between 2004 and 2012 at the Gynecological Cancer Unit of the Comprehensive Cancer Center at the Medical University of Vienna according to standardized procedures, as previously described in detail [15], were included in the study. Inclusion criteria were: (1) available digitally stored CT scans taken within 30 days pre-operatively for initial pretherapeutic staging and suitable for adequate image analysis; (2) presence of a frozen serum sample collected at time of diagnosis (IRB approval number 366/ 2003 -Medical University of Vienna; written informed consent was given). In a total of 140 EOC patients, both, preoperative blood samples and digital CT images of adequate quality were available. Clinical and histo-pathological data, laboratory parameters associated with cancer associated cachexia, as well as follow-up data was collected by experienced clinicians and documented prior to therapy. Laboratory analyses were performed as described below. Overall survival (OS) was defined as the time interval between diagnosis and tumor associated death and progression free survival (PFS) as the time between diagnosis and disease progression or death. Overall observation time was the time interval between diagnosis and last contact or date of death. Patients without recurrence, disease progression or non-cancer related death were censored at the time of last follow-up visit. The body mass index (BMI) was calculated as patient weight (kg) / height (m) 2 , as assessed within 30 days pre-operatively. The BMI was categorized as > 30.0 kg/m 2 , 25.0 to 29.9 kg/m 2 , 18.5 to 24.9 kg/m 2 , and <18.5 kg/m 2 , reflecting obese, overweight, normal weight and underweight patients, respectively.

Luminex assay description
Multiplexed Luminex-based assays are reliable to investigate the role of selected markers in cancer [16]. We measured circulating levels of cytokine and inflammation markers (IL-1β, IL- Blood samples were collected prior to treatment initiation and stored at -80°C. A multiplex panel was used according to the manufacturer`s protocol. The Luminex assay was analyzed using a Bio-Plex 200 array reader (Bio-rad). A quantitative determination of the respective analytes was achieved by comparing the raw data obtained from the patient samples with a standard curve. A total of six cytokines (IL-1β, IL-5, IL-7, IL-15, IL-17, and MIG) had to be excluded because of 25% values below detection limit.

CT Image Analysis
CT images were acquired using a Siemens Sensation 16 or a Siemens Sensation Cardiac 64 (both Siemens Healthcare, Germany) with 4mm or 3mm slice thickness and 120kVp tube voltage, respectively. CT images were analyzed according to a previously described protocol by using Slice-O-Matic software (v5.0, Tomovision, Montreal, Quebec, Canada) [3]. In CT images, each tissue has a specific radiation attenuation which is measured in absolute numbers in Hounsfield Units (HU) with a range of -1000 HU (air) to +1000 HU (dense bone), e.g. fat ranges from -190 to -30, muscle ranges from -29 to +150 [17]. Thus, the differentiation of fat and muscle is possible and also a quantification of fatty muscle infiltration via calculation of the mean MA. The lower the value of the MA, the higher the fatty infiltration of the muscle. The following muscles were determined and analyzed by two radiologists in consensus on two adjacent axial cross-sectional CT images at the L3 level as estimation from this cross-sectional area and total-body measurements are highly correlated (4): rectus abdominus, abdominal (lateral and oblique), psoas, and paraspinal (quadratus lumborum, erector spinae). Total crosssectional muscle area (cm 2 ) was measured within a range of -29 to +150 HU for the identified muscles on both slices as depicted in Fig 1. Quantified muscle area and calculated mean MA on both slices were averaged for each patient, respectively. The muscle area was normalized for height (meters squared; m 2 ) and reported as lumbar skeletal muscle index (SMI) (cm 2 /m 2 ).

Statistical analysis
Statistical analyses were performed using SPSS software version 19 (IBM Corporation, Armonk, New York, United States) and R (R: A Language and Environment for Statistical Computing, R Core Team, 2014, Vienna, Austria). Differences were analyzed using independent ttests for continuous, χ2-tests for categorical variables, and Fisher's exact test as appropriate. Luminex data were log-transformed to reach normal distribution. Univariate and multivariate analyses for overall survival (OS) and progression free survival (PFS) were conducted using the Cox proportional hazards model; hazard ratios (HRs) and corresponding 95% CIs are shown. To assess also the independent impact of factors not significant in the univariate Cox regression analyses, all factors were included in the multiple models according to suggestions from Harrell [18] and Sun et al. [19]. The optimal cut-off for the MA factor was obtained using the optimal non-linear association of MA with overall survival as determined by the Multivariable Fractional Polynomials method (mfp, R-package) [20]. Correlations and multivariate linear regression analyses with model selection maximizing the Akaike Information Criterion (AIC) were performed to identify independent relationships between BCMs and clinical pathological, laboratory parameters, as well as chemokine and cytokine concentrations. P values of less than 0.05 were considered statistically significant.

