Hot-spot microvessel density (MVD) and vascular patterns have been reported as histopathologic factors that influence prognosis in retrospective series of malignant gliomas. To investigate clinical utility, the authors systematically studied observer agreement on MVD and vascular patterns and the influence of repeatedly assessed data on patient outcomes in 2 independent glioblastoma series.

METHODS

MVD and vascular patterns were assessed retrospectively by 5 observers in 1) a retrospectively compiled glioblastoma series that included 110 patients and 2) a glioblastoma series that included 233 patients who were treated within a randomized trial. MVD was determined in the field of greatest density (“hot-spot”). Predominantly classic or bizarre vascular patterns were determined by using a previously defined algorithm.

RESULTS

Observer agreement on MVD was highly variable (range of κ values, 0.464-0.901). The worst observer agreement was achieved when both the selection of hot-spots and MVD counts were performed independently. Survival analysis did not show a consistent association between repeatedly assessed MVD and patient outcome. Observer agreement on vascular patterns was poor (κ = 0.297). Survival analysis did not show a consistent association between repeatedly assessed vascular patterns and patient outcome.

CONCLUSIONS

Observer agreement on hot-spot MVD and vascular patterns in patients with glioblastoma was poor in independent assessments. MVD and vascular patterns were not associated consistently with patient outcome. Based on these findings, the authors concluded that poor observer agreement limits the clinical utility of histopathologically assessed hot-spot MVD and vascular patterns as prognostic factors in patients with glioblastoma. Improved methodologies for morphologic assessment of glioblastoma vascularization need to be identified. Cancer 2006. © 2006 American Cancer Society.

Glioblastoma is the most common malignant type of primary brain tumor. Recently, it was shown that radiochemotherapy followed by adjuvant chemotherapy with temozolomide prolonged the survival of 45% to 50% of patients with glioblastoma who had tumor tissues that showed methylation of the O-6-methylguanine DNA methyltransferase (MGMT) promoter.1-3 Therefore, the view of an irremediable, dismal prognosis for patients with glioblastoma multiforme finally has been broken open.

The identification and assessment of prognostic and predictive factors in glioblastoma is important with regard to patient stratification in clinical trials4 and patient management in the routine clinical setting. Recently, hot-spot microvessel density (MVD) and vascular patterns have been described as histopathologic factors that influence prognosis in retrospective series of patients with malignant gliomas.5, 6 However, the clinical utility of MVD and vascular patterns has not been studied systematically to date. Clinical utility requires agreement7-10 on these parameters in independent assessments performed by different observers in independent patient cohorts. To this end, we systematically studied observer agreement on MVD and vascular patterns and the influence of repeatedly assessed data on patient outcomes.

MATERIALS AND METHODS

Patients

Glioblastoma Series 1

One hundred ten patients who underwent initial surgery for primary glioblastoma between 1995 and 1999 at the University Hospital of Vienna were included in this retrospective series. Prior to this surgery, none of the patients had received chemotherapy, radiotherapy, or surgery for their glioblastoma. All patients belonged to a previously published cohort of 114 patients with glioblastoma.6 Four of those 114 patients were not included in the current study, because the amount of tissue available was considered too small for the analysis of vascular parameters at reassessment. For evaluation of MVD, 13 of 110 patients were excluded, because the area of previous MVD assessment was not documented. The series does not include recurrent tumors or patients with a history of a previous, low-grade astrocytoma. The median follow-up was 1106 days (range, 5-1509 days).

Brief summary of postoperatively applied adjuvant therapy

Radiotherapy was administered to a total dose of 66 grays (Gy) (2 Gy per fraction) or 51 Gy (3 Gy per fraction) within 5 to 6 weeks after surgery.6 Nitrosurea-based chemotherapy was started 10 to 14 days after surgery. Patients age >50 years and those who had a Karnofsky performance status ≥60 received 1-(2-chloroethyl)-3-cyclohexyl-1-nitrosourea orally at a dose of 100 mg/m² in 6-week to 8-week intervals or received 1-(4-amino-2-methyl-5-pyrimidinyl)-methyl-3-(2-chloroethyl)-3-nitrosourea intravenously at a dose of 100 mg/m² in 6-week intervals. Patients age <50 years and those who had a Karnofsky performance status ≥60 received a combination of dacarbacine (200 mg/m²) and fotemustine (100 mg/m²) every 3 weeks intravenously.

For patients who were recruited retrospectively, no informed consent could be obtained. However, the local ethical committee gave approval for the study.

