The MALOVA study ("MALignant OVArian cancer study") is a multidisciplinary Danish study covering epidemiology, lifestyle factors, biochemistry and molecular biology with the purpose of identifying risk factors and prognostic factors for OC. The design of the MALOVA study is described in details elsewhere [27]. Briefly, preoperative blood samples and tumor tissue samples were obtained from the majority of the 681 OC patients and from the 235 women with borderline ovarian tumor were included in the MALOVA study. FIGO stages were obtained from clinical records and reviewed by two gynecologists specialized in OC. Patients were either classified as radically operated with no macroscopic residual tumor or non-radically operated with macroscopic residual tumor after surgery. On the retrospectively collected paraffin embedded tissues, histologic grading were performed individually by two persons.

In Denmark all inhabitants have a unique personal (10-digit) identification number (CPR number), which is used universally in the Danish society. These identification numbers, which comprise information on date of birth and sex, are registered in the computerized Danish National Central Population Register. The register contains information on e.g. dates of death and emigration. All cases in this study were traced in the register and date of death, emigration or August 23^rd, 2007, whichever came first, were registered. The relevant hospital files were collected and scrutinized and information on treatment (surgery and chemotherapy), was established. At the end of follow-up, a total of 490 OC patients had died (median follow-up time: 23 months, range: 1–111), and 191 OC patients were still alive (median follow-up time: 91 months, range: 66–121).

The paraffin embedded tissue from the tumors was used for tissue array analyses. Forty-eight OC and 24 borderline ovarian tumors were excluded because of poor tissue quality or lack of tumor tissue in the collected blocks, and 132 OC and 16 borderline ovarian tumors were excluded due to unreadable YKL-40 expression results, lack of tumor cells in the selected TA cylinders, loss of tissue during the staining procedure or folding/tearing of sections by the microtome. Finally, 42 non-epithelial tumors (28 OC and 14 borderline ovarian tumors) were excluded because epithelial and non-epithelial ovarian tumors differ in embryologic and pathologic characteristics. Together, these exclusions left 473 OC and 181 borderline ovarian tumors, which were suitable for the YKL-40 tissue expression and survival analyses.

For the tissue array production, one to two representative areas were marked on a freshly cut H&E stained section from each selected block. Tissue cylinders with a diameter of 2 mm were punched from corresponding areas in the donor tumor block and brought into 2–4 individual recipient paraffin blocks (tissue array blocks), using a custom-made manual Tissue Array Instrument (Beecher Instruments, Silver Spring, MD, USA). Once the cores were laid into the recipient block, the paraffin was slightly melted in order to bind the cores into the block. This treatment secured the cores during sectioning. Melting was achieved by placing the blocks in an oven at 37°C for 10 min.

A total of 2 to 4 tissue arrays consisting of corresponding viable and representative tissues from each ovarian tumor were constructed in order to minimize the risk of intra-tumor variability. The tissue arrays were sorted with respect to histological subtype and FIGO stage. Four cores of control tissue (2 from kidney, 2 from liver) were placed strategically in each tissue array to ensure unique orientation of every tissue array together with a maximum of 31 different patient tumor samples.

Two μm sections (Section Transfer System, STS, Ergostar HM200, MICROM International GmbH, Walldorf, Germany) from the tissue array blocks were transferred to glass slides (DAKO Chem-Mate Capillary Cap Microscope Slides, 57 mm, DAKO A/S, Glostrup, Denmark). Slides were stored at 4°C for a maximum of 8 days until staining for YKL-40.

Prior to staining, the sections were deparaffinized in xylene and rehydrated in graded dilutions of alcohol. In order to demask antigens within the tissue, sections were pretreated in TEG buffer, pH 9.0 (Tris 10 mmol/l and EGTA 0.5 mmol/l), followed by heating in a microwave oven to 98°C for 15 min. Sections were then left to cool in the buffer at room temperature for an additional 15 min. To block endogen peroxidase activity the sections were treated with 0.03% w/v hydrogen peroxide for 5 min. To avoid background staining the sections were then pre-incubated with 5% w/v purified bovine serum albumin (BSA) (Dade Behring, Liederbach, Germany) diluted in TRIS buffer pH 7.6 for 10 min. The primary specific mouse monoclonal antibody against human YKL-40 (201.F9, 3.8 mg/ml, isotype IgG2b, κ, epitope GAWRGTTGHHS, corresponding to the amino acids 210–220) was diluted 1:100 (38 ng/ml) in 1% w/v purified BSA (Dade Behring) in TRIS buffer pH 7.6 and incubated for 60 min. The secondary antibody EnVisionTM+System-HRP (DAB) (Dako, Glostrup, Denmark, Code K 400711-2, lot no. 104.225) was used for 30 min. The colour was developed with DAB added with chromogen for 10 min. Sections were rinsed in water, counterstained in Mayers Hematoxylin for 3 min and finally dehydrated, mounted and coverslipped with Pertex. Washings with 5 mM Tris buffer, pH 7.6 with NaCl 0.9% w/v and Tween 0.1% v/v (TBS) were used between all steps in the procedure. After heating all steps were performed at room temperature in a humidity chamber to avoid air-drying of the sections. The staining process was done manually.