Patient characteristics
A total of 140 EOC patients undergoing surgical tumor debulking followed by platinum-based chemotherapy between 2004 and 2012 were included. Baseline clinic-pathological characteristics and BCM values are depicted in Table 1. Mean (standard deviation) age at time of diagnosis was 60 (13) years. A total of 103 tumors (73.6%) were of serous histology. In 72.1% (n = 101), tumors were poorly differentiated (high grade), followed by moderately and well differentiated tumors in 19.3% and 8.6% (n = 27 and n = 12, respectively). According to the dualistic model of EOC [21], patients were grouped into high grade serous (type II) versus low grade serous and non-serous tumors (type I; Table 1). Complete cytoreductive surgery with no evidence of residual tumor could be achieved in 54.3% (n = 76). Out of the 64 patients with macroscopically visual residual tumor, the tumor load was < 5mm in 15 patients (10.7%) and >2cm in 33 patients (23.6%). A total of 56 patients (40%) died of disease and 93 patients (66%) experienced tumor recurrence. Median (interquartile range) follow up time was 56 (33-73) months.

Body composition measurements
A wide variation of the BCMs could be observed in our study population. The mean BMI was 24.9 (4.8) kg/m 2 . At time of diagnosis, only 2.9% (n = 4) of patients were classified as underweight. On the other hand, a total of 38.6% (n = 54) was classified as obese or overweight. The mean lumbar total muscle cross-sectional area was 120.3 cm 2 (18.6). The mean SMI was 44.9 cm 2 /m 2 (7.4). For statistical analyses, the cut-off for high and low SMI was set at 41 cm 2 / m 2 , previously reported as significantly associated with survival in female cancer patients. This threshold was defined according to sex and BMI revealing the cutoff <41 (cm2/m2)-stable among all four BMI groups. [3]. Thecut-off for high and low MA was newly defined and set at 39 HU according to the relative risk of death related to the MA values (Fig 2). We decided to define the optimal MA threshold for this population, as a previously reported threshold differed according to the patient's BMI [3]. The majority of patients with low MA (57.1%) had residual tumor after cytoreduction, whereas optimal cytoreduction could be archieved in 60.4% of patients with high MA (p = 0.046; Table 2). Interestingly, a low BMI did not correspond to a low MA. In our cohort, not only normal weight, but also underweight patients showed a significantly higher MA compared to overweight and obese EOC patients (

Correlation of inflammatory parameters with MA
To understand the biological factors associated with low MA, basic laboratory parameters available in clinical routine and know to be associated with low nutrition status and tumor associated inflammation-namely albumin, total protein, leukocytes, and C-reactive protein (CRP)-were correlated with MA. The correlation was strongest for albumin, with a significantly positive correlation of higher albumin levels with a higher median MA (p = 0.013), Likewise, higher protein levels were related to higher MA, though not reaching statistical significance (p = 0.064). The other parameters showed no statistically significant correlation with MA (data not shown).

Correlation of cytokines with MA
Numerous cytokines seem to play a role in the metabolic changes associated with wasting and cancer cachexia. To determine the relationship between inflammatory processes and low MA, the serum levels of a set of 25 cytokines were evaluated, including the pro-cachectic cytokines TNFα, IL-6, IL-1 and IFN-y as well as anti-cachectic cytokines such as IL-10, IL-4, IL-13, and IL-1RA [22] and important markers previously analyzed in ovarian cancer [23]. Multivariate linear regression analysis revealed that Eotaxin and IL-10 were independent predictors of MA. IL-13, IL-2R, and IL-4 remained as correcting parameters in the final model. Results of the analyses are depicted in Table 4.