Glioblastoma Series 2

In total, 233 biopsy specimens from patients with glioblastoma were available for analysis. All patients in this study group belong to a previously published glioblastoma cohort of 573 patients who were recruited prospectively in a collaborative, multicenter approach by the European Organization for Research and Treatment of Cancer and the National Cancer Institute of Canada.1, 3 To exclude a selection bias, we analyzed our study cohort of 233 patients as follows: In 130 of 233 patients, methylation status of the MGMT promoter could be determined. Of those 130 patients, 62 patients (47.7%) had detectable methylation of the MGMT promoter, and 68 patients (52.3%) had no detectable methylation. Survival analysis of the 233 patients showed that MGMT promoter methylation was an independent, favorable prognostic factor irrespective of treatment (P<.0001; log-rank test; 95% confidence interval [95% CI], 0.207-0.479; hazard ratio [HR], 0.315). Among patients who had tumors that contained a methylated MGMT promoter (n = 62 patients), a survival benefit was observed in patients who received temozolomide plus radiotherapy (P = .0147, log-rank test; 95% CI, 0.26-0.88; HR, 0.475). In the absence of methylation of the MGMT promoter, there was no statistically significant difference in survival between the treatment groups (P = .5063, log-rank test; 95% CI, 0.510-1.395; HR, 0.843). These findings are congruent with the observations in the complete cohort of 573 patients.1, 3 Thus, there was no evidence for a selection bias. None of the patients had received preoperative chemotherapy, radiotherapy, or surgery for their glioblastoma.

Brief summary of postoperatively applied adjuvant therapy

Patients were enrolled in a randomized trial of chemoradiotherapy (temozolomide plus radiotherapy) versus radiotherapy alone.1 Temozolomide was administered at a dose of 75 mg/m² body surface area daily during standard fractionated radiotherapy (60 Gy) for 6 to 7 weeks and at a dose of 150 to 200 mg/m² daily for 5 days of every 28-day cycle after radiotherapy for up to 6 cycles. In patients who had tumor progression, salvage or second-line therapy was administered at the investigators' discretion. The median follow-up was 825 days (range, 11-1197 days).

Methylation-Specific Polymerase Chain Reaction

Methylation status of the MGMT promoter was determined by using bisulfite modification and methylation-specific polymerase chain reaction (MSP).3 Informed consent was obtained from all patients who were recruited prospectively.

Immunohistochemistry

Sections were cut at a thickness of 4 μm from routinely formalin fixed and paraffin embedded tumor tissue blocks. Antigen retrieval was performed by boiling sections in citrate buffer (pH 6.0). For assessment of MVD and vascular patterns, endothelial cells were visualized by anti-CD34 immunostaining (clone QBEnd/10; Novocastra Laboratories Ltd., Newcastle-Upon-Tyne, U.K.) (dilution, 1:10; incubation for 30 minutes). Detection of immunostaining was performed by using the ChemMate® kit (DAKO, Glostrup, Denmark), and diaminobenzidine was used as the chromogen.

Analysis

MVD

Assessment of MVD comprised 2 steps. First, anti-CD34-immunostained tissue sections were scanned at low magnification, and the tumor area with the highest density of distinctly highlighted microvessels (“hot-spot”) was selected. Selected areas were free of surgical wound changes (e.g., bleedings, traumatic damage, cauterization damage). In a second step, all immunolabeled vessels were counted manually at 200-fold magnification within an examination area of 0.25 mm² using an eye grid. Each stained lumen was regarded as a single, countable microvessel. If no lumen but only a single CD34-positive cell was visible, then that cell also was interpreted as a single microvessel. MVD was assessed by multiple observers in various constellations with regard to field selection and counting (Table 1). The hot-spot method is a common approach for MVD assessment in solid tumors11-13 and also has been used in the previous study of MVD in glioblastoma by Birner et al.6

Table 1. Assessment of Hot-Spot Microvessel Density in Glioblastoma Series 1 (n = 97) and Glioblastoma Series 2 (n = 233)*