Two observers, both experienced in evaluating immunohistochemical stained tissues, simultaneously assessed the patterns of YKL-40 protein expression of each tumor sample. The observers had no knowledge of clinical parameters and study endpoint. Standardization of scoring was achieved by comparison of the scores, and any discrepancies were resolved by consensus. Scoring for YKL-40 protein expression was based on the proportion of cells in a given tumor specimen exhibiting distinct cytoplasmic immunopositivity as well as the intensity of staining (percentage scale: 0, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, and 100). Secondly, the YKL-40 scoring results were transformed into two scales: a two-tiered scale (1: negative; 2: ≥ 5% positive tumor cells) and a four-tiered scale (1: negative; 2: 5% positive tumor cells; 3: 10% positive tumor cells; 4: ≥ 20% positive tumor cells). The four tissue control cores of kidney and liver, respectively, showed consistent staining results. For the general description and the prognostic evaluation both scales were used.

Blood samples were obtained no earlier than two weeks prior to surgery from the 95 patients operated at Herlev Hospital and Rigshospitalet, University of Copenhagen. EDTA-blood samples were left on the blood cells at room temperature for less than 8 hours. Following the samples were separated by centrifugation at 2000 g for 10 min. at room temperature and plasma EDTA samples stored in aliquots at -80°C until analysis. EDTA plasma samples collected from patients included in the MALOVA study, who had been treated at other hospitals were not used for plasma YKL-40 measurements, since the plasma obtained from these centers had not been separated from the blood cells within 8 hours after sampling. It is known that EDTA plasma YKL-40 is only stable when samples are processed within 8 hours [28].

Plasma levels of YKL-40 were determined in duplicates by a commercial two-site, sandwich-type enzyme-linked immunosorbent assay (Quidel, Santa Clara, CA, USA) using streptavidin coated microplate wells, a biotinylated-Fab monoclonal capture antibody, and an alkaline phosphatase-labeled polyclonal detection antibody. The detection limit was 20 μg/l. The intra-assay coefficient of variation (CV) was ≤ 5.0% and inter-assay CVs ≤ 10.2%. The samples were analyzed blinded to clinical parameters and study endpoint.

The reference interval for plasma YKL-40 was determined in 144 healthy women (median age 51 years, range 18–79 years) characterized by not being on medication and having no signs of pre-existing disorders such as joint, liver, metabolic or endocrine disease or malignancy.

Written informed consent was obtained from all patients. The study has been approved by the scientific ethical committee in the study area (KF01-384/95).

Statistical comparisons between groups were carried out using Chi-square or rank sum tests. Association between YKL-40 tissue expression levels and plasma YKL-40 levels was assessed using the Spearman correlation. The clinical endpoint in the OC patients was survival determined as the time from baseline blood sample before operation to time of death updated at August 23^rd, 2007. Patients who died from non-related OC were censored in the survival analyses at the date of death. Cases in which patients were alive by this date were censored. Survival probabilities were estimated by the Kaplan-Meier method and tests for differences between strata were done using the log-rank statistic. Graphical presentation using Kaplan-Meier estimates of survival were shown grouping the patients by YKL-40 expression in the tumor tissue dichotomized at 5%. Graphical presentation using Kaplan-Meier estimates of survival were shown grouping the patients by plasma YKL-40 levels dichotomized as normal or elevated compared to age-matched controls. In the univariate survival analyses the YKL-40 expression scale in tumor tissue was dichotomized as < 5% vs. ≥ 5% and the plasma YKL-40 levels by the actual value on the log scale (natural) (log transformed). The multivariate Cox regression multivariate analysis was based on FIGO stage, residual tumor after primary surgery, serum CA-125, age and plasma YKL-40 levels by the actual value on the log scale (natural) (log transformed). Confidence intervals (95% CI) were based on Wald's test statistic for the corresponding parameters in the Cox regression model, i.e. on the log-scale for the hazard ratios (HR). Model assessment was done using graphical methods, Schoenfeld and martingale residuals. P-values less than 5% were considered significant. All calculations were performed using SAS (version 9.1, SAS Institute, Cary, NC, USA).

Tissue arrays for YKL-40 immunohistochemical analysis were available from 181 patients with borderline ovarian tumors (median age 54 years, range 33 – 80) and from 473 patients with OC (median age 59 years, range 26 – 80).