Discussion
The importance of cancer cachexia and the nutritional state in ovarian cancer prognosis has been recognized. This study demonstrates that MA, which seems to be associated with cancer cachexia, is an independent prognostic parameter easily ascertained by routine pre-operative CT in EOC patients. Trying to measure cancer related metabolic and inflammatory changes, an increased infiltration of skeletal muscle by inter-and intramuscular fat can be quantified by analysis of BCMs such as MA in CT scans [4,24]. Still, little is known about the prognostic relevance of BCMs determined by CT scan analysis, especially in EOC, although it is well known that malnutrition and muscle depletion affect surgical outcome and postoperative morbidity and mortality in cancer patients [25].
We found a significantly reduced overall survival in EOC patients with low MA-information of importance in the course of individualized treatment planning. A reduced SMI showed no significant impact on survival in EOC. Additionally, previous studies showed that the value of adipose tissue measurement was lacking clinical significance [26]. A significant association of low MA with reduced overall or progression-free survival has been described in patients with tumors of the gastrointestinal tract, the lung, in renal cell carcinoma, and in melanoma patients [3,26,27]. Despite significant improvements in treatment and understanding of metabolic and inflammatory pathways, recent studies reported that malnutrition still continues to be a significant challenge in ovarian cancer management [3,26,27].
One could hypothesize, that the BMI of EOC patients might correspond to the measured MA. However, in our patient population not only normal weight, but also underweight patients showed a significantly higher MA compared to overweight and obese EOC patients. This underlines, that BMI alone is a relatively inaccurate measurement of muscle composition and cancer cachexia in EOC. This is comparable to findings in patients with lung-or GI cancer [3]. Furthermore, we observed that MA was significantly lower in physically restricted patients, whereas the patients' BMI was not associated with the ECOG performance status.
MA was also significantly associated with the presence of residual tumor after cytoreduction. The presence of any post-operative residual tumor is a powerful factor with a tremendous impact on OS. [28]. Thus, CT scan analysis might help to identify those patients, in whom optimal cytoreduction is unlikely to be achieved.
In the present study we also investigated the biochemical processes corresponding to this condition.
We found that patients with low MA presented in a worse nutritional and systemic inflammatory status reflected among others by lower protein and albumin levels. Pre-operative serum  albumin is an accepted marker of poor nutrition in EOC [29]. In solid tumors, albumin seems to accumulate in the tumor microenvironment, and cancer cells seem to have the ability to utilize albumin as source of energy [30,31]. The rate of albumin synthesis is not only influenced by nutrition but also inflammation, reflected in the observation that albumin is a negative acute phase protein. Considering inflammation, we tried to understand the biochemical processes related to a low MA by measuringthe concentration of a multiplexed inflammatory marker panel of 25 cytokines, revealing Eotaxin-1 together with the anti-inflammatory cytokine IL-10 as significantly associated with MA.  Cancer is highly connected to the patient's immune response and inflammation plays an important role in muscle function. The imbalance between pro-and anti-tumor immunological profiles, as reflected by cytokines, increases with tumor progression. The potent antiinflammatory factor IL-10 induces systemic tumor-specific immunity and plays an important role in the control of tumor-promoting inflammation [32,33]. The IL-10 network is probably the most relevant link between cancer and inflammation [34]. In EOC, an impaired anti-tumor immune response seems to be associated with increased IL-10 levels [35], also reflected in higher levels in high grade compared to low grade tumors [36]. Similar to IL-10, Eotaxin-1 has been mainly studied in inflammatory disorders [37]. In EOC, cancer-dependent changes in Eotaxin-1 levels were demonstrated [38]. In gastric cancer, Eotaxin-1 levels were even proposed as biomarker for early diagnosis [39].
The measurement of MA on routine pre-operative CT is easy to perform, reproducible, ready available, and cheap. Aubrey et al. already mentioned a possible merit by including the quantification of attenuation in the repertoire of radiologists [17]. Our results showed impaired overall survival and unfavorable surgical outcome in patients with EOC and low MA in a multivariate survival analysis, highlighting the importance of this easy-accessible marker.
Limitations of the study include its retrospective design, resulting in a lack of information on important co-variates such as the amount of tumor load and the amount of ascites. Furthermore, research would benefit from standardized diagnostic criteria for low MA that could be used for individualized care-such as nutritional management-of patients with EOC.

Conclusion
This study demonstrates that without much effort, the measurement of MA in pre-operative CT scans of EOC patients could be utilized to become aware of the systemic muscle condition and the related predictive and prognostic implications that should be addressed in course of individual treatment planning.
Supporting Information S1