MVD Assessment	Area Defined By	MVD Counted By	Range	Median	P†	HR	95%CI
GS-1: Overall (n = 97)
MVD-1	IF	IF	10–407	79	.0060	0.527	0.331–0.839
MVD-1				84	.0060	0.527	0.331–0.839
MVD-2	EG and MP	MP	4–254	54	.1434	0.711	0.449–1.126
MVD-2				79	.1098	0.662	0.398–1.102
MVD-2				84	.1464	0.684	0.407–1.147
MVD-3	EG and MP	EG	4–328	56	.1849	0.733	0.461–1.164
MVD-3				79	.3713	0.799	0.488–1.309
MVD-3				84	.2634	0.753	0.457–1.241
MVD-4	IF	EG	15–329	85	.4559	0.842	0.535–1.325
MVD-4				79	.1770	0.731	0.462–1.155
MVD-4				84	.1529	0.721	0.456–1.140
GS-2: Overall (n = 233)
MVD-5	EG and MP	EG	0–236	38	.4532	1.113	0.841–1.474
MVD-5				79	.0818	0.705	0.475–1.048
MVD-5				84	.1753	0.762	0.513–1.131
GS-2: Methylated (n = 62)
MVD-5a	EG and MP	EG	7–204	34.5	.9235	1.030	0.562–1.886
MVD-5a				79	.2608	0.642	0.294–1.402
MVD-5a				84	.5916	0.811	0.376–1.751
GS-2: Unmethylated (n = 68)
MVD-5b	EG and MP	EG	6–148	43	.0033	2.139	1.275–3.587
MVD-5b				79	.733	1.120	0.584–2.149
MVD-5b				84	.733	1.120	0.584–2.149
GS-2: Methylated, RTX (n = 29)
MVD-5c	EG and MP	EG	13–145	34	.5189	1.282	0.581–2.829
MVD-5c				79	.2911	0.580	0.192–1.722
MVD-5c				84	.2911	0.580	0.195–1.722
GS-2: Methylated, RTX and TMZ (n = 33)
MVD-5d	EG and MP	EG	7-204	35	.8266	1.109	0.439–2.799
MVD-5d				79	.5951	0.736	0.237–2.286
MVD-5d				84	.9217	1.057	0.349–3.207
GS-2: Unmethylated, RTX (n = 37)
MVD-5e	EG and MP	EG	6–148	46	.0201	2.375	1.120–5.039
MVD-5e				79	.5016	1.362	0.550–3.373
MVD-5e				84	.5016	1.362	0.550–3.373
GS-2: Unmethylated, RTX and TMZ (n = 31)
MVD-5f	EG and MP	EG	6–144	37	.0715	2.006	0.929–4.331
MVD-5f				79	.9043	1.061	0.401–2.808
MVD-5f				84	.9043	1.061	0.401–2.808
GS-2: Unmethylated (n = 68)
MVD-6	EG and MP	MP	11–157	40.5	.4657	1.199	0.735–1.957
MVD-6				79	.6432	1.161	0.616–2.189
MVD-6				84	.6432	1.161	0.616–2.189

MVD indicates microvessel density; HR, hazards ratio; 95% CI, 95% confidence interval; GS, glioblastoma series; IF, Inge Fischer; EG, Ellen Gelpi; MP, Matthias Preusser; RTX, radiotherapy; TMZ, temozolomide.
* Definitions of area and counting of MVD were performed by 3 observers in various constellations. For overall survival analyses, multiple cutoff points were used: the median in the subcohort (first P value, HR, and 95% CI); the median of assessment MVD-1 (second value); and the median value according to Birner et al., 20036 (third value).
† Log-rank test.

Vascular patterns

Vascular patterns were assessed on anti-CD34-immunostained sections according to a previously published algorithm.14 This algorithm identifies glioblastomas with either a predominantly capillary-like, classic vascular pattern or with a predominantly bizarre vascular pattern (predominance of glomeruloid vascular proliferations, garland-like vascular proliferations, vascular clusters). Two observers assessed vascular patterns together on a multiheaded microscope. In Glioblastoma Series 1, assessment was performed by 2 observer pairs. In Glioblastoma Series 2, only 1 observer pair assessed vascular patterns.

Statistical Analysis

Continuous data were summarized as median and range, and categorical data were described by counts and frequencies. For continuous variables, a Spearman rank-correlation coefficient was used to measure observer agreement: Continuous variables were dichotomized at their respective median values, and the Cohen κ statistic was used to measure observer agreement (κ values <0.5 signified poor observer agreement). For κ values, 95% CIs also were calculated.

Overall survival was measured from the day of initial surgery in Glioblastoma Series 1 and from the day of randomization in Glioblastoma Series 2 until the death of the patient. Survival until the end of the observation period was considered a censoring event. Survival probabilities were computed according to the Kaplan–Meier method. Log-rank tests and Cox proportional hazards regression models were used to assess the prognostic and predictive effect of covariates on overall survival. HRs and corresponding Wald-type 95% CIs calculated given. Note that, in small-scale studies, the results of the log-rank test (which is a score test) and the Wald 95% CIs may differ slightly, which happened in the current analyses.

No correction for multiple testing was done, because the statistical tests are performed for demonstration purposes only. A 2-tailed significance level of 5% was assumed. The statistical software packages SAS (SAS Institute Inc., Cary, NC) and SPSS (SPSS Inc., Chicago, IL) were used for statistical calculations.