The distribution of the YKL-40 score in the borderline ovarian tumors was: level 1 (negative): 44 (24%); level 2: 74 (41%); level 3: 35 (19%) and level 4: 28 (16%). The distribution of the YKL-40 score in the OC samples was: level 1 (negative): 112 (24%); level 2: 101 (21%); level 3: 138 (29%) and level 4: 122 (26%). Figure 1 illustrates representative examples of the different levels of YKL-40 staining in borderline ovarian tumors and OC. The cancer cells showed a diffuse granular staining of the cytoplasm. Staining of membranes and nuclei were not observed. Borderline ovarian tumors appeared with the same staining pattern as the carcinomas, but with reduced staining intensity (Figure 1, b–d). YKL-40 protein expression was found in all the different histological types of OC (Figure 1). The YKL-40 protein was also expressed in inflammatory cells such as macrophages, mast cells and neutrophils. These results have been confirmed by double-labeling methods verifying co-expression of YKL-40 and tryptase (mast cells) and CD68 (macrophages), respectively (data not shown).

The association between clinico-pathological variables and YKL-40 score is shown in Table 1 and Table 2. The YKL-40 protein expression (staining percentage scale) in OC was associated with FIGO stage (p < 0.0001) and histological type of tumor (p < 0.0001), and the YKL-40 expression (staining percentage scale) was higher in the OC than in the borderline tumors (p = 0.0019). Also a tendency towards a correlation between reduced YKL-40 protein expression and radicality of surgery was observed (p = 0.05).

The highest frequency of any level of positivity of YKL-40 protein expression was observed in serous adenocarcinomas, where 123 out of 249 tissues were scored positive (49%).

Plasma samples for YKL-40 analysis were available from 19 patients with borderline ovarian tumors (median age 45 years, range 38 – 76 years) and from 76 patients with OC (median age 63 years, range 37 – 79 years). The median preoperative plasma level of YKL-40 in the OC patients was 125 μg/l (range 20 – 2650 μg/l) and significantly higher than the level in 144 healthy women (median 32 μg/l, range 20 – 143 μg/l; p < 0.001). 67% (51/76) of the OC patients had a plasma YKL-40 level above the age-adjusted 95^th percentile of the healthy women. Patients with borderline ovarian tumors did not have elevated plasma YKL-40 (median 42 μg/l, range 20 – 180 μg/l) compared to healthy women. 21% (4/19) of the patients with borderline tumors had a plasma YKL-40 level above the age-adjusted 95^th percentile of the healthy women. Table 1 and Table 2 gives the relationship between preoperative plasma YKL-40 and standard clinical-pathological variables. Plasma YKL-40 increased with increasing FIGO stage among OC patients (p = 0.0001). Furthermore, plasma YKL-40 was correlated with serum CA-125 (Spearmans rho = 0.47, p < 0.0001) and age (rho = 0.59, p < 0.0001). No correlation was found between plasma YKL-40 and the YKL-40 tissue expression percentage score (rho = 0.032, p = 0.081).

Univariate survival analysis demonstrated that YKL-40 tissue expression (negative or ≥ 5% positive tumor cells) was not associated with survival (HR = 1.2; 95% CI: 0.95 – 1.53) (Figure 2). Similarly, using the four-tiered scale YKL-40 tissue expression was confirmed of no evidence of separation (data not shown). Including YKL-40 tissue expression (negative or ≥ 5% positive tumor cells) in a multivariate Cox regression analysis with FIGO stage, patient age, radicality of surgery and serum CA-125 as other parameters, showed no prognostic impact of YKL-40 tissue expression (HR = 0.97; 95% CI: 0.99 – 1.01) (data not shown).

During follow-up 55 (72%) of the 76 OC patients died. Univariate Cox analysis of plasma YKL-40 (log transformed and treated as a continuous covariate) showed significant association with overall survival (HR = 2.85, 95% CI: 2.04 – 3.97, p < 0.0001). Univariate Cox analysis of plasma YKL-40 (dichotomized as normal vs. high age-corrected plasma YKL-40 level) showed also significant association with overall survival (HR = 3.51, 95% CI: 1.82 – 6.75, p < 0.0001). Figure 3 illustrates the Kaplan-Meier estimates of survival stratified by baseline plasma YKL-40 dichotomized as normal vs. high age-corrected plasma YKL-40 level.

Performing a multivariate Cox analysis including plasma YKL-40 (log transformed and treated as a continuous covariate), serum CA-125 (log transformed and treated as a continuous covariate), FIGO stage (I-IV), residual tumor after surgery (yes or no), and age showed that only plasma YKL-40 (HR = 2.13, 95% CI: 1.40 – 3.25, p = 0.0004) was found of independent prognostic value of overall survival. Age at surgery (HR = 1.02, 95% CI: 0.99 – 1.06), radicality after primary surgery (HR = 1.20, 95% CI: 0.44 – 3.33), serum CA-125 (HR = 1.08, 95% CI: 0.94 – 1.23) and FIGO stage (p = 0.63, FIGO 1 vs 4: HR = 0.61, 95% CI: 0.11 – 3.34, FIGO 2 vs 4: HR = 0.25, 95% CI: 0.03–2.37, FIGO 3 vs 4: HR = 0.66, 95% CI: 0.28–1.58) were found of no independent prognostic value in OC patients.