RESULTS

Assessment of MVD in Glioblastoma Series 1

MVD was assessed 4 times in the overall Glioblastoma Series 1 by 3 observers in various constellations with regard to field selection and counting (Table 1). Among 4 MVD assessments (MVD-1 through MVD-4) (Table 1), there was considerable variation of the range and median MVD.

Spearman correlation coefficients and Cohen κ values are shown in Table 2 and in Figure 1. Observer agreement was variable (κ range, 0.464-0.901). The highest observer agreement was achieved when different observers counted MVD in the same area (MVD-2 and MVD-3) (Tables 1 and 2) (Fig. 1C). Observer agreement was lowest when the area of assessment was defined independently by different observers (MVD-2 and MVD-4) (Tables 1 and 2).

Table 2. Observer Agreement on Hot-Spot Microvessel Density in Glioblastoma Series 1*

MVD Assessment	MVD-1	MVD-2	MVD-3	MVD-4	MVD-6
MVD-1
κ (95% CI)	1	0.485 (0.31–0.66)	0.505 (0.34–0.67)	0.670 (0.52–0.82)
Coefficient†	1	0.411	0.552	0.907
MVD-2
κ (95% CI)		1	0.901 (0.82–0.99)	0.464 (0.29–0.64)
Coefficient†		1	0.927	0.541
MVD-3
κ (95% CI)			1	0.485 (0.31–0.66)
Coefficient†			1	0.600
MVD-4
κ (95% CI)				1
Coefficient†				1
MVD-5b
κ (95% CI)					0.647 (0.47–0.83)
Coefficient†					0.794

MVD indicates microvessel density; 95% CI, 95% confidence interval.
* Definitions of area and counting of MVD were performed by 3 observers in various constellations.
† Spearman correlation coefficient.

Details are in the caption following the image — **Figure 1**
Open in figure viewer PowerPoint

These scatter plots illustrate observer agreement, and Kaplan–Meier curves show the influence of hot-spot microvessel density (MVD) on patient outcome. Glioblastoma Series 1 (A-F): (A) Correlation of MVD data sets MVD-1 through MVD-3 (see Table 1): When selection of the microscopic area and MVD counting were performed by different observers independently, there was only poor congruency of MVD values. (B) Correlation of MVD data sets MVD-3 and MVD-4 (see Table 1): When the microscopic field for MVD assessment was defined by different observers independently, and MVD counting was performed by 1 observer, the congruency of MVD values remained poor. (C) Correlation of MVD data sets MVD-2 and MVD-3 (see Table 1): When the microscopic field was defined by 2 observers on a multiheaded microscope, and independent counting was performed by both observers, the congruency of MVD values was greater. (D) In 1 count, the MVD showed a significant influence on patient outcome (MVD-1; see Table 1). (E) When the MVD was recounted in the same microscopic field by an independent observer (MVD-4; see Table 1), there was good congruency of MVD values (MVD-1 through MVD-4; see Table 1). (F) However, reassessed MVD values (MVD-4; see Table 1) did not show a statistically significant association with patient outcome. Glioblastoma Series 2 (G-I): (G) In the subcohort of glioblastomas with unmethylated O-6-methylguanine DNA methyltransferase promoter levels, low MVD was associated with favorable patient outcomes (MVD-5b; see Table 1). (H) Reassessed MVD (MVD-6; see Table 1) values showed moderate congruency with the original data set (MVD-5b and MVD-6; see Table 1). (I) There was no statistical influence of reassessed MVD values (MVD-6; see Table 1) on patient outcome.

Data set MVD-1 (the data set from the study published by Birner et al.6) showed a significant association of high MVD with favorable patient outcomes. In 3 reassessments (MVD-2 through MVD-4) of the original MVD evaluation (MVD-1), high MVD (cut off at the median) was not associated with a statistically significant, favorable patient outcome (Table 1) (Fig. 1).

Assessment of MVD in Glioblastoma Series 2

In Glioblastoma Series 2, the area of MVD counting was determined by 2 observers (M.P. and E.G.) on a multiheaded microscope (area of highest MVD), and counting was performed by 1 of those 2 observers. Generally, the MVD range was smaller, and the median values were lower than the values in Glioblastoma Series 1. Other than pure chance, a plausible explanation for this discrepancy may be the application of selection criteria for patient recruitment in the prospective series. Another explanation may be that anti-CD34 immunostaining was performed immediately after the tissue sections were cut in Glioblastoma Series 1; whereas, in Glioblastoma Series 2, there was a delay of up to 2 years. However, in general, the observers found that the quality of CD34 immunostaining was good.

In the overall cohort, MVD did not correlate with patient outcome (Table 1). In a second step, the influence of MVD on patient outcome was analyzed in subcohorts of this series defined according to MGMT promoter methylation status and adjuvant therapy arm (Table 1). The entire subcohort of patients who had glioblastomas with unmethylated MGMT promoter (MVD-5b) and its subfraction of patients who received radiotherapy only (MVD-5e) showed significantly more favorable outcomes among those patients who had low MVD (Table 1) (Fig. 1G). These results contradict the finding of a favorable influence of high MVD in Glioblastoma Series 1. To test stability of this finding, the area of counting was redetermined by the same 2 observers, and 1 of those 2 observers (M.P.) recounted MVD. The congruency of MVD values determined in these 2 assessments was rather low (κ = 0.647) (Table 2) (Fig. 1H), and the survival analysis showed no significant association between reassessed MVD values and patient outcomes (Table 1) (Fig. 1I).

Assessment of Vascular Patterns in Glioblastoma Series 1

Overall, in Glioblastoma Series 1 (n = 110 patients), observer agreement on vascular patterns was poor (κ = 0.297) (Table 3). There was a significant association of classic vascular pattern with favorable patient outcome in 1 assessment (VP-1) (Table 3) (previously published data by Birner et al.6), whereas an additional assessment (VP-2) did not show any correlation with patient outcome in univariate (Table 3) (Fig. 2A and 2B) or multiple survival analyses (P = .477; HR, 0.840; 95% CI, 0.520-1.358) after adjusting for patient age (<60 years vs. >60 years), Karnofsky performance status (<80 vs. >80), and extent of resection. In the subfraction of patients for whom the 2 assessments yielded concordant results (73 of 110 patients) (Table 3), no influence of vascular patterns on patient outcome was evident (Fig. 2C).

Table 3. Assessment of Vascular Patterns in Glioblastoma Series 1 (n = 110) and Glioblastoma Series 2 (n = 233)

Cohort	VP Assessment	Assessed By*	C:B Ratio	P†	HR	95% CI‡	κ (95% CI)
GS-1: Overall (n = 110)	VP-1	JH and PB	0.36	.0488§	0.597	0.355–1.033
GS-1: Overall (n = 110)	VP-2	MP and EG	0.83	.9405	1.016	0.664–1.556	0.297 (0.13–0.46)
GS-2: Overall (n = 233)	VP-3	MP and EG	0.49	.7593	1.407	0.779–1.408	—
GS-2: Methylated (n = 61)	VP-3a	MP and EG	0.55	.3347	1.359	0.726–2.544	—
GS-2: Unmethylated (n = 68)	VP-3b	MP and EG	0.56	.4509	0.841	0.508–1.394	—
GS-2: Methylated, RTX (n = 29)	VP-3c	MP and EG	0.45	.4048	1.463	0.593–3.610	—
GS-2: Methylated, RTX and TMZ (n = 33)	VP-3d	MP and EG	0.73	.9632	1.022	0.409–2.553	—
GS-2: Unmethylated, RTX (n = 37)	VP-3e	MP and EG	0.48	.5831	1.230	0.585–2.586	—
GS-2: Unmethylated, RTX and TMZ (n = 31)	VP-3f	MP and EG	0.63	.3523	0.700	0.329–1.490	—

VP indicates vascular pattern; C:B Ratio, ratio of classic VP to bizarre VP; HR, hazards ratio; 95% CI, 95% confidence interval; GS, glioblastoma series; JH, Johannes A. Hainfellner; PB, Peter Birner; MP, Matthias Preusser; RTX, radiotherapy; TMZ, temozolomide;
* Assessments were performed by 2 observer pairs.
† Log-rank test.
‡ Wald-type confidence intervals.
§ In the report by Birner et al., 2003,6 the log-rank P value was mistyped as .0048.

Assessment of Vascular Patterns in Glioblastoma Series 2

In Glioblastoma Series 2, vascular patterns were assessed only by 1 observer pair, because poor observer agreement already was documented in Glioblastoma Series 1. Vascular patterns did not show a statistically significant influence on patient outcome in the overall cohort in univariate or multivariate survival analyses (Table 3, Fig. 2D). Furthermore, no significant association with patient outcome was observed in the subcohorts of the series defined by MGMT promoter methylation status and treatment arm (Table 3).

DISCUSSION

A significant correlation of hot-spot MVD and vascular patterns with patient outcome was reported in previous retrospective studies of malignant gliomas.5, 6 To translate these factors into clinical application, agreement of different observers on these factors and consistency of prognostic impact is required.15 Furthermore, MVD and vascular patterns need to be validated.

Observer Agreement

We systematically studied observer agreement on MVD and vascular patterns. The main finding of this investigation was that observer agreement is variable. Observer agreement was particularly poor in independent assessments. In breast cancer, interobserver variation of MVD assessment was tested in a manner similar to that used in the current study.16, 17 Those breast cancer studies yielded results similar to our findings in glioblastoma: Observer agreement was poor in general, and interobserver variation was greater when the observers had to identify the hot-spots independently. These data indicate that lack of observer agreement in the assessment of hot-spot MVD may be a problem irrespective of the particular tumor type.

Consistency of Prognostic Influence

We studied the consistency of the prognostic influence of repeatedly assessed MVD and vascular patterns in 2 independent glioblastoma series in overall cohorts and in patient subgroups. MVD and vascular patterns were not associated consistently with patient outcomes in our series. These findings also indicated the limitations of the clinical utility of MVD and vascular patterns as prognostic factors. However, some words of caution must be added with regard to the analysis of prognostic influence in our study. In the instance of a statistically nonsignificant finding, the statistical power may be questioned. In Glioblastoma Series 1, the statistical power was rather low because of the small sample size (97 patients). However, in Glioblastoma Series 2, which included 233 observations, moderate HRs from 1.7 to 1.8 would be detectable with sufficient power of ≈ 85%. Another issue of caution concerns the splitting of the glioblastoma cohort into 2 distinct risk groups according to a continuous factor (MVD value). Flexible modeling of the continuous factor MVD (e.g., by cubic splines or fractional polynomials18, 19) may provide more insight into the underlying risk relation. However, these attempts still may fail when the factor interacts with therapy. This leads to an important distinction between a prognostic versus a predictive factor. We define a prognostic factor as a marker that has an association with some clinical outcome, typically a time-to-event outcome (e.g., survival).20 We define a predictive factor as a marker that is used to make specific choices between treatment options.20 In our Glioblastoma Series 1, the therapy protocols were heterogeneous; and, in Glioblastoma Series 2, 2 therapy arms were used that were not identical to the therapy arms used in the retrospective series. Therefore, we could analyze the prognostic influence of MVD and vascular patterns in both glioblastoma series; however, the relative predictive influence of these parameters could be assessed only in Glioblastoma Series 2 (in which 2 well defined therapy arms were used; whereas, in Glioblastoma Series 1, heterogeneous therapies were administered). Analyzing the predictive effect in Glioblastoma Series 2, we could not show any difference between the 2 therapy arms with regard to MVD or vascular patterns.

Validity

Glioblastomas usually present with a regionally heterogeneous vascularization. Therefore, it remains uncertain whether the hot-spot method for MVD assessment and the algorithm for the assessment of vascular patterns that were used in the current study indeed are representative of these parameters with regard to the whole tumor: Both markers are measured on tissue sections and may not represent the 3-dimensional structure of the vascular network. Variation of MVD values in subsequent assessments may be reduced by the use of alternative hot-spot techniques, such as multihot-spot assessment,21 or strict stereologic principles,22, 23 or the Chalkley method of MVD assessment.24, 25 In addition, the development of algorithms for automated scanning microscopy may be a feasible approach to reduce interobserver variability in the assessment of MVD in malignant glioma.26, 27

In summary, the current data showed that poor observer agreement limits the clinical utility of histopathologically assessed hot-spot MVD and vascular patterns as prognostic factors in patients with glioblastoma. Nevertheless, in our opinion, the assessment of glioblastoma vascularization still harbors potential clinical relevance. However, the challenge for neurooncologists is to identify more appropriate methodologies for the assessment of valid vascular parameters. Such methodologies may include, for example, the use of stereologic methods on tissue specimens22, 23 or neuroradiologic in vivo visualization of whole tumor vascularization by ligands directed against endothelium (e.g., lectin nanoparticles).28 The assessment of vascular parameters by neuromorphologists will remain important and cannot be abolished. However, all efforts must be made to reduce interobserver variation. Tight cooperation and interaction between medical experts, systematic training, and the implementation of telemedical applications29, 30 may be helpful.

Acknowledgements

We thank Helga Flicker and Gerda Ricken for expert technical assistance; and are grateful to Drs. Herbert Budka, Roger Stupp, Karl Rossler, Karin Dieckmann, and Inge Fischer for their continuous support and critical evaluations of the article.

REFERENCES

1 Stupp R, Mason WP, van den Bent MJ, et al. Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. N Engl J Med. 2005; 352: 987–996.
10.1056/NEJMoa043330
CASPubMedWeb of Science®Google Scholar
2 Hegi ME, Diserens AC, Godard S, et al. Clinical trial substantiates the predictive value of O-6-methylguanine-DNA methyltransferase promoter methylation in glioblastoma patients treated with temozolomide. Clin Cancer Res. 2004; 10: 1871–1874.
10.1158/1078-0432.CCR-03-0384
CASPubMedWeb of Science®Google Scholar
3 Hegi ME, Diserens AC, Gorlia T, et al. MGMT gene silencing and benefit from temozolomide in glioblastoma. N Engl J Med. 2005; 352: 997–1003.
10.1056/NEJMoa043331
CASPubMedWeb of Science®Google Scholar
4 Lamborn KR, Chang SM, Prados MD. Prognostic factors for survival of patients with glioblastoma: recursive partitioning analysis. Neurooncol. 2004; 6: 227–235.
PubMedWeb of Science®Google Scholar
5 Leon SP, Folkerth RD, Black PM. Microvessel density is a prognostic indicator for patients with astroglial brain tumors. Cancer. 1996; 77: 362–372.
10.1002/(SICI)1097-0142(19960115)77:2<362::AID-CNCR20>3.0.CO;2-Z
CASPubMedWeb of Science®Google Scholar
6 Birner P, Piribauer M, Fischer I, et al. Vascular patterns in glioblastoma influence clinical outcome and associate with variable expression of angiogenic proteins: evidence for distinct angiogenic subtypes. Brain Pathol. 2003; 13: 133–143.
10.1111/j.1750-3639.2003.tb00013.x
CASPubMedWeb of Science®Google Scholar
7 Scholten AN, Smit VT, Beerman H, van Putten WL, Creutzberg CL. Prognostic significance and interobserver variability of histologic grading systems for endometrial carcinoma. Cancer. 2004; 100: 764–772.
10.1002/cncr.20040
PubMedWeb of Science®Google Scholar
8 Marie D, Liu Y, Moore SA, et al. Interobserver variability associated with the MIB-1 labeling index: high levels suggest limited prognostic usefulness for patients with primary brain tumors. Cancer. 2001; 92: 2720–2726.
10.1002/1097-0142(20011115)92:10<2720::AID-CNCR1626>3.0.CO;2-Z
PubMedWeb of Science®Google Scholar
9 Lytwyn A, Salit IE, Raboud J, et al. Interobserver agreement in the interpretation of anal intraepithelial neoplasia. Cancer. 2005; 103: 1447–1456.
10.1002/cncr.20927
CASPubMedWeb of Science®Google Scholar
10 Chhieng DC, Talley LI, Roberson J, Gatscha RM, Jhala NC, Elgert PA. Interobserver variability: comparison between liquid-based and conventional preparations in gynecologic cytology. Cancer. 2002; 96: 67–73.
10.1002/cncr.10477
PubMedWeb of Science®Google Scholar
11 Bottini A, Berruti A, Bersiga A, et al. Changes in microvessel density as assessed by CD34 antibodies after primary chemotherapy in human breast cancer. Clin Cancer Res. 2002; 8: 1816–1821.
CASPubMedWeb of Science®Google Scholar
12 Weidner N, Semple JP, Welch WR, Folkman J. Tumor angiogenesis and metastasis—correlation in invasive breast carcinoma. N Engl J Med. 1991; 324: 1–8.
10.1056/NEJM199101033240101
CASPubMedWeb of Science®Google Scholar
13 McCarthy NJ, Yang X, Linnoila IR, et al. Microvessel density, expression of estrogen receptor alpha, MIB-1, p53, and c-erbB-2 in inflammatory breast cancer. Clin Cancer Res. 2002; 8: 3857–3862.
CASPubMedWeb of Science®Google Scholar
14 Preusser M, Birner P, Hainfellner JA. Algorithm for the standardized assessment of vascular patterns in glioblastoma specimens. Clin Neuropathol. 2004; 23: 238–240.
CASPubMedWeb of Science®Google Scholar
15 Bast RC Jr., Lilja H, Urban N, et al. Translational crossroads for biomarkers. Clin Cancer Res. 2005; 11: 6103–6108.
10.1158/1078-0432.CCR-04-2213
PubMedWeb of Science®Google Scholar
16 Ingeholm P, Pedersen L, Holck S. Quantification of microvessel density of breast carcinoma: an assessment of the inter- and intraobserver variation. Breast. 1999; 8: 251–256.
10.1054/brst.1999.0057
CASPubMedWeb of Science®Google Scholar
17 Hansen S, Grabau DA, Rose C, Bak M, Sorensen FB. Angiogenesis in breast cancer: a comparative study of the observer variability of methods for determining microvessel density. Lab Invest. 1998; 78: 1563–1573.
CASPubMedWeb of Science®Google Scholar
18 Sauerbrei W, Royston P, Bojar H, Schmoor C, Schumacher M. Modelling the effects of standard prognostic factors in node-positive breast cancer. German Breast Cancer Study Group (GBSG). Br J Cancer. 1999; 79 ( 11–12): 1752–1760.
10.1038/sj.bjc.6690279
CASPubMedWeb of Science®Google Scholar
19 Heinzl H, Kaider A. Gaining more flexibility in Cox proportional hazards regression models with cubic spline functions. Comput Methods Programs Biomed. 1997; 54: 201–208.
10.1016/S0169-2607(97)00043-6
CASPubMedWeb of Science®Google Scholar
20 McShane LM, Altman DG, Sauerbrei W, Taube SE, Gion M, Clark GM. REporting recommendations for tumour MARKer prognostic studies (REMARK). Eur J Cancer. 2005; 41: 1690–1696.
10.1016/j.ejca.2005.03.032
PubMedWeb of Science®Google Scholar
21 Sharma S, Sharma MC, Sarkar C. Morphology of angiogenesis in human cancer: a conceptual overview, histoprognostic perspective and significance of neoangiogenesis. Histopathology. 2005; 46: 481–489.
10.1111/j.1365-2559.2005.02142.x
CASPubMedWeb of Science®Google Scholar
22 Gundersen HJ, Bendtsen TF, Korbo L, et al. Some new, simple and efficient stereological methods and their use in pathological research and diagnosis. APMIS. 1988; 96: 379–394.
10.1111/j.1699-0463.1988.tb05320.x
CASPubMedWeb of Science®Google Scholar
23 Nyengaard JR, Bendtsen TF, Gundersen HJ. Stereological estimation of the number of capillaries, exemplified by renal glomeruli. APMIS Suppl. 1988; 4: 92–99.
CASPubMedGoogle Scholar
24 Hansen S, Grabau DA, Sorensen FB, Bak M, Vach W, Rose C. The prognostic value of angiogenesis by Chalkley counting in a confirmatory study design on 836 breast cancer patients. Clin Cancer Res. 2000; 6: 139–146.
CASPubMedWeb of Science®Google Scholar
25 Hansen S, Sorensen FB, Vach W, Grabau DA, Bak M, Rose C. Microvessel density compared with the Chalkley count in a prognostic study of angiogenesis in breast cancer patients. Histopathology. 2004; 44: 428–436.
10.1111/j.1365-2559.2004.01848.x
CASPubMedWeb of Science®Google Scholar
26 Fox SB, Leek RD, Weekes MP, Whitehouse RM, Gatter KC, Harris AL. Quantitation and prognostic value of breast cancer angiogenesis: comparison of microvessel density, Chalkley count, and computer image analysis. J Pathol. 1995; 177: 275–283.
10.1002/path.1711770310
CASPubMedWeb of Science®Google Scholar
27 Cruz D, Valenti C, Dias A, Seixas M, Schmitt F. Microvessel density counting in breast cancer. Slides versus digital images. Anal Quant Cytol Histol. 2001; 23: 15–20.
CASPubMedWeb of Science®Google Scholar
28 Paschkunova-Martic I, Kremser C, Mistlberger K, et al. Design, synthesis, physical and chemical characterisation, and biological interactions of lectin-targeted latex nanoparticles bearing Gd-DTPA chelates: an exploration of magnetic resonance molecular imaging (MRMI). Histochem Cell Biol. 2005; 123: 283–301.
10.1007/s00418-005-0780-7
CASPubMedWeb of Science®Google Scholar
29 Dietel M, Nguyen-Dobinsky TN, Hufnagl P. The UICC Telepathology Consultation Center. International Union Against Cancer. A global approach to improving consultation for pathologists in cancer diagnosis. Cancer. 2000; 89: 187–191.
10.1002/1097-0142(20000701)89:1<187::AID-CNCR25>3.0.CO;2-X
CASPubMedWeb of Science®Google Scholar
30 Callas PW, Leslie KO, Mattia AR, et al. Diagnostic accuracy of a rural live video telepathology system. Am J Surg Pathol. 1997; 21: 812–819.
10.1097/00000478-199707000-00009
CASPubMedWeb of Science®Google Scholar

Citing Literature

Volume107, Issue1

1 July 2006

Pages 162-170

Histopathologic assessment of hot-spot microvessel density and vascular patterns in glioblastoma: Poor observer agreement limits clinical utility as prognostic factors

A translational research project of the European Organization for Research and Treatment of Cancer Rrain Tumor